Collagen fibers affect metastasis in two opposing ways, by supporting invasive cells but also by generating a barrier to invasion. We hypothesized that these functions might be performed by different isoforms of type I collagen. Carcinomas are reported to contain α1(I)3 homotrimers, a type I collagen isoform normally not present in healthy tissues, but the role of the homotrimers in cancer pathophysiology is unclear. In this study, we found that these homotrimers were resistant to all collagenolytic matrix metalloproteinases (MMP). MMPs are massively produced and used by cancer cells and cancer-associated fibroblasts for degrading stromal collagen at the leading edge of tumor invasion. The MMP-resistant homotrimers were produced by all invasive cancer cell lines tested, both in culture and in tumor xenografts, but they were not produced by cancer-associated fibroblasts, thereby comprising a specialized fraction of tumor collagen. We observed the homotrimer fibers to be resistant to pericellular degradation, even upon stimulation of the cells with proinflammatory cytokines. Furthermore, we confirmed an enhanced proliferation and migration of invasive cancer cells on the surface of homotrimeric versus normal (heterotrimeric) type I collagen fibers. In summary, our findings suggest that invasive cancer cells may use homotrimers for building MMP-resistant invasion paths, supporting local proliferation and directed migration of the cells whereas surrounding normal stromal collagens are cleaved. Because the homotrimers are universally secreted by cancer cells and deposited as insoluble, MMP-resistant fibers, they offer an appealing target for cancer diagnostics and therapy. Cancer Res; 70(11); 4366–74. ©2010 AACR.

Overexpression of matrix metalloproteinases (MMP) is a hallmark of invasive cancers (1, 2). This enzyme family has multiple functions, including cleaving various extracellular matrix components (3, 4). MMPs capable of cleaving type I collagen triple helices, the main matrix component in many tissues, are collagenases (MMP-1, MMP-8, MMP-13, and MMP-14) and gelatinase A (MMP-2) (5). Membrane-bound MMP-14 might play a particularly important role, e.g., MMP-14 knockout mice have severe collagen turnover deficiency (6). One of the critical functions of collagenolytic MMPs in cancer is clearing an invasion path through the barrier of type I collagen fibers in stroma and blood vessel walls.

However, collagen fibers do not only hinder cancer invasion, they also promote invasion by inducing epithelial-mesenchymal transition and supporting the invasiveness, proliferation, and migration of cancer cells (7, 8). Consistent with this dual role, invasive cancer cells and cells recruited into tumors produce both an increased amount of MMPs and an increased amount of collagen (8, 9). Remarkably, the invasion-supporting function may be performed, at least in part, by a type I collagen isoform normally not present in healthy tissues.

The normal isoform of type I collagen is a heterotrimer of two α1(I) chains and one α2(I) chain (α12α2). A fetal, homotrimeric isoform (α13) with three α1(I) chains was found in carcinomas (1014) and in cultures of chemically transformed cells (15, 16) and cancer cells (1721). It was suggested that the homotrimers might enhance the proliferation (22) and migration (23) of cancer cells, but their role in cancer pathophysiology has not been established.

In the present study, we hypothesized that the homotrimers might be selectively produced by cancer and not by surrounding cells, forming MMP-resistant roadways for tissue invasion and providing the necessary supporting function of collagen without the inhibitory barrier effect. We took a cue from an observation that homotrimers refolded in vitro from denatured α1(I) chains were resistant to MMP-1 and MMP-8 (24), although the refolding could result in improper chain register, disrupting the normal collagenase cleavage site. Testing our hypothesis revealed that naturally produced homotrimers were indeed resistant to cleavage by all collagenolytic MMPs (MMP-1, 2, 8, 13, 14) and to degradation by fibroblasts and cancer cells. In culture, the homotrimers comprised 15% to 40% of type I collagens secreted by invasive melanoma, adenocarcinoma, fibrosarcoma, and neuroblastoma cells, but they were not produced by normal fibroblasts. In xenograft tumors, the homotrimers comprised ∼50% of type I collagen produced by the same human cancer cells, but no homotrimers were produced by mouse cells recruited into the tumors. Finally, we confirmed faster proliferation and migration of cancer cells on matrix reconstituted from the homotrimers compared with heterotrimers.

Cell culture

CRL-2127 fibroblasts and HT-1080 fibrosarcoma cells were purchased from American Type Culture Collection; PacMet cells were generously supplied by Dr. Linda A. deGraffenried, University of Texas at Austin (Austin, TX). All other cells were obtained from the National Cancer Institute Drug Screen and characterized by analysis of short tandem repeats (Supplementary Table S1). Cells were cultured at 37°C, 5% CO2 in DMEM with 2 mmol/L of GlutaMAX (Invitrogen), and 10% fetal bovine serum (Gemini Bioproducts). After 70% to 90% confluence, the cells were incubated for 24 hours in DMEM/GlutaMAX with 0.1% fetal bovine serum. The harvested medium was buffered with 100 mmol/L of Tris-HCl (pH 7.4), protected with protease inhibitors, and used for purification of collagen as described in ref. (25). In some experiments, collagen synthesis was stimulated with 50 μg/mL of ascorbate or 5 ng/mL of transforming growth factor-β1 (PeproTech). The cells were released from the flask surface by 0.05% trypsin-EDTA (Invitrogen) and counted.

Type I collagen purification, labeling, and characterization

Mouse heterotrimers and homotrimers were extracted from tail and spinal tendons of wild-type and homozygous osteogenesis imperfecta murine (oim) mice, respectively, with 0.5 mol/L of acetic acid (acid soluble) or with 0.1 mg/mL of pepsin in 0.5 mol/L of acetic acid (pepsin treated) and purified by selective precipitation with 0.7 mol/L of NaCl (26). Human heterotrimers were isolated from cultured CRL-2127 fibroblasts (25) and homotrimers were isolated from cultured fibroblasts with two nonfunctional COL1A2 alleles (27) generously provided by Dr. Peter Byers, University of Washington (Seattle, WA). Collagen concentrations were measured by circular dichroism (CD). Collagen was labeled with amino-reactive Alexa Fluor 488 (AF488) carboxylic acid succinimidyl ester (Invitrogen), DyLight 549 or 649 NHS esters (Pierce), or Cy5 NHS ester (GE Healthcare) (25). The labeling efficiency was adjusted to one dye per 3 to 10 collagen molecules. Labeled collagen was characterized by gel electrophoresis on precast 3-8% gradient Tris-acetate mini gels (Invitrogen); the gels were scanned on an FLA5000 fluorescence scanner (Fuji Medical Systems) and analyzed with Multi-Gauge software supplied with the scanner. The identity of polypeptides migrating at the α(I) and α2(I) collagen chain positions was confirmed by CNBr digestion and analysis of CNBr peptides on precast 12% Tris-Gly mini gels (Invitrogen).

