Datasets:
File size: 253,257 Bytes
0f2f2d3 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 | {
"v1_col_introduction": "introduction : To function correctly, a cell needs a real-time assessment of its environment. A key component of this running tally is cell density \u2013 a parameter known to alter both the rates of cell proliferation and the level of cell differentiation. Loss of cell density awareness has long been associated with the onset of cancer and the ability of malignant cells to overgrow the monolayer when grown in cell culture(Aaronson & Todaro 1968). Similarly, in normal cells the slowdown in growth as cells become dense is associated with the induction of differentiated function(Kaldis & Richardson 2012). Despite its importance to normal function, how a cell \u201csees\u201d its neighbors remains an enigma. The sensing mechanism is known to be influenced by high concentrations of growth factors, for instance high serum levels, but this has not shed light on how a cell actually detects the presence of its neighbors(Pardee 1987; Vogel et al. 1980). Therefore, to better comprehend tissue morphogenesis and stability requires an understanding of how cells signal their presence to each other. Tendon morphogenesis is highly dependent on cell density signaling. This is because the tissue is over 90% type I collagen and the two parameters controlling local collagen deposition \u2013 the rate of cellular collagen production and the cell number \u2013 are cell density regulated(Rowe & Schwarz 1983; Schwarz & Bissell 1977). Cell density regulation of cell proliferation and apoptosis causes a growth plate to form, allowing the cells to deposit an even distribution of collagen along the longitudinal axis of the fibril \u2013 a hallmark of this tissue(Schwarz 1996). In cell culture primary avian tendon (PAT) cells from 16 day embryos devote ~50% of their total protein production to procollagen(Schwarz et al. 1979). This high level is identical to that seen in ovo where tendon development occurs rapidly (~ 11 days) enabling the newly hatched chick the ability to walk. To produce high levels of procollagen from a single copy gene and allow rapid regulation puts restrictions on where this pathway can be controlled. Transcription is an unlikely candidate because induction is slow from a single copy gene requiring 3 days to fully induce procollagen mRNA levels in PAT cells(Rowe & Schwarz 1983). Moreover, the procollagen mRNA is stable (24h half-life) so returning to the uninduced state can take over 2 days(Lyons & Schwarz 1984). Instead, PAT cells regulate procollagen at a posttranslational step. Translation and secretion rates are both tied to formation of a triple helical molecule(Rowe & Schwarz 1983; Schwarz 1985) and this in turn requires hydroxylation of prolines to stabilize this conformation. The enzyme, prolyl hydroxlase, responsible for this rapid regulatory control has two subunits and the level of the alpha subunit is dependent on cell density(Kao et al. 1985; Lee et al. 2001). To begin to understand cell density regulation, PAT cells were grown as a 6 mm island in the middle of a 60 mm dish(Schwarz 1991). The cells in middle of the island grow to be confluent while cells at the edge of the island would grow outwards to be at low cell density. In this way, cells at multiple cell densities could be studied at the same time. Indeed, when one probed the levels of procollagen mRNA by in situ hybridization, the level dramatically increased from the edge of the island to the confluent center(Schwarz 1991). Growing cells as an island was useful but it made a change that turned out to be much more significant: the ratio of medium to cells was increased 100 fold. While gently agitating PAT cells confluent over the whole dish had no affect, gently agitating PAT cells grown as an island caused dramatic changes by increasing cell proliferation and decreasing procollagen production(Schwarz 1991; Zayas & Schwarz 1992). The reason for this difference is that cells confluent over the whole dish could condition the medium such that shaking only increased the rate at which the factor exchanged locations between the medium and the cell layer and not its concentration. When grown as an island of cells, there was too much medium and the factor released from the cell layer was basically lost. The lower concentration of this factor strongly altered the ability of PAT cells to\n7\n8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54\nPeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013)\nR ev ie w in g M an\nus cr ip t\nsense the cells around them. Adding shaken conditioned medium (CM) from confluent cultures would strongly stimulate an island of PAT cells to grow and this became an assay to purify this factor. Using 400 liters of CM as starting material, this was serially applied to 4 columns. Analyzing the 30 fractions from the last column (reverse phase), one set was run on SDS gels and the second set was analyzed for biological activity. The intensity of the band running at 16 kD tracked best with activity. This band was sequenced and yielded: EPLAVVDLTEKTIS(Schwarz 2002). This chicken sequence does not have an immediate correspondence to the human equivalent for two reasons. First, as with all N-terminal Edman sequence data, with each cycle the noise level increases reducing the confidence of each additional read. Not unexpectedly, the chicken genome shows an exact homology only with the first 8 AA. Second, the human genome has a variant in the fifth position from V to E. Nevertheless, this 8 letter AA sequence in the chicken comes from a much larger gene with a gene product of 424 AA from seven exons. The homology between the chicken and the human genes at the amino acid level is 89% identical and 94% positive. For this large gene to be correct requires further processing of the protein product to a smaller form: a specific cleavage (there is no methionine for an alternative start site) at the EPLAVVDL site would give the right starting sequence; however, this sequence is found near the carboxy end and would only produce a protein with 94 AA, ~10 kD. So to run at 16 kD on SDS gels would require that the protein be modified or that the protein bind a cofactor. In addition the 94 AA cleavage product does not contain highly hydrophobic regions that could facilitate the secretion to the outside of the cell. In this paper, a candidate gene for the diffusible cell density signal is tagged with myc and his tags and ectopically expressed in another cell type from another species. We probe whether the expressed gene gets cleaved, whether the cleaved product gets secreted, and the modification or cofactor that allows a 94 AA protein run at 16 kD. Finally, we want to test whether the CM containing the recombinant factor when applied to an island of PAT cells acts like the original diffusible cell density signal. This classic test is complicated by the finding that the cell type producing the 94 AA protein binds a unique tissue-specific lipid cofactor and this active cell density signal imparts a tissue-specific response. Material and Methods",
"v2_col_introduction": "introduction : To function correctly, a cell needs a real-time assessment of its environment. A key component of this running tally is cell density \u2013 a parameter known to alter both the rates of cell proliferation and the level of cell differentiation. Loss of cell density awareness has long been associated with the onset of cancer and the ability of malignant cells to overgrow the monolayer when grown in cell culture(Aaronson & Todaro 1968). Similarly, in normal cells the slowdown in growth as cells become dense is associated with the induction of differentiated function(Kaldis & Richardson 2012). Despite its importance to normal function, how a cell \u201csees\u201d its neighbors remains an enigma. The sensing mechanism is known to be influenced by high concentrations of growth factors, for instance high serum levels, but this has not shed light on how a cell actually detects the presence of its neighbors(Pardee 1987; Vogel et al. 1980). Therefore, to better comprehend tissue morphogenesis and stability requires an understanding of how cells signal their presence to each other. Tendon morphogenesis is highly dependent on cell density signaling. This is because the tissue is over 90% type I collagen and the two parameters controlling local collagen deposition \u2013 the rate of cellular collagen production and the cell number \u2013 are cell density regulated(Rowe & Schwarz 1983; Schwarz & Bissell 1977). Cell density regulation of cell proliferation and apoptosis causes a growth plate to form, allowing the cells to deposit an even distribution of collagen along the longitudinal axis of the fibril \u2013 a hallmark of this tissue(Schwarz 1996). In cell culture primary avian tendon (PAT) cells from 16 day embryos devote ~50% of their total protein production to procollagen(Schwarz et al. 1979). This high level is identical to that seen in ovo where tendon development occurs rapidly (~ 11 days) enabling the newly hatched chick the ability to walk. To produce high levels of procollagen from a single copy gene and allow rapid regulation puts restrictions on where this pathway can be controlled. Transcription is an unlikely candidate because induction is slow from a single copy gene requiring 3 days to fully induce procollagen mRNA levels in PAT cells(Rowe & Schwarz 1983). Moreover, the procollagen mRNA is stable (24h half-life) so returning to the uninduced state can take over 2 days(Lyons & Schwarz 1984). Instead, PAT cells regulate procollagen at a posttranslational step. Translation and secretion rates are both tied to formation of a triple helical molecule(Rowe & Schwarz 1983; Schwarz 1985) and this in turn requires hydroxylation of prolines to stabilize this conformation. The enzyme, prolyl hydroxlase, responsible for this rapid regulatory control has two subunits and the level of the alpha subunit is dependent on cell density(Kao et al. 1985; Lee et al. 2001). To begin to understand cell density regulation, PAT cells were grown as a 6 mm island in the middle of a 60 mm dish(Schwarz 1991). The cells in middle of the island grow to be confluent while cells at the edge of the island would grow outwards to be at low cell density. In this way, cells at multiple cell densities could be studied at the same time. Indeed, when one probed the levels of procollagen mRNA by in situ hybridization, the level dramatically increased from the edge of the island to the confluent center(Schwarz 1991). Growing cells as an island was useful but it made a change that turned out to be much more significant: the ratio of medium to cells was increased 100 fold. While gently agitating PAT cells confluent over the whole dish had no affect, gently agitating PAT cells grown as an island caused dramatic changes by increasing cell proliferation and decreasing procollagen production(Schwarz 1991; Zayas & Schwarz 1992). The reason for this difference is that cells confluent over the whole dish could condition the medium such that shaking only increased the rate at which the factor exchanged locations between the medium and the cell layer and not its concentration. When grown as an island of cells, there was too much medium and the factor released from the cell layer was basically lost. The lower concentration of this factor strongly altered the ability of PAT cells to\n7\n8 9\n10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54\nPeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013)\nR ev ie w in g M an\nus cr ip t\nsense the cells around them. Adding shaken conditioned medium (CM) from confluent cultures would strongly stimulate an island of PAT cells to grow and this became an assay to purify this factor. Using 400 liters of CM as starting material, this was serially applied to 4 columns. Analyzing the 30 fractions from the last column (reverse phase), one set was run on SDS gels and the second set was analyzed for biological activity. The intensity of the band running at 16 kD tracked best with activity. This band was sequenced and yielded: EPLAVVDLTEKTIS(Schwarz 2002). This chicken sequence does not have an immediate correspondence to the human equivalent for two reasons. First, as with all N-terminal Edman sequence data, with each cycle the noise level increases reducing the confidence of each additional read. Not unexpectedly, the chicken genome shows an exact homology only with the first 8 AA. Second, the human genome has a variant in the fifth position from V to E. Nevertheless, this 8 letter AA sequence in the chicken comes from a much larger gene with a gene product of 424 AA from seven exons. The homology between the chicken and the human genes at the amino acid level is 89% identical and 94% positive. For this large gene to be correct requires further processing of the protein product to a smaller form: a specific cleavage (there is no methionine for an alternative start site) at the EPLAVVDL site would give the right starting sequence; however, this sequence is found near the carboxy end and would only produce a protein with 94 AA, ~10 kD. So to run at 16 kD on SDS gels would require that the protein be modified or that the protein bind a cofactor. In addition the 94 AA cleavage product does not contain highly hydrophobic regions that could facilitate the secretion to the outside of the cell. In this paper, a candidate gene for the diffusible cell density signal is tagged with myc and his tags and ectopically expressed in another cell type from another species. We probe whether the expressed gene gets cleaved, whether the cleaved product gets secreted, and the modification or cofactor that allows a 94 AA protein run at 16 kD. Finally, we want to test whether the CM containing the recombinant factor when applied to an island of PAT cells acts like the original diffusible cell density signal. This classic test is complicated by the finding that the cell type producing the 94 AA protein binds a unique tissue-specific lipid cofactor and this active cell density signal imparts a tissue-specific response. 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83\nPeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013)\nR ev ie w in g M an\nus cr ip t",
"v3_col_introduction": "introduction : To function correctly, a cell needs a real-time assessment of its environment. A key component of this running tally is cell density \u2013 a parameter known to alter both the rates of cell proliferation and the level of cell differentiation. Loss of cell density awareness has long been associated with the onset of cancer and the ability of malignant cells to overgrow the monolayer when grown in cell culture(Aaronson & Todaro 1968). Similarly, in normal cells the slowdown in growth as cells become dense is associated with the induction of differentiated function(Kaldis & Richardson 2012). Despite its importance to normal function, how a cell \u201csees\u201d its neighbors remains an enigma. The sensing mechanism is known to be influenced by high concentrations of growth factors, for instance high serum levels, but this has not shed light on how a cell actually detects the presence of its neighbors(Pardee 1987; Vogel et al. 1980). Therefore, to better comprehend tissue morphogenesis and stability requires an understanding of how cells signal their presence to each other. Tendon morphogenesis is highly dependent on cell density signaling. This is because the tissue is over 90% type I collagen and the two parameters controlling local collagen deposition \u2013 the rate of cellular collagen production and the cell number \u2013 are cell density regulated(Rowe & Schwarz 1983; Schwarz & Bissell 1977). Cell density regulation of cell proliferation and apoptosis causes a growth plate to form, allowing the cells to deposit an even distribution of collagen along the longitudinal axis of the fibril \u2013 a hallmark of this tissue(Schwarz 1996). In cell culture primary avian tendon (PAT) cells from 16 day embryos devote ~50% of their total protein production to procollagen(Schwarz et al. 1979). This high level is identical to that seen in ovo where tendon development occurs rapidly (~ 11 days) enabling the newly hatched chick the ability to walk. To produce high levels of procollagen from a single copy gene and allow rapid regulation puts restrictions on where this pathway can be controlled. Transcription is an unlikely candidate because induction is slow from a single copy gene. When PAT cells are stimulated to make high levels of collagen, there is a 12h lag before the cells linearly increase the procollagen mRNA levels over a 2.5 day period(Rowe & Schwarz 1983). Moreover, to maintain the pool of procollagen mRNA, it has to be stable with a half-life of 24h(Lyons & Schwarz 1984). So manipulating procollagen mRNA levels is not feasible when the cells are required to make high levels of procollagen. Instead the cell regulates procollagen at a post-translational step. Translation and secretion rates are both tied to formation of a triple helical molecule and this in turn requires hydroxylation of prolines to stabilize this conformation. This can be manipulated by the addition of ascorbate whose only known role in the process is being a reducing agent and this rapidly (<30 min) induces a 6 fold increase in procollagen secretion(Schwarz 1985) and a 2 fold increase in procollagen translation (rates reach 6 fold but require a partial increase in mRNA)(Rowe & Schwarz 1983). In the fully induced state, inhibiting prolyl hydroxylase by chelating ferrous ion causes a rapid drop in procollagen translation (>50% in 2h)(Rowe & Schwarz 1983). The enzyme, prolyl hydroxlase, responsible for this rapid regulatory control requires ascorbate to keep its catalytic ferrous ion in a reduced state. The enzyme has two subunits and the level of the alpha subunit is dependent on cell density(Lee et al. 2001). How high cell density induces prolyl hydroxylase has not been defined except that the pool of this enzyme increases 5-6 fold while the level of its mRNA remains unchanged(Lee et al. 2001).\n2\n1 2 3 4 5 6 7 8 9\n10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46\n12 34 56 78 91011 12 13\nPeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013)\nR ev ie w in g M an\nus cr ip t\nWith 5 cysteines the alpha subunit of prolyl hydroxylase would be sensitive to the redox potential of the cell and this has been postulated as a mechanism for signal transduction(Lee et al. 2001). Our long-term goal is to resolve how cells sense cell density and then tie this signaling to the induction of prolyl hydroxylase. To begin to understand cell density regulation, PAT cells were grown as a 6 mm island in the middle of a 60 mm dish(Schwarz 1991). The cells in middle of the island grow to be confluent while cells at the edge of the island would grow outwards to be at low cell density. In this way, cells at multiple cell densities could be studied at the same time. Indeed, when one probed the levels of procollagen mRNA by in situ hybridization, the level dramatically increased from the edge of the island to the confluent center(Schwarz 1991). Growing cells as an island was useful but it made a change that turned out to be much more significant: the ratio of medium to cells was increased 100 fold. While gently agitating PAT cells confluent over the whole dish had no affect, gently agitating PAT cells grown as an island caused a dramatic change in both cell proliferation and procollagen production(Schwarz 1991; Zayas & Schwarz 1992). The reason for this difference is that cells confluent over the whole dish could condition the medium such that shaking only increased the rate at which the factor exchanged locations between the medium and the cell layer and not its concentration. When grown as an island of cells, there was too much medium and the factor released from the cell layer was basically lost. The lower concentration of this factor strongly altered the ability of PAT cells to sense the cells around them. To identify the molecular nature of this factor is the focus of this paper. However, one needs to point out that the complexity of the cell density response requires two factors. The one described and a second that is not diffusible but interacts with the first and changes the cellular response(Schwarz 1996). With the diffusible factor there was an assay and so its molecular characterization was approached first. In this paper, a candidate gene for the diffusible cell density signal is put through the classic test to validate its activity: ectopic expression in another cell type from another species, purification of the recombinant factor, and testing back on PAT cells to see if it acts like the original diffusible cell density signal. This classic approach is complicated by the finding that the cell type producing this protein binds a unique tissue-specific lipid cofactor and this composite molecule imparts a tissue-specific response.",
"v1_text": "results : Expression in U2OS cells causes differentiation A replication defective retroviral vector containing a neomycin resistance gene was used to transfect the chicken gene plus tags into U2OS cells(Koh et al. 2002; Lee 2007). Because this study was started when the chicken genome was first sequenced, the chicken equivalent to the first exon was not present. So the insert contained only the last six exons (pESY1c). U2OS cells, derived from a human osteosarcoma tumor, showed stable expression of this gene after infection when grown in low serum media (0.2%)(Valmassoi & Schwarz 1988). Despite the low serum levels, U2OS cells transfected with pESY1c would start to overgrow the monolayer as small ridges. The ridges would coalesce into circles sharing boundaries with other circles that resembled spongy bone in a two dimensional space (Fig. 2b). Uninfected U2OS cells showed some ability to overgrow the monolayer but could not maintain the growth needed to form completed structures (Fig.2a). The multicell ridges produced alkaline phosphophatase on the cell surface, indicating increased expression of the osteoblast phenotype (Fig. 2c). Pattern of molecular expression intracellularly and extracellularly A monoclonal myc antibody and Western blotting was used to probe the expression and processing of this myc tagged gene. Two specific bands were seen intracellularly: ~44 kD band which is expected for the full size gene product; ~26 kD band that could be a cleavage product (Fig. 3). However, this band is larger that the predicted 20 kD protein \u2013 16 kD + tags. To observe the secreted forms a simple purification was used whereby the CM was put through a 30 kD ultrafilter. Both the retentate and the filtrate were analyzed and both had to be further concentrated on a 10 kD ultrafilter in order to load sufficient protein on the gel for Western analysis (Fig. 3). The retentate contained many bands: the smallest was a weak band at ~28 kD, the strongest was at ~60 kD, and the largest was over 200 kD. The filtrate showed basically a single band at ~60 kD. This was unusual since 60 kD proteins are excluded to a high degree from passing through a 30 kD ultrafilter. One explanation is that by concentrating the sample ~20 fold causes the tagged protein to aggregate into stable forms that resist the denaturing effects of heat, \u03b2-mercaptoethanol, and SDS. aggregation is due to a cofactor : The multiple bands seen in the Western blot complicates the interpretation of how this tagged protein is working in the extracellular space. But more specifically, even the smallest MW band (~28 kD) seen outside the cell is higher than predicted. Is this larger size due to a different amount or type of modification in bone cells versus tendon cells? Or is this increase in apparent MW a reflection of an interaction with another molecule? To begin to answer these questions, we needed to resolve how the unmodified secreted form of the protein factor with two tags (94 AA + tags) migrates on a Western. To accomplish this, the tagged protein was expressed in E. coli, purified on a nickel column, and then used as a standard in Western blots. The tagged secreted form of the protein ran with a MW of 16 kD (Fig. 4, lane 6). This confirms that there is either a covalent modification or a binding to another molecule that gives rise to the 28 kD form in U2OS cells. Further aggregation could then occur when concentrating the factor to give rise to larger MW forms. This raises the question of whether different physical or chemical environments could alter the size of the bands observed on Western blots. CM from U2OS-pESY1c was put through a 30 kD ultrafilter and the filtrate was then concentrated on a 10 kD ultrafilter. This was divided into 2 parts. One was further concentrated on a 10 kD ultrafilter for loading on a gel but in this case, 30% DMSO was added. To the other sample, salt was added to raise the concentration by 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t 0.5M; concentrated with an ultrafilter; then diluted to low salt; and then concentrated to apply to a gel (Fig. 4). The effect of DMSO was to reduce the number of bands, especially those at high MW, but the smallest band still remained 28 kD. In contrast, the treatment with high salt, low salt and ultrafiltration yielded a much simpler picture where the bands were reduced to the 16 kD band and its dimer at 32 kD. One can conclude that the 28 kD band is due to a binding of a cofactor and not a covalent modification. In addition, when the cofactor is separated from the secreted form of the protein factor, the MW of the protein is the expected 16 kD (94AA + tags). This shows that U2OS cells are capable of cleaving the large gene product to the 94 AA + tags and then secrete it. Localization and characterization of the factor/cofactor in the extracellular space The previous experiment raises the question of how the untagged 16 kD form in tendon cells compares to the 28 kD tagged form seen in bone cells. Even taking into account a 4 kD tag, there is an additional 8 kD shift in MW. Are these forms variations with common characteristics or is the bone form acting in a different manner? In tendon cells the 16 kD form is loosely bound to the cell layer. This gives a key feature to this form in that gently shaking a flask of tendon cells causes them to loose their ability to sense high cell density, resulting in increased proliferation. First, the localization was examined with immunofluorescence using a rabbit polyclonal antibody specific to the tagged secreted form of the chicken protein expressed in E. coli. During tests to optimize the signal, procedures that disrupted membranes decreased the signal. Using an optimized protocol, immunofluorescence staining was localized to areas where the cells have overgrown the monolayer (hills) and expression was suppressed in monolayer areas (valleys) (Fig. 5a). Second, the cells were subjected to gentle shaking (100 rpm) for 3h per day with the medium changed each hour. After 3 days the hills had spread into the valleys and the valleys had additional growth. The net effect was to even out the growth over the flask (Fig. 5c). One can conclude that some of the major physical properties of the bone and tendon factor/cofactor are the same. They are both localized to the cell layer \u2013 most likely bound to the cell membrane. However, this interaction is not strong, allowing significant diffusion that can be enhanced by gentle agitation of the medium. Characterization of the bone cofactor with degradative enzymes The cofactor is tightly bound to the factor and gives a consistent mobility pattern on Western blots. This pattern could then be used to test whether specific degradative enzymes could cleave the cofactor, alter the migration profile, and reveal the partial composition of the cofactor. One possible candidate for the cofactor would be a variation on the anchoring structure for membrane bound proteins: a glycosylphosphatidylinositol lipid(Paulick & Bertozzi 2008). To test for a glycolipid, the factor/cofactor was treated with a variety of polysaccharide degradative enzymes \u2013 chondroitinase ABC, hyaluronidase, heparinase \u2013 and the pattern on a Western blot was compared to that of an untreated control. None of the treated samples showed a distinct change in the pattern (data not shown). At the same time, a variety of lipid degrading enzymes were tested \u2013 lipase, phospholipase A2, phospholipase C, and sphingomyelinase. Lipase, phospholipase C, and sphingomyelinase caused dramatic shifts in the pattern seen on Westerns (Fig 6). Lipase was the most dramatic causing bands to shift to lower MW including the 16 kD, free protein form. The lipase was a crude preparation used at a high concentration; so while its primary target would be a glycerol fatty acid linkage in the cofactor, this is probably not the only cleavage product. The phospholipase C used in this experiment has overlapping activities with the sphingomyelinase. The sphingomyelinase is more specific to cleaving moities on ceramide phosphates. In contrast, phospholipase A2 did not show a significant change in mobility or 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t intensity (data not shown). A negative result is hard to interpret because there are too many potential causes. The positive results, while not yielding a definitive structure, implicate lipid components containing a phosphoceramide and a glycerol fatty acid as part of the cofactor. The lipid nature of the cofactor can explain why the factor/cofactor has a strong tendency to aggregate when concentrated due to hydrophobic interactions. Purification and further characterization of the bone cofactor using mass spectrometry In order to characterize the bone cofactor in more detail, procedures were