To quantify collagen secretion, AF488-labeled MMP-1 fragments of type I mouse-tail-tendon collagen (6.6 ng/mL) were added to a media aliquot before collagen purification, as an internal standard. After the purification, the ratio of full-length α1(I) chains to 3/4-MMP-1 fragments of α1(I) chains was measured by Cy5-labeling and gel electrophoresis. The initial concentration of secreted collagen in cell culture medium was determined by comparing the measured α1(I):3/4-α1(I) ratio to mock experiments, in which the 3/4 fragments were mixed in a known proportion with purified procollagen.

Collagen cleavage with soluble MMPs

Recombinant human proMMP-1 and proMMP-13 were prepared, activated, and purified as described in ref. (28). Active human neutrophil MMP-8 and recombinant human MMP-2 were purchased from EMD Biosciences. Binary mixtures of human or mouse type I collagen homotrimers and heterotrimers, one component labeled with AF488 and the other with Cy5, were prepared in 50 mmol/L of Tris-HCl (pH 7.5), 0.15 mol/L of NaCl, 5 mmol/L of CaCl2, 0.05% Brij 35 (TNC buffer) and incubated with MMP-1 (2 nmol/L), MMP-2 (75 nmol/L), MMP-8 (∼50 nmol/L), or MMP-13 (10 nmol/L). Aliquots taken at different time intervals were mixed with a lithium dodecyl sulfate gel-loading buffer (Invitrogen) and 20 mmol/L of EDTA, denatured and analyzed by gel electrophoresis.

Homotrimer assays

AF488-labeled heterotrimers were mixed 1:1 with Cy5-labeled unknown samples or with Cy5-labeled calibration samples with known homotrimer fractions. The mixtures (0.05–0.1 mg/mL) were incubated with 2 nmol/L of activated MMP-1 at 25°C. Aliquots collected after different time intervals were analyzed by gel electrophoresis. To disrupt pepsin-resistant intermolecular cross-links, tissue-derived samples were treated with 0.05 mg/mL of Pronase (EMD Biosciences) at 4°C before electrophoresis. Comparison of the fraction of uncleaved, Cy5-labeled α1(I) chains in the unknown sample with the calibration mixtures allowed detection of 1% homotrimers and measurement of the homotrimer fraction with relative accuracy better than 10%. The best accuracy was achieved at 90% to 95% cleavage of the AF488-labeled heterotrimers. In addition, we measured the α1(I)/α2(I) chain ratio in initial mixtures and as a function of the incubation time.

Pericellular degradation of fibrillar collagen

The fibrillar matrix was reconstituted from acid-soluble homotrimeric and heterotrimeric mouse-tail-tendon type I collagens. Ice-cold solutions of DyLight 649–labeled homotrimers and heterotrimers in 2 mmol/L of HCl (0.4 mg/mL) were mixed 9:1 with ice-cold 10× PBS and placed into wells of precooled multiwell, chambered coverslips (Invitrogen). Collagen fibril formation was induced by incubating the coverslips at 32°C for at least 1 hour followed by overnight incubation at 37°C. The resulting fibril gels were dried into films under a flow of air in a cell culture hood. The films were sterilized with 70% ethanol and imaged in the FLA5000 fluorescence scanner at 10 to 25 μm resolution. Wild-type and MMP-14 knockout mouse fibroblasts (6) were stimulated with 1 nmol/L of interleukin-1β (IL-1β; PeproTech) and 10 nmol/L of tumor necrosis factor-α (TNF-α; PeproTech) for 24 hours as described in ref. (29). The stimulated cells were seeded in the middle of the reconstituted collagen films (10,000–20,000 cells/well) and allowed to attach for 6 to 16 hours. The medium was then replaced with DMEM, 0.5% fetal bovine serum, 1 nmol/L IL-1β, 10 nmol/L TNF-α. After 3 to 4 days of incubation at 37°C under 5% CO2, the films were fixed with 2% formaldehyde (1 hour at 37°C), labeled with 4′,6-diamidino-2-phenylindole, reexamined in the FLA5000 scanner, and imaged in a wide field fluorescence microscope (Olympus BX-51), or a confocal microscope (Zeiss LSM 510 Meta) at different resolutions.

Cell adhesion, proliferation, and migration

Heterotrimer and homotrimer films were prepared from pepsin-treated mouse-tail-tendon collagen in 96-well plates as described above. In cell adhesion assays, 100,000 cells/well were seeded on these films and incubated for 30 minutes at 37°C. Nonadherent cells were washed out with media. The remaining cells were labeled with calcein (Invitrogen) and total fluorescence intensity of each film was measured in the FLA5000 fluorescence scanner. The number of adherent cells was determined based on calibration curves for films with known cell density. The proliferation assay was essentially similar, but 100 to 1,000 cells/well were seeded, allowed to adhere and proliferate for 4 days. Cell migration was measured using Platypus 96-well plate (Platypus). A central circle in each collagen film was blocked with a stopper and cells were seeded outside the stopper. Once the cells adhered, the stopper was removed and cells were allowed to move into the central circle for 1 to 2 days. The cells were then labeled with calcein, the outside area was masked and the fluorescence intensity from the cells, which migrated into the central circle, was measured.

Xenograft tumors

Tumors generated in athymic nude mice (nu/nu NCr) by injection of LOX-IMVI melanoma, PC-3 prostate cancer, and MDA-MB-231 breast cancer cells (at least three tumors for each cell line) were obtained from Dr. Melinda Hollingshead, National Cancer Institute (Frederick, MD). The LOX-IMVI and MDA-MB-231 tumors were collected from the first passage in mice; PC-3 tumors were collected from the second passage. The tumors were minced and collagen was extracted in 0.1 mg/mL of pepsin, 0.5 mol/L of acetic acid, and 0.5% Brij 35 at 4°C for 24 to 48 hours. Extraction was repeated four to five times until all collagen was solubilized. Collagen was purified by several rounds of precipitation with 0.7 mol/L of NaCl and examined by gel electrophoresis after Cy5 labeling. Thermal denaturation thermograms in 0.2 mol/L of sodium phosphate and 0.5 mol/L of glycerol (pH 7.4) were measured by differential scanning calorimetry (DSC) in a Nano III calorimeter (TA Instruments) or by CD in a J810 spectropolarimeter (Jasco, Inc.) with a thermoelectric device (30). A sample of collagen from each tumor was analyzed with the MMP-1 cleavage assay as described above. The remaining collagen was heated in the TNC buffer to 39°C for 10 to 20 minutes, cooled to 20°C and treated with 2.5 mg/mL of trypsin, 2.5 mg/mL of chymotrypsin for 10 minutes. The reaction was stopped by adding 0.5 mol/L of acetic acid; and undigested collagen was reprecipitated with 0.7 mol/L of NaCl. This procedure resulted in complete digestion of the less stable collagen triple helices produced by mouse cells whereas ∼50% of collagen produced by human cancer cells remained intact. The resulting human collagen was reanalyzed by gel electrophoresis, MMP-1 digestion, and CD.

MMP resistance

We tested the MMP resistance of α1(I) homotrimers produced by human fibroblasts with nonfunctional COL1A2 alleles (27) as well as α1(I) homotrimers from tail tendons of homozygous oim mice with nonfunctional α2(I) chains (31). We mixed fluorescently labeled homotrimers with the corresponding human or mouse heterotrimers labeled with a different dye. Each mixture was then processed with recombinant human MMP-1, MMP-2, MMP-8, or MMP-13 at temperatures from 20°C to 35°C. Gel electrophoresis of aliquots collected at different times revealed cleavage of type I homotrimers and heterotrimers at the same sites, producing the expected 3/4 and 1/4 fragments (Fig. 1). Additional fragments were observed only upon cleavage with MMP-2.

Figure 1.

Collagen cleavage by interstitial MMPs in binary homotrimer/heterotrimer mixtures. Each pair of images shows the same gel lane in two fluorescent colors, corresponding to heterotrimer and homotrimer labeling: α1(I)C and α2(I)C are the 3/4 fragments produced by MMP cleavage at the expected site after Gly-775 (see Supplementary Fig. S1). Only 1% to 5% homotrimers were cleaved (faint α1(I)C bands in α13 lanes) at the incubation time with ∼50% heterotrimer cleavage. The same results were observed for human collagen (shown here for MMP-1 and MMP-2) and mouse collagen (shown here for MMP-8 and MMP-13).

Figure 1.

Collagen cleavage by interstitial MMPs in binary homotrimer/heterotrimer mixtures. Each pair of images shows the same gel lane in two fluorescent colors, corresponding to heterotrimer and homotrimer labeling: α1(I)C and α2(I)C are the 3/4 fragments produced by MMP cleavage at the expected site after Gly-775 (see Supplementary Fig. S1). Only 1% to 5% homotrimers were cleaved (faint α1(I)C bands in α13 lanes) at the incubation time with ∼50% heterotrimer cleavage. The same results were observed for human collagen (shown here for MMP-1 and MMP-2) and mouse collagen (shown here for MMP-8 and MMP-13).

Close modal

The homotrimer cleavage rates by all MMPs at all temperatures were more than 5 to 10 times slower than heterotrimer cleavage rates, both for human and mouse collagens (Fig. 1; Supplementary Fig. S1). The cleavage kinetics of naturally produced homotrimers by MMP-1 was consistent with the homotrimers refolded in vitro (24). Confocal imaging showed that the reconstituted homotrimer fibers were also much more resistant to cleavage at 37°C.

Next, we compared pericellular cleavage of collagen matrix by fibroblasts from wild-type and MMP-14 knockout mice (6). The cells were seeded on 2-μm-thick, fluorescently labeled matrix reconstituted from heterotrimeric or homotrimeric type I collagen fibers. Cells were stimulated with TNF-α and IL-1β to speed up matrix degradation (29). After 3 to 4 days of culturing, matrix and cells were fixed and imaged (Fig. 2). Wild-type cells made holes in type I heterotrimer matrix, clearly visible both at low and high resolution. The degradation of type I heterotrimer films by knockout cells was strongly suppressed (6). We observed no detectable pericellular degradation of type I homotrimer films by either wild-type or knockout cells. The normal appearance of nuclei suggested that the cells were viable. Thus, type I homotrimer fibers were resistant to pericellular degradation in general and to MMP-14, in particular. Similarly, we found that homotrimer fibers were resistant to pericellular cleavage by HT-1080 fibrosarcoma cells with and without the stimulation with TNF-α and IL-1β as well as to pericellular cleavage by LOX-IMVI melanoma cells (Supplementary Fig. S2).

Figure 2.

Pericellular degradation of collagen matrix by MMP-14. A, matrix degradation by mouse fibroblasts stimulated with TNF-α/IL-1β. DyLight 649–labeled collagen matrix (red) was imaged before and after 3 days of incubation with the fibroblasts seeded in the center of the film. Cell nuclei were imaged after 4′,6-diamidino-2-phenylindole (DAPI) labeling. The gray scale image of matrix degradation was obtained by subtraction of the matrix fluorescence intensity after 3 days of incubation with the cells from the matrix fluorescence intensity before incubation; darker areas represent larger loss of fluorescence. The loss of fluorescence was caused by mechanically damaging the film with a pipette tip (e.g., black spot in the bottom left quadrant), bleaching of the fluorescence (overall background, see B), and matrix degradation by cells (round area in the middle, surrounding cell nuclei). B to D, fluorescence bleaching without cells (B) and matrix degradation by wild-type (C) and MMP-14 knockout (D) fibroblasts. Low-resolution difference images of matrix degradation (top) were obtained as described in A. High-resolution images are projections of confocal stacks within the areas populated by cells (bottom). The cyan channel shows 4′,6-diamidino-2-phenylindole–labeled cell nuclei and magenta channel shows DyLight 649–labeled collagen. The dark areas in the confocal image of the heterotrimer matrix with wild-type (WT) fibroblasts (C, bottom left) are holes cut in the matrix by the cells. No holes were observed in the homotrimer matrix (C, bottom right) or with MMP-14 knockout fibroblasts (D).

Figure 2.