needed to separate both the factor from the cofactor and then the cofactor from other lipid-like molecules. The first step in the purification \u2013 the separation of the cofactor from the factor \u2013 could benefit from standard lipid procedures using a methanol/chloroform extraction. In this procedure the lipids are found in the chloroform layer and proteins are found as a precipitate between the layers. There is no assay for the cofactor but the proteins in the precipitate can be solubilized and then Western analysis can be used to see if the protein factor runs as the free form (16 kD) or as a complex with the cofactor (28 kD). The amount of free factor is an indicator of cofactor release. However, redissolving the protein precipitates at the interface and running Western blots only showed the 28 kD form and higher order aggregates (Fig. 7). In this extraction protocol, the cofactor remained bound to the factor, and as a consequence, precipitated at the interface. Despite attempts to weaken the interaction between the factor and the cofactor by changes in pH, salt, solvent, chaotropic agents, temperature, and chelating agents, Western blots showed little or no conversion to the free form (data not shown). The only successful method required a combination of a shift in salt concentration and ultrafiltration (Fig. 4). Using a simplified version of this procedure whereby the CM was made high in salt (~0.6 M) and then concentrated on a 10 kD ultrafilter. This was then diluted 10 fold with water and concentrated to run on Westerns (Fig. 8). By intensity of the bands, most of the tagged factor runs as the free form. Thus, the corresponding free cofactor was expected to have passed through the ultrafilter as part of the filtrate. Using a small C18 resin column to concentrate, differential solvent elution to purify, and then comparing the untreated and the enzyme treated samples via mass spectrometry one can identify candidate cofactor peaks (Fig. 9). Only molecular weights between 530 and 580 are shown because there was no significant differences between the samples at either higher or lower values (mass range = m/z 300 \u2013 m/z 1500). Two doubly-charged ions at m/z 543.4273 and m/z 557.4429 were dramatically reduced by treatment with either enzyme and these peaks were not observed in the buffer only control and the intensity of the m/z 557 ion was highly correlated with the intensity of the free protein factor in the retentate. The observed difference between the doubly-charged ions was 14 representing a 28 Da difference in MW commonly seen for lipids. In the rest of the analysis, we focused on the 557 species with a MW of 1114.8858. By using a high mass accuracy Q-TOF mass spectrometer we generated potential molecular formulas with a list of atoms commonly found in organic biomolecules (i.e., carbon, hydrogen, nitrogen, oxygen, and phosphorous). The most likely formula C63H123N2O11P (exact mass = 557.4432252+; 0.6 ppm from the measured mass) had nearly a 1:2 Carbon:Hydrogen ratio common to lipids and the nitrogen, oxygen and phosphorous indicative of a hydrophilic region. Support for the proposed cofactor structure derived from MS/MS fragmentation of m/z 557. Two fragment ions, shown in Fig. 10, at m/z of 323 and m/z 378 were consistent with a ceramide phosphate and a linker moiety, supporting the results from the enzymatic reactions. From our analysis, the cofactor was susceptible to three lipid enzymes \u2013 lipase, phospholipase C, and sphingomylenase, and a peak observed at m/z 557 that has a strong lipid composition. This peak was lost when treated with lipase, phopholipase C, and sphingomylenase. Using our knowledge of which enzymes cleave the cofactor and the molecular formula, a 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t working model can be postulated that includes a ceramide phosphate, a glycerol lipid, and a linker consisting of a short modification of a proline residue between them (Fig. 10). Tendon cofactor characterization using enzymes and mass spectrometry Having analyzed the bone factor/cofactor, the focus returned to the original aim of characterizing the tendon factor, but now knowing that it binds a cofactor. The bone and tendon cofactors must be different because they impart a unique shift in mobility to the factor/cofactor complex on SDS gels. This raises important questions about the nature of the tendon cofactor and how it compares to its bone counterpart. As a first step in the characterization, the same panel of lipid degradative enzymes were tested on PAT cell CM. In this case, concentrated CM was used that was over 6 months old. This results in a stable aggregate at ~60 kD being the dominant band. As with bone, lipase and phospholipase C were active on the PAT cell factor/cofactor while phospholipase A2 was inactive (Fig. 11a). Unlike the bone cofactor, sphingomyelinase was inactive on the PAT cell factor/cofactor (Fig. 11b). To insure that the inactivity of the enzyme was not due to the factor/cofactor being in a specific aggregate, the treatment with sphingomylinase was repeated with fresh CM (Fig. 11c) and the result remained unchanged. This indicates a lack of a ceramide phosphate moiety in the tendon cofactor. The cofactor was purified for mass spectrometry by binding it to C18 columns followed by washing with methanol, acetonitrile, and elution with chloroform. Via mass spectrometry we compared preparations that were treated with either phosphospholipase C or lipase while the cofactor was bound to the C18 column to the untreated cofactor samples. No differences were seen in the methanol or acetonitrile washes; however, the chloroform fraction did reveal several peaks that changed between the untreated and treated samples. Two singly charged peaks, at 668.61 and 696.65, were reduced to background levels by the enzyme treatments. Similar to the bone cofactor mass spectrometric analysis, these peaks differed by 28 MW and reflect variation a lipid chain length by two carbons. Additionally, there was no doubly charged ion observed at m/z 557 as was seen in the bone cofactor sample (data not shown). Using MS/MS to further characterize the 668.61 ion yielded a rather simple spectrum with two peaks of roughly equal intensity and with MW of 283.26 and 311.29. These ions differ by 28 and accurate mass measurements correspond to saturated fatty acids with 18 and 20 carbons, respectively.Both chains are required to be in the model in order to achieve the high lipid content indicated by the parent molecule\u2019s MW.Since the band is degraded by phospholipase C, there is a phosphate group yet the total mass of the intact ion is insufficient to include a commonly observed glycerol phosphate. Consequently, our proposed structure includes anhydride linkage of the fatty aciddirectly to the phosphate.The mass difference between the total mass and the fatty acid and phosphate groups corresponds to a cyano group, thus our working model adds a cyano group to the phosphate and is consistent with the enzymology, the mass spectrometry data, and the gel mobility data (Fig. 12). Whether the cyano group was part of the original molecule or it displaced another group during the purification remains to be determined. cells and cell culture : PAT cells were isolated from 16 day chick embryos by a modification of the procedures described by Dehm and Prockop(Dehm & Prockop 1971). Specifically, a higher serum level (3%) was used in the dissociation medium and the use of a nylon mesh (20 micron; TETKO) instead of filter paper to separate out single cells. In these experiments the medium was changed from F12 to a 50/50 mix of F12/DMEM in order to be consistent with the medium used to grow the U2OS cells. The medium contained 0.2% fetal bovine serum deactivated for 30 m at 56o C. U2OS cells (from frozen stocks that have been in the laboratory for over a dozen years(Campeau et al. 2009)) were grown in F12/DMEM medium with 0.2% fetal bovine serum. In 75 cm2 flasks the medium, 50-75 mls, was changed every other day. When the cells first overgrow the monolayer, they are very sensitive to medium volume and will apoptose if too much medium is added. CM was produced by growing cells in 225 cm2 flasks. Medium was changed to serumfree medium and the volume reduced to 12 mls. The flasks were placed in an environmental rotator at 100 rpm, 37o C for 1 h. The CM was collected and new medium added, and the process was repeated for the third hour. After the 3rd collection medium with serum was added (50-75 mls) and the flask returned to the regular incubator. 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t The 16 day chick embryos were exempt from requiring an animal use protocol as determined by LBL Animal Welfare and Research Committee. Similarly, the legs from 5 week mice (BALB/c) came from mice already euthanized for other approved protocols. This is also an exempt use. discussion : In 1976, we showed that freshly isolated chicken embryonic tendon cells were exquisitely sensitive to cell density as reflected in the levels of procollagen production(Schwarz et al. 1976). Despite this dramatic regulation of the most important protein in tendon morphogenesis, the paper went on to say that the \u201cmechanism by which one cell communicates its presence to another cell is unclear.\u201d A few years and many experiments later the signaling mechanism is now shown to use a tight complex between a protein and a lipid. The protein is a small cleavage product from the carboxy end of a larger gene product and this small protein is subsequently secreted from the cell. The protein itself appears to have little or no activity but requires the binding of a lipid cofactor. Remarkably, the composition of the lipid is tissuespecific. While only two types have been characterized, they are each unique in both the lipid structure and the triggered response. SNZR 1P is also unique in composition, processing, and function. By being able to bind unique phospholipids, SNZR 1P allows diffusion into the medium of a lipid that would otherwise be expected to be tightly embedded in the membrane. In earlier papers, this trait was described as being \u201cloosely bound to the cell layer\u201d an explanation for its ability to localize to the cell layer but be partially removable by gentle agitation(Schwarz 1991; Stoker 1973; Zayas & Schwarz 1992). Being able to identify SNZR 1P and its lipid partners gives this complex a molecular identity that explains its unusual characteristics. In addition, as more has become known about the SNZR 1P1L, similarities can be seen to other signaling molecules \u2013 in particular, sonic hedgehog. There is no sequence homology but they are both made as large proteins that are processed to a smaller form and then secreted from the cell(Weed et al. 1997). In the case of sonic hedgehog, it is the N-terminus that is secreted and in this case, it is the C-terminus. The activity of sonic hedgehog is greatly enhanced by the covalent linkage of palmitate and/or cholesterol(Taylor et al. 2001); with the cell density signaling the protein itself has little activity but this is greatly enhanced by binding to a lipid cofactor. Because of the lipid attachments, sonic hedgehog runs as aggregates on SDS gels(Goetz et al. 2006). The binding of the cofactor causes a similar effect on the SNZR 1P. On the other hand, the differences are equally important: SNZR 1P can bind a variety of lipids and in doing so can change its signaling function in a tissue specific manner. Besides how this complicates the analysis of gene function, the result is that the cell has the ability to sense both the type and number of cells in its immediate vicinity and thereby react accordingly. Given the nature of the complex, the simplest hypothesis is that the protein holds the lipid in a specific conformation that can then bind to the cell membrane and organize it(Simons & Toomre 2000). With more cells there would be additional binding resulting in a more organized plasma membrane and this would trigger a stronger signal transduction. But we should point out that our focus has been on the carboxy end of the transcript yet the rest of the protein is equally well conserved between the chick and the human. Another group searching for homology to other tubulin stabilizing proteins found a 20% overlap with this gene(Bartolini et al. 2005). By Northern analysis, many tissues appeared to transcribe this gene at low levels while testis produced it in larger amounts. Over expressing this gene in HeLa cells caused tubulin to be destabilized in contrast to their initial selection criteria. Because they tagged the transcript at the amino terminus, the secreted form would have gone undetected. Moreover, they pointed out that they could not distinguish whether the destabilization of tubulin was a direct 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t or an indirect effect. If the effect on tubulin was an indirect effect of cell density signaling in HeLa cells, then the conservation of sequence could be as simple as conserving the 3-dimensional structure of the cleavage site, or another alternative could be as an intracellular part of the cell density signaling process. The role of the larger, N-terminus protein, will require further analysis to resolve why it is so highly conserved. Our focus has been on molecular characterization of the sensing mechanism for cell density regulation that generates a proliferative signal, but the cell usually requires a more sophisticated response. For instance, in U2OS-pESY1c cells expressing SNZR 1P1L one might postulate that as the cells overgrow the monolayer they begin to produce a secreted inhibitor to prevent the cells in the valleys from producing SNZR 1P1L and dividing: a classic Turing model(Murray 1988; Turing 1952). In tendon cells, more is known and with the knowledge of the cofactor one can adjust a previous model to reveal how cells form a growth plate and then fill a growing tube (fibril) with collagen. To do this the cells need a timer, a growth stimulator, a growth inhibitor and a trigger for apoptosis. Mathematical modeling has shown that this can be accomplished with two factors and the cell(Schwarz 1996). The first factor stimulates growth and the production of the second factor. The second factor, which does not diffuse into the medium; has no activity on its own but by interacting with the first factor changes it to an inhibitor of growth, and a stimulator of procollagen synthesis. Cells left at high cell density cause a drop in the production of the first factor triggering apoptosis. The first factor can now be seen as SNZR 1P2L that we have described. The second factor has until now been hypothetical but a good candidate would be the free cofactor (SNZR 2L). Their interactions occur in the plasma membrane and their consequences are predicted to further organize the membrane structure that in turn can alter the internal state of the cell. In a permissive cell culture environment for tendon cells \u2013 very low serum(0.2%) and addition of ascorbate \u2013 these factors come to the forefront and results in the \u201cpeculiar\u201d behavior of primary cells in culture(Schwarz 1991; Schwarz 1996; Schwarz & Bissell 1977; Schwarz et al. 1976; Schwarz et al. 1979; Valmassoi & Schwarz 1988; Zayas & Schwarz 1992). The cells isolated from 16 day chick embryos were programmed to rapidly make a tendon by way of a growth plate in order that the hatched chick is able to walk to find food and water. In the research laboratory, however, we wanted tendon cells to conform to normal cell culture protocol: grow on a 2-dimensional plastic surface from low density to high density; then be trypsinized off the plastic and the process repeated. Meanwhile the cells were trying to interpret these changes in terms of its own programming for generating a tendon \u2013 a collagen rope -- by regulating collagen production and cell proliferation. As a result we found that: PAT cells only made high levels of procollagen at high cell density (~50% of total protein synthesis). Left at high cell density for several days procollagen production declines. Splitting high density cultures resulted in cells that grew to a lower confluent density and produced a lower level of procollagen. Seeding cells a low cell density resulted in uneven cell growth. Seeding cells at low cell density as a small island in the middle of a dish resulted in the whole island apoptosing after a few hours. Gentle agitation of confluent cells seeded as an island in the middle of dish \u2013 high medium to cell ratio \u2013 caused cells to grow. Adding back CM to an island of confluent cells caused cells to grow. Taking all this information and developing a 2 factor model to explain it was the first step; the second step was to mathematically model the behavior of the cells in 3-dimensional space. Their programming generated a growth plate that was confirmed in vivo at the tendon/muscle junction(Schwarz 1996). At the growth plate one can postulate how tendon cells normally respond: at the leading edge cells at moderate cell density are growing driven by their production of the SNZR 1P2L. As one moves towards the trailing edge the cell density increases as does the level of SNZR 2L. This slows proliferation, increases procollagen production and begins a decline in the level of SNZR 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t 1P. At the trailing edge the cells have been at high cell density producing high levels of procollagen for a longer time and the buildup of SNZR 2L causes a continued decrease in the SNZR 1P. This drop eventually triggers apoptosis although some cells at the periphery of the tube (fibril) survive. This is only an outline with the principal players identified (Fig. 15). Nevertheless, a \u201cpeculiar\u201d behavior in one setting, becomes \u201cnormal\u201d in another. Analyzing molecular changes raises the question of why does the cell need an additional method for signal transduction? Why not use a protein to activate a receptor? Or put another way, what is the advantage of using a diffusible lipid? In the case of activating a receptor, the cell responds in a graded manner in a set direction. Adding a growth factor can stimulate a cascade of events that results in more rapid cell proliferation. However, the participation of a cell within a growth plate is a far more complex process requiring the cell to respond to parameters related to both space and time. By detecting the levels of SNZR 1P2L and SNZR 2L, the cell can determine its location within the growth plate, whether it should be growing or slowing down, producing high levels of procollagen, or apoptosing. Having a diffusible lipid be a major signal makes the system sensitive to subtle and rapid changes; and with all of the sensors and modifiers acting in one place creates the biological equivalent of an analog logic board allowing the cell to respond to a complex environment correctly. Key elements to this \u201ccircuit\u201d require unique lipids yet this is not a common type. Because they are present at low concentrations, their detection is difficult. We were only able to analyze the two unique lipids because of their tight binding to a protein to which we had both tag and polyclonal antibodies. On the other hand, the tight binding of a protein made purification by standard lipid extraction methods impossible. But the critical observation is that production of a chick protein made in tendon with high homology to the human equivalent causes a unique human bone lipid to be produced in U2OS cells. This makes the protein look like a cassette that can bind a variety of different lipids. But this raises the question of how many kinds are there? If they are in tendon that uses a growth plate to form tubes filled with collagen, do we expect to see unique lipids in other tissues that form tubes \u2013 filled (structural tissues) and unfilled (ductal glands) \u2013 using a growth plate? Similarly we have seen one case where two cell types share a common protein. But how many cell types share this protein and are there alternative proteins for other cell types? Taking into account the unknown number of unique lipids, one can ask how many unique cell density signals exist in an organism? Finally, even though our initial aim in selecting an osteosarcoma cell line (U2OS) was simply to use its cellular machinery to express the chicken cell density signal, these cells were not passive responders to the expression of a chicken cDNA. Instead they used this expression to make more of their own cell density signal, and in turn structures resembling spongy bone with the production of an early osteoblast specific marker. This complex differentiation program was retained by these cells even after becoming malignantly transformed and spending 49 years in cell culture. This indicates that the right signaling molecules can push the malignant cell in new and potentially stable directions. This is not unique to U2OS cells. Using malignant mammary cells(Weaver et al. 1997), \u201cthese studies show that despite a number of prominent mutations, amplifications, and deletions, signaling events which are linked to the maintenance of normal tissue architecture are sufficient to abrogate malignancy and to repress the tumor phenotype.\u201d Again, melanoma cells(Postovit et al. 2006) grown on matrices preconditioned by human embryonic stem cells \u201cinduced melanoma spheroid formation, promoted the re-expression of Melan-A, and inhibited melanoma cell invasion.\u201d With cell density, the signaling mechanism is beginning to be defined at the molecular level. This raises the question of whether these molecules can be used in a tissue-specific manner to effectively manipulate the differentiated state in order to stimulate the healing of injured tendons and bones, and as a new therapeutic approach to reversing the uncontrolled proliferation of cancer cells by directing cells to normal 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t morphogenesis. The latter would be particularly ironic if studies focused on the differentiated state of tendon cells \u2013 the cell type with perhaps the lowest rate of malignancy \u2013 provided new insights into controlling this disease. Acknowledgments RS would like to thank three people at Amgen for their knowledge, experience and hard work in purifying factors \u2013 Bill Kenney, Harvey Yamane, and Mike Bendor. Bill\u2019s encouragement that \u201cthis project is hard because all the easy ones have been done\u201d turned out to be true. Eva Lee for pulling lots of embryonic chick tendons, collecting 400 liters of conditioned medium, and working through many problems. The members of the Bissell group past and present who gave advice and encouragement to an eccentric biologist investigating how cells \u201csee\u201d their neighbors. Recently, Ana Correia, Cyrus Ghajar, Jamie Inman, Hidetoshi Mori, Ramray Bhat, Joel Hyman and Sabine Becker-Weiman knowingly and unknowingly helped get this work to the finish line. Finally, Mina Bissell for her contagious enthusiasm for good biology. We would also like to thank Peter Rademacher for his help in analyzing the mass spectrometry data and giving us many useful interpretations. in contrast, when the filtrate was treated with high salt (0.5 m) and then low salt by 10 fold : dilution and then concentrated on a 10 kD ultrafilter, the bands almost all went to a 16 kD form and its 32 kD dimmer (lane 3). This is strong evidence that whatever was causing the increased MW to 28kD and the further aggregation was not covalently bound PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 5 Localization of the bone factor/cofactor in the extracellular space and disruption by gentle agitation U2OS-pESY1c cells were grown in culture for 9 days and then fixed with 4% paraformaldehyde in a high calcium buffer. A rabbit polyclonal antibody was used as a primary and goat anti-rabbit alexa 488 was used as a secondary (panel a). For comparison, the same field is shown in phase contrast (panel b). Areas where the cells are overgrowing the monolayer show strong staining whereas cells remaining in the monolayer appear to be suppressed. This leads to the strong localized differences in the expression of this gene. When CM was collected from these cells, the flasks are gently rotated (100 rpm) for 1h; then the medium is collected and new medium is added. After 3 collections the flasks are stationary for the next 21 h. After collecting medium for 3 days a dramatic change is observed in the culture \u2013 cells that had overgrown the monolayer begin to spread into regions that had remained a monolayer (panel c). PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 6 Treating U2OS-pESY1c CM with various lipid degradative enzymes to identify the makeup of the bone cofactor The only assay for the cofactor is its binding to the factor and shifting its mobility. The CM was treated with lipase, phospholipase C, and sphingomyelinase for 30, 60, and 180 min and analyzed by Westerns using a myc monoclonal antibody(panel a, lanes 1,2,3; panel a, lanes molecular weight markers (lane 2) are superimposed on the other lanes. The secreted form of the factor made in E. coli (94 AA+ tags) was run in lane 1 and the concentrated CM was run in lanes 3, 4. To release the cofactor the CM was made ~0.6 M NaCl and then concentrated on a 10 kD YM-10 ultrafilter and then diluted with water and refiltered using the same filter (lanes 5, 6). In this case, the retentate shows a strong shift to the secreted form of the factor without cofactor, 16 kD. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 9 Identification of the bone cofactor on mass spectrometry by using lipid degradative enzymes and differential elution from a C-18 spin column After separation from the factor, the bone cofactor was in the dilute filtrate. To concentrate the lipid containing cofactor, the filtrate was run over a C-18 column and washed with water, acetonitrile, and eluted with chloroform. To identify the cofactor, the cofactor was treated separately with phospholipase C (panel b) and lipase (panel c). These were compared to untreated cofactor (panel a) and a lipase buffer only control (no cofactor; panel d) using mass spectrometry. The only regions that showed a strong peak in the untreated sample and little or no intensity in the other panels were at an m/z of 543.42 and 557.44. These peaks had the characteristic pattern of a doubly charged molecule and this gave the cofactor a MW of towards the middle. The polyclonal antibody staining for the factor (using a secondary antibody, goat anti rabbit alexa 568), showed intense staining compared to the muscle bundles. Staining the same section with DAPI shows that this part of the tendon is full of nuclei reflecting that this is a growing end (panel b). The connective tissue surrounding the muscle shows nuclei staining with DAPI that looks like a line of nuclei. The light staining of the muscle connective tissue with the polyclonal antibody suggests another cell density signal with its own cofactor. As a control, a preimmune serum was used at the same concentration and exposed for the same time, was used to detect non specific signal (panel c). The same section was DAPI stained and the stained nuclei show a growing tendon rising steeply from the lower left corner to the top of the picture (panel d). One has to take into account that the tendon does not have to be in exactly the same plane as the tissue slice and therefore, the nuclei density can vary as the tendon weaves through the muscle. Scale bar =100 microns PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 15 Schematic representation of tube formation \u2013 a critical process in diverse tissues Glandular tissues clearly need ducts (tubes) to secrete their products. Tendons, a structural tissue, also form tubes but fill them with collagen to make fibrils. This solves the problem of how to make a collagen thread that can be long but maintain a precise diameter for uniform strength. The key element is a growth plate where cells are proliferating and moving forward on one side (right) while they are mostly dying on the opposite side (left). To accomplish this requires two factors: SNZR PL (blue) and SNZR L (dark yellow). Growth is stimulated at moderate cell density by the production by many cells of SNZR PL \u2013 the factor is found in the fluid surrounding the cells as well as bound to the plasma membrane. The binding of SNZR PL not only stimulates growth but the production of a non-diffusible second factor that we postulate is SNZR L. This causes another change in the cell membrane and some cells slow growth and become more differentiated which we show as oval cells becoming retangular in the transition zone. As cells become denser, they accumulate more SNZR PL and SNZR L and become highly differentiated. The exception is cells at the outer edge of the tube that due to their position never reach high cell density (narrow ovals). The high level of SNZR L causes a drop in the production of SNZR P and this triggers apoptosis. In tendon, the differentiated cells produce half of their total protein synthesis as procollagen and they fill the tube with collagen and become a fibril. The cells at the edge are not necessarily contiguous or ordered. One can extrapolate that in a gland, the differentiated zone would produce a variety of mature phenotypes that could escape apoptosis and form a tight single or multi-cell layered duct. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t to test whether the standard extraction method of purifying lipids would yield a free cofactor, : a simplified procedure using ultrafiltration and high salt releases the bone cofactor : The same approach as previously used with the bone cofactor yielded two peaks at 668.61 and 696.65 separated by 28. This difference is commonly seen for lipid chains by addition of two CH2 groups. Since the 668 band was stronger, we focused the model on this band. The ms/ms fragmentation pattern was fairly simple yielding two bands of roughly equal intensity corresponding to an 18 and 20 carbon fatty acids. Since the molecule is degraded by phospholipase C, a phosphate was added to the model. However, there was insufficient MW for a glycerol so the fatty acids were attached directly to the phosphate. The remaining MW fits a cyano group. The working model is presented to give a framework to the type of structure expected. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 13 The cofactor has a unique lipid makeup depending on the producing cell type but would SNZR 1P1L and SNZR 1P2L give the same or different response if added to the same cell type In this case we used the control cell line U2OS and added a 5X concentrated condition medium from PAT cells or from U2OS-pESY1c cells. An additional test was done to see if the free factor made in E. coli had any activity on U2OS cells if applied at a high concentration (5 mg/ml). In comparison to expressing pESY1c in U2OS cells which causes hills and valleys to form, adding 5X CM from these cells to uninfected U2OS cells shows only a slight growth effect after 6 days as the cells appeared to be more compact (panel b) vs medium only control (panel a). Growing the cells for an additional 9 days showed a strong growth effect (panel f) vs the medium only control (panel e). Moreover, the increased growth is even over the culture \u2013 no hills or valleys. In contrast, adding 5X CM from PAT cell cultures caused the U2OS cells to grow more slowly and the presence of more rounded up cells (panel d). Adding high concentrations of the secreted form of the factor alone (expressed in E. coli, 94 AA + tags) did not appear to increase growth after 6 days (panel c). PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 14 Localization of the factor using a rabbit polyclonal antibody in 5 week old mice at the tendon/muscle junction mass spectrometry analysis : The diluted lipid extract samples were analyzed on an ABSciex TripleTOF 5600 (AB Sciex, Foster City, CA). Samples were introduced to the mass spectrometer via syringe pump by using a Nanospray III source (ABSciex) with a nano-tip emitter (New Objective, Woburn, MA) operating in positive-ion mode (2400 V). The data were acquired with Analyst TF 1.5.1 by averaging the signal from 400 m/z to 1250 m/z over several minutes. MS/MS spectra were collected in \"high sensitivity\" mode with the quadrupole set to UNIT resolution and collision energy set to 30 to optimize fragmentation. MS/MS spectra were scanned from 100 m/z to 1600 m/z and were collected for a total accumulation time of 500 ms. 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t tissue localization in mice : One aim of this paper \u2013 producing the recombinant form of the tendon cell density signal in another cell type and then applying it to tendon cells to show it has the original activity \u2013 is not compatible with our data. Since the cofactor that is essential for activity is unique (in two cases) to the cell type producing it, another test would be useful in showing that the protein factor is the correct candidate for being part of the cell density signaling mechanism. There is the similarity that both the tendon SNZR 1P2L and bone SNZR 1P1L are loosely bound to the cell layer and cause the cells to differentiate. In addition, the polyclonal antibody to the 94 AA protein does detect a band at 16 kD in the PAT cell CM (Fig. 11C). A more functional test would further confirm the identity. To accomplish this, the polyclonal antibody was used to check whether the factor was at the tendon/muscle junction of 5 week old mice. This is the growing stage for mice and one would predict that the factor would be at high concentration at the growing ends of the tendon; where there should also be a high density of fibroblasts(Schwarz 1996). To detect the protein factor, a polyclonal rabbit antibody was used with a secondary antibody labeled with alexa 568; to detect the nuclei, the sections were stained with DAPI (Fig. 14). At the growing end of the tendon there is a strong fluorescence signal for the factor (Fig. 14a) as well as numerous nuclei labeled with DAPI (Fig. 14b). There is variation in the signal over the length of the tendon but this can result from the plane of the tendon not being in exactly the same plane as the section. The strong staining for the factor at the growth plate of the tendon supports its role as a cell density signal. As a control, a preimmune serum was used at an equal concentration and exposed for equal time and it is basically negative (Fig. 14c) and the DAPI staining shows the tendon nuclei starting at the lower left corner and rising steeply to the top of the picture (Fig. 14d). 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t One unexpected observation is the light staining of the muscle connective tissue with its single file nuclei (Fig. 14a,b). This is an indication that a third cell type may be producing this protein factor along with its own unique cofactor. 10 kd ultrafilter so that a sufficient level of protein could be loaded onto the gel. the retentate : shows a wide spectrum of bands from the smallest at 28 kD, the most intense at ~60 kD and the highest at ~250 kD (lanes 5,6). The filtrate showed one band at ~60 kD (lanes 7,8). Since the filtrate initially went through a 30 kD ultrafilter, this large band is an indication that concentrating the sample causes aggregation. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 4 Determining whether the apparent aggregation of the factor in the CM is due to covalent modifications or binding of a cofactor When the CM was concentrated using a 10 kD ultrafilter and this sample was analyzed by Western using a myc antibody (clone 4A6), one sees a lane of multiple bands, the smallest being 28 KD (lane1). In contrast, if the secreted form of the factor with tags (94 AA + tags) is expressed in E. coli, one sees a single band at 16 kD (lane 6). Lane 5 was reserved for MW markers. The CM was put through a 30 kD ultrafilter and the retentate used as a control (lane immunofluorescence : Cell culture: cells were fixed with 4% paraformaldehyde for 10m in a high calcium buffer (.05M TES pH 7.2, 0.1M NaCl, .05M CaCl2). The fixation was quenched with two, five minute washes in buffer A (.05M TES pH 7.2, 0.1M NaCl, .01M CaCl2 ) containing 0.1M glycine. A blocking agent (1% casein Hammerstein [EM Science] in PBS) was used for 30 m followed by a rinse with buffer A. The primary antibody was a polyclonal made in rabbits against the secreted form of pESY1c plus tags expressed in E. coli. This was added at 1:40 (the antibody was diluted by 25% with stabilizer [Candor]) for 2h with agitation. Then, washed 3x with buffer A with agitation and then a fluorescently tagged goat anti-rabbit secondary was added for 1h with agitation. Washed 3X with buffer A and then once with a DAPI buffer ( 500nM in PBS). Finally, a wash with buffer A. As a control, a preimmune serum from the same rabbits was used at 1:40 (the serum was diluted 25% with stabilizer) and used as described above. Tissue sections: hind legs from 5 week old mice were dissected to reveal the muscle and the attached tendon. These were embeded in OCT (Sakura) for 10 to 30m and then frozen in small embedding molds in a dry ice/ethanol bath. Frozen tissue was sliced into 16 \u00b5 sections and attached to superfrost plus slides (VWR). The sections on the slide were outlined with a pap pen and then analyzed as described above. In this case, because the tendon had less autofluoresence in the red, this was the preferred color for the secondary emission. Because the plane of the knife and the plane of the tendon were infrequently in agreement, the analysis required many legs and sections. alkaline phophatase assay : The Sigma kit was used (85L2-1KT) and the included procedures were followed with only slight modifications for labeling tissue culture flasks instead of slides. genes, cloning, and expression : RNA was extracted from confluent cultures of PAT cells using Trizol (Invitrogen). The RNA was reversed transcribed using superscript II (Invitrogen). The cDNA was amplified by PCR using proof reading taq polymerase (Invitrogen) using primers from the 5' end of exon 2 plus a methionine start site and the 3' end of exon 7. Forward primer: GCTGGATCCATGAAAGACCGTCTCAACCTTCCA Reverse primer: ACAGAATTCGACTTCATCCTGGGCTCC This was cloned into the vector pcDNA3.1/Myc-His(+)A,B,C (Invitrogen)and sequenced. The sequence agreed with the predicted expression of this gene (minus exon 1; accession: XP_427094). Using primers for standard T7 promoter/priming site and the BGH Reverse priming site with the addition of a Sal1 site, the gene with tags was amplified using PCR. It was TOPO cloned (Invtrogen) to enhance Sal1 activity and then cut with BamH1 and Sal1. This was ligated to pESY-Neo-II(Koh et al. 2002; Lee 2007), a retroviral vector, that had also been cut with BamH1 and Sal1. In summary, at the time the sequence for the chicken exon 1 was not known, so a 3 AA sequence containing methonine was added to the 5' end (above the dashed line below) and myc and his tags added to the 3' end (below the dashed line below). This was inserted into a retroviral vector pESY-Neo-II \u2013 retroviral pEYK 3.1 was modified to remove the GFP and bleomycin DNA and replace it with Neo/Kan antibiotic resistance gene and a multi-cloning site(Lee 2007) and renamed pESY1c: M K D -------------------------------------------------------------------------------------- R L N L P S V L V L N S C G I T C A G D E N E I A A F C A H V S E L D L S D N K L E D W H E V S K I V S N V P H L E F L N L S S N P L S L S V L E R R C A G S F A G V R K L V L N N S K A S W E T V H T I L Q E L P D L E E L F L C L N D Y E T V S C S P V C C Q S L K L L H I T D N N L Q D W T E I R K L G I M F P S L D T L I L A N N N L T T I E E S E D S L A R L F P N L R S I N L H K S G L H C W E D I D K L N S F P K L E E V K L L G I P L L Q S Y T T E E R R K L L I A R L P S I I K L N G S I V G D G E R E D S E R F F I R Y Y M E F P E E E V P F R Y H E L I T K Y G K L E P L A V V D L R P Q S S V K V E V H F Q D K V E E M S I R L D Q T V A E L K K H L K T V V Q L S T S N M L L F Y L D Q E A P F G P E E M K Y S S R A L H S Y G I R D G D K I Y V E P R M K -------------------------------------------------------------------------------------- S N S A D I Q H S G G R S S L E G P R F E Q K L I S E E D L N M H T G H H H H H H Alternating normal and bold/italic fonts indicate exons and the underlined AAs in last exon indicate the secreted form of the gene (94 AA). The vector pESY1c (4 \u00b5g/60 mm2 dish) was used to transfect phoenix cells using lipofectamine 2000 using the manufactures protocol (Invitrogen). The cells were incubated overnight and the next day medium was collected, filtered (0.45 \u00b5), and used to transfect pESY1c into U2OS cells in the presence of 8 \u00b5g/ml polybrene. The transfected cells were selected using G418 (100 \u00b5g/ml). 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t polyclonal antibody : The secreted form of pEsy1c (underlined AA above, 94AA) with a methionine at the Nterminus (above dashed line) was inserted into the E. coli expression vector pTrcHis2-TOPO (Invitrogen). The vector included myc and his tags at the C-terminus (below lower dashed line). M ------------------------------------------------------------------------------------- E P L A V V D L R P Q S S V K V E V H F Q D K V E E M S I R L D Q T V A E L K K H L K T V V Q L S T S N M L L F Y L D Q E A P F G P E E M K Y S S R A L H S Y G I R D G D K I Y V E P R M K ------------------------------------------------------------------------------------- D D D D K K G E F E A Y V E Q K L I S E E D L N S A V D H H H H H H Forward primer: GTGACCATGGAGCCCTTGGCAGTCGTGGAT Reverse primer: CTTATCGTCATCGTCCTTCATCCTGGGCTC The procaryotic expression vector yielded an expressed protein that was 7 AA smaller in the linker between the protein and the tags (the variation is a function of the type of multiple cloning site and the primers used in cloning) and 1 AA larger at the N-terminus. This difference was too small to detect on gels and the linker region was not important for antibody production. The protein was expressed in E. coli and purified and nickel columns (Invitrogen). This protein (4 mg) was used to immunize rabbits and for affinity purification (Biosynthesis). This protein was also used as a standard for the secreted form of the factor with tags. western analysis : For cytoplasmic proteins: cells were lysed (400 \u00b5l/75 cm2) with a detergent buffer (150 mM NaCl, 1% NP-40, and 50mM Tris [pH 8.0]). This was diluted 2X sample buffer (Invitrogen) 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t and treated as described below. For extracellular protein: samples were concentrated on 10 kD spin ultrafilters (Millipore) diluted with 2X sample buffer and concentrated again to <200 \u00b5l. The samples were made 5% \u03b2mercaptoethanol and heated at 80o C for 10 m. 40 \u00b5l or less was applied and run on an SDS gel (4-20%, 1.5mm, 10 well, Tris-Glycine [Invitrogen]). The gels were blotted onto nitrocellulose (0.45 \u00b5), washed with water, and dried overnight. Next, the blots were rehydrated with water and then blocked with 1% Hammerstein casein in PBS for \u00bd h. After 2X water washes the primary antibody (monoclonal myc (Millipore) or polyclonal rabbit, see above) was added at 1:1000 in PBS with 0.2% casein Hammerstein for 2 h. The blot was then washed 3X for 5m in PBS with . 02% Tween 20. After which the rabbit secondary antibody conjugated to horseradish peroxidase (Thermo) was added at 1:1000 in PBS with 0.2% casein. Again the blot was washed 3X for 5m in PBS with .02% Tween 20 and then developed with a Dura or Femto kits (Thermo). The blots were photographed with a FluoroChem HD2 (ProteinSimple). antibody selection and specificity : Antibody affinity to a protein could be strongly affected by a lipid cofactor. This became an issue when the monoclonal antibody (clone 4A6, Millipore) to the myc tag stopped working. Over the course of these experiments Millipore changed the storage buffer several times and finally to a liquid stored at 4o C. This latter change reduced the affinity of the antibody so that only the free protein factor was detected (data not shown). The benefit of the antibody from the 4A6 clone is that it \u201crecognizes and is specific for recombinant proteins containing the Myc epitope tag in a variety of sequence contexts.\u201d Other clones are known to show strong affinity differences depending on the protein attached to the myc tag. Add a lipid cofactor to a protein with a myc tag and it is unclear how a context sensitive antibody would bind. To find an adequate replacement, three alternatives were tested: a rabbit polyclonal antibody to the secreted form of the factor (94 AA + tags) expressed in E. coli; a monoclonal antibody to the myc tag from Applied Biological Materials (clone A7) and one from GenScript (clone 2G8D5). Using the same CM sample for each antibody yields a Western blot that is unique to the probing antibody (Fig. 1). Critical to our purification procedure for the cofactor, only the polyclonal antibody can detect the free factor at our test concentration of ~0.5 ng as well as bands with the factor/cofactor. Most importantly, one cannot compare the intensity of one band to that of another because each antibody has its preferred conformations; instead, one can only compare the relative amount of a specific band after different treatments. cofactor purification : The CM was filtered through a 0.2\u00b5 filter and then the NaCl level was increased by adding 1/10 the volume with 5M NaCl. The high salt CM, 10 to 50 ml, was concentrated to 1ml in a stirred cell (10 ml, Millipore), at 700 rpm and pressurized (14 psi, N2). The filter was a YM10 with a nominal cutoff at 10 kD (Millipore). The retentate was diluted to 10 ml with water and reconcentrated to 0.5 ml. The retentate was analyzed on Westerns to see the level of free factor released. The filtrate containing the cofactor was concentrated using a Pierce C18 spin column (Thermo). The column was wetted with 1 ml of 50% acetonitrile/water and then spun for 1m at moderate speed in a clinical centrifuge. Then, 1ml of the filtrate was added and spun and repeated until all the filtrate was bound to the column. After which the column was washed twice with 1ml of water, and twice with 1 ml of acetonitrile. Then, the washing was repeated: 1ml of 50% acetonitrile, 1 ml water, 2X 750 \u00b5l acetonitrile, and eluted with 2X 750 \u00b5l of chloroform. 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Samples were stored in chloroform and concentrated before analysis by blowing down with nitrogen to ~100 \u00b5l. For mass spectrometry, 10 \u00b5l of sample was diluted with 80 \u00b5l of acetonitrile and 10 \u00b5l water. All solvents were lc,ms, or hplc grade. With recently purchased YM-10 filters (in comparison to using older filters from the original maker Amicon), the cofactor would not separate from the factor unless the following changes were made: the filters had to soak at 37o C in 20% acetonitrile for 48h; after concentrating the high salt CM to 1 ml, it was put into 10ml of buffer (10 mM EDTA[pH 8.0], 2 ml acetonitrile, 3 ml isopropyl alcohol, 5 ml water) and heated to 80o C for 2h and then concentrated to 0.5 ml using the treated YM-10 filter. The filtrate was further concentrated and purified as described above. The cofactor for bone and the one for tendon are unique. The SNZR 1P can bind either SNZR 1L or SNZR 2L and drive a unique differentiation pathway dependent on the type of cofactor produced by the cell. This raises the question of whether the SNZR 1P2L when added to U2OS cells is capable of driving bone differentiation or does it elicit a new or null response? In addition, one can ask whether the protein factor, by itself, has any activity? To answer these questions, U2OS cells were grown in regular media as a control, or with 5X concentrated condition media from U2OS-pESY1c cells expressing the SNZR 1P1L, and from 5X concentrated condition media from PAT cells expressing SNZR 1P2L. In addition, SNZR 1P purified from E. coli, was added at 5 \u00b5g/ml \u2013 3 orders of magnitude greater than the estimated level in CM (Fig. 13a,b,c,d). These levels were chosen because early work with PAT cells have shown that 1X CM is insufficient to stimulate growth when added back to confluent cultures(Zayas & Schwarz 1992). The high level of the protein factor was chosen because it seemed unlikely that the factor without the cofactor would have a lot of activity and small amounts of activity are difficult to detect by visualization. After a week in the various media, the most dramatic effect was that the tendon CM was inhibitory to growth when added to the U2OS cells yielding a subconfluent culture. The addition of the SNZR 1P alone did not appear to change the cultures compared to the control even at this high concentration. Adding 5X CM from U2OS cells expressing the SNZR 1P1L, showed some increased in growth as the cells appeared to be more compact. However, to confirm this observation we allowed this flask and its control to grow for additional 9 days, at which point it was obvious that the CM had caused the cells to completely overgrow the monolayer (Fig. 13e,f). What is particularly striking is uniformity of growth when the CM is added to cells versus the lattice patterns that are formed when the cells themselves express the SNZR 1P1L (Fig. 2b). lipid enzyme treatment : Lipid degradative enzymes were first used to characterize the makeup of the cofactor. In this case, we were looking for a change in the mobility of factor/cofactor on a Western blot after being treated with these enzymes for various times 30m, 90m, 180m, and a no enzyme control. CM (3ml) was concentrated to 0.5ml (one sample for each time point). Then the enzymes were added as detailed below and incubated for the specific time, 2X sample buffer was added to stop the reaction, concentrated on a 10 kD spin ultrafilter, and analyzed by Western blotting. Lipase: Sigma (L8525); 380 units/\u00b5l; added 50 \u00b5l; incubated 37o C in buffer B (see below) Phospholipase C: Sigma (P7633); .038 units/\u00b5l; added 50 \u00b5l; incubated 37o C in buffer B Phospholipase A2: Sigma (P9279); 6 units/\u00b5l; added 50 \u00b5l; 20 mM triethanolamine pH 8.9 substituted for TES in buffer B ; incubated room temperature Sphingomyelinase: Sigma (S9396); 0.54 units/\u00b5l; added 2 \u00b5l; incubated 37o C in buffer B with MgCl2 substituted for CaCl2 The second use of lipid enzymes was to identify the peaks on mass spectrometry that were sensitive the enzyme treatment. In the cofactor purification above, between the wash steps, the column material was resuspended in buffer B (20 mM TES pH 7.2; 10 mM NaCl; 1 mM CaCl2), enzyme added, and incubated for 30 min at 37o C with agitation. This was spun down and the protocol above continued. In the buffer, when using sphingomyelinase, CaCl2 was replaced with MgCl2. nomenclature : With two different cofactors and the potential for more in the future, there is a need to give them distinct names. For the bone factor/cofactor will be named SNZR(sensor)1P1L; for tendon factor/cofactor will be named SNZR 1P2L. Similarly, the protein factor will be SNZR 1P and the bone cofactor, SNZR 1L and the tendon cofactor, SNZR 2L. At present, because of the extremely high homology between species at the protein level and no knowledge of the conservation of lipid structure between species, the delineation is only on cell type. 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t abbreviations : PAT - primary avian tendon CM - conditioned medium SNZR 1P1L - cell density factor with a bone cofactor SNZR 1P2L - cell density factor with a tendon cofactor SNZR 1P - protein factor SNZR 1L - bone cofactor SNZR 2L - tendon cofactor 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t 4) and the filtrate was divided into two parts. the first was further concentrated on 10 kd : ultrafilter and then made 30% DMSO in order to see if this solvent could affect the mobility of the factor (lane 2). The DMSO did limit the aggregation in that more of the factor ran at 28kD. 7,8,9; and panel b, lanes 2,3,4; respectively). the e. coli expressed secreted form of the : factor (94 AA + tags) was run as a control (panel a, lane 5 showing a monomer band at 16 kD and a dimer band at 32 kD) as well as untreated CM (panel a, lane 6; and panel b, lane 1). Panel a, lane 4 contained MW markers (data not shown). The lipase was the least specific enzyme and it degrades the cofactor and frees most of the factor to run at 16 kD. This type of phospholipase C can cleave at both glycerol phosphates and ceramide phosphates. The sphigomyelinase is specific for ceramide phosphates. Both of these enzymes cause single breaks and this alters the binding/mobility but does not free the cofactor from the factor. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 7 Chloroform/methanol extraction does not free the lipid bone cofactor from the protein factor cm from u2os-pesy1c was extracted with chloroform/methanol. the protein fraction which : precipitates at the interface between the two solvents was examined by Western blotting using a myc monoclonal antibody (lane 1 is half the sample volume of lane 2). Despite unusual sample preparation, the SDS gels and Western blotting worked reasonably well. Nevertheless, there was no evidence of a 16 kD free protein factor. Instead the bands resembled those of the factor/cofactor running together with a minimum size of 28 kD. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 8 1086 or 1114.88. : PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 10 Working model is shown for the bone cofactor Using our knowledge of the lipid enzyme preferences, the molecular weight of the whole molecule, the predicted molecular formula, and the fragmentation pattern (MS/MS), a preliminary model has been drawn with a ceramide phosphate, a single chain glycerol lipid, and a small linker. This model also explains why this molecule is soluble from aqueous buffers to chloroform: a structure with a hydrophilic end and hydrophobic tails. The model also shows how the fragmentation pattern of two breaks (MS/MS) is consistent with the model. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 11 Characterizing the tendon cofactor by its sensitivity to specific lipid degradative enzymes PAT cell CM was put through a 30 kD filter and then concentrated with a 10 kD filter. This concentrated CM after several months at 4o C forms a dominant aggregate that runs at ~60 kD. This CM was run on Westerns (using a polyclonal antibody) as follows: untreated (panel a, lanes 2,6); lipase treated (panel a, lanes 3,7); phospholipase A2 (panel a, lanes 4,8); and phospholipase C (panel a, lanes 5,9). The data indicates a strong sensitivity to lipase and phospholipase C but not to phospholipase A2. This is identical sensitivity pattern as seen with the bone cofactor. Next sphingomylinase was tested (panel b, lanes 5,6,7) and compared to the untreated control (panel b, lanes 2,3,4). In contrast to the bone cofactor, sphingomyelinase did not affect the tendon cofactor. To rule out that the sphingomyelinase was being inhibited by the aggregation caused by \u201caging\u201d the CM, fresh cultures of PAT cells were made and the CM treated with sphingomyelinase (panel c, lanes 4,5) and untreated control (panel c, lanes 2,3). Again, this enzyme does not alter the mobility of the tendon factor/cofactor. PeerJ reviewing PDF | (v2013:06:612:2:0:NEW 3 Oct 2013) R ev ie w in g M an us cr ip t Figure 12",