Pericellular degradation of collagen matrix by MMP-14. A, matrix degradation by mouse fibroblasts stimulated with TNF-α/IL-1β. DyLight 649–labeled collagen matrix (red) was imaged before and after 3 days of incubation with the fibroblasts seeded in the center of the film. Cell nuclei were imaged after 4′,6-diamidino-2-phenylindole (DAPI) labeling. The gray scale image of matrix degradation was obtained by subtraction of the matrix fluorescence intensity after 3 days of incubation with the cells from the matrix fluorescence intensity before incubation; darker areas represent larger loss of fluorescence. The loss of fluorescence was caused by mechanically damaging the film with a pipette tip (e.g., black spot in the bottom left quadrant), bleaching of the fluorescence (overall background, see B), and matrix degradation by cells (round area in the middle, surrounding cell nuclei). B to D, fluorescence bleaching without cells (B) and matrix degradation by wild-type (C) and MMP-14 knockout (D) fibroblasts. Low-resolution difference images of matrix degradation (top) were obtained as described in A. High-resolution images are projections of confocal stacks within the areas populated by cells (bottom). The cyan channel shows 4′,6-diamidino-2-phenylindole–labeled cell nuclei and magenta channel shows DyLight 649–labeled collagen. The dark areas in the confocal image of the heterotrimer matrix with wild-type (WT) fibroblasts (C, bottom left) are holes cut in the matrix by the cells. No holes were observed in the homotrimer matrix (C, bottom right) or with MMP-14 knockout fibroblasts (D).

Close modal

Homotrimer synthesis by cancer cells in culture

Type I collagen homotrimers found in carcinomas and carcinoma cell cultures could be selectively synthesized by cancer cells. However, they could also be products of cancer-associated fibroblasts and/or byproducts of residual homotrimer synthesis by all cells and selective heterotrimer degradation by collagenases massively produced within tumors. To clarify the homotrimer origin, we analyzed procollagen secreted into cell culture medium by normal dermal fibroblasts (CRL-2127) and nine cell lines from different types of cancer (Table 1).

Table 1.

Secretion of type I collagen by normal and cancer cells

Cell lineCell originCollagen (fg)/cell*Homotrimers (%)*
MMP-1 assayα1/α2 assay
CRL-2127 Foreskin, normal fibroblast 7.5 
LOX-IMVI Melanoma 2.7 25 30 
MCF-7 Mammary adenocarcinoma 0.6 27 25 
MDA-MB-231 Mammary adenocarcinoma 1.1 37 32 
MDA-MB-435 Mammary adenocarcinoma 0.3 n.d. 20 
LNCaP Prostate adenocarcinoma 0.5 25 30 
PacMet Prostate adenocarcinoma 0.4 17 25 
PC-3 Prostate adenocarcinoma 1.1 25 35 
HT-1080 Fibrosarcoma 1.7 37 28 
SY5Y Neuroblastoma 1.6 14 25 
Cell lineCell originCollagen (fg)/cell*Homotrimers (%)*
MMP-1 assayα1/α2 assay
CRL-2127 Foreskin, normal fibroblast 7.5 
LOX-IMVI Melanoma 2.7 25 30 
MCF-7 Mammary adenocarcinoma 0.6 27 25 
MDA-MB-231 Mammary adenocarcinoma 1.1 37 32 
MDA-MB-435 Mammary adenocarcinoma 0.3 n.d. 20 
LNCaP Prostate adenocarcinoma 0.5 25 30 
PacMet Prostate adenocarcinoma 0.4 17 25 
PC-3 Prostate adenocarcinoma 1.1 25 35 
HT-1080 Fibrosarcoma 1.7 37 28 
SY5Y Neuroblastoma 1.6 14 25 

*The collagen yield and homotrimer fraction varied by up to a factor of 2 from experiment to experiment. Hence, the average values reported in this table are intended to serve only as estimates.

Within 24 hours, CRL-2127 fibroblasts secreted ∼7 fg collagen per cell, in which no homotrimers were detected. Under the same conditions, cancer cells secreted 0.3 to 3 fg/cell type I collagen, of which 15% to 40% were α1(I) homotrimers. Figure 3 shows the presence of an MMP-1–resistant, homotrimeric type I collagen in medium from LNCaP cells (A and B) and illustrates assays for the homotrimer fraction based on the α1(I)/α2(I) chain ratio (C) and on the fraction of uncleaved α1(I) chains after 4 to 8 hours of incubation with MMP-1 (D). The homotrimers were not byproducts of selective heterotrimer degradation by collagenases because such degradation was negligible: (a) we observed no change in the homotrimer/heterotrimer ratio when a mixture of fluorescently labeled procollagens was prepared in a cell culture medium and incubated with cells for 24 hours (Supplementary Fig. S3). (b) We did not detect any collagen degradation upon mixing collagen solution 1:1 with the conditioned medium and only minimal degradation when MMPs in the conditioned medium were activated at 37°C for 1 hour with 1 mmol/L of p-aminophenylmercuric acetate (Supplementary Fig. S3).

Figure 3.

Measurement of the homotrimer fraction in type I collagen secreted into cell culture media. A, gel electrophoresis of Cy5-labeled-LNCaP-collagen mixed with AF488-labeled-CRL-2127-collagen before and after 4 h incubation with activated MMP-1. Each pair of images represents the same gel lane scanned for AF488 (2127) and Cy5 (LNCaP) fluorescence, respectively. B, gel electrophoresis of peptides obtained by CNBr cleavage of α1(I) bands cut out from the gels in A. The 2127 peptide pattern corresponds to the expected cleavage products of α1(I) chains (e.g., CB7 is the CNBr peptide 7 of the α1(I) chain and CB3+7+6 is the partial cleavage product with intact bonds between CB7 and adjacent CB3 and CB6). The LNCaP peptide pattern shows only minor (<10%) contamination of the α1(I) gel band with chain(s) of unknown origin (the band above CB7). C, the α1(I)/α2(I) ratios of fluorescence intensities of the gel bands for LNCaP collagen (dashed line) and homotrimer/heterotrimer mixtures with known compositions (open circles and solid regression curve). D, the fraction of unprocessed α1(I) chains after 4 h of cleavage with MMP-1 for LNCaP collagen (dashed line) and homotrimer/heterotrimer mixtures with known compositions (filled circles and solid regression curve). The x-coordinates of the intersections between the dashed and solid lines in C and D give the type I homotrimer fraction in LNCaP collagen (∼30% in both assays).

Figure 3.

Measurement of the homotrimer fraction in type I collagen secreted into cell culture media. A, gel electrophoresis of Cy5-labeled-LNCaP-collagen mixed with AF488-labeled-CRL-2127-collagen before and after 4 h incubation with activated MMP-1. Each pair of images represents the same gel lane scanned for AF488 (2127) and Cy5 (LNCaP) fluorescence, respectively. B, gel electrophoresis of peptides obtained by CNBr cleavage of α1(I) bands cut out from the gels in A. The 2127 peptide pattern corresponds to the expected cleavage products of α1(I) chains (e.g., CB7 is the CNBr peptide 7 of the α1(I) chain and CB3+7+6 is the partial cleavage product with intact bonds between CB7 and adjacent CB3 and CB6). The LNCaP peptide pattern shows only minor (<10%) contamination of the α1(I) gel band with chain(s) of unknown origin (the band above CB7). C, the α1(I)/α2(I) ratios of fluorescence intensities of the gel bands for LNCaP collagen (dashed line) and homotrimer/heterotrimer mixtures with known compositions (open circles and solid regression curve). D, the fraction of unprocessed α1(I) chains after 4 h of cleavage with MMP-1 for LNCaP collagen (dashed line) and homotrimer/heterotrimer mixtures with known compositions (filled circles and solid regression curve). The x-coordinates of the intersections between the dashed and solid lines in C and D give the type I homotrimer fraction in LNCaP collagen (∼30% in both assays).

Close modal

We also tested the effects of factors known to regulate collagen production. In particular, ascorbic acid (50 μg/mL) increased the total collagen yield from CRL-2127 fibroblasts by 200 to 300 times, but it did not induce homotrimer synthesis and it had almost no effect on other cells. Transforming growth factor-β1 (5 ng/mL) increased the collagen yield from CRL-2127 fibroblasts up to 10 to 15 times, also without inducing the homotrimer synthesis. Exogenous transforming growth factor-β1 had more complex, condition-dependent effects on cancer cells, in some cases, increasing (but never decreasing) the homotrimer synthesis.

Homotrimer synthesis by cancer cells in vivo

For in vivo testing of type I homotrimer synthesis, we selected xenograft tumors, in which we could distinguish type I collagen produced by host and engrafted cells based on different denaturation temperatures (Tm) of mouse and human collagens. We purified type I collagen from tumors produced in athymic nude mice by LOX-IMVI, PC-3, and MDA-MB-231 cells with pepsin digestion and selective salt fractionation. Analysis by DSC and CD (30), gel electrophoresis, and MMP-1 digestion revealed that normal heterotrimers of murine origin comprised >95% of total type I tumor collagen (Fig. 4; Supplementary Fig. S4).

Figure 4.

Collagen synthesis in xenograft tumors. A, gel electrophoresis of total collagen and human collagen fractions in PC-3 xenograft tumors (note slightly slower migration of human collagen). B, DSC denaturation thermogram of total xenograft collagen (bold dashed line), CD denaturation thermogram of human xenograft collagen (bold line), and DSC thermograms of mouse heterotrimers and homotrimers from spinal tendons (thin dashed line) and human heterotrimers and homotrimers produced in cell culture (thin line). DSC thermograms of xenograft collagen were compared with mouse collagen from different tissues in Supplementary Fig. S4.

Figure 4.

Collagen synthesis in xenograft tumors. A, gel electrophoresis of total collagen and human collagen fractions in PC-3 xenograft tumors (note slightly slower migration of human collagen). B, DSC denaturation thermogram of total xenograft collagen (bold dashed line), CD denaturation thermogram of human xenograft collagen (bold line), and DSC thermograms of mouse heterotrimers and homotrimers from spinal tendons (thin dashed line) and human heterotrimers and homotrimers produced in cell culture (thin line). DSC thermograms of xenograft collagen were compared with mouse collagen from different tissues in Supplementary Fig. S4.

Close modal

To separate a small fraction of human collagen produced by cancer cells from the murine collagen, we briefly incubated collagen solutions above the Tm of heterotrimeric mouse collagen but below the Tm of human collagen, cooled the solution to room temperature, degraded denatured mouse collagen with a trypsin/chymotrypsin mixture, and reprecipitated intact collagen, which comprised 1% to 2% of the initial sample. The CD denaturation thermogram of this collagen (Fig. 4B, bold dashed line) revealed two denaturation peaks; the more stable fraction consistent with human homotrimers and a less stable fraction consistent with human heterotrimers and/or murine homotrimers. From the relative intensities of the two peaks, we measured the fraction of human homotrimers (30). Comparison of this fraction with the α1(I)/α2(I) chain ratio in gel electrophoresis suggested that the reprecipitated collagen was predominantly human, i.e., virtually all mouse collagen seemed to be degraded. This conclusion was supported by differences in electrophoretic mobilities of the α1(I) and α2(I) chains in the initial tumor sample and reprecipitated collagen (Fig. 4A). Comparison with calibration mixtures of known composition (after the same treatment) allowed us to correct for partial degradation of human heterotrimers and estimate the tumor collagen composition.

Thus, we deduced that recruited (cancer associated) mouse cells produced 98% to 99% of tumor type I collagen, all of which was heterotrimeric. In contrast, transplanted human cancer cells synthesized 1% to 2% of tumor type I collagen, ∼50% of which was homotrimeric.

Proliferation and migration of cancer cells on homotrimer matrix

Previous studies revealed faster proliferation (22) and migration (23) of 8701-BC breast carcinoma cells on homotrimeric type I collagen deposited from 0.5 mol/L of acetic acid by incubation at 37°C or air drying. To compare cancer cell interactions with heterotrimer and homotrimer collagen fibers prepared under physiologic conditions, we reconstituted thin (1–2 μm), dense films of mouse-tail-tendon type I collagen fibers by in vitro fibrillogenesis. Consistent with the results for 8701-BC cells on acid-deposited human collagen (22, 23, 32), we observed no significant difference in cell adhesion, but ∼50% faster proliferation of HT-1080 cells and ∼40% faster migration of LOX-IMVI cells on the homotrimer matrix (Fig. 5; Supplementary Fig. S5).

Figure 5.

Proliferation and migration of cancer cells on heterotrimer (black) and homotrimer (gray) matrices. Cells were seeded on dense, thin films reconstituted by in vitro fibrillogenesis from wild-type (heterotrimers) and oim (homotrimers) mouse-tail-tendon type I collagen. The number of cells was evaluated by calcein fluorescence after 4 days of proliferation within whole matrix area or 1 to 2 days of migration into an initially blocked area (Supplementary Fig. S5). The results were normalized based on the average fluorescence of cells on heterotrimer matrix.