"v2_text": "results : Expression in U2OS cells causes differentiation A replication defective retroviral vector containing a neomycin resistance gene was used to transfect the chicken gene plus tags into U2OS cells(Koh et al. 2002; Lee 2007). Because this study was started when the chicken genome was first sequenced, the chicken equivalent to the first exon was not present. So the insert contained only the last six exons (pESY1c). U2OS cells, derived from a human osteosarcoma tumor, showed stable expression of this gene after infection when grown in low serum media (0.2%)(Valmassoi & Schwarz 1988). Despite the low serum levels, U2OS cells transfected with pESY1c would start to overgrow the monolayer as small ridges. The ridges would coalesce into circles sharing boundaries with other circles that resembled spongy bone in a two dimensional space (Fig. 2b). Uninfected U2OS cells showed some ability to overgrow the monolayer but could not maintain the growth needed to form completed structures (Fig.2a). The multicell ridges also started to show the onset of bone production as reflected in the production of alkaline phosphatase on the cell surface (Fig. 2c). Pattern of molecular expression intracellularly and extracellularly A monoclonal myc antibody and Western blotting was used to probe the expression and processing of this myc tagged gene. Two specific bands were seen intracellularly: ~44 kD band which is expected for the full size gene product; ~26 kD band that could be a cleavage product (Fig. 3). However, this band is larger that the predicted 20 kD protein \u2013 16 kD + tags. To observe the secreted forms a simple purification was used whereby the CM was put through a 30 kD ultrafilter. Both the retentate and the filtrate were analyzed and both had to be further concentrated on a 10 kD ultrafilter in order to load sufficient protein on the gel for Western analysis (Fig. 3). The retentate contained many bands: the smallest was a weak band at ~28 kD, the strongest was at ~60 kD, and the largest was over 200 kD. The filtrate showed basically a single band at ~60 kD. This was unusual since 60 kD proteins are excluded to a high degree from passing through a 30 kD ultrafilter. One explanation is that by concentrating the sample ~20 fold causes the tagged protein to aggregate into stable forms that resists the denaturing effects of heat, \u03b2-mercaptoethanol, and SDS. aggregation is due to a cofactor : The multiple bands seen in the Western blot complicates the interpretation of how this tagged protein is working in the extracellular space. But more specifically, even the smallest MW band (~28 kD) seen outside the cell is higher than predicted. Is this larger size due to a different amount or type of modification in bone cells versus tendon cells? Or is this increase in apparent MW a reflection of an interaction with another molecule? To begin to answer these questions, we needed to resolve how the unmodified secreted form of the protein factor with two tags (94 AA + tags) migrates on a Western. To accomplish this, the tagged protein was expressed in E. coli, purified on a nickel column, and then used as a standard in Western blots. The tagged secreted form of the protein ran with a MW of 16 kD (Fig. 4, lane 6). This confirms that there is either a covalent modification or a binding to another molecule that gives rise to the 28 kD form in U2OS cells. Further aggregation could then occur when concentrating the factor to give rise to larger MW forms. This raises the question of whether different physical or chemical environments could alter the size of the bands observed on Western blots. CM from U2OS-pESY1c was put through a 30 kD ultrafilter and the filtrate was then concentrated on a 10 kD ultrafilter. This was divided into 2 parts. One was further concentrated on a 10 kD ultrafilter for loading on a gel but in this case, 30% DMSO was added. To the other sample, salt was added to raise the concentration by 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t 0.5M; concentrated with an ultrafilter; then diluted to low salt; and then concentrated to apply to a gel (Fig. 4). The effect of DMSO was to reduce the number of bands, especially those at high MW, but the smallest band still remained 28 kD. In contrast, the treatment with high salt, low salt and ultrafiltration yielded a much simpler picture where the bands were reduced to the 16 kD band and its dimer at 32 kD. One can conclude that the 28 kD band is due to a binding of a cofactor and not a covalent modification. In addition, when the cofactor is separated from the secreted form of the protein factor, the MW of the protein is the expected 16 kD (94AA + tags). This shows that U2OS cells are capable of cleaving the large gene product to the 94 AA + tags and then secrete it. Localization and characterization of the factor/cofactor in the extracellular space The previous experiment raises the question of how the untagged 16 kD form in tendon cells compares to the 28 kD tagged form seen in bone cells. Even taking into account a 4 kD tag, there is an additional 8 kD shift in MW. Are these forms variations with common characteristics or is the bone form acting in a different manner? In tendon cells the 16 kD form is loosely bound to the cell layer. This gives a key feature to this form in that gently shaking a flask of tendon cells causes them to loose their ability to sense high cell density, resulting in increased proliferation. First, the localization was examined with immunofluorescence using a rabbit polyclonal antibody specific to the tagged secreted form of the chicken protein expressed in E. coli. During tests to optimize the signal, procedures that disrupted membranes decreased the signal. Using an optimized protocol, immunofluorescence staining was localized to areas where the cells have overgrown the monolayer (hills) and expression was suppressed in monolayer areas (valleys) (Fig. 5a). Second, the cells were subjected to gentle shaking (100 rpm) for 3h per day with the medium changed each hour. After 3 days the hills had spread into the valleys and the valleys had additional growth. The net effect was to even out the growth over the flask (Fig. 5c). One can conclude that some of the major physical properties of the bone and tendon factor/cofactor are the same. They are both localized to the cell layer \u2013 most likely bound to the cell membrane. However, this interaction is not strong, allowing significant diffusion that can be enhanced by gentle agitation of the medium. Characterization of the bone cofactor with degradative enzymes The cofactor is tightly bound to the factor and gives a consistent mobility pattern on Western blots. This pattern could then be used to test whether specific degradative enzymes could cleave the cofactor, alter the migration profile, and revealing the partial composition of the cofactor. One possible candidate for the cofactor would be a variation on the anchoring structure for membrane bound proteins: a glycosylphosphatidylinositol lipid(Paulick & Bertozzi 2008). To test for a glycolipid, the factor/cofactor was treated with a variety of polysaccharide degradative enzymes \u2013 chondroitinase ABC, hyaluronidase, heparinase \u2013 and the pattern on a Western blot was compared to that of an untreated control. None of the treated samples showed a distinct change in the pattern (data not shown). At the same time, a variety of lipid degrading enzymes were tested \u2013 lipase, phospholipase A2, phospholipase C, and sphingomyelinase. Lipase, phospholipase C, and sphingomyelinase caused dramatic shifts in the pattern seen on Westerns (Fig 6). Lipase was the most dramatic causing bands to shift to lower MW including the 16 kD, free protein form. The lipase was a crude preparation used at a high concentration; so while its primary target would be a glycerol fatty acid linkage in the cofactor, this is probably not the only cleavage product. The phospholipase C used in this experiment has overlapping activities with the sphingomyelinase. The sphingomyelinase is more specific to cleaving moities on ceramide 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t phosphates. In contrast, phospholipase A2 did not show a significant change in mobility or intensity (data not shown). A negative result is hard to interpret because there are too many potential causes. The positive results, while not yielding a definitive structure, implicate lipid components containing a phosphoceramide and a glycerol fatty acid as part of the cofactor. The lipid nature of the cofactor can explain why the factor/cofactor has a strong tendency to aggregate when concentrated due to hydrophobic interactions. Purification and further characterization of the bone cofactor using mass spectrometry In order to characterize the bone cofactor in more detail, procedures were needed to separate both the factor from the cofactor and then the cofactor from other lipid-like molecules. The first step in the purification \u2013 the separation of the cofactor from the factor \u2013 could benefit from standard lipid procedures using a methanol/chloroform extraction. In this procedure the lipids are found in the chloroform layer and proteins are found as a precipitate between the layers. There is no assay for the cofactor but the proteins in the precipitate can be solubilized and then Western analysis can be used to see if the protein factor runs as the free form (16 kD) or as a complex with the cofactor (28 kD). The amount of free factor is an indicator of cofactor release. However, redissolving the protein precipitates at the interface and running Western blots only showed the 28 kD form and higher order aggregates (Fig. 7). In this extraction protocol, the cofactor remained bound to the factor, and as a consequence, precipitated at the interface. Despite attempts to weaken the interaction between the factor and the cofactor by changes in pH, salt, solvent, chaotropic agents, temperature, and chelating agents, Western blots showed little or no conversion to the free form (data not shown). The only successful method required a combination of a shift in salt concentration and ultrafiltration (Fig. 4). Using a simplified version of this procedure whereby the CM was made high in salt (~0.6 M) and then concentrated on a 10 kD ultrafilter. This was then diluted 10 fold with water and concentrated to run on Westerns (Fig. 8). By intensity of the bands, most of the tagged factor runs as the free form. Thus, the corresponding free cofactor was expected to have passed through the ultrafilter as part of the filtrate. Using a small C18 resin column to concentrate, differential solvent elution to purify, and then comparing the untreated and the enzyme treated samples via mass spectrometry one can identify candidate cofactor peaks (Fig. 9). Only molecular weights between 530 and 580 are shown because there was no significant differences between the samples at either higher or lower values (mass range = m/z 300 \u2013 m/z 1500). Two doubly-charged ions at m/z 543.4273 and m/z 557.4429 were dramatically reduced by treatment with either enzyme and these peaks were not observed in the buffer only control and the intensity of the m/z 557 ion was highly correlated with the intensity of the free protein factor in the retentate. The observed difference between the doubly-charged ions was 14 representing a 28 Da difference in MW commonly seen for lipids. In the rest of the analysis, we focused on the 557 species with a MW of 1114.8858. By using a high mass accuracy Q-TOF mass spectrometer we generated potential molecular formulas with a list of atoms commonly found in organic biomolecules (i.e., carbon, hydrogen, nitrogen, oxygen, and phosphorous). The most likely formula C63H123N2O11P (exact mass = 557.4432252+; 0.6 ppm from the measured mass) had nearly a 1:2 Carbon:Hydrogen ratio common to lipids and the nitrogen, oxygen and phosphorous indicative of a hydrophilic region. Support for the proposed cofactor structure derived from MS/MS fragmentation of m/z 557. Two fragment ions, shown in Fig. 10, at m/z of 323 and m/z 378 are consistent with a ceramide phosphate and a linker moiety, supporting the results from the enzymatic reactions. From our analysis, the cofactor was susceptible to three lipid enzymes \u2013 lipase, phospholipase C, and sphingomylenase, and a peak observed at m/z 557 that has a strong lipid composition. This peak was lost when treated with lipase, phopholipase C, and sphingomylenase. 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Using our knowledge of which enzymes cleave the cofactor and the molecular formula, a working model can be postulated that includes a ceramide phosphate, a glycerol lipid, and a linker consisting of a short modification of a proline residue between them (Fig. 10). Tendon cofactor characterization using enzymes and mass spectrometry Having analyzed the bone factor/cofactor, the focus returned to the original aim of characterizing the tendon factor, but now knowing that it binds a cofactor. The bone and tendon cofactors must be different because they impart a unique shift in mobility to the factor/cofactor complex on SDS gels. This raises important questions about the nature of the tendon cofactor and how it compares to its bone counterpart. As a first step in the characterization, the same panel of lipid degradative enzymes were tested on PAT cell CM. In this case, concentrated CM was used that was over 6 months old. This results in a stable aggregate at ~60 kD being the dominant band. As with bone, lipase and phospholipase C were active on the PAT cell factor/cofactor while phospholipase A2 was inactive (Fig. 11a). Unlike the bone cofactor, sphingomyelinase was inactive on the PAT cell factor/cofactor (Fig. 11b). To insure that the inactivity of the enzyme was not due to the factor/cofactor being in a specific aggregate, the treatment with sphingomylinase was repeated with fresh CM (Fig. 11c) and the result remained unchanged. This indicates a lack of a ceramide phosphate moiety in the tendon cofactor. The cofactor was purified for mass spectrometry by binding it to C18 columns followed by washing with methanol, acetonitrile, and elution with chloroform. Via mass spectrometry we compared preparations that were treated with either phosphospholipase C or lipase while the cofactor was bound to the C18 column to the untreated cofactor samples. No differences were seen in the methanol or acetonitrile washes; however, the chloroform fraction did reveal several peaks that changed between the untreated and treated samples. Two singly charged peaks, at 668.61 and 696.65, were reduced to background levels by the enzyme treatments. Similar to the bone cofactor mass spectrometric analysis, these peaks differed by 28 MW and reflect variation a lipid chain length by two carbons. Additionally, there was no doubly charged ion observed at m/z 557 as was seen in the bone cofactor sample (data not shown). Using MS/MS to further characterize the 668.61 ion yielded a rather simple spectrum with two peaks of roughly equal intensity and with MW of 283.26 and 311.29. These ions differ by 28 and accurate mass measurements correspond to saturated fatty acids with 18 and 20 carbons, respectively. Both chains are required to be in the model in order to achieve the high lipid content indicated by the parent molecule\u2019s MW. Since the band is degraded by phospholipase C, there is a phosphate group yet the total mass of the intact ion is insufficient to include a commonly observed glycerol phosphate. Consequently, our proposed structure includes anhydride linkage of the fatty acid directly to the phosphate. The mass difference between the total mass and the fatty acid and phosphate groups corresponds to a cyano group, thus our working model adds a cyano group to the phosphate and is consistent with the enzymology, the mass spectrometry data, and the gel mobility data (Fig. 12). Whether the cyano group was part of the original molecule or it displaced another group during the purification remains to be determined. cells and cell culture : PAT cells were isolated from 16 day chick embryos by a modification of the procedures described by Dehm and Prockop(Dehm & Prockop 1971). Specifically, a higher serum level (3%) was used in the dissociation medium and the use of a nylon mesh (20 micron; TETKO) instead of filter paper to separate out single cells. In these experiments the medium was changed from F12 to a 50/50 mix of F12/DMEM in order to be consistent with the medium used to grow the U2OS cells. The medium contained 0.2% fetal bovine serum deactivated for 30 m at 56o C. U2OS cells (from frozen stocks that have been in the laboratory for over a dozen years) were grown in F12/DMEM medium with 0.2% fetal bovine serum. In 75 cm2 flasks the medium, 50-75 mls, was changed every other day. When the cells first overgrow the monolayer, they are very sensitive to medium volume and will apoptose if too much medium is added. CM was produced by growing cells in 225 cm2 flasks. Medium was changed to serumfree medium and the volume reduced to 12 mls. The flasks were placed in an environmental rotator at 100 rpm, 37o C for 1 h. The CM was collected and new medium added, and the process was repeated for the third hour. After the 3rd collection medium with serum was added (50-75 mls) and the flask returned to the regular incubator. The 16 day chick embryos were exempt from requiring an animal use protocol as determined by LBL Animal Welfare and Research Committee. Similarly, the legs from 5 week mice (BALB/c) came from mice already euthanized for other approved protocols. This is also an exempt use. discussion : In 1976, we showed that freshly isolated chicken embryonic tendon cells were exquisitely sensitive to cell density as reflected in the levels of procollagen production(Schwarz et al. 1976). Despite this dramatic regulation of the most important protein in tendon morphogenesis, the paper went on to say that the \u201cmechanism by which one cell communicates its presence to another cell is unclear.\u201d A few years and many experiments later the signaling mechanism is now shown to use a tight complex between a protein and a lipid. The protein is a small cleavage product from the carboxy end of a larger gene product and this small protein is subsequently secreted from the cell. The protein itself appears to have little or no activity but requires the binding of a lipid cofactor. Remarkably, the composition of the lipid is tissuespecific. While only two types have been characterized, they are each unique in both the lipid structure and the triggered response. SNZR 1P is also unique in composition, processing, and function. By being able to bind unique phospholipids, SNZR 1P allows diffusion into the medium of a lipid that would otherwise be expected to be tightly embedded in the membrane. In earlier papers, this trait was described as being \u201cloosely bound to the cell layer\u201d an explanation for its ability to localize to the cell layer but be partially removable by gentle agitation(Schwarz 1991; Stoker 1973; Zayas & Schwarz 1992). Being able to identify SNZR 1P and its lipid partners gives this complex a molecular identity that explains its unusual characteristics. In addition, as more has become known about the SNZR 1P1L, similarities can be seen to other signaling molecules \u2013 in particular, sonic hedgehog. There is no sequence homology but they are both made as large proteins that are processed to a smaller form and then secreted from the cell(Weed et al. 1997). In the case of sonic hedgehog, it is the N-terminus that is secreted and in this case, it is the C-terminus. The activity of sonic hedgehog is greatly enhanced by the covalent linkage of palmitate and/or cholesterol(Taylor et al. 2001); with the cell density signaling the protein itself has little activity but this is greatly enhanced by binding to a lipid cofactor. Because of the lipid attachments, sonic hedgehog runs as aggregates on SDS gels(Goetz et al. 2006). The binding of the cofactor causes a similar effect on the SNZR 1P. On the other hand, the differences are equally important: SNZR 1P can bind a variety of lipids and in doing so can change its signaling function in a tissue specific manner. Besides how this complicates the analysis of gene function, the result is that the cell has the ability to sense both the type and number of cells in its immediate vicinity and thereby react accordingly. Given the nature of the complex, the simplest hypothesis is that the protein holds the lipid in a specific conformation that can then bind to the cell membrane and organize it(Simons & Toomre 2000). With more cells there would be additional binding resulting in a more organized plasma membrane and this would trigger a stronger signal transduction. But we should point out that our focus has been on the carboxy end of the transcript yet the rest of the protein is equally well conserved between the chick and the human. Another group searching for homology to other tubulin stabilizing proteins found a 20% overlap with this gene(Bartolini et al. 2005). By Northern analysis, many tissues appeared to transcribe this gene at low levels while testis produced it in larger amounts. Over expressing this gene in HeLa cells caused tubulin to be destabilized in contrast to their initial selection criteria. Because they tagged the transcript at the amino terminus, the secreted form would have gone undetected. Moreover, they pointed out that they could not distinguish whether the destabilization of tubulin was a direct or an indirect effect. If the effect on tubulin was an indirect effect of cell density signaling in HeLa cells, then the conservation of sequence could be as simple as conserving the 3-dimensional structure of the cleavage site, or another alternative could be as an intracellular part of the cell density signaling process. The role of the larger, N-terminus protein, will require further analysis 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t to resolve why it is so highly conserved. Our focus has been on molecular characterization of the sensing mechanism for cell density regulation that generates a proliferative signal, but the cell usually requires a more sophisticated response. For instance, in U2OS-pESY1c cells expressing SNZR 1P1L one might postulate that as the cells overgrow the monolayer they begin to produce a secreted inhibitor to prevent the cells in the valleys from producing SNZR 1P1L and dividing: a classic Turing model(Murray 1988; Turing 1952). In tendon cells, more is known and with the knowledge of the cofactor one can adjust a previous model to reveal how cells form a growth plate and then fill a growing tube (fibril) with collagen. To do this the cells need a timer, a growth stimulator, a growth inhibitor and a trigger for apoptosis. Mathematical modeling has shown that this can be accomplished with two factors and the cell(Schwarz 1996). The first factor stimulates growth and the production of the second factor. The second factor, which does not diffuse into the medium; has no activity on its own but by interacting with the first factor changes it to an inhibitor of growth, and a stimulator of procollagen synthesis. Cells left at high cell density cause a drop in the production of the first factor triggering apoptosis. The first factor can now be seen as SNZR 1P2L that we have described. The second factor has until now been hypothetical but a good candidate would be the free cofactor (SNZR 2L). Their interactions occur in the plasma membrane and their consequences are predicted to further organize the membrane structure that in turn can alter the internal state of the cell. In a permissive cell culture environment for tendon cells \u2013 very low serum(0.2%) and addition of ascorbate \u2013 these factors come to the forefront and results in the \u201cpeculiar\u201d behavior of primary cells in culture(Schwarz 1991; Schwarz 1996; Schwarz & Bissell 1977; Schwarz et al. 1976; Schwarz et al. 1979; Valmassoi & Schwarz 1988; Zayas & Schwarz 1992). The cells isolated from 16 day chick embryos were programmed to rapidly make a tendon by way of a growth plate in order that the hatched chick is able to walk to find food and water. In the research laboratory, however, we wanted tendon cells to conform to normal cell culture protocol: grow on a 2-dimensional plastic surface from low density to high density; then be trypsinized off the plastic and the process repeated. Meanwhile the cells were trying to interpret these changes in terms of its own programming for generating a tendon \u2013 a collagen rope -- by regulating collagen production and cell proliferation. As a result we found that: PAT cells only made high levels of procollagen at high cell density (~50% of total protein synthesis). Left at high cell density for several days procollagen production declines. Splitting high density cultures resulted in cells that grew to a lower confluent density and produced a lower level of procollagen. Seeding cells a low cell density resulted in uneven cell growth. Seeding cells at low cell density as a small island in the middle of a dish resulted in the whole island apoptosing after a few hours. Gentle agitation of confluent cells seeded as an island in the middle of dish \u2013 high medium to cell ratio \u2013 caused cells to grow. Adding back CM to an island of confluent cells caused cells to grow. Taking all this information and developing a 2 factor model to explain it was the first step; the second step was to mathematically model the behavior of the cells in 3-dimensional space. Their programming generated a growth plate that was confirmed in vivo at the tendon/muscle junction(Schwarz 1996). At the growth plate one can postulate how tendon cells normally respond: at the leading edge cells at moderate cell density are growing driven by their production of the SNZR 1P2L. As one moves towards the trailing edge the cell density increases as does the level of SNZR 2L. This slows proliferation, increases procollagen production and begins a decline in the level of SNZR 1P. At the trailing edge the cells have been at high cell density producing high levels of procollagen for a longer time and the buildup of SNZR 2L causes a continued decrease in the SNZR 1P. This drop eventually triggers apoptosis although some cells at the periphery of the tube (fibril) survive. This is only an outline with the principal players identified (Fig. 15). 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Nevertheless, a \u201cpeculiar\u201d behavior in one setting, becomes \u201cnormal\u201d in another. Analyzing molecular changes raises the question of why does the cell need an additional method for signal transduction? Why not use a protein to activate a receptor? Or put another way, what is the advantage of using a diffusible lipid? In the case of activating a receptor, the cell responds in a graded manner in a set direction. Adding a growth factor can stimulate a cascade of events that results in more rapid cell proliferation. However, the participation of a cell within a growth plate is a far more complex process requiring the cell to respond to parameters related to both space and time. By detecting the levels of SNZR 1P2L and SNZR 2L, the cell can determine its location within the growth plate, whether it should be growing or slowing down, producing high levels of procollagen, or apoptosing. Having a diffusible lipid be a major signal makes the system sensitive to subtle and rapid changes; and with all of the sensors and modifiers acting in one place creates the biological equivalent of an analog logic board allowing the cell to respond to a complex environment correctly. Key elements to this \u201ccircuit\u201d require unique lipids yet this is not a common type. Because they are present at low concentrations, their detection is difficult. We were only able to analyze the two unique lipids because of their tight binding to a protein to which we had both tag and polyclonal antibodies. On the other hand, the tight binding of a protein made purification by standard lipid extraction methods impossible. But the critical observation is that production of a chick protein made in tendon with high homology to the human equivalent causes a unique human bone lipid to be produced in U2OS cells. This makes the protein look like a cassette that can bind a variety of different lipids. But this raises the question of how many kinds are there? If they are in tendon that uses a growth plate to form tubes filled with collagen, do we expect to see unique lipids in other tissues that form tubes \u2013 filled (structural tissues) and unfilled (ductal glands) \u2013 using a growth plate? Similarly we have seen one case where two cell types share a common protein. But how many cell types share this protein and are there alternative proteins for other cell types? Taking into account the unknown number of unique lipids, one can ask how many unique cell density signals exist in an organism? Finally, even though our initial aim in selecting an osteosarcoma cell line (U2OS) was simply to use its cellular machinery to express the chicken cell density signal, these cells were not passive responders to the expression of a chicken cDNA. Instead they used this expression to make more of their own cell density signal, and in turn structures resembling spongy bone with the production of an early osteoblast specific marker. This complex differentiation program was retained by these cells even after becoming malignantly transformed and spending 48 years in cell culture. This indicates that the right signaling molecules can push the malignant cell in new and potentially stable directions. This is not unique to U2OS cells. Using malignant mammary cells(Weaver et al. 1997), \u201cthese studies show that despite a number of prominent mutations, amplifications, and deletions, signaling events which are linked to the maintenance of normal tissue architecture are sufficient to abrogate malignancy and to repress the tumor phenotype.\u201d Again, melanoma cells(Postovit et al. 2006) grown on matrices preconditioned by human embryonic stem cells \u201cinduced melanoma spheroid formation, promoted the re-expression of Melan-A, and inhibited melanoma cell invasion.\u201d With cell density, the signaling mechanism is beginning to be defined at the molecular level. This raises the question of whether these molecules can be used in a tissue-specific manner to effectively manipulate the differentiated state in order to stimulate the healing of injured tendons and bones, and as a new therapeutic approach to reversing the uncontrolled proliferation of cancer cells by directing cells to normal morphogenesis. The latter would be particularly ironic if studies focused on the differentiated state of tendon cells \u2013 the cell type with perhaps the lowest rate of malignancy \u2013 provided new insights into controlling this disease. 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t in contrast, when the filtrate was treated with high salt (0.5 m) and then low salt by 10 fold : dilution and then concentrated on a 10 kD ultrafilter, the bands almost all went to a 16 kD form and its 32 kD dimmer (lane 3). This is strong evidence that whatever was causing the increased MW to 28kD and the further aggregation was not covalently bound PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 5 Localization of the bone factor/cofactor in the extracellular space and disruption by gentle agitation U2OS-pESY1c cells were grown in culture for 9 days and then fixed with 4% paraformaldehyde in a high calcium buffer. A rabbit polyclonal antibody was used as a primary and goat anti-rabbit alexa 488 was used as a secondary (panel a). For comparison, the same field is shown in phase contrast (panel b). Areas where the cells are overgrowing the monolayer show strong staining whereas cells remaining in the monolayer appear to be suppressed. This leads to the strong localized differences in the expression of this gene. When CM was collected from these cells, the flasks are gently rotated (100 rpm) for 1h; then the medium is collected and new medium is added. After 3 collections the flasks are stationary for the next 21 h. After collecting medium for 3 days a dramatic change is observed in the culture \u2013 the cells no longer accurately detect the cells around them and they grow all over the dish as seen in phase contrast (panel c). PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 6 Treating U2OS-pESY1c CM with various lipid degradative enzymes to identify the makeup of the bone cofactor The only assay for the cofactor is its binding to the factor and shifting its mobility. The CM was treated with lipase, phospholipase C, and sphingomyelinase for 30, 60, and 180 min and analyzed by Westerns using a myc monoclonal antibody(panel a, lanes 1,2,3; panel a, lanes molecular weight markers (lane 2) are superimposed on the other lanes. The secreted form of the factor made in E. coli (94 AA+ tags) was run in lane 1 and the concentrated CM was run in lanes 3, 4. To release the cofactor the CM was made ~0.6 M NaCl and then concentrated on a 10 kD YM-10 ultrafilter and then diluted with water and refiltered using the same filter (lanes 5, 6). In this case, the retentate shows a strong shift to the secreted form of the factor without cofactor, MW of 16 kD. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 9 Identification of the bone cofactor on mass spectrometry by using lipid degradative enzymes and differential elution from a C-18 spin column After separation from the factor, the bone cofactor was in the dilute filtrate. To concentrate the lipid containing cofactor, the filtrate was run over a C-18 column and washed with water, acetonitrile, and eluted with chloroform. To identify the cofactor, the cofactor was treated separately with phospholipase C (panel b) and lipase (panel c). These were compared to untreated cofactor (panel a) and a lipase buffer only control (no cofactor; panel d) using mass spectrometry. The only regions that showed a strong peak in the untreated sample and little or no intensity in the other panels were at an m/z of 543.42 and 557.44. These peaks had the characteristic pattern of a doubly charged molecule and this gave the cofactor a MW of towards the middle. The polyclonal antibody staining for the factor (using a secondary antibody, goat anti rabbit alexa 568 (Invitrogen), showed intense staining compared to the muscle bundles. Staining the same section with DAPI shows that this part of the tendon is full of nuclei reflecting that this is a growing end (panel b). The connective tissue surrounding the muscle shows nuclei staining with DAPI that looks like a line of nuclei. The light staining of the muscle connective tissue with the polyclonal antibody suggests another cell density signal with its own cofactor. As a control, a preimmune serum used at the same concentration and exposed for the same time, was used to detect non specific signal (panel c). The same section was DAPI stained and the stained nuclei show a growing tendon rising steeply from the lower left corner to the top of the picture (panel d). One has to take into account that the tendon does not have to be in exactly the same plane as the tissue slice and therefore, the nuclei density can vary as the tendon weaves through the muscle. Scale bar =100m PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 15 Schematic representation of tube formation \u2013 a critical process in diverse tissues Glandular tissues clearly need ducts (tubes) to secrete their products. Tendons, a structural tissue, also form tubes but fill them with collagen to make fibrils. This solves the problem of how to make a collagen thread that can be long but maintain a precise diameter for uniform strength. The key element is a growth plate where cells are proliferating and moving forward on one side (right) while they are mostly dying on the opposite side (left). To accomplish this requires two factors: SNZR PL (blue) and SNZR L (dark yellow). Growth is stimulated at moderate cell density by the production by many cells of SNZR PL \u2013 the factor is found in the fluid surrounding the cells as well as bound to the plasma membrane. The binding of SNZR PL not only stimulates growth but the production of a non-diffusible second factor that we postulate is SNZR L. This causes another change in the cell membrane and some cells slow growth and become more differentiated which we show as oval cells becoming retangular in the transition zone. As cells become denser, they accumulate more SNZR PL and SNZR L and become highly differentiated. The exception is cells at the outer edge of the tube that due to their position never reach high cell density (narrow ovals). The high level of SNZR L causes a drop in the production of SNZR P and this triggers apoptosis. In tendon, the differentiated cells produce half of their total protein synthesis as procollagen and they fill the tube with collagen and become a fibril. The cells at the edge are not necessarily contiguous or ordered. One can extrapolate that in a gland, the differentiated zone would produce a variety of mature phenotypes that could escape apoptosis and form a tight single or multi-cell layered duct. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t to test whether the standard extraction method of purifying lipids would yield a free cofactor, : a simplified procedure using ultrafiltration and high salt releases the bone cofactor : The same approach as previously used with the bone cofactor yielded two peaks at 668.61 and 696.65 separated by 28. This difference is commonly seen for lipid chains by addition of two CH2 groups. Since the 668 band was stronger, we focused the model on this band. The ms/ms fragmentation pattern was fairly simple yielding two bands of roughly equal intensity corresponding to an 18 and 20 carbon fatty acids. Since the molecule is degraded by phospholipase C, a phosphate was added to the model. However, there was insufficient MW for a glycerol so the fatty acids were attached directly to the phosphate. The remaining MW fits a cyano group. The working model is presented to give a framework to the type of structure expected. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 13 The cofactor has a unique lipid makeup depending on the producing cell type but would SNZR 1P1L and SNZR 1P2L give the same or different response if added to the same cell type In this case we used the control cell line U2OS and added a 5X concentrated condition medium from PAT cells or from U2OS-pESY1c cells. An additional test was done to see if the free factor made in E. coli had any activity on U2OS cells if applied at a high concentration (5 mg/ml). In comparison to expressing pESY1c in U2OS cells which causes hills and valleys to form, adding 5X CM from these cells to uninfected U2OS cells shows only a slight growth effect after 6 days as the cells appeared to be more compact (panel b) vs medium only control (panel a). Growing the cells for an additional 9 days showed a strong growth effect (panel f) vs the medium only control (panel e). Moreover, the increased growth is even over the culture \u2013 no hills or valleys. In contrast, adding 5X CM from PAT cell cultures caused the U2OS cells to grow more slowly and the presence of more rounded up cells (panel d). Adding high concentrations of the secreted form of the factor alone (expressed in E. coli, 94 AA + tags) did not appear to increase growth after 6 days (panel c). PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 14 Localization of the factor using a rabbit polyclonal antibody in 5 week old mice at the tendon/muscle junction material and methods : abbreviations : PAT - primary avian tendon CM - conditioned medium SNZR 1P1L - cell density sensor with a bone cofactor SNZR 1P2L - cell density sensor with a tendon cofactor SNZR 1P - protein factor SNZR 1L - bone cofactor SNZR 2L - tendon cofactor 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t mass spectrometry analysis : The diluted lipid extract samples were analyzed on an ABSciex TripleTOF 5600 (AB Sciex, Foster City, CA). Samples were introduced to the mass spectrometer via syringe pump by using a Nanospray III source (ABSciex) with a nano-tip emitter (New Objective, Woburn, MA) operating in positive-ion mode (2400 V). The data were acquired with Analyst TF 1.5.1 by averaging the signal from 400 m/z to 1250 m/z over several minutes. MS/MS spectra were collected in \"high sensitivity\" mode with the quadrupole set to UNIT resolution and collision energy set to 30 to optimize fragmentation. MS/MS spectra were scanned from 100 m/z to 1600 m/z and were collected for a total accumulation time of 500 ms. 264 265 266 267 268 269 270 271 272 273 274 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t acknowledgments : RS would like to thank three people at Amgen for their knowledge, experience and hard work in purifying factors \u2013 Bill Kenney, Harvey Yamane, and Mike Bendor. Bill\u2019s encouragement that \u201cthis project is hard because all the easy ones have been done\u201d turned out to be true. Eva Lee for pulling lots of embryonic chick tendon, collecting 400 liters of conditioned medium, and working through many problems. The members of the Bissell group past and present who gave advice and encouragement to an eccentric biologist investigating how cells \u201csee\u201d their neighbors. Recently, Ana Correia, Cyrus Ghajar, Jamie Inman, Hidetoshi Mori, Ramray Bhat, Joel Hyman and Sabine Becker-Weiman knowingly and unknowingly helped get this work to the finish line. Finally, Mina Bissell for her contagious enthusiasm for good biology. We would also like to thank Peter Rademacher for his help in analyzing the mass spectrometry data and giving us many useful interpretations. western analysis : For cytoplasmic proteins: cells were lysed (400 \u00b5l/75 cm2) with a detergent buffer (150 mM NaCl, 1% NP-40, and 50mM Tris [pH 8.0]). This was diluted 2X sample buffer (Invitrogen) and treated as described below. For extracellular protein: samples were concentrated on 10 kD spin ultrafilters (Millipore) diluted with 2X sample buffer and concentrated again to <200 \u00b5l. The samples were made 5% \u03b2mercaptoethanol and heated at 80o C for 10 m. 40 \u00b5l or less was applied and run on an SDS gel (4-20%, 1.5mm, 10 well, Tris-Glycine [Invitrogen]). The gels were blotted onto nitrocellulose (0.45 \u00b5), washed with water, and dried overnight. Next, the blots were rehydrated with water and then blocked with 1% Hammerstein casein in PBS for \u00bd h. After 2X water washes the primary antibody (monoclonal myc (Millipore) or polyclonal rabbit, see above) was added at 1:1000 in PBS with 0.2% casein Hammerstein for 2 h. The blot was then washed 3X for 5m in PBS with . 02% Tween 20. After which the rabbit secondary antibody conjugated to horseradish peroxidase (Thermo) was added at 1:1000 in PBS with 0.2% casein. Again the blot was washed 3X for 5m in PBS with .02% Tween 20 and then developed with a Dura or Femto kits (Thermo). The blots were photographed with a FluoroChem HD2 (ProteinSimple). 4) and the filtrate was divided into two parts. the first was further concentrated on 10 kd : ultrafilter and then made 30% DMSO in order to see if this solvent could affect the mobility of the factor (lane 2). The DMSO did limit the aggregation in that more of the factor ran at 28kD. tissue localization in mice : One aim of this paper \u2013 producing the recombinant form of the tendon cell density signal in another cell type and then applying it to tendon cells to show it has the original activity \u2013 is not compatible with our data. Since the cofactor that is essential for activity is unique (in two cases) to the cell type producing it, another test would be useful in showing that the protein factor is the correct candidate for being part of the cell density signaling mechanism. There is the similarity that both the tendon SNZR 1P2L and bone SNZR 1P1L are loosely bound to the cell layer and cause the cells to differentiate. In addition, the polyclonal antibody to the 94 AA protein does detect a band at 16 kD in the PAT cell CM (Fig. 11C). A more functional test would further confirm the identity. To accomplish this, the polyclonal antibody was used to check whether the factor was at the tendon/muscle junction of 5 week old mice. This is the growing stage for mice and one would predict that the factor would be at high concentration at the growing ends of the tendon; where there should also be a high density of fibroblasts(Schwarz 1996). To detect the protein factor, a polyclonal rabbit antibody was used with a secondary antibody labeled with alexa 568; to detect the nuclei, the sections were stained with DAPI (Fig. 14). At the growing end of the tendon there is a strong fluorescence signal for the factor (Fig. 14a) as well as numerous nuclei labeled with DAPI (Fig. 14b). There is variation in the signal over the length of the tendon but this can result from the plane of the tendon not being in exactly the same plane as the section. The strong staining for the factor at the growth plate of the tendon supports its role as a cell density signal. As a control, a preimmune serum was used at an equal concentration and exposed for 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t equal time and it is basically negative (Fig. 14c) and the DAPI staining shows the tendon nuclei starting at the lower left corner and rising steeply to the top of the picture (Fig. 14d). One unexpected observation is the light staining of the muscle connective tissue with its single file nuclei (Fig. 14a,b). This is an indication that a third cell type may be producing this protein factor along with its own unique cofactor. 510 511 512 513 514 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t 10 kd ultrafilter so that a sufficient level of protein could be loaded onto the gel. the retentate : shows a wide spectrum of bands from the smallest at 28 kD, the most intense at ~60 kD and the highest at ~250 kD (lanes 5,6). The filtrate showed one band at ~60 kD (lanes 7,8). Since the filtrate initially went through a 30 kD ultrafilter, this large band is an indication that concentrating the sample causes aggregation. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 4 Determining whether the apparent aggregation of the factor in the CM is due to covalent modifications or binding of a cofactor When the CM was concentrated using a 10 kD ultrafilter and this sample was analyzed by Western using a myc antibody (clone 4A6), one sees a lane of multiple bands, the smallest being 28 KD (lane1). In contrast, if the secreted form of the factor with tags (94 AA + tags) is expressed in E. coli, one sees a single band at 16 kD (lane 6). Lane 5 was reserved for MW markers. The CM was put through a 30 kD ultrafilter and the retentate used as a control (lane immunofluorescence : Cell culture: cells were fixed with 4% paraformaldehyde for 10m in a high calcium buffer (.05M TES pH 7.2, 0.1M NaCl, .05M CaCl2). The fixation was quenched with two, five minute washes in buffer A (.05M TES pH 7.2, 0.1M NaCl, .01M CaCl2 ) containing 0.1M glycine. A blocking agent (1% casein Hammerstein [EM Science] in PBS) was used for 30 m followed by a rinse with buffer A. The primary antibody was a polyclonal made in rabbits against the secreted 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t form of pESY1c plus tags expressed in E. coli. This was added at 1:40 (the antibody was diluted by 25% with stabilizer [Candor]) for 2h with agitation. Then, washed 3x with buffer A with agitation and then a fluorescently tagged goat anti-rabbit secondary was added for 1h with agitation. Washed 3X with buffer A and then once with a DAPI buffer ( 500nM in PBS). Finally, a wash with buffer A. As a control, a preimmune serum from the same rabbits was used at 1:40 (the serum was diluted 25% with stabilizer) and used as described above. Tissue sections: hind legs from 5 week old mice were dissected to reveal the muscle and the attached tendon. These were embeded in OCT (Sakura) for 10 to 30m and then frozen in small embedding molds in a dry ice/ethanol bath. Frozen tissue was sliced into 16 \u00b5 sections and attached to superfrost plus slides (VWR). The sections on the slide were outlined with a pap pen and then analyzed as described above. In this case, because the tendon had less autofluoresence in the red, this was the preferred color for the secondary emission. Because the plane of the knife and the plane of the tendon were infrequently in agreement, the analysis required many legs and sections. alkaline phophatase assay : The Sigma kit was used (85L2-1KT) and the included procedures were followed with only slight modifications for labeling tissue culture flasks instead of slides. genes, cloning, and expression : RNA was extracted from confluent cultures of PAT cells using Trizol (Invitrogen). The RNA was reversed transcribed using superscript II (Invitrogen). The cDNA was amplified by PCR using proof reading taq polymerase (Invitrogen) using primers from the 5' end of exon 2 plus a methionine start site and the 3' end of exon 7. Forward primer: GCTGGATCCATGAAAGACCGTCTCAACCTTCCA Reverse primer: ACAGAATTCGACTTCATCCTGGGCTCC This was cloned into the vector pcDNA3.1/Myc-His(+)A,B,C (Invitrogen)and sequenced. The sequence agreed with the predicted expression of this gene (minus exon 1; accession: XP_427094). Using primers for standard T7 promoter/priming site and the BGH Reverse priming site with the addition of a Sal1 site, the gene with tags was amplified using PCR. It was TOPO cloned (Invtrogen) to enhance Sal1 activity and then cut with BamH1 and Sal1. This was ligated to pESY-Neo-II(Koh et al. 2002; Lee 2007), a retroviral vector, that had also been cut with BamH1 and Sal1. In summary, at the time the sequence for the chicken exon 1 was not known, so a 3 AA sequence containing methonine was added to the 5' end (above the dashed line below) and myc and his tags added to the 3' end (below the dashed line below). This was inserted into a retroviral vector pESY-Neo-II \u2013 retroviral pEYK 3.1 was modified to remove the GFP and bleomycin DNA and replace it with Neo/Kan antibiotic resistance gene and a multi-cloning site(Lee 2007) and renamed pESY1c: 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t M K D -------------------------------------------------------------------------------------- R L N L P S V L V L N S C G I T C A G D E N E I A A F C A H V S E L D L S D N K L E D W H E V S K I V S N V P H L E F L N L S S N P L S L S V L E R R C A G S F A G V R K L V L N N S K A S W E T V H T I L Q E L P D L E E L F L C L N D Y E T V S C S P V C C Q S L K L L H I T D N N L Q D W T E I R K L G I M F P S L D T L I L A N N N L T T I E E S E D S L A R L F P N L R S I N L H K S G L H C W E D I D K L N S F P K L E E V K L L G I P L L Q S Y T T E E R R K L L I A R L P S I I K L N G S I V G D G E R E D S E R F F I R Y Y M E F P E E E V P F R Y H E L I T K Y G K L E P L A V V D L R P Q S S V K V E V H F Q D K V E E M S I R L D Q T V A E L K K H L K T V V Q L S T S N M L L F Y L D Q E A P F G P E E M K Y S S R A L H S Y G I R D G D K I Y V E P R M K -------------------------------------------------------------------------------------- S N S A D I Q H S G G R S S L E G P R F E Q K L I S E E D L N M H T G H H H H H H Alternating normal and bold/italic fonts indicate exons and the underlined AAs in last exon indicate the secreted form of the gene (94 AA). The vector pESY1c (4 \u00b5g/60 mm2 dish) was used to transfect phoenix cells using lipofectamine 2000 using the manufactures protocol (Invitrogen). The cells were incubated overnight and the next day medium was collected, filtered (0.45 \u00b5), and used to transfect pESY1c into U2OS cells in the presence of 8 \u00b5g/ml polybrene. The transfected cells were selected using G418 (100 \u00b5g/ml). polyclonal antibody : The secreted form of pEsy1c (underlined AA above, 94AA) with a methionine at the Nterminus (above the dashed line) was inserted into the E. coli expression vector pTrcHis2-TOPO (Invitrogen). The vector included myc and his tags at the C-terminus (below the dashed line). M ------------------------------------------------------------------------------------- E P L A V V D L R P Q S S V K V E V H F Q D K V E E M S I R L D Q T V A E L K K H L K T V V Q L S T S N M L L F Y L D Q E A P F G P E E M K Y S S R A L H S Y G I R D G D K I Y V E P R M K ------------------------------------------------------------------------------------- D D D D K K G E F E A Y V E Q K L I S E E D L N S A V D H H H H H H Forward primer: GTGACCATGGAGCCCTTGGCAGTCGTGGAT Reverse primer: CTTATCGTCATCGTCCTTCATCCTGGGCTC The procaryotic expression vector yielded an expressed protein that was 7 AA smaller in the linker between the protein and the tags (the variation is a function of the type of multiple cloning site and the primers used in cloning) and 1 AA larger at the N-terminus. This difference was too small to detect on gels and the linker region was not important for antibody production. The protein was expressed in E. coli and purified and nickel columns (Invitrogen). This protein (4 mg) was used to immunize rabbits and for affinity purification (Biosynthesis). This protein was also used as a standard for the secreted form of the factor with tags. antibody selection and specificity : Antibody affinity to a protein could be strongly affected by a lipid cofactor. This became an issue when the monoclonal antibody (clone 4A6, Millipore) to the myc tag stopped working. Over the course of these experiments Millipore changed the storage buffer several times and finally to a liquid stored at 4o C. This latter change reduced the affinity of the antibody so that only the free protein factor was detected (data not shown). The benefit of the antibody from the 4A6 clone is that it \u201crecognizes and is specific for recombinant proteins containing the Myc epitope tag in a variety of sequence contexts.\u201d Other clones are known to show strong affinity differences depending on the protein attached to the myc tag. Add a lipid cofactor to a protein with a myc tag and it is unclear how a context sensitive antibody would bind. To find an adequate replacement, three alternatives were tested: a rabbit polyclonal antibody to the secreted form of the factor (94 AA + tags) expressed in E. coli; a monoclonal antibody to the myc tag from Applied Biological Materials (clone A7) and one from GenScript (clone 2G8D5). Using the same CM sample for each antibody yields a Western blot that is unique to the probing antibody (Fig. 1). Critical to our purification procedure for the cofactor, only the polyclonal antibody can 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t detect the free factor at our test concentration of ~0.5 ng as well as bands with the factor/cofactor. Most importantly, one cannot compare the intensity of one band to that of another because each antibody has its preferred conformations; instead, one can only compare the relative amount of a specific band after different treatments. cofactor purification : The CM was filtered through a .22\u00b5 filter and then the NaCl level was increased by adding 1/10 the volume with 5M NaCl. The high salt CM, 10 to 50 ml, was concentrated to 1ml in a stirred cell (10 ml, Millipore), at 700 rpm and pressurized (14 psi, N2). The filter was a YM10 with a nominal cutoff at 10 kD (Millipore). The retentate was diluted to 10 ml with water and reconcentrated to 0.5 ml. The retentate was analyzed on Westerns to see the level of free factor released. The filtrate containing the cofactor was concentrated using a Pierce C18 spin column (Thermo). The column was wetted with 1 ml of 50% acetonitrile/water and then spun for 1m at moderate speed in a clinical centrifuge. Then, 1ml of the filtrate was added and spun and repeated until all the filtrate was bound to the column. After which the column was washed twice with 1ml of water, and twice with 1 ml of acetonitrile. Then, the washing was repeated: 1ml of 50% acetonitrile, 1 ml water, 2X 750 \u00b5l acetonitrile, and eluted with 2X 750 \u00b5l of chloroform. Samples were stored in chloroform and concentrated before analysis by blowing down with nitrogen to ~100 \u00b5l. For mass spectrometry, 10 \u00b5l of sample was diluted with 80 \u00b5l of acetonitrile and 10 \u00b5l water. All solvents were lc,ms, or hplc grade. With recently purchased YM-10 filters (in comparison to using older filters from the original maker Amicon), the cofactor would not separate from the factor unless the following changes were made: the filters had to soak at 37o C in 20% acetonitrile for 48h; after concentrating the high salt CM to 1 ml, it was put into 10ml of buffer (10 mM EDTA[pH 8.0], 2 ml acetonitrile, 3 ml isopropyl alcohol, 5 ml water) and heated to 80o C for 2h and then concentrated to 0.5 ml using the treated YM-10 filter. The filtrate was further concentrated and purified as described above. The cofactor for bone and the one for tendon are unique. The SNZR 1P can bind either SNZR 1L or SNZR 2L and drive a unique differentiation pathway dependent on the type of cofactor produced by the cell. This raises the question of whether the SNZR 1P2L when added to U2OS cells is capable of driving bone differentiation or does it elicit a new or null response? In addition, one can ask whether the protein factor, by itself, has any activity? To answer these questions, U2OS cells were grown in regular media as a control, or with 5X concentrated condition media from U2OS-pESY1c cells expressing the SNZR 1P1L, and from 5X concentrated condition media from PAT cells expressing SNZR 1P2L. In addition, SNZR 1P purified from E. coli, was added at 5 \u00b5g/ml \u2013 3 orders of magnitude greater than the estimated level in CM (Fig. 13a,b,c,d). These levels were chosen because early work with PAT cells have shown that 1X CM is insufficient to stimulate growth when added back to confluent cultures(Zayas & Schwarz 1992). The high level of the protein factor was chosen because it seemed unlikely that the factor without the cofactor would have a lot of activity and small amounts of activity are difficult to detect by visualization. After a week in the various media, the most dramatic effect was that the tendon CM was inhibitory to growth when added to the U2OS cells yielding a subconfluent culture. The addition of the SNZR 1P alone did not appear to change the cultures compared to the control even at this high concentration. Adding 5X CM from U2OS cells expressing the SNZR 1P1L, showed some increased in growth as the cells appeared to be more compact. However, to confirm this observation we allowed this flask and its control to grow for additional 9 days, at which point it was obvious that the CM had caused the cells to completely overgrow the monolayer (Fig. 13e,f). What is particularly striking is uniformity of growth when the CM is added to cells versus the lattice patterns that are formed when the cells themselves express the SNZR 1P1L (Fig. 2b). lipid enzyme treatment : Lipid degradative enzymes were first used to characterize the makeup of the cofactor. In this case, we were looking for a change in the mobility of factor/cofactor on a Western blot after being treated with these enzymes for various times 30m, 90m, 180m, and a no enzyme control. CM (3ml) was concentrated to 0.5ml (one sample for each time point). Then the enzymes were added as detailed below and incubated for the specific time, 2X sample buffer was added to stop the reaction, concentrated on a 10 kD spin ultrafilter, and analyzed by Western blotting. Lipase: Sigma (L8525); 380 units/\u00b5l; added 50 \u00b5l; incubated 37o C in buffer B (see below) Phospholipase C: Sigma (P7633); .038 units/\u00b5l; added 50 \u00b5l; incubated 37o C in buffer B Phospholipase A2: Sigma (P9279); 6 units/\u00b5l; added 50 \u00b5l; 20 mM triethanolamine pH 8.9 substituted for TES in buffer B; incubated room temperature Sphingomyelinase: Sigma (S9396); 0.54 units/\u00b5l; added 2 \u00b5l; incubated 37o C in buffer B with MgCl2 substituted for CaCl2 The second use of lipid enzymes was to identify the peaks on mass spectrometry that were sensitive the enzyme treatment. In the cofactor purification above, between the wash steps, the column material was resuspended in buffer (20 mM TES pH 7.2; 10 mM NaCl; 1 mM CaCl2), enzyme added, and incubated for 30 min at 37o C with agitation. This was spun down and the 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t protocol above continued. In the buffer, when using sphingomyelinase, CaCl2 was replaced with MgCl2. nomenclature : With two different cofactors and the potential for more in the future, there is a need to give them distinct names. For the bone factor/cofactor will be named SNZR 1P1L; for tendon factor/cofactor will be named SNZR 1P2L. Similarly, the protein factor will be SNZR 1P and the bone cofactor, SNZR 1L and the tendon cofactor, SNZR 2L. At present, because of the extremely high homology between species at the protein level and no knowledge of the conservation of lipid 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t structure between species, the delineation is only on cell type. 7,8,9; and panel b, lanes 2,3,4; respectively). the e. coli expressed secreted form of the : factor (94 AA + tags) was run as a control (panel a, lane 5 showing a monomer band at 16 kD and a dimer band at 32 kD) as well as untreated CM (panel a, lane 6; and panel b, lane 1). Panel a, lane 4 contained MW markers (data not shown). The lipase was the least specific enzyme and it degrades the cofactor and leaves most of the factor running at 16 kD. This type of phospholipase C can cleave at both glycerol phosphates and ceramide phosphates. The sphigomyelinase is specific for ceramide phosphates. Both of these enzymes cause single breaks and this alters the binding/mobility but does not free the cofactor from the factor. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 7 Chloroform/methanol extraction does not free the lipid bone cofactor from the protein factor cm from u2os-pesy1c was extracted with chloroform/methanol. the protein fraction which : precipitates at the interface between the two solvents was examined by Western blotting using a myc monoclonal antibody (lane 1 is half the sample volume of lane 2). Despite unusual sample preparation, the SDS gels and Western blotting worked reasonably well. Nevertheless, there was no evidence of a 16 kD free protein factor. Instead the bands resembled those of the factor/cofactor running together with a minimum size of 28 kD. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 8 1086 or 1114.88. : PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 10 Working model is shown for the bone cofactor Using our knowledge of the lipid enzyme preferences, the molecular weight of the whole molecule, the predicted molecular formula, and the fragmentation pattern (MS/MS), a preliminary model has been drawn with a ceramide phosphate, a single chain glycerol lipid, and a small linker. This model also explains why this molecule is soluble from aqueous buffers to chloroform: a structure with a hydrophilic end and hydrophobic tails. The model also shows how the fragmentation pattern of two breaks (MS/MS) is consistent with the model. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 11 Characterizing the tendon cofactor by its sensitivity to specific lipid degradative enzymes PAT cell CM was put through a 30 kD filter and then concentrated with a 10 kD filter. This concentrated CM after several months at 4o C forms a dominant aggregate that runs at ~60 kD. This CM was run on Westerns (using a polyclonal antibody) as follows: untreated (panel a, lanes 2,6); lipase treated (panel a, lanes 3,7); phospholipase A2 (panel a, lanes 4,8); and phospholipase C (panel a, lanes 5,9). The data indicates a strong sensitivity to lipase and phospholipase C but not to phospholipase A2. This is identical sensitivity pattern as seen with the bone cofactor. Next sphingomylinase was tested (panel b, lanes 5,6,7) and compared to the untreated control (panel b, lanes 2,3,4). In contrast to the bone cofactor, sphingomyelinase did not affect the tendon cofactor. To rule out that the sphingomyelinase was being inhibited by the aggregation caused by \u201caging\u201d the CM, fresh cultures of PAT cells were made and the CM treated with sphingomyelinase (panel c, lanes 4,5) and untreated control (panel c, lanes 2,3). Again, this enzyme does not alter the mobility of the tendon factor/cofactor. PeerJ reviewing PDF | (v2013:06:612:1:0:NEW 4 Sep 2013) R ev ie w in g M an us cr ip t Figure 12",