Figure 5.

Proliferation and migration of cancer cells on heterotrimer (black) and homotrimer (gray) matrices. Cells were seeded on dense, thin films reconstituted by in vitro fibrillogenesis from wild-type (heterotrimers) and oim (homotrimers) mouse-tail-tendon type I collagen. The number of cells was evaluated by calcein fluorescence after 4 days of proliferation within whole matrix area or 1 to 2 days of migration into an initially blocked area (Supplementary Fig. S5). The results were normalized based on the average fluorescence of cells on heterotrimer matrix.

Close modal

Collagen fibers in the stroma and vasculature as well as fibers produced by cancer-associated (recruited) fibroblasts form an invasion barrier. At the same time, collagen fibers are essential for maintaining the invasive phenotype and supporting migration and proliferation of cancer cells. Massive production of both collagen and MMPs that degrade collagen, mostly by cancer-associated fibroblasts, might be one solution to this dilemma. However, the present study suggests that invasive cells may benefit even more from producing fibers of MMP-resistant homotrimeric type I collagen. Not only are the homotrimers resistant to all collagenolytic MMPs in solution (MMP-1, 2, 8, and 13; Fig. 1), but the homotrimer matrix is resistant to cleavage by fibroblasts (Fig. 2) and cancer cells (Supplementary Fig. S2), even when the cells are stimulated with proinflammatory cytokines.

Because the homotrimer fibers seem to be produced only by cancer cells and not by cancer-associated fibroblasts (Table 1; Figs. 3 and 4), they comprise a small fraction of tumor collagens. As a result, cancer cells might use these fibers as MMP-resistant roadways for invasion rather than as building materials for the tumor stroma. Our data suggest that these collagenase-resistant fiber roadways may provide the necessary support for proliferation and migration of the cancer cells without forming invasion barriers (Supplementary Fig. S6). They may also promote better organization and migration of all tumor cells and more directed and efficient degradation of surrounding stroma.

In addition to MMP resistance, type I collagen homotrimer fibers have distinct mechanical properties (33, 34). Matrix mechanics plays an important role in malignancy, significantly affecting the behavior of cancer cells (3538). For instance, higher rigidity of the homotrimer fibers (34) might contribute to the faster proliferation and migration of cancer cells. Potential differences in binding of proteoglycans, cytokines, and other matrix molecules to homotrimer fibers may also affect cancer microenvironment.

Note that collagenase-resistant fibers could still be degraded by other enzymes that cleave nonhelical terminal peptides. Such cleavage is likely involved in normalizing collagen turnover in Col1a1r/r and oim mice; which make collagenase-resistant type I collagen but exhibit only mild localized fibrosis (refs. 3942; type I collagen has an altered MMP cleavage site in Col1a1r/r and is homotrimeric in oim mice). Alternative collagen cleavage does not prevent severe, generalized collagen turnover deficiency in MMP-14 knockout mice (6); although this might be related to other functions of the multifunctional (43) MMP-14. Alternative cleavage may also be less important in cancer progression.

Consistent with our hypothesis for the role of type I collagen homotrimers in cancer invasion, the same cancer cells seem to produce a higher fraction of the homotrimers in vivo (∼50%; Fig. 4) than in vitro (25–35%, Table 1). The higher homotrimer content of cancer cell–derived collagen in xenograft tumors may result from selective degradation of the heterotrimers by collagenases. However, it may also be caused by selective proliferation of cancer cell subpopulations producing more homotrimers. Indeed, a several fold reduction in the relative amount of the α2(I) chain mRNA compared with the α1(I) chain mRNA was observed after several cycles of selection for more aggressive TC-1 cells in C57BL/6 mice, as reported in the Gene Expression Omnibus database, National Center for Biotechnology Information, accession no. GSE2774 (44).

What are the factors that enable cancer cells to produce homotrimeric type I collagen? These cells are exposed to the same environment as cancer-associated fibroblasts, which do not make the homotrimers. Thus, the answer likely lies within the cells themselves. One possibility is insufficient expression of the α2(I) chain, e.g., due to methylation of the α2(I) gene (45). However, similar homotrimer synthesis by different cancer cells (Table 1) would then mean similar α2(I) chain expression deficiency. Such similarity between different cancers seems unlikely.

Another clue to answering this question might be contained in comparing different types of cells producing the homotrimers. From analysis of the literature, we found no convincing evidence for homotrimer production by cells normally responsible for collagen synthesis, except for cases with deficient α2(I) chain synthesis caused by rare mutations. In our own measurements, we also observed no detectable homotrimers in (a) fibroblast and osteoblast cultures (human and murine); (b) murine skin, tendons, and bone; and (c) normal and fibrotic human skin. In contrast, the homotrimer synthesis by embryonic cells (46), dedifferentiated cells (47), nonosteogenic bone marrow cells (48), chemically transformed cells (15, 16), cancer cells (1721), and stressed mesangial cells (49) has been well documented. These observations suggest that mature collagen-producing cells may have a mechanism for preventing the α1(I) homotrimer formation when a sufficient number of the α2(I) chains is synthesized, e.g., a specialized chaperone that promotes association and folding of two α1(I) with one α2(I) C-propeptide chains. This mechanism might be absent in cells that normally produce little or no type I collagen, in which the corresponding chaperone might not be expressed. This mechanism might also be absent or not fully functional in fetal cells.

Regardless of the underlying mechanism, homotrimeric type I collagen seems to be produced only in fetal or pathologic tissues. This property may be used for diagnostic and therapeutic targeting of cancer and other pathologies involving the homotrimer synthesis, e.g., for visualizing peripheral areas invaded by cancer cells during surgery. Indeed, insoluble, collagenase-resistant homotrimer fibers might present an ideal target, provided that a molecule that selectively binds to the homotrimers but not heterotrimers can be designed. This task may be challenging because the homotrimers do not contain unique peptide sequences. However, some triple helix regions are much less stable within the homotrimers than within the heterotrimers (26). Targeting these regions by molecules that recognize unfolded but not folded chains may present one possible solution.

No potential conflicts of interest were disclosed.

We thank Drs. Peter H. Byers, Linda A. deGraffenried, Melinda Hollingshead, Daniel McBride, James Pace, and Ulrike Schwarze for generously providing tissues and cells used for this study. Confocal microscopy was performed at the Microscopy & Imaging Core of the NICHD with the assistance of Dr. Vincent Schram.

Grant Support: Intramural Research Programs of NICHD (S. Leikin) and NIDCR (K. Holmbeck), NIH; Wellcome Trust Programme grant 075473 (H. Nagase); and NIH/NIDDK grant DK069522 (C.L. Phillips).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1
Liotta
LA
,
Stetler-Stevenson
WG
. 
Tumor invasion and metastasis: an imbalance of positive and negative regulation
.
Cancer Res
1991
;
51
:
5054s
9s
.
2
Mignatti
P
,
Rifkin
DB
. 
Biology and biochemistry of proteinases in tumor invasion
.
Physiol Rev
1993
;
73
:
161
95
.
3
Deryugina
EI
,
Quigley
JP
. 
Matrix metalloproteinases and tumor metastasis
.
Cancer Metastasis Rev
2006
;
25
:
9
34
.
4
Egeblad
M
,
Werb
Z
. 
New functions for the matrix metalloproteinases in cancer progression
.
Nat Rev Cancer
2002
;
2
:
161
74
.
5
Nagase
H
,
Visse
R
,
Murphy
G
. 
Structure and function of matrix metalloproteinases and TIMPs
.
Cardiovasc Res
2006
;
69
:
562
73
.
6
Holmbeck
K
,
Bianco
P
,
Caterina
J
, et al
. 
MT1-MMP-deficient mice develop dwarfism, osteopenia, arthritis, and connective tissue disease due to inadequate collagen turnover
.
Cell
1999
;
99
:
81
92
.
7
Grzesiak
JJ
,
Ho
JC
,
Moossa
AR
,
Bouvet
M
. 
The integrin-extracellular matrix axis in pancreatic cancer
.
Pancreas
2007
;
35
:
293
301
.
8
Yilmaz
M
,
Christofori
G
. 
EMT, the cytoskeleton, and cancer cell invasion
.
Cancer Metastasis Rev
2009
;
28
:
15
33
.
9
Kalluri
R
,
Zeisberg
M
. 
Fibroblasts in cancer
.
Nat Rev Cancer
2006
;
6
:
392
401
.
10
Pucci-Minafra
I
,
Andriolo
M
,
Basirico
L
, et al
. 
Absence of regular α2(I) collagen chains in colon carcinoma biopsy fragments
.
Carcinogenesis
1998
;
19
:
575
84
.
11
Shapiro
FD
,
Eyre
DR
. 
Collagen polymorphism in extracellular matrix of human osteosarcoma
.
J Natl Cancer Inst
1982
;
69
:
1009
16
.
12
Yamagata
S
,
Yamagata
T
. 
FBJ virus-induced osteosarcoma contains type I, type I trimer, type III as well as type V collagens
.
J Biochem
1984
;
96
:
17
26
.
13
Moro
L
,
Smith
BD
. 
Identification of collagen α1(I) trimer and normal type I collagen in a polyoma virus-induced mouse tumor
.
Arch Biochem Biophys
1977
;
182
:
33
41
.
14
Asokan
R
,
Puvanakrishnan
R
,
Ravichandran
LV
,
Kokila
V
,
Reddy
GK
,
Dhar
SC
. 
Purification and characterization of collagens from rat fibrosarcoma induced by 3-methylcholanthrene
.
Mol Cell Biochem
1993
;
121
:
99
107
.
15
Marsilio
E
,
Sobel
ME
,
Smith
BD
. 
Absence of procollagen α 2(I) mRNA in chemically transformed rat liver epithelial cells
.
J Biol Chem
1984
;
259
:
1401
4
.
16
Smith
BD
,
Niles
R
. 
Characterization of collagen synthesized by normal and chemically transformed rat liver epithelial cell lines
.
Biochemistry
1980
;
19
:
1820
5
.
17
DeClerck
YA
,
Bomann
ET
,
Spengler
BA
,
Biedler
JL
. 
Differential collagen biosynthesis by human neuroblastoma cell variants
.
Cancer Res
1987
;
47
:
6505
10
.
18
Little
CD
,
Church
RL
,
Miller
RA
,
Ruddle
FH
. 
Procollagen and collagen produced by a teratocarcinoma-derived cell line, TSD4: evidence for a new molecular form of collagen
.
Cell
1977
;
10
:
287
95
.
19
Minafra
S
,
Luparello
C
,
Rallo
F
,
Pucci-Minafra
I
. 
Collagen biosynthesis by a breast carcinoma cell strain and biopsy fragments of the primary tumour
.
Cell Biol Int Rep
1988
;
12
:
895
905
.
20
Rupard
JH
,
Dimari
SJ
,
Damjanov
I
,
Haralson
MA
. 
Synthesis of type I homotrimer collagen molecules by cultured human lung adenocarcinoma cells
.
Am J Pathol
1988
;
133
:
316
26
.
21
Ishikoh
AU
,
Hayashi
A
,
Tokimitsu
I
,
Tajima
S
,
Nishikawa
T
. 
Coordinate modulation of melanogenesis and type I trimer collagen secretion by type I collagen substratum during reversible conversion between melanotic and amelanotic cells in mouse B16 melanoma
.
J Biochem
1994
;
116
:
610
4
.
22
Schillaci
R
,
Luparello
C
,
Minafra
S
. 
Type I and I-trimer collagens as substrates for breast carcinoma cells in culture. Effect on growth rate, morphological appearance and actin organization
.
Eur J Cell Biol
1989
;
48
:
135
41
.
23
Luparello
C
,
Sheterline
P
,
Pucci-Minafra
I
,
Minafra
S
. 
A comparison of spreading and motility behaviour of 8701-BC breast carcinoma cells on type I, I-trimer and type V collagen substrata. Evidence for a permissive effect of type I-trimer collagen on cell locomotion
.
J Cell Sci
1991
;
100
:
179
85
.
24
Narayanan
AS
,
Meyers
DF
,
Page
RC
,
Welgus
HG
. 
Action of mammalian collagenases on type I trimer collagen
.
Coll Relat Res
1984
;
4
:
289
96
.
25
Makareeva
E
,
Cabral
WA
,
Marini
JC
,
Leikin
S