"v3_text": "discussion : In 1976, we showed that freshly isolated chicken embryonic tendon cells were exquisitely sensitive to cell density as reflected in the levels of procollagen production(Schwarz et al. 1976). Despite this dramatic regulation of the most important protein in tendon morphogenesis, the paper went on to say that the \u201cmechanism by which one cell communicates its presence to another cell is unclear.\u201d A few years and many experiments later the signaling mechanism is now shown to use a tight complex between a protein and a lipid. The protein is a small cleavage product from the carboxy end of a larger gene product and this small protein is subsequently secreted from the cell. The protein itself appears to have little or no activity but requires the binding of a lipid cofactor. Remarkably, the composition of the lipid is tissue-specific. While only two types have been characterized, they are each unique in both the lipid structure and the triggered response. SNZR 1P is also unique in composition, processing, and function. By being able to bind unique phospholipids, SNZR 1P allows diffusion into the medium of a lipid that would otherwise be expected to be tightly embedded in the membrane. In earlier papers, this trait was described as being \u201cloosely bound to the cell layer\u201d an explanation for its ability to localize to the cell layer but be partially removable by gentle agitation(Schwarz 1991; Stoker 1973; Zayas & Schwarz 1992). Being able to identify SNZR 1P and its lipid partners gives 14 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t this complex a molecular identity that explains its unusual characteristics. In addition, as more has become known about the SNZR 1P1L, similarities can be seen to other signaling molecules \u2013 in particular, sonic hedgehog. There is no sequence homology but they are both made as large transcripts that are processed to a smaller protein and then secreted from the cell(Weed et al. 1997). In the case of sonic hedgehog, it is the N-terminus that is secreted and in this case, it is the C-terminus. The activity of sonic hedgehog is greatly enhanced by the covalent linkage of palmitate and/or cholesterol(Taylor et al. 2001); with the cell density signaling the protein itself has little activity but this is greatly enhanced by binding to a lipid cofactor. Because of the lipid attachments, sonic hedgehog runs as an aggregate on SDS gels(Goetz et al. 2006). The binding of the cofactor causes a similar effect on the SNZR 1P. On the other hand, the differences are equally important: SNZR 1P can bind a variety of lipids and in doing so can change its signaling function in a tissue specific manner. Besides how this complicates the analysis of gene function, the result is that the cell has the ability to sense both the type and number of cells in its immediate vicinity and thereby react accordingly. Given the nature of the complex, the simplest hypothesis is that the protein holds the lipid in a specific conformation that can then bind to the cell membrane and organize it(Simons & Toomre 2000). With more cells there would be additional binding resulting in a more organized plasma membrane and this would trigger a stronger signal transduction. But we should point out that our focus has been on the carboxy end of the transcript yet the rest of the protein is equally well conserved between the chick and the human. Another group searching for homology to other tubulin stabilizing proteins found a 20% overlap with this gene(Bartolini et al. 2005). By Northern analysis, many tissues appeared to transcribe this gene at low levels while testis produced it in larger amounts. Over expressing this gene in HeLa cells caused tubulin to be destabilized in contrast to their initial selection criteria. Because they tagged the transcript at the amino terminus, the secreted form would have gone undetected. Moreover, they pointed out that they could not distinguish whether the destabilization of tubulin was a direct or an indirect effect. If the effect on tubulin was an indirect effect of cell density signaling in HeLa cells, then the conservation of sequence could be as simple as conserving the 3-dimensional structure of the cleavage site, or another alternative could be as an intracellular part of the cell density signaling process. The role of the larger, N-terminus protein, will require further analysis to resolve why it is so highly conserved. Our focus has been on molecular characterization of the sensing mechanism for cell density regulation that generates a proliferative signal, but the cell usually requires a more sophisticated response. For instance, in U2OS-pESY1c cells expressing SNZR 1P1L one might postulate that as the cells overgrow the monolayer they begin to produce a secreted inhibitor to prevent the cells in the valleys from producing SNZR 1P1L and dividing: a classic Turing model(Murray 1988; Turing 1952). In tendon cells, more is known and with the knowledge of the cofactor one can adjust a previous model to reveal how cells form a growth plate and then fill a growing tube (fibril) with collagen. To do this the cells need a timer, a growth stimulator, a growth inhibitor and a trigger for apoptosis. Mathematical modeling has shown that this can be accomplished with two factors and the cell(Schwarz 1996). The first factor stimulates growth and the production of the second factor. The second factor, which does not diffuse into the medium; has no activity on its own but by interacting with the first factor changes it to an inhibitor of growth, and a stimulator of 15 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t procollagen synthesis. Cells left at high cell density cause a drop in the production of the first factor triggering apoptosis. The first factor can now be seen as SNZR 1P2L that we have described. The second factor has until now been hypothetical but a good candidate would be the free cofactor (SNZR 2L). Their interactions occur in the plasma membrane and their consequences are predicted to further organize the membrane structure that in turn can alter the internal state of the cell. In a permissive cell culture environment for tendon cells \u2013 very low serum(0.2%) and addition of ascorbate \u2013 these factors come to the forefront and results in the \u201cpeculiar\u201d behavior of primary cells in culture(Schwarz 1991; Schwarz 1996; Schwarz & Bissell 1977; Schwarz et al. 1976; Schwarz et al. 1979; Valmassoi & Schwarz 1988; Zayas & Schwarz 1992). The cells isolated from 16 day chick embryos were programmed to rapidly make a tendon by way of a growth plate in order that the hatched chick is able to walk to find food and water. In the research laboratory, however, we wanted tendon cells to conform to normal cell culture protocol: grow on a 2-dimensional plastic surface from low density to high density; then be trypsinized off the plastic and the process repeated. Meanwhile the cells were trying to interpret these changes in terms of its own programming for generating a tendon \u2013 a collagen rope -- by regulating collagen production and cell proliferation. As a result we found that: PAT cells only made high levels of procollagen at high cell density (~50% of total protein synthesis). Left at high cell density for several days procollagen production declines. Splitting high density cultures resulted in cells that grew to a lower confluent density and produced a lower level of procollagen. Seeding cells a low cell density resulted in uneven cell growth. Seeding cells at low cell density as a small island in the middle of a dish resulted in the whole island apoptosing after a few hours. Gentle agitation of confluent cells seeded as an island in the middle of dish \u2013 high medium to cell ratio \u2013 caused cells to grow. Adding back CM to an island of confluent cells caused cells to grow. Taking all this information and developing a 2 factor model to explain it was the first step; the second step was to mathematically model the behavior of the cells in 3-dimensional space. Their programming generated a growth plate that was confirmed in vivo at the tendon/muscle junction(Schwarz 1996). At the growth plate one can postulate how tendon cells normally respond: at the leading edge cells at moderate cell density are growing driven by their production of the SNZR 1P2L. As one moves towards the trailing edge the cell density increases as does the level of SNZR 2L. This slows proliferation, increases procollagen production and begins a decline in the level of SNZR 1P. At the trailing edge the cells have been at high cell density producing high levels of procollagen for a longer time and the buildup of SNZR 2L causes a continued decrease in the SNZR 1P. This drop eventually triggers apoptosis although some cells at the periphery of the tube (fibril) survive. This is only an outline with the principal players identified (Fig. 15). Nevertheless, a \u201cpeculiar\u201d behavior in one setting, becomes \u201cnormal\u201d in another. Analyzing molecular changes raises the question of why does the cell need an additional method for signal transduction? Why not use a protein to activate a receptor? Or put another way, what is the advantage of using a diffusible lipid? In the case of activating a receptor, the cell responds in a graded manner in a set direction. Adding a growth factor can stimulate a cascade of events that results in more rapid cell proliferation. However, the participation of a cell within a growth plate is a far more complex process requiring the cell to respond to parameters related to both space and time. By detecting the levels of SNZR 16 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t 1P2L and SNZR 2L, the cell can determine its location within the growth plate, whether it should be growing or slowing down, producing high levels of procollagen, or apoptosing. Having a diffusible lipid be a major signal makes the system sensitive to subtle and rapid changes; and with all of the sensors and modifiers acting in one place creates the biological equivalent of an analog logic board allowing the cell to respond to a complex environment correctly. Key elements to this \u201ccircuit\u201d require unique lipids yet this is not a common type. Because they are present at low concentrations, their detection is difficult. We were only able to analyze the two unique lipids because of their tight binding to a protein to which we had both tag and polyclonal antibodies. On the other hand, the tight binding of a protein made purification by standard lipid extraction methods impossible. But the critical observation is that production of a chick protein made in tendon with high homology to the human equivalent causes a unique human bone lipid to be produced in U2OS cells. This makes the protein look like a cassette that can bind a variety of different lipids. But this raises the question of how many kinds are there? If they are in tendon that uses a growth plate to form tubes filled with collagen, do we expect to see unique lipids in other tissues that form tubes \u2013 filled (structural tissues) and unfilled (ductal glands) \u2013 using a growth plate? Similarly we have seen one case where two cell types share a common protein. But how many cell types share this protein and are there alternative proteins for other cell types? Taking into account the unknown number of unique lipids, one can ask how many unique cell density signals exist in an organism? Finally, even though our initial aim in selecting an osteosarcoma cell line (U2OS) was simply to use its cellular machinery to express the chicken cell density signal, these cells were not passive responders to the expression of a chicken cDNA. Instead they used this expression to make more of their own cell density signal, and in turn structures resembling spongy bone with the production of an early osteoblast specific marker. This complex differentiation program was retained by these cells even after becoming malignantly transformed and spending 48 years in cell culture. This indicates that the right signaling molecules can push the malignant cell in new and potentially stable directions. This is not unique to U2OS cells. Using malignant mammary cells(Weaver et al. 1997), \u201cthese studies show that despite a number of prominent mutations, amplifications, and deletions, signaling events which are linked to the maintenance of normal tissue architecture are sufficient to abrogate malignancy and to repress the tumor phenotype.\u201d Again, melanoma cells(Postovit et al. 2006) grown on matrices preconditioned by human embryonic stem cells \u201cinduced melanoma spheroid formation, promoted the re-expression of Melan-A, and inhibited melanoma cell invasion.\u201d With cell density, the signaling mechanism is beginning to be defined at the molecular level. This raises the question of whether these molecules can be used in a tissue-specific manner to effectively manipulate the differentiated state in order to stimulate the healing of injured tendons and bones, and as a new therapeutic approach to reversing the uncontrolled proliferation of cancer cells by directing cells to normal morphogenesis. The latter would be particularly ironic if studies focused on the differentiated state of tendon cells \u2013 the cell type with perhaps the lowest rate of malignancy \u2013 provided new insights into controlling this disease. 17 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t in addition, a polyclonal antibody made in rabbits to the chicken gene product expressed in e. coli. : was tested (lanes 11-16). For each case, the same samples were used: MW markers (lanes 1,6,7) the retentate from trying to separate the factor from the cofactor at 1X (lanes 2,7,12) and 2X (lanes lanes. The free factor made in E. coli was run in lane 1 and the concentrated CM was run in lanes 3, 4. the middle. The antibody staining for the factor shows fairly intense staining compared to the muscle bundles but the connective tissue that surrounds the muscle stains lightly as well. Staining the same section with DAPI shows that this part of the tendon is full of nuclei reflecting that this is a growing end (panel b). One has to take into account that the tendon does not have to be in exactly the same plane as the tissue slice and therefore, the nuclei density can vary as the tendon weaves through the muscle. The connective tissue surrounding the muscle shows nuclei staining with DAPI that looks like a line of nuclei. The light staining of the muscle connective tissue with the polyclonal antibody is suggestive of yet another cell density signal but the low stringency of the washes in this experiment requires further analysis to confirm its specificity.Scale bar =100 microns PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 15 Schematic representation of tube formation \u2013 a critical process in diverse tissues. Glandular tissues clearly need ducts (tubes) to secrete their products. Tendons, a structural tissue, also forms tubes but fills them with collagen to make fibrils. This solves the problem of how to make a collagen thread that can be long but maintain a precise diameter for uniform strength. The key element is a growth plate where cells are proliferating and moving forward on one side (right) while they are mostly dying on the opposite side (left). To accomplish this requires two factors: SNZR PL (blue) and SNZR L (dark yellow). Growth is stimulated at moderate cell density by the production by many cells of SNZR PL \u2013 the factor is found in the fluid surrounding the cells as well as bound to the plasma membrane. The binding of SNZR PL not only stimulates growth but the production of a non diffusible second factor that we postulate is SNZR L. This causes another change in the cell membrane and some cells slow growth and become more differentiated which we show as oval cells becoming retangular in the transition zone. As cells become more dense, they accumulate more SNZR PL and SNZR L and become highly differentiated. The exception is cells at the outer edge of the tube that due to their position never reach high cell density (narrow ovals). The high level of SNZR L causes a drop in the production of SNZR P and this triggers apoptosis. In tendon, the differentiated cells produce half of their total protein synthesis as procollagen and they fill the tube with collagen and become a fibril. The cells at the edge are not necessarily contiguous or ordered. One can extrapolate that in a gland, the differentiated zone would produce a variety of mature phenotypes that could escape apoptosis and form a tight single or multi-cell layered duct. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t a simplified procedure using ultrafiltration and high salt releases the bone cofactor. : mass spectrometry analysis : The diluted lipid extract samples were analyzed on an ABSciex TripleTOF 5600 (AB Sciex, Foster City, CA). Samples were introduced to the mass spectrometer via syringe pump by using a Nanospray III source (ABSciex) with a nano\u00a9tip emitter (New Objective, Woburn, MA) operating in positive\u00a9ion mode (2400 V). The data were acquired with Analyst TF 1.5.1 by averaging the signal from 400 m/z to 1250 m/z over several minutes. MS/MS spectra were collected in \"high sensitivity\" mode with the quadrupole set to UNIT resolution and collision energy set to 30 to optimize fragmentation. MS/MS spectra were scanned from 100 m/z to 1600 m/z and were collected for a total accumulation time of 500 ms. Results 7 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t The gene and potential protein products Previously 400 liters of conditioned media (CM; using gentle agitation) from PAT cells were put through 4 columns and the band on an SDS gel running at 16 kD tracked best with activity (stimulating an island of PAT cells to proliferate) was sequenced and yielded: EPLAVVDLTEKTIS(Schwarz 2002). This chicken sequence does not have an immediate correspondence to the human equivalent for two reasons. First, as with all N-terminal Edman sequence data, with each cycle the noise level increases reducing the confidence of each additional read. Not unexpectedly, the chicken genome shows an exact homology only with the first 8 AA. Second, the human genome has a variant in the fifth position from V to E. Nevertheless, this 8 letter AA sequence in the chicken comes from a much larger gene with a gene product of 424 AA from seven exons. The homology between the chicken and the human genes at the amino acid level is 89% identical and 94% positive. For this large gene to be correct requires further processing of the protein product to a smaller secreted form: a specific cleavage at the EPLAVVDL site would give the right starting sequence; however, this sequence is found near the carboxy end and would only produce a protein with 94 AA, ~10 kD. In summary, the high homology between chicken and human is striking for such a distant relationship from a common ancestor but this, by itself, does not confirm a significant role for this protein. Moreover, the potential 94 AA protein is small for the observed molecule weight on SDS gels of 16 kD inferring that if this is the correct gene that some modification or tightly bound cofactor may be involved. The next step is to test whether this is the correct gene. Expression in U2OS cells causes differentiation The classic proof that a gene codes for a given function is to clone it, express it in a different cell type, and then show that the purified recombinant protein has the same activity on the original cells. In this case, the chicken gene could be cloned and expressed in a non chicken cell line. The secreted form could be purified from shaken CM and tested back on PAT cells. To simplify detection and purification of the recombinant protein, a myc and his tag were added to the carboxy end. A replication defective retroviral vector containing a neomycin resistance gene was used to transfect the chicken gene into U2OS cells(Koh et al. 2002; Lee 2007). Because this study was started when the chicken genome was first sequenced, the chicken equivalent to the first exon was not present. So the insert contained only the last six exons (pESY1c). U2OS cells, derived from a human osteosarcoma tumor, showed stable expression of this gene after infection when grown in low serum media (0.2%)(Valmassoi & Schwarz 1988). Despite the low serum levels, U2OS cells transfected with pESY1c would start to overgrow the monolayer as small ridges. The ridges would coalesce into circles sharing boundaries with other circles that resembled spongy bone in a two dimensional space (Fig. 2b). Uninfected U2OS cells showed some ability to overgrow the monolayer but could not maintain the growth needed to form completed structures (Fig.2a). The multicell ridges also started to show the onset of bone production as reflected in the production of alkaline phosphatase on the cell surface (Fig. 2c). Pattern of molecular expression intracellularly and extracellularly The expression of pESY1c in U2OS cells directed a unique phenotype in human 8 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t bone cells derived from an osteosarcoma tumor but the putative processing of a large transcript to obtain a much smaller protein outside the cell requires further confirmation. Because the transfected U2OS cells expressed a myc and his tagged chicken cDNA, the processing of this large protein could be observed by Western blotting. In PAT cells the secreted protein ran with an apparent MW of 16 kD on SDS gels; extrapolating, the tagged protein would be predicted to have an apparent MW of ~ 20 kD. A monoclonal myc antibody and Western blotting was used to probe the expression and processing of this myc tagged gene. Two specific bands were seen intracellularly: ~44 kD band which is expected for the full size gene product; ~26 kD band that could be a cleavage product (Fig. 3). However, this band is larger that the predicted 20 kD protein. To observe the secreted forms a simple purification was used whereby the CM was put through a 30 kD ultrafilter. Both the retentate and the filtrate were analyzed and both had to be further concentrated on a 10 kD ultrafilter in order to load sufficient protein on the gel for Western analysis (Fig. 3). The retentate contained many bands: the smallest was a weak band at ~28 kD, the strongest was at ~60 kD, and the largest was over 200 kD. The filtrate showed basically a single band at ~60 kD. This was unusual since 60 kD proteins are excluded to a high degree from passing through a 30 kD ultrafilter. One explanation is that by concentrating the sample ~20 fold this caused the tagged protein to aggregate into stable forms that resist the denaturing effects of heat, beta mercaptoethanol, and SDS. In summary, this data shows both positive and negative findings. Intracellularly, the large protein does get processed to a smaller form but this appears larger than predicted. A smaller tagged protein does get secreted outside the cell but this protein is larger than expected and it appears to aggregate to stable forms when concentrated. acknowledgments : RS would like to thank three people at Amgen for their knowledge, experience and hard work in purifying factors \u2013 Bill Kenney, Harvey Yamane, and Mike Bendor. Bill\u2019s encouragement that \u201cthis project is hard because all the easy ones have been done\u201d turned out to be true. Eva Lee for pulling lots of embryonic chick tendon, collecting 400 liters of conditioned medium, and working through many problems. The members of the Bissell group past and present who gave advice and encouragement to an eccentric biologist investigating how cells \u201csee\u201d their neighbors. Recently, Ana Correia, Cyrus Ghajar, Jamie Inman, Hidetoshi Mori, Ramray Bhat, Joel Hyman and Sabine Becker-Weiman knowingly and unknowingly helped get this work to the finish line. Finally, Mina Bissell for her contagious enthusiasm for good biology. We would also like to thank Peter Rademacher for his help in analyzing the mass spectrometry data and giving us many useful interpretations. 18 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t 19 1 2 3 4 5 6 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t conclusions : Cell density signaling is accomplished in part through a protein/lipid complex that allows the lipid to influence the plasma membrane of the cell but still enables the complex to diffuse in the extracellular space. As the cell density increase more of the complex binds to the membrane and this is postulated to organize the membrane and allow for signal transduction. The lipid has been shown to be cell type specific and to trigger a cell type specific response. Cell density regulation allows for more intricate regulatory schemes. In tendon, this results in the formation of a growth plate that fills a tube with collagen forming a fibril that is the building block of this tissue. tissue localization in mice : The initial aim of this paper \u2013 producing the active form of the tendon cell density signal in another cell type and then applying it to tendon cells to show it has the original activity \u2013 is not compatible with our data. Since the cofactor that is essential for activity is unique (in two cases) to the cell type producing it, another test would be useful in showing that the protein factor is the correct candidate for being part of the cell density signaling mechanism. There is the similarity that both the tendon SNZR 1P2L and bone SNZR 1P1L are loosely bound to the cell layer and cause the cells to differentiate. In addition, the polyclonal antibody to the 94 AA protein does detect a band at 16 kD in the PAT cell CM (Fig. 11C). A more functional test would further confirm the identity. 