. 
Molecular mechanism of α 1(I)-osteogenesis imperfecta/Ehlers-Danlos syndrome: unfolding of an N-anchor domain at the N-terminal end of the type I collagen triple helix
.
J Biol Chem
2006
;
281
:
6463
70
.
26
Kuznetsova
NV
,
McBride
DJ
,
Leikin
S
. 
Changes in thermal stability and microunfolding pattern of collagen helix resulting from the loss of α2(I) chain in osteogenesis imperfecta murine
.
J Mol Biol
2003
;
331
:
191
200
.
27
Schwarze
U
,
Hata
R
,
McKusick
VA
, et al
. 
Rare autosomal recessive cardiac valvular form of Ehlers-Danlos syndrome results from mutations in the COL1A2 gene that activate the nonsense-mediated RNA decay pathway
.
Am J Hum Genet
2004
;
74
:
917
30
.
28
Chung
L
,
Dinakarpandian
D
,
Yoshida
N
, et al
. 
Collagenase unwinds triple-helical collagen prior to peptide bond hydrolysis
.
EMBO J
2004
;
23
:
3020
30
.
29
Birkedal-Hansen
H
,
Yamada
S
,
Windsor
J
, et al
. 
Matrix metalloproteinases
.
Curr Protoc Cell Biol
2008
.
Chapter 10: Unit 10.8
.
30
Makareeva
E
,
Mertz
EL
,
Kuznetsova
NV
, et al
. 
Structural heterogeneity of type I collagen triple helix and its role in osteogenesis imperfecta
.
J Biol Chem
2008
;
283
:
4787
98
.
31
Chipman
SD
,
Sweet
HO
,
McBride
DJ
, et al
. 
Defective proα2(I) collagen synthesis in a recessive mutation in mice: a model of human osteogenesis imperfecta
.
Proc Natl Acad Sci U S A
1993
;
90
:
1701
5
.
32
Ghersi
G
,
La Fiura
AM
,
Minafra
S
. 
Direct adhesion to type I and homotrimer collagens by breast carcinoma and embryonic epithelial cells in culture: a comparative study
.
Eur J Cell Biol
1989
;
50
:
279
84
.
33
Misof
K
,
Landis
WJ
,
Klaushofer
K
,
Fratzl
P
. 
Collagen from the osteogenesis imperfecta mouse model (oim) shows reduced resistance against tensile stress
.
J Clin Invest
1997
;
100
:
40
5
.
34
Han
S
,
McBride
DJ
,
Losert
W
,
Leikin
S
. 
Segregation of type I collagen homo- and heterotrimers in fibrils
.
J Mol Biol
2008
;
383
:
122
32
.
35
Alexander
NR
,
Branch
KM
,
Parekh
A
, et al
. 
Extracellular matrix rigidity promotes invadopodia activity
.
Curr Biol
2008
;
18
:
1295
9
.
36
Kostic
A
,
Lynch
CD
,
Sheetz
MP
. 
Differential matrix rigidity response in breast cancer cell lines correlates with the tissue tropism
.
PLoS One
2009
;
4
:
e6361
.
37
Paszek
MJ
,
Zahir
N
,
Johnson
KR
, et al
. 
Tensional homeostasis and the malignant phenotype
.
Cancer Cell
2005
;
8
:
241
54
.
38
Ulrich
TA
,
de Juan Pardo
EM
,
Kumar
S
. 
The mechanical rigidity of the extracellular matrix regulates the structure, motility, and proliferation of glioma cells
.
Cancer Res
2009
;
69
:
4167
74
.
39
Beare
AH
,
O'Kane
S
,
Krane
SM
,
Ferguson
MW
. 
Severely impaired wound healing in the collagenase-resistant mouse
.
J Invest Dermatol
2003
;
120
:
153
63
.
40
Liu
X
,
Wu
H
,
Byrne
M
,
Jeffrey
J
,
Krane
S
,
Jaenisch
R
. 
A targeted mutation at the known collagenase cleavage site in mouse type I collagen impairs tissue remodeling
.
J Cell Biol
1995
;
130
:
227
37
.
41
Zhao
W
,
Byrne
MH
,
Wang
Y
,
Krane
SM
. 
Osteocyte and osteoblast apoptosis and excessive bone deposition accompany failure of collagenase cleavage of collagen
.
J Clin Invest
2000
;
106
:
941
9
.
42
Brodeur
AC
,
Wirth
DA
,
Franklin
CL
,
Reneker
LW
,
Miner
JH
,
Phillips
CL
. 
Type I collagen glomerulopathy: postnatal collagen deposition follows glomerular maturation
.
Kidney Int
2007
;
71
:
985
93
.
43
Barbolina
MV
,
Stack
MS
. 
Membrane type 1-matrix metalloproteinase: substrate diversity in pericellular proteolysis
.
Semin Cell Dev Biol
2008
;
19
:
24
33
.
44
Lin
KY
,
Lu
D
,
Hung
CF
, et al
. 
Ectopic expression of vascular cell adhesion molecule-1 as a new mechanism for tumor immune evasion
.
Cancer Res
2007
;
67
:
1832
41
.
45
Sengupta
PK
,
Smith
EM
,
Kim
K
,
Murnane
MJ
,
Smith
BD
. 
DNA hypermethylation near the transcription start site of collagen α2(I) gene occurs in both cancer cell lines and primary colorectal cancers
.
Cancer Res
2003
;
63
:
1789
97
.
46
Leheup
BP
,
Federspiel
SJ
,
Guerry-Force
ML
, et al
. 
Extracellular matrix biosynthesis by cultured fetal rat lung epithelial cells. I. Characterization of the clone and the major genetic types of collagen produced
.
Lab Invest
1989
;
60
:
791
807
.
47
Mayne
R
,
Vail
MS
,
Miller
EJ
. 
Analysis of changes in collagen biosynthesis that occur when chick chondrocytes are grown in 5-bromo-2′-deoxyuridine
.
Proc Natl Acad Sci U S A
1975
;
72
:
4511
5
.
48
Diduch
DR
,
Coe
MR
,
Joyner
C
,
Owen
ME
,
Balian
G
. 
Two cell lines from bone marrow that differ in terms of collagen synthesis, osteogenic characteristics, and matrix mineralization
.
J Bone Joint Surg Am
1993
;
75
:
92
105
.
49
Haralson
MA
,
Jacobson
HR
,
Hoover
RL
. 
Collagen polymorphism in cultured rat kidney mesangial cells
.
Lab Invest
1987
;
57
:
513
23
.

Supplementary data