13 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t To accomplish this, the polyclonal antibody was used to check whether the factor was at the tendon/muscle junction of 5 week old mice. This is the growing stage for mice and one would predict that the factor would be at high concentration at the growing ends of the tendon; there should also be a high density of fibroblasts(Schwarz 1996). To detect the protein factor, a polyclonal rabbit antibody was used with a secondary antibody labeled with alexa 568; to detect the nuclei, the sections were stained with DAPI (Fig. 14). At the growing end of the tendon there is a strong fluorescence signal for the factor (Fig. 14a) as well as numerous nuclei labeled with DAPI (Fig. 14b). There is variation in the signal over the length of the tendon but this can result from the plane of the tendon not being in exactly the same plane as the section. The strong staining for the factor at the growth plate of the tendon supports its role as a cell density signal. One unexpected observation is the light staining of the muscle connective tissue with its single file nuclei (Fig. 14a,b). This could be an indication that a third cell type is producing this protein factor along with their own unique cofactor. However, to be confident that the signal is specific, alternative methods for improving the stringency of the washes need to be developed that are compatible with a lipid cofactor. material and methods : Cells, cell culture, and animals PAT cells were isolated from 16 day chick embryos by a modification of the 3 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t procedures described by Dehm and Prockop(Dehm & Prockop 1971). Specifically, a higher serum level (3%) was used in the dissociation medium and the use of a nylon mesh (20 micron; TETKO) instead of filter paper to separate out single cells. In these experiments the medium was changed from F12 to a 50/50 mix of F12/DMEM in order to be consistent with the medium used to grow the U2OS cells. The medium contained 0.2% fetal bovine serum deactivated for 30 m at 56o C. U2OS cells (from frozen stocks that have been in the laboratory for over a dozen years) were grown in F12/DMEM medium with 0.2% fetal bovine serum. In 75 cm2 flasks the medium, 50-75 mls, was changed every other day. When the cells first overgrow the monolayer, they are very sensitive to medium volume and will apoptose if too much medium is added. CM was produced by growing cells in 225 cm2 flasks. Medium was changed to serum-free medium and the volume reduced to 12 mls. The flasks were placed in an environmental rotator at 100 rpm, 37o C for 1 h. The CM was collected and new medium added, and the process was repeated for the third hour. After the 3rd collection medium with serum was added and the flask returned to the regular incubator. The 16 day chick embryos were exempt from requiring an animal use protocol as determined by LBL Animal Welfare and Research Committee. Similarly, the legs from 5 week mice (BALB/c) came from mice already euthanized for other approved protocols. This is also an exempt use. Alkaline Phophatase Assay The Sigma kit was used (85L2-1KT) and the included procedures were followed with only slight modifications for labeling tissue culture flasks instead of slides. genes, cloning, and expression : RNA was extracted from confluent cultures of PAT using Trizol (Invitrogen). The RNA was reversed transcribed using superscript II (Invitrogen). The cDNA was amplified by PCR using primers from the 5' end of exon 2 and the 3' end of exon 6. This was cloned and sequenced yielding the expected sequence in agreement with chicken genome predicted expression of this gene (accession: XP_427094). At the time the sequence for the chicken exon 1 was not known. So a 3 AA sequence containing methonine was added to the 5' end (above the dashed line below) and myc and his tags added to the 3' end (below the dashed line below). This was inserted into a retroviral vector pESY-Neo-II \u2013 retroviral pEYK 3.1(Koh et al. 2002) was modified to remove the GFP and bleomycin dna and replace it with Neo/Kan antibiotic resistance gene and a multi-cloning site(Lee 2007) -- and renamed pESY1c: M K D ---------------------------------------------------------------------------------- ---- R L N L P S V L V L N S C G I T C A G D E N E I A A F C A H V S E L D L S D N K L E D W H E V S K I V S N V P H L E F L N L S S N P L S L S V L E R R C A G S F A G V R K L V L N N S K A S W E T V H T I L Q E L P D L E E L F L C L N D Y E T V S C S P V C C Q S L K L L H I T D N N L Q D W T E I R K L G I M F P S L D T L I L A N N N L T T I E E S E D S L A R L F P N L R S I N L H K S G L H C W E D I D K L N S F P K L E E V K L L G I P L L Q S Y T T E E R R K L L I A R L P S I I K L N G S I V G D G E R E D S E R F F I R Y Y M E F P E E E V P F R Y H E L I T K Y G K L E P L A V V D L R P Q S S V K V E V H F Q D K V E E M S I R L D Q T V A E L K K H L K T V V Q L S T S N M L L F Y L D Q E A P F G P E E M K Y S S R A L H S Y G I R D G D K I Y V E P R M K 4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t ---------------------------------------------------------------------------------- ---- S N S A D I Q H S G G R S S L E G P R F E Q K L I S E E D L N M H T G H H H H H H Alternating normal and bold/italic fonts indicate exons and the underlined AAs in last exon indicate the secreted form of the gene. The vector pESY1c (4 \u00b5g/60 mm2 dish) was used to transfect phoenix cells using lipofectamine 2000 using the manufactures protocol (Invitrogen). The cells were incubated overnight and the next day medium was collected, filtered (0.45 \u00b5), and used to transfect pESY1c into U2OS cells in the presence of 8 \u00b5g/ml polybrene. The transfected cells were selected using G418 (100 \u00b5g/ml). polyclonal antibody : The secreted form of pEsy1c (underlined AA above) with a methionine at the N-terminus was inserted into the E. coli expression vector pTrcHis2-TOPO (Invitrogen). The vector included myc and his tags at the C-terminus. This yielded an expressed protein that was 7 AA smaller in the linker between the protein and the tags. This difference was too small to detect on gels and the linker region was not important for antibody production. The protein was expressed in E. coli and purified and nickel columns (Invitrogen). This protein (4 mg) was used to immunize rabbits and for affinity purification (Biosynthesis). immunofluorescence : Cell culture: cells were fixed with 4% paraformaldehyde for 10m in a high calcium buffer (.05M TES pH 7.2, 0.1M NaCl, .05M CaCl2). The fixation was quenched with two, five minute washes in buffer A (.05M TES pH 7.2, 0.1M NaCl, .01M CaCl2 ) containing 0.1M glycine. A blocking agent (1% casein Hammerstein [EM Science]) was used for 30 m followed by a rinse with buffer A. The primary antibody was a polyclonal made in rabbits against the secreted form of pESY1c plus tags expressed in E. coli. This was added at 1:40 (the antibody was diluted by 25% with stabilizer [Candor]) for 2h with agitation. Then, washed 3x with buffer A with agitation and then a fluorescently tagged goat anti-rabbit secondary was added for 1h with agitation. Washed 3X with buffer A and then once with a DAPI buffer ( 500nM in PBS). Finally, a wash with buffer A. Tissue sections: hind legs from 5 week old mice were dissected to reveal the muscle and the attached tendon. These were embeded in OCT (Sakura) for 10 to 30m and then frozen in small embedding molds in a dry ice/ethanol bath. Frozen tissue was sliced into 16 \u00b5 sections and attached to superfrost plus slides (VWR). The sections on the slide were outlined with a pap pen and then analyzed as described above. In this case, because the tendon had less autofluoresence in the red, this was the preferred color for the secondary emission. Because the plane of the knife and the plane of the tendon were infrequently in agreement, the analysis required many legs and sections. western analysis : Samples were concentrated on 10 kD spin ultrafilters (Millipore) diluted with 2X sample buffer and concentrated again to <200 \u00b5l and 40 \u00b5l or less was applied and run on an SDS gel (4-20%, 1.5mm, 10 well, Tris-Glycine [Invitrogen]). The gels were blotted onto 5 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t nitrocellulose (0.45 \u00b5), washed with water, and dried overnight. Next, the blots were rehydrated with water and then blocked with 1% Hammerstein casein for \u00bd h. After 2X water washes the primary antibody (monoclonal myc (Millipore) or polyclonal rabbit, see above) was added at 1:1000 in PBS with 0.25% casein Hammerstein for 2 h. The blot was then washed 3X for 5m in PBS with .02% Tween 20. After which the rabbit secondary antibody conjugated to horseradish peroxidase (Thermo) was added at 1:1000 in PBS with 0.25% casein. Again the blot was washed 3X for 5m in PBS with .02% Tween 20 and then developed with a Dura or Femto kits (Thermo). The blots were photographed with a FluoroChem HD2 (ProteinSimple). antibody selection and specificity : Antibody affinity to a protein could be strongly affected by a lipid cofactor. This became an issue when the monoclonal antibody (Millipore) to the myc tag stopped working. Over the course of these experiments Millipore changed the storage buffer several times and finally to a liquid stored at 4o C. This latter change reduced the affinity of the antibody so that only the free protein factor was detected (data not shown). To find an adequate replacement, three alternatives were tested: a rabbit polyclonal antibody to the factor expressed in E. coli; a monoclonal antibody to the myc tag from Applied Biological Materials (clone A7) and one from GenScript (clone 2G8D5). Using the same CM sample for each antibody yields a Western blot that is unique to the probing antibody (Fig. 1). Critical to our purification procedure for the cofactor, only the polyclonal antibody can detect the free factor as well as bands with the factor/cofactor. Most importantly, one cannot compare the intensity of one band to that of another because each antibody has its preferred conformations; instead, one can only compare the relative amount of a specific band after different treatments. cofactor purification : The CM was filtered through a .22\u00b5 filter and then the NaCl level was increased by adding 1/10 the volume with 5M NaCl. The CM, 10 to 50 ml, was concentrated to 1ml in a stirred cell (10 ml, Millipore), at 700 rpm and pressurized (14 psi, N2). The filter was a YM-10 with a nominal cutoff at 10 kD (Millipore). The retentate was diluted to 10 ml with water and reconcentrated to 0.5 ml. The retentate was analyzed on Westerns to see the level of free factor released. The filtrate containing the cofactor was concentrated using a Pierce C18 spin column (Thermo). The column was wetted with 1 ml of 50% acetonitrile/water and then spun for 1m at moderate speed in a clinical centrifuge. Then, 1ml of the filtrate was added and spun and repeated until all the filtrate was bound to the column. After which the column was washed twice with 1ml of water, and twice with 1 ml of acetonitrile. Then, the washing was repeated: 1ml of 50% acetonitrile, 1 ml water, 2X 750 \u00b5l acetonitrile, and elute with 2X 750 \u00b5l of chloroform. Samples were stored in chloroform and concentrated before analysis by blowing down with nitrogen to ~100 \u00b5l. For mass spectrometry, 10 \u00b5l of sample was diluted with 80 \u00b5l of acetonitrile and 10 \u00b5l water. All solvents were lc,ms, or hplc grade. With recently purchased YM-10 filters (in comparison to using older filters from the original maker Amicon), the cofactor would not separate from the factor unless the following changes were made: the filters had to soak at 37o C in 20% acetonitrile for 48h; 6 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t after concentrating the high salt CM to 1 ml, it was put into 10ml of buffer (10 mM EDTA, 2 ml acetonitrile, 3 ml isopropyl alcohol, 5 ml water) and heated to 80o C for 2h and then concentrated to 0.5 ml. The filtrate was further concentrated and purified as described above. The cofactor for bone and the one for tendon are unique. The SNZR 1P can bind either SNZR 1L or SNZR 2L and drive a unique differentiation pathway dependent on the type of cofactor produced by the cell. A critical question is whether the SNZR 1P2L when added to U2OS cells is capable of driving bone differentiation or does it elicit a new or null response? In addition, one can ask whether the protein factor, by itself, has any activity? To answer these questions, U2OS cells were grown in regular media as a control, or with 5X concentrated condition media from U2OS-pESY1c cells expressing the SNZR 1P1L, and from 5X concentrated condition media from PAT cells expressing SNZR 1P2L. In addition, SNZR 1P purified from E. coli, was added at 5 \u00b5g/ml \u2013 3 orders of magnitude greater than the estimated level in CM (Fig. 13a,b,c,d). These levels were chosen because early work with PAT cells have shown that 1X CM is insufficient to stimulate growth when added back to confluent cultures(Zayas & Schwarz 1992). The high level of the protein factor was chosen because it seemed unlikely that the factor without the cofactor would have a lot of activity and small amounts of activity are difficult to detect by visualization. After a week in the various mediums, the most dramatic effect was that the tendon CM was inhibitory to growth when added to the U2OS cells yielding a subconfluent culture. The addition of the SNZR 1P alone did not appear to change the cultures compared to the control even at this high concentration. Adding 5X CM from U2OS cells expressing the SNZR 1P1L, showed some increased in growth as the cells appeared to be more compact. However, to confirm this observation we allowed this flask and its control to grow for additional 9 days, at which point it was obvious that the CM had caused the cells to completely overgrow the monolayer (Fig. 13e,f). What is particularly striking is uniformity of growth when the CM is added to cells versus the lattice patterns that are formed when the cells themselves express the SNZR 1P1L (Fig. 2b). lipid enzyme treatment : Lipid degradative enzymes were first used to identify the makeup of the cofactor. In this case, we were looking for a change in the mobility of factor/cofactor on a SDS gel after being treated with these enzymes for various times 30m, 90m, 180m, and a no enzyme control. CM (3ml) was concentrated to 0.5ml. Then the enzymes were added as detailed below and incubated for the specific time, 2X sample buffer was added to stop the reaction, concentrated on a 10 kD ultrafilter, and analyzed by Western blotting. Lipase: Sigma (L8525); 380 units/\u00b5l; added 50 \u00b5l; incubated 37o C Phospholipase C: Sigma (P7633); .038 units/\u00b5l; added 50 \u00b5l; incubated 37o C Phospholipase A2: Sigma (P9279); 6 units/\u00b5l; added 50 \u00b5l; made 20 mM triethanolamine pH 8.9; incubated room temperature Sphingomyelinase: Sigma (S9396); 0.54 units/\u00b5l; added 2 \u00b5l; incubated 37o C The second use of lipid enzymes was to identify the peaks on mass spectrometry that were sensitive the enzyme treatment. In the cofactor purification above, between the wash steps, the column material was resuspended in buffer (20 mM TES pH 7.2; 10 mM NaCl; 1 mM CaCl2) and incubated for 30 min at 37o C with agitation. This was spun down and the protocol above continued. In the buffer, when using sphingomyelinase, CaCl2 was replaced with MgCl2. aggregation is due to a cofactor : The multiple bands seen in the Western blot complicates the interpretation of how this tagged protein is working in the extracellular space. But more specifically, even the smallest MW band (~28 kD) seen outside the cell is higher than predicted. Is this larger size due to a different amount or type of modification in bone cells versus tendon cells? Or is this increase in apparent MW a reflection of an interaction with another molecule? To begin to answer these questions, we needed to resolve how the unmodified protein factor with two tags migrates on a Western. To accomplish this, the tagged protein was expressed in E. coli, purified on a nickel column, and then used as a standard in Western blots. The tagged protein ran with a MW of 16 kD (Fig. 4, lane 6). This confirms that there is either a covalent modification or a binding to another molecule that gives rise to the 28 kD form in U2OS cells. Further aggregation could then occur when concentrating the factor to give rise to larger MW forms. This raises the question of whether different physical or chemical environments could alter the size of the bands observed on Western blots. CM from U2OS-pESY1c was put through a 30 kD ultrafilter and the filtrate was then concentrated on a 10 kD ultrafilter. This was divided into 2 parts. One was further concentrated on a 10 kD ultrafilter for loading on a gel but in this case, 30% DMSO was added. To the other sample, salt (0.5 M) was added; concentrated with an ultrafilter; then diluted to low salt; and then concentrated to apply to a gel (Fig. 4). The effect of DMSO was to reduce the number of bands, especially those at high MW, but the smallest band still remained 28 kD. In contrast, the treatment with high salt, low salt and ultrafiltration yielded a much simpler picture where the bands 9 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t were reduced to the 16 kD band and its dimer at 32 kD. One can conclude that the 28 kD band is due to a binding of a cofactor and not a covalent modification. Localization and characterization of the factor/cofactor in the extracellular space The previous experiment raises the question of how the untagged 16 kD form in tendon cells compares to the 28 kD tagged form seen in bone cells. Even taking into account a 4 kD tag, there is an additional 8 kD shift in MW. Are these forms variations with common characteristics or is the bone form acting in a different manner? In tendon cells the 16 kD form is loosely bound to the cell layer. This gives a key feature to this form in that gently shaking a flask of tendon cells causes them to loose their ability to sense high cell density, resulting in increased proliferation. First, the localization was examined with immunofluorescence using a rabbit polyclonal antibody specific to the tagged secreted form of the chicken protein expressed in E. coli. During tests to optimize the signal, procedures that disrupted membranes decreased the signal. Using an optimized protocol, immunofluorescence staining was localized to areas where the cells have overgrown the monolayer (hills) and expression was suppressed in monolayer areas (valleys) (Fig. 5a). Second, the cells were subjected to gentle shaking (100 rpm) for 3h per day with the medium changed each hour. After 3 days the hills had spread into the valleys and the valleys had additional growth. The net effect was to even out the growth over the flask (Fig. 5c). One can conclude that some of the major physical properties of the bone and tendon factor/cofactor are the same. They are both localized to the cell layer \u2013 most likely bound to the cell membrane. However, this interaction is not strong, allowing significant diffusion that can be enhanced by gentle agitation of the medium. Characterization of the bone cofactor with degradative enzymes The cofactor is tightly bound to the factor and gives a consistent mobility pattern on Western blots. This pattern could then be used to test whether specific degradative enzymes could cleave the cofactor and alter the migration profile. One possible candidate for the cofactor would be a variation on the anchoring structure for membrane bound proteins: a glycosylphosphatidylinositol lipid(Paulick & Bertozzi 2008). To test for a glycolipid, the factor/cofactor was treated with a variety of polysaccharide degradative enzymes \u2013 chondroitinase ABC, hyaluronidase, heparinase \u2013 and the pattern on a Western blot was compared to that of an untreated control. None of the treated samples showed a distinct change in the pattern (data not shown). At the same time, a variety of lipid degrading enzymes were tested \u2013 lipase, phospholipase A2, phospholipase C, and sphingomyelinase. Lipase, phospholipase C, and sphingomyelinase caused dramatic shifts in the pattern seen on Westerns (Fig 6). Lipase was the most dramatic causing bands to shift to lower MW including the 16 kD, free protein form. The lipase was a crude preparation used at a high concentration; so while its primary target would be a glycerol fatty acid linkage in the cofactor, this is probably not the only cleavage product. The phospholipase C used in this experiment has overlapping activities with the sphingomyelinase. The sphingomyelinase is more specific to cleaving moities on ceramide phosphates. In contrast, phospholipase A2 did not show a significant change in mobility or intensity (data not shown). A negative result is hard to interpret because there are too many potential causes. The positive results, while not yielding a definitive structure, implicate 10 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t lipid components containing a phosphoceramide and a glycerol fatty acid as part of the cofactor. The lipid nature of the cofactor can explain why the factor/cofactor has a strong tendency to aggregate when concentrated due to hydrophobic interactions. Purification and further characterization of the bone cofactor using mass spectrometry In order to characterize the bone cofactor in more detail, procedures were needed to separate both the factor from the cofactor and then the cofactor from other lipid-like molecules. The first step in the purification \u2013 the separation of the cofactor from the factor \u2013 could benefit from standard lipid procedures using a methanol/chloroform extraction. In this procedure the lipids are found in the chloroform layer and proteins are found as a precipitate between the layers. There is no assay for the cofactor but the proteins in the precipitate can be solubilized and then Western analysis can be used to see if the protein factor runs as the free form (16 kD) or as a complex with the cofactor (28 kD). The amount of free factor is an indicator of cofactor release. However, redissolving the protein precipitates at the interface and running Western blots only showed the 28 kD form and higher order aggregates (Fig. 7). In this extraction protocol, the cofactor remained bound to the factor, and as a consequence, precipitated at the interface. Despite attempts to weaken the interaction between the factor and the cofactor by changes in pH, salt, solvent, chaotropic agents, temperature, and chelating agents, Western blots showed little or no conversion to the free form (data not shown). The only successful method required a combination of a shift in salt concentration and ultrafiltration (Fig. 4). Using a simplified version of this procedure whereby the CM was made high in salt (~0.6 M) and then concentrated on a 10 kD ultrafilter. This was then diluted 10 fold with water and concentrated to run on Westerns (Fig. 8). By intensity of the bands, most of the tagged factor runs as the free form. Thus, the corresponding free cofactor was expected to have passed through the ultrafilter as part of the filtrate. Using a small C18 resin column to concentrate, differential solvent elution to purify, and then comparing the untreated and the enzyme treated samples via mass spectrometry one can identify candidate cofactor peaks (Fig. 9). Only molecular weights between 530 and 580 are shown because there was no significant differences between the samples at either higher or lower values (mass range = m/z 300 \u2013 m/z 1500). Two doubly-charged ions at m/z 543.4273 and m/z 557.4429 were dramatically reduced by treatment with either enzyme and these peaks were not observed in the control and the intensity of the m/z 557 ion was highly correlated with the intensity of the free protein factor in the retentate. The observed difference between the doubly-charged ions was 14 representing a 28 Da difference in MW commonly seen for lipids. In the rest of the analysis, we focused on the 557 species with a MW of 1114.8858. By using a high mass accuracy Q-TOF mass spectrometer we generated potential molecular formulas with a list of atoms commonly found in organic biomolecules (i.e., carbon, hydrogen, nitrogen, oxygen, and phosphorous). The most likely formula C63H123N2O11P (exact mass = 557.4432252+; 0.6 ppm from the measured mass) had nearly a 1:2 Carbon:Hydrogen ratio common to lipids and the nitrogen, oxygen and phosphorous indicative of a hydrophilic region. Support for the proposed cofactor structure derived from MS/MS fragmentation of m/z 557. Two fragment ions, shown in Fig. 10, at m/z of 323 and m/z 378 are consistent with a ceramide phosphate and a linker moiety, supporting the results from the enzymatic reactions. From our analysis, the cofactor was susceptible to three lipid enzymes \u2013 lipase, 11 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t phospholipase C, and sphingomylenase, and a peak observed at m/z 557 that has a strong lipid composition. This peak was lost when treated with lipase, phopholipase C, and sphingomylenase. Using our knowledge of which enzymes cleave the cofactor and the molecular formula, a working model can be postulated that includes a ceramide phosphate, a glycerol lipid, and a linker consisting of a short modification of a proline residue between them (Fig. 10). Tendon cofactor characterization using enzymes and mass spectrometry Having analyzed the bone factor/cofactor, the focus returned to the original aim of characterizing the tendon factor, but now knowing that it binds a cofactor. The bone and tendon cofactors must be different because they impart a unique shift in mobility to the factor/cofactor complex on SDS gels. This raises important questions about the nature of the tendon cofactor and how it compares to its bone counterpart. As a first step in the characterization, the same panel of lipid degradative enzymes were tested on PAT cell CM. In this case, concentrated CM was used that was over 6 months old. This results in a stable aggregate at ~60 kD being the dominant band. As with bone, lipase and phospholipase C were active on the PAT cell factor/cofactor while phospholipase A2 was inactive (Fig. 11a). Unlike the bone cofactor, sphingomyelinase was inactive on the PAT cell factor/cofactor (Fig. 11b). To insure that the inactivity of the enzyme was not due to the factor/cofactor being in a specific aggregate, the treatment with sphingomylinase was repeated with fresh CM (Fig. 11c) and the result remained unchanged. This indicates a lack of a ceramide phosphate moiety in the tendon cofactor. The cofactor was purified for mass spectrometry by binding it to C18 columns followed by washing with methanol, acetonitrile, and elution with chloroform. Via mass spectrometry we compared preparations that were treated with either phosphospholipase C or lipase while the cofactor was bound to the C18 column to the untreated cofactor samples. No differences were seen in the methanol or acetonitrile washes; however, the chloroform fraction did reveal several peaks that changed between the untreated and treated samples. Two singly charged peaks, at 668.61 and 696.65, were reduced to background levels by the enzyme treatments. Similar to the bone cofactor mass spectrometric analysis, these peaks differ by 28 MW and reflect variation a lipid chain length by two carbons. Additionally, there was no doubly charged ion observed at m/z 557 as was seen in the bone cofactor sample (data not shown). Using MS/MS to further characterize the 668.61 ion yielded a rather simple spectrum with two peaks of roughly equal intensity and with MW of 283.26 and 311.29. These ions differ by 28 and accurate mass measurements correspond to saturated fatty acids with 18 and 20 carbons, respectively. Both chains are required to be in the model in order to achieve the high lipid content indicated by the parent molecule\u2019s MW. Since the band is degraded by phospholipase C, there is a phosphate group yet the total mass of the intact ion is insufficient to include a commonly observed glycerol phosphate. Consequently, our proposed structure includes anhydride linkage of the fatty acid directly to the phosphate. The mass difference between the total mass and the fatty acid and phosphate groups corresponds to a cyano group, thus our working model adds a cyano group to the phosphate and is consistent with the enzymology, the mass spectrometry data, and the gel mobility data (Fig. 12). Whether the cyano group is part of the original molecule or it displaced another group during the purification remains to be determined. 12 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 12 34 56 78 91011 12 13 PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t nomenclature : With two different cofactors and the potential for more in the future, there is a need to give them distinct names. For the bone factor/cofactor will be named SNZR 1P1L; for tendon factor/cofactor will be named SNZR 1P2L. Similarly, the protein factor will be SNZR 1P and the bone cofactor, SNZR 1L and the tendon cofactor, SNZR 2L. At present, because of the extremely high homology between species at the protein level and no knowledge of the conservation of lipid structure between species, the delineation is only on cell type. 2g8d5 (lanes 1-5) and abm clone a7 (lanes 6-10) -- to see if they could be an adequate replacement. : 3,8,13,14) and the free factor with tags made in e. coli at 2x (lanes 4,9,15) and 1x (lanes 5,10,16). : The monoclonal antibodies show no overlap in what they detect and neither of them detects the free factor made in E. Coli. The ABM monoclonal is striking in how strong it reacts to only one form at ~75 kD with almost no background. The polyclonal does prefer the ~50 kD form, shows some affinity for the 75kD form as well as weak affinity for some of the lower MW forms. Most importantly, it does detect the free factor made in E. coli and can be used in an assay for the separation of the factor from the cofactor. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 2 Expression of the cDNA, pESY1c, causes human osteosarcoma cells (U2OS) to differentiate. Control U2OS cells do overgrow the monolayer but cannot sustain growth necessary to form structures (2a). U2OS-pESY1c cells, after becoming confluent (end of week 1), the cells begin to overgrow the monolayer. First as linear short stripes but these coalesce into small circles and then finally some of the common borders recede and the circles become larger (week 3, phase contrast, 2b). When U2OS-pESY1c cells get to the stage of differentiation where they are making large circles of overgrown cells, they start to produce alkaline phosphatase, an early marker of bone development. In this assay using brightfield microscopy, the overgrowing cells turn dark, an indicator of alkaline phosphatase activity, while the confluent cells inside the circle express little or no activity (2c).The bars = 100 microns PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 3 Western analysis of U2OS-pESY1c expression both inside and outside the cell. The monoclonal antibody to the myc tag detects c-myc inside the cell so control U2OS cells were used to identify the background bands (lanes 1,2). For each condition, two lanes were run on an SDS gel where the second was loaded with twice the sample. U2OS cells expressing pESY1c showed two additional bands \u2013 a full length protein at 44 kD and a smaller band at 26 kD (lanes 3,4). To analyze what was secreted from the cell, the CM was put through a 30 kD ultrafilter and the retentate and the filtrate were further concentrated on a 10 kD ultrafilter so that a sufficient level of protein could be loaded onto the gel. The retentate shows a wide spectrum of bands from the smallest at 28 kD, the most intense at ~60 kD and the highest at ~250 kD (lanes 5,6). The filtrate showed one band at ~60 kD (lanes 7,8). Since the filtrate initially went through a 30 kD ultrafilter, this large band is an indication that concentrating the sample causes aggregation. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 4 Determining whether the apparent aggregation of the factor in the CM is due to covalent modifications or binding of a cofactor. When the CM was concentrated using a 10 kD ultrafilter and this sample was analyzed by Western using a myc antibody, one sees a lane of multiple bands, the smallest being 28 KD (lane1). In contrast, if the same protein is expressed in E. coli, one sees a single band at 16 kD (lane 6). Lane 5 was reserved for MW markers. The CM was put through a 30 kD ultrafilter and the retentate used as a control (lane 4) and the filtrate was divided into two parts. The first was further concentrated on 10 kD ultrafilter and then made 30% DMSO in order to see if this solvent could affect the mobility of the factor (lane 2). The DMSO did seem to limit the aggregation in that more of the factor ran at 28kD. In contrast, when the filtrate was treated with high salt (0.5 M) and then low salt by 10 fold dilution and then concentrated on a 10 kD ultrafilter, the bands all went to a 16 kD form and its 32 kD dimmer (lane 3). This is a strong evidence that whatever was causing the increased MW to 28kD and the further aggregation was not covalently bound. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 5 Localization of the bone factor/cofactor in the extracellular space and disruption by gentle agitation. U2OS-pESY1c cells were grown in culture for 9 days and then fixed with 4% paraformaldehyde in a high calcium buffer. A rabbit polyclonal antibody was used as a primary and goat anti-rabbit was used as a secondary (panel a). For comparison, the same field is shown in phase contrast (panel b). Areas where the cells are overgrowing the monolayer show strong staining whereas cells remaining in the monolayer appear to be suppressed. This leads to the strong localized differences in the expression of this gene. When CM was collected from these cells, the flasks are gently rotated (100 rpm) for 1h; then the medium is collected and new medium is added. After 3 collections the flasks are stationary for the next 21 h. After collecting medium for 3 days a dramatic change is observed in the culture \u2013 the cells no longer accurately detect the cells around them and they grow all over the dish as seen in phase contrast (panel c). PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 6 Treating U2OS-pESY1c CM with various lipid degradative enzymes to identify the makeup of the bone cofactor. The only assay for the cofactor is its binding to the factor and shifting its mobility. The CM was treated with lipase, phospholipase C, and sphingomyelinase for 30, 60, and 180 min and analyzed by Westerns (panel a, lanes 1,2,3; panel a, lanes 7,8,9; and panel b, lanes 2,3,4; respectively). The E. coli expressed protein was run as a control (panel a, lane 5 showing a monomer band at 16 kD and a dimer band at 32 kD) as well as untreated CM (panel a, lane 6; and panel b, lane 1). Panel a, lane 4 contained mw markers (data not shown). the lipase was the least specific enzyme and it degrades the cofactor : and leaves most of the factor running at 16 kD. This type of phospholipase C can cleave at both glycerol phosphates and ceramide phosphates. The sphigomyelinase is specific for ceramide phosphates. Both of these enzymes cause single breaks and this probably alters the binding/mobility but does not free the cofactor from the factor. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 7 Chloroform/methanol extraction does not free the lipid bone cofactor from the protein factor. to test whether the standard extraction method of purifying lipids would yield a free cofactor, cm : from U2OS-pEsy1c was extracted with chloroform/methanol. The protein fraction which precipitates at the interface between the two solvents was examined by Western blotting (lane 1 is half the sample volume of lane 2). Despite unusual sample preparation, the bands ran reasonably well yet there was no evidence of a 16 kD free protein factor. Instead the bands resembled those of the factor/cofactor running together with a minimum size of 28 kD. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 8 ultrafilter and then diluted with water and refiltered on the same filter (lanes 5, 6). In this case, the retentate shows a strong shift to the free factor MW of 16 kD. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 9 Identification of the bone cofactor on mass spectrometry by using lipid degradative enzymes and differential elution from a C-18 spin column. After separation from the factor, the bone cofactor was in the dilute filtrate. To concentrate the lipid containing cofactor, the filtrate was run over a C-18 column and washed with water, acetonitrile, and eluted with chloroform. To identify the cofactor, the cofactor was treated separately with phospholipase C (panel b) and lipase (panel c). These were compared with an untreated sample (panel a) and a lipase buffer only control (panel d) using mass spectrometry. The only regions that showed a strong peak in the untreated sample and little or no intensity in the other panels were at an m/z of 543 and 557.44. these peaks had the characteristic pattern of a doubly charged molecule and this : gave the cofactor a MW of 1086.84 or 1114.88. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 10 Working model of the bone cofactor. Using our knowledge of the lipid enzyme preferences, the molecular weight of the whole molecule, the predicted molecular formula, and the fragmentation pattern (MS/MS), a preliminary model has been drawn with a ceramide phosphate, a single chain glycerol lipid, and a small linker. This model also explains why this molecule is soluble from aqueous buffers to chloroform: a structure with a hydrophilic end and hydrophobic tails. The model also shows how the fragmentation pattern of two breaks (MS/MS) is consistent with the model. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 11 Characterizing the tendon cofactor using lipid degradative enzymes. PAT cell CM was put through a 30 kD filter and then concentrated with a 10 kD filter. This concentrated CM after several months at 4o C forms a dominant aggregate that runs at ~60 kD. This cm was run on westerns as follows: untreated (panel a, lanes 2,6); lipase treated (panel a, lanes 3,7); : phospholipase A2 (panel a, lanes 4,8); and phospholipase C (panel a, lanes 5,9). The data indicates a strong sensitivity to lipase and phospholipase C but not to phospholipase A2. This is identical sensitivity pattern as seen with the bone cofactor. Next sphingomylinase was tested (panel b, lanes 5,6,7) and compared to the untreated control (panel b, lanes 2,3,4). in contrast to the bone cofactor, : sphingomyelinase did not seem to affect the tendon cofactor. To rule out that the sphingomyelinase was being inhibited by the aggregation caused by \u201caging\u201d the CM, fresh cultures of PAT cells were made and the CM treated with sphingomyelinase (panel c, lanes 4,5) and untreated control (panel c, lanes 2,3). Again, this enzyme does not alter the mobility of the tendon factor/cofactor. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 12 Working model for the tendon cofactor. . The same approach as previously used with the bone cofactor yielded two peaks at 668.61 and 696.65 separated by 28. This difference is commonly seen for lipid chains by addition of two CH2 groups. Since the 668 band was stronger, we focused the model on this band. The ms/ms fragmentation pattern was fairly simple yielding two bands of roughly equal intensity corresponding to an 18 and 20 carbon fatty acids. Since the molecule is degraded by phospholipase C, a phosphate was added to the model. However, there was insufficient MW for a glycerol so the fatty acids were attached directly to the phosphate. The remaining MW fits a cyano group. The working model is presented to give a framework to the type of structure expected. PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 13 The cofactor has a unique lipid makeup depending on the producing cell type but also each cofactor triggers a unique response depending on the target cell type. In this case we used the control cell line U2OS and added a 5X concentrated condition medium from PAT cells or from U2OS-pESY1c cells. An additional test was done to see if the free factor made in E. coli had any activity on U2OS cells if applied at a high concentration (5 mg/ml). In comparison to expressing pESY1c in U2OS cells which causes hills and valleys to form, adding 5X CM from these cells to uninfected U2OS cells shows only a slight growth effect after 6 days as the cells appeared to be more compact (panel b) vs medium only control (panel a). Growing the cells for an additional 9 days showed a strong growth effect (panel f) vs the medium only control (panel e). Moreover, the increased growth is even over the culture \u2013 no hills or valleys. In contrast, adding 5X CM from PAT cell cultures caused the cells to grow more slowly and the presence of more rounded up cells (panel d). Adding high concentrations of the protein factor alone (expressed in E. coli) did not appear to increase growth after 6 days (panel c). PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t PeerJ reviewing PDF | (v2013:06:612:0:1:NEW 25 Jun 2013) R ev ie w in g M an us cr ip t Figure 14 Localization of the factor using a rabbit polyclonal antibody in 5 week old mice at the tendon/muscle junction.",
"url": "https://peerj.com/articles/193/reviews/",
"review_1": "Jean Bousquet \u00b7 Oct 8, 2013 \u00b7 Academic Editor\nACCEPT\nexcellent paper. Thank you for submitting the paper to PeerJ",
"review_2": "Jean Bousquet \u00b7 Sep 14, 2013 \u00b7 Academic Editor\nMINOR REVISIONS\nSorry for the delay but it was difficult to find a reviewer",
"review_3": "Christine Cole Johnson \u00b7 Sep 12, 2013\nBasic reporting\nNo comments.\nExperimental design\nThis manuscript reports on the relationship between exposures to traffic related air pollutants (TRAP) and asthma and allergic outcomes in Germany. It is clearly written and the methods are well described, with the exception that the statistical analyses could use more detail for the cross sectional analyses if they remain in the paper (see next section). The paper is laudable in that the authors have harmonized data across several birth cohorts all recruited around the same years in order to increase sample size and geographic variability. The sample sizes, while large, still result in low numbers of diagnoses in LISA East, and asthma in all locations. Nevertheless the confidence limits are relatively precise. This paper is also important because of its use of sophisticated pollutant models that reveal, bottom line, relatively conclusive negative results.\nValidity of the findings\nPotential confounders should be included in a model because they are related to both the exposure and the outcome. Therefore it is unclear why some of the variables were included, such as sex of the child or presence of an older sibling. Are variables such as these, and others such as gas stove use and parental history of atopy, related to outdoor pollutant exposure in any way? Did the authors test to see whether the addition of a confounder materially changed the risk estimates for the pollutants and outcomes, such that it warranted keeping the variable in the model?\n\nThe authors could consider the possibility of effect modification. Could the relationship between pollutants and outcomes be different for subgroups such as males, or those with a positive family history or ETS exposure?\n\nMore information is needed for the non-German reader about the characteristics of the geographic areas of the cohorts. It appears from the cohort characteristics that the North area is of lower socioeconomic status, the South, the highest and East somewhere in between. Are socioeconomic data less crude than one parent with more than 10 years of education (yes/no) needed, or available? Presumably socioeconomic status is highly related to pollutant level and perhaps the outcomes? Would there be any differences in medical care use and therefore probability of obtaining a doctor diagnosis, between the areas? What are the pollutant sources in the areas?\n\nMost importantly, what has happened to pollution levels over time in the three areas, and does this differ by area? The Escape models used for North and South are based on data collected 10-13 years after the births occurred. The model used for the East population is based on data collected just a few years after the births, so even though the model is noted to have limitations that the Escape models do not, the measurements are much closer in time to the births. Non-systematic exposure misclassification, which may be especially a problem in the North and South area if pollution has decreased over time (which is indicated in the Discussion for the South, lines 341-342), would tend to drive associations towards the null, and most of the associations are consistent with 1.0. These are critical limitations that need to be included in the Discussion.\n\nResults, page 9, line 238. These results for the total population are completely driven by GINI/LISA South, so I don\u2019t think they are worth mentioning except perhaps in that context. The results mentioned for GINI/LISA North (also in Discussion lines 321-322) are also not really notable. They are not statistically significant even though it is easy to obtain significance with the large sample and high prevalence of allergic rhinitis and eye/nose symptoms. Given the large number of comparisons, the large sample sizes for most estimates, the variation across geographies for the same associations, and the relatively weak estimates with confidence limits that include or are very close to 1.0, in this reviewer\u2019s view, Table 4 is basically a table of null results.\n\nThe paper is focused on exposure to pollutants at birth and outcomes at age 10, but in Figure 2 suddenly switches to cross-sectional analyses of pollutants estimated by addresses at ages between 3 and 10 and the same outcomes. These analyses are less clear. If a child was diagnosed at age 4 with asthma and then moves, is he included in the asthma prevalence estimates at age 10 with the pollution measures at his new address? I see no reason to include cross-sectional analyses when longitudinal data are available.\n\nFigure 1: These graphs appear to be cumulative incidence data (ever history of a doctor diagnosis), so it is unclear to me why the prevalence of asthma (and once for allergic rhinitis) goes down in some years.\n\nAssuming most of the children with asthma were also atopic, it would be useful to repeat the analyses for the specific phenotype of atopic asthmatics to reduce misclassification, presuming this could not be done due to sample size for non-atopic asthmatics.\n\nAre the data such that analyses can be repeated for the combined category of doctor diagnosed and undiagnosed hay fever since many children do not receive a diagnosis for these conditions (or is that not the case in Germany)?\n\n\nMinor Comments:\n\nIntro, second paragraph. I think validity of exposure is just as difficult, if not more so, than accuracy of outcome measurement.\n\nSince the authors couldn\u2019t evaluate exposure timing due to the high correlation across time points in most cases, and couldn\u2019t decipher a trend, they may just want to mention this in the Discussion (and maybe the Methods) but not in the Introduction.\n\nDiscussion, line 368, except for the cross-sectional analyses (see my concern above), they were not really able to use the pollutant data at three time points successfully.\n\nDiscussion: line 360: I am not sure what is meant by the concern of recall bias. Individuals are not providing pollution data, only home addresses, which should be easily collected via bi-annual questionnaires and not vary in validity by those with positive outcomes versus no outcome.\n\nI am not sure that Table 1 needs to include source data that is not used in the paper.\nAdditional comments\nNo comments\nCite this review as\nCole Johnson C (2013) Peer Review #1 of \"A longitudinal analysis of associations between traffic-related air pollution with asthma, allergies and sensitization in the GINIplus and LISAplus birth cohorts (v0.1)\". PeerJ https://doi.org/10.7287/peerj.193v0.1/reviews/1",
"review_4": "Bente Oftedal \u00b7 Sep 4, 2013\nBasic reporting\nNo comments.\nExperimental design\nLine 122: \"For the seven to 10 year follow-ups of GINIplus and for the entire LISAplus follow-up, this information ...\" As there is no follow-up between six and 10 years, it seems like this part should be left out: \"the seven to\"\n\nLine 138: \"A test was defined as positive if the immunoglobulin E value was greater or equal to 0.35 kU/L.\" This definition needs a reference to document its validity.\n\nLine 141: The size of the subpopulation with sensitization tests should be added, separately for 6 and 10 years of age.\n\nLine 174: The definition of parental education is rather coarse. Is it possible to apply a three category definition instead? Furthermore, socioeconomic factors (SES) are usually related to environmental variables such as air pollution. Thus, individual SES should be properly adjusted for when studying health effects of air pollution. Is information about parental marital status, household income or ethnic origin available?\nValidity of the findings\nLine 293: The authors discuss whether residual confounding may be influencing the results, although several individual level confounders were considered. This discussion is appropriate. I suggest to mention individual SES especially (see my comment above). Generally, individual SES variables are more important than SES variables at neighborhood level.\n\nLine 359: \u201cAt the 10 year follow-up, the retention rate was 55.9% for both cohorts combined.\u201d How is 55.9% achieved?\n\nLine 360: \u201c\u2026 , but objective measures of aeroallergen sensitization were available for 55.3% of the study population.\u201d How is 55.3% achieved? (n/N) should be added.\nAdditional comments\nThe study is a nicely designed and very well written paper about the relationship between air pollution and allergic outcomes in a birth cohort followed up to 10 years. Publication is recommended.\nCite this review as\nOftedal B (2013) Peer Review #2 of \"A longitudinal analysis of associations between traffic-related air pollution with asthma, allergies and sensitization in the GINIplus and LISAplus birth cohorts (v0.1)\". PeerJ https://doi.org/10.7287/peerj.193v0.1/reviews/2",
"pdf_1": "https://peerj.com/articles/193v0.2/submission",
"pdf_2": "https://peerj.com/articles/193v0.1/submission",
"all_reviews": "Review 1: Jean Bousquet \u00b7 Oct 8, 2013 \u00b7 Academic Editor\nACCEPT\nexcellent paper. Thank you for submitting the paper to PeerJ\nReview 2: Jean Bousquet \u00b7 Sep 14, 2013 \u00b7 Academic Editor\nMINOR REVISIONS\nSorry for the delay but it was difficult to find a reviewer\nReview 3: Christine Cole Johnson \u00b7 Sep 12, 2013\nBasic reporting\nNo comments.\nExperimental design\nThis manuscript reports on the relationship between exposures to traffic related air pollutants (TRAP) and asthma and allergic outcomes in Germany. It is clearly written and the methods are well described, with the exception that the statistical analyses could use more detail for the cross sectional analyses if they remain in the paper (see next section). The paper is laudable in that the authors have harmonized data across several birth cohorts all recruited around the same years in order to increase sample size and geographic variability. The sample sizes, while large, still result in low numbers of diagnoses in LISA East, and asthma in all locations. Nevertheless the confidence limits are relatively precise. This paper is also important because of its use of sophisticated pollutant models that reveal, bottom line, relatively conclusive negative results.\nValidity of the findings\nPotential confounders should be included in a model because they are related to both the exposure and the outcome. Therefore it is unclear why some of the variables were included, such as sex of the child or presence of an older sibling. Are variables such as these, and others such as gas stove use and parental history of atopy, related to outdoor pollutant exposure in any way? Did the authors test to see whether the addition of a confounder materially changed the risk estimates for the pollutants and outcomes, such that it warranted keeping the variable in the model?\n\nThe authors could consider the possibility of effect modification. Could the relationship between pollutants and outcomes be different for subgroups such as males, or those with a positive family history or ETS exposure?\n\nMore information is needed for the non-German reader about the characteristics of the geographic areas of the cohorts. It appears from the cohort characteristics that the North area is of lower socioeconomic status, the South, the highest and East somewhere in between. Are socioeconomic data less crude than one parent with more than 10 years of education (yes/no) needed, or available? Presumably socioeconomic status is highly related to pollutant level and perhaps the outcomes? Would there be any differences in medical care use and therefore probability of obtaining a doctor diagnosis, between the areas? What are the pollutant sources in the areas?\n\nMost importantly, what has happened to pollution levels over time in the three areas, and does this differ by area? The Escape models used for North and South are based on data collected 10-13 years after the births occurred. The model used for the East population is based on data collected just a few years after the births, so even though the model is noted to have limitations that the Escape models do not, the measurements are much closer in time to the births. Non-systematic exposure misclassification, which may be especially a problem in the North and South area if pollution has decreased over time (which is indicated in the Discussion for the South, lines 341-342), would tend to drive associations towards the null, and most of the associations are consistent with 1.0. These are critical limitations that need to be included in the Discussion.\n\nResults, page 9, line 238. These results for the total population are completely driven by GINI/LISA South, so I don\u2019t think they are worth mentioning except perhaps in that context. The results mentioned for GINI/LISA North (also in Discussion lines 321-322) are also not really notable. They are not statistically significant even though it is easy to obtain significance with the large sample and high prevalence of allergic rhinitis and eye/nose symptoms. Given the large number of comparisons, the large sample sizes for most estimates, the variation across geographies for the same associations, and the relatively weak estimates with confidence limits that include or are very close to 1.0, in this reviewer\u2019s view, Table 4 is basically a table of null results.\n\nThe paper is focused on exposure to pollutants at birth and outcomes at age 10, but in Figure 2 suddenly switches to cross-sectional analyses of pollutants estimated by addresses at ages between 3 and 10 and the same outcomes. These analyses are less clear. If a child was diagnosed at age 4 with asthma and then moves, is he included in the asthma prevalence estimates at age 10 with the pollution measures at his new address? I see no reason to include cross-sectional analyses when longitudinal data are available.\n\nFigure 1: These graphs appear to be cumulative incidence data (ever history of a doctor diagnosis), so it is unclear to me why the prevalence of asthma (and once for allergic rhinitis) goes down in some years.\n\nAssuming most of the children with asthma were also atopic, it would be useful to repeat the analyses for the specific phenotype of atopic asthmatics to reduce misclassification, presuming this could not be done due to sample size for non-atopic asthmatics.\n\nAre the data such that analyses can be repeated for the combined category of doctor diagnosed and undiagnosed hay fever since many children do not receive a diagnosis for these conditions (or is that not the case in Germany)?\n\n\nMinor Comments:\n\nIntro, second paragraph. I think validity of exposure is just as difficult, if not more so, than accuracy of outcome measurement.\n\nSince the authors couldn\u2019t evaluate exposure timing due to the high correlation across time points in most cases, and couldn\u2019t decipher a trend, they may just want to mention this in the Discussion (and maybe the Methods) but not in the Introduction.\n\nDiscussion, line 368, except for the cross-sectional analyses (see my concern above), they were not really able to use the pollutant data at three time points successfully.\n\nDiscussion: line 360: I am not sure what is meant by the concern of recall bias. Individuals are not providing pollution data, only home addresses, which should be easily collected via bi-annual questionnaires and not vary in validity by those with positive outcomes versus no outcome.\n\nI am not sure that Table 1 needs to include source data that is not used in the paper.\nAdditional comments\nNo comments\nCite this review as\nCole Johnson C (2013) Peer Review #1 of \"A longitudinal analysis of associations between traffic-related air pollution with asthma, allergies and sensitization in the GINIplus and LISAplus birth cohorts (v0.1)\". PeerJ https://doi.org/10.7287/peerj.193v0.1/reviews/1\nReview 4: Bente Oftedal \u00b7 Sep 4, 2013\nBasic reporting\nNo comments.\nExperimental design\nLine 122: \"For the seven to 10 year follow-ups of GINIplus and for the entire LISAplus follow-up, this information ...\" As there is no follow-up between six and 10 years, it seems like this part should be left out: \"the seven to\"\n\nLine 138: \"A test was defined as positive if the immunoglobulin E value was greater or equal to 0.35 kU/L.\" This definition needs a reference to document its validity.\n\nLine 141: The size of the subpopulation with sensitization tests should be added, separately for 6 and 10 years of age.\n\nLine 174: The definition of parental education is rather coarse. Is it possible to apply a three category definition instead? Furthermore, socioeconomic factors (SES) are usually related to environmental variables such as air pollution. Thus, individual SES should be properly adjusted for when studying health effects of air pollution. Is information about parental marital status, household income or ethnic origin available?\nValidity of the findings\nLine 293: The authors discuss whether residual confounding may be influencing the results, although several individual level confounders were considered. This discussion is appropriate. I suggest to mention individual SES especially (see my comment above). Generally, individual SES variables are more important than SES variables at neighborhood level.\n\nLine 359: \u201cAt the 10 year follow-up, the retention rate was 55.9% for both cohorts combined.\u201d How is 55.9% achieved?\n\nLine 360: \u201c\u2026 , but objective measures of aeroallergen sensitization were available for 55.3% of the study population.\u201d How is 55.3% achieved? (n/N) should be added.\nAdditional comments\nThe study is a nicely designed and very well written paper about the relationship between air pollution and allergic outcomes in a birth cohort followed up to 10 years. Publication is recommended.\nCite this review as\nOftedal B (2013) Peer Review #2 of \"A longitudinal analysis of associations between traffic-related air pollution with asthma, allergies and sensitization in the GINIplus and LISAplus birth cohorts (v0.1)\". PeerJ https://doi.org/10.7287/peerj.193v0.1/reviews/2\nReview 5: nan\nReview 6: \n",
"LLM_extracted_review": "['1. Review 1: \"excellent paper. Thank you for submitting the paper to PeerJ\"\\n\\n2. Review 2: \"Sorry for the delay but it was difficult to find a reviewer\"\\n\\n3. Review 3: \\n - Basic reporting: No comments.\\n - Experimental design: The statistical analyses could use more detail for the cross-sectional analyses. The sample sizes result in low numbers of diagnoses in LISA East and asthma in all locations. The confidence limits are relatively precise. The paper uses sophisticated pollutant models that reveal relatively conclusive negative results.\\n - Validity of the findings: Potential confounders should be included in a model. It is unclear why some variables were included. The authors could consider the possibility of effect modification. More information is needed about the geographic areas of the cohorts. Socioeconomic data may be needed. The paper should discuss pollution levels over time in the three areas. Non-systematic exposure misclassification could drive associations towards the null. Results for the total population are driven by GINI/LISA South and may not be worth mentioning. The paper switches to cross-sectional analyses without clear justification. The graphs in Figure 1 are unclear. It would be useful to repeat analyses for specific phenotypes of atopic asthmatics. The validity of exposure measurement is difficult. The authors should mention the inability to evaluate exposure timing in the Discussion. The concern of recall bias is unclear. Table 1 may include unnecessary source data.\\n - Additional comments: No comments.\\n\\n4. Review 4: \\n - Basic reporting: No comments.\\n - Experimental design: The mention of \"the seven to\" should be removed. A reference is needed for the definition of a positive test. The size of the subpopulation with sensitization tests should be added. The definition of parental education is coarse and could be improved. Individual SES should be adjusted for when studying health effects of air pollution.\\n - Validity of the findings: The discussion of residual confounding is appropriate. Clarification is needed on how the retention rate of 55.9% was achieved. The percentage of the study population with objective measures of aeroallergen sensitization should include (n/N).\\n - Additional comments: The study is well designed and written. Publication is recommended.']"
} |