Research ArticleFibrosis

Thrombospondin-1 promotes matrix homeostasis by interacting with collagen and lysyl oxidase precursors and collagen cross-linking sites

See allHide authors and affiliations

Science Signaling  29 May 2018:
Vol. 11, Issue 532, eaar2566
DOI: 10.1126/scisignal.aar2566

Thrombospondin-1 controls collagen homeostasis directly

Collagens, components of the extracellular matrix important for tissue architecture and function, undergo a complex series of processing events before and after secretion. Deposition of excessive amounts of collagens and abnormal cross-linking of collagen fibrils are associated with fibrosis. Using knockout mice, primary human dermal fibroblasts, and in vitro binding assays, Rosini et al. found that thrombospondin-1 (TSP1) bound to and inhibited the processing of the precursor form of the collagen cross-linking enzyme lysyl oxidase. TSP1 also bound to collagen molecules intracellularly and at conserved cross-linking sites. Blocking the latter interaction extracellularly stimulated fibroblasts to undergo inappropriate conversion to myofibroblasts. Although the mechanism by which the TSP1-collagen interaction prevents fibroblast-to-myofibroblast conversion is unclear, these findings identify a direct role for TSP1 as a modulator of the collagen matrix.

Abstract

Fibrillar collagens of the extracellular matrix are critical for tissue structure and physiology; however, excessive or abnormal deposition of collagens is a defining feature of fibrosis. Regulatory mechanisms that act on collagen fibril assembly potentially offer new targets for antifibrotic treatments. Tissue weakening, altered collagen fibril morphologies, or both, are shared phenotypes of mice lacking matricellular thrombospondins. Thrombospondin-1 (TSP1) plays an indirect role in collagen homeostasis through interactions with matrix metalloproteinases and transforming growth factor–β1 (TGF-β1). We found that TSP1 also affects collagen fibril formation directly. Compared to skin from wild-type mice, skin from Thbs1−/− mice had reduced collagen cross-linking and reduced prolysyl oxidase (proLOX) abundance with increased conversion to catalytically active LOX. In vitro, TSP1 bound to both the C-propeptide domain of collagen I and the highly conserved KGHR sequences of the collagen triple-helical domain that participate in cross-linking. TSP1 also bound to proLOX and inhibited proLOX processing by bone morphogenetic protein-1. In human dermal fibroblasts (HDFs), TSP1 and collagen I colocalized in intracellular vesicles and on extracellular collagen fibrils, whereas TSP1 and proLOX colocalized only in intracellular vesicles. Inhibition of LOX-mediated collagen cross-linking did not prevent the extracellular association between collagen and TSP1; however, treatment of HDFs with KGHR-containing, TSP1-binding, triple-helical peptides disrupted the collagen-TSP1 association, perturbed the collagen extracellular matrix, and increased myofibroblastic differentiation in a manner that depended on TGF-β receptor 1. Thus, the extracellular KGHR-dependent interaction of TSP1 with fibrillar collagens contributes to fibroblast homeostasis.

INTRODUCTION

Fibrillar collagens are abundant and essential components of the extracellular matrix (ECM) of connective tissues. The processing and secretion of collagen molecules and their initial assembly into fibrils have been studied extensively, yet knowledge of the cell and molecular mechanisms that regulate the organization of collagen fibrils within the ECM remains limited. This is of major importance for human health because fibrosis, the excessive and disorderly deposition of collagenous ECM by myofibroblasts as a result of tissue injury and repair, is a central and currently untreatable pathology in many chronic human diseases (1). Although the collagen family is quite extensive, the fibrillar collagens I, II, and III are the most abundant. Collagens II and III are homotrimers, with three identical α chains and separate gene products for each type. Each assembles as a right-handed triple helix. Collagen I, the most widespread, differs in that it is composed of two α1 chains and a single α2 chain. Collagen II is the principal collagen of cartilage, whereas collagen I predominates in the bone and tendon (2). Collagens I and III both occur in skin and blood vessel walls, but collagen I is the most abundant and is the main focus of the present work.

The synthesis of fibrillar collagens involves multiple steps of processing and posttranslational assembly (2). Each α chain is translated as a preproprotein from which the secretory signal peptide is cleaved upon cotranslational entry into the lumen of the endoplasmic reticulum (ER). This is followed by steps important for stable assembly of the triple helix: posttranslational hydroxylation of lysine and proline residues and glycosylation of lysines (3). The C-propeptide domain guides α chain assembly into procollagen molecules, forming an in-register triple helix and preventing premature intracellular fibril nucleation by restricting lateral packing (4). Upon secretion, the C-propeptide domain is cleaved off mainly by bone morphogenetic protein-1 (BMP-1), and the N-propeptide is cleaved off by a disintegrin and metalloproteinase with thrombospondin motifs 2 (ADAMTS2), generating mature collagen molecules, consisting of the N- and C-telopeptides and the central triple-helical domain, that are competent for fibril assembly. In parallel, secreted prolysyl oxidase (proLOX) is also cleaved by BMP-1 to yield active mature LOX (mLOX), which catalyzes oxidative deamination of lysine residues in the collagen N- and C-telopeptides (3, 5). The resulting reactive lysine aldehydes target lysines or hydroxylysines of KGHR motifs within the triple helix to cross-link and stabilize the assembled fibrils (6).

In common with collagens, thrombospondins (TSPs) are among the most ancient and conserved components of the ECM (7). TSPs are matricellular proteins that have context- and tissue-specific roles through interactions with cell-surface receptors, growth factors, and other ECM proteins (8, 9). There are five TSPs in mammals [TSP1 to TSP5; TSP5 is also known as COMP (cartilage oligomeric matrix protein)], and in both humans and mice, tissue abundance of TSP1 and TSP2 increases with aging, during which the ECM becomes less elastic and more rigid (10, 11). Rare polymorphisms in TSP1 have also been implicated in increased risk of cardiovascular disease (12, 13). TSP1 acts as an inhibitor of tumor growth (14) because it inhibits angiogenesis; however, the presence of TSP1 in tumor stroma may promote cell migration, ECM turnover, or both, leading to increased tumor metastasis (15). Increased abundance, extracellular deposition of TSPs, or both, also correlate with fibrotic events (1618). Knockout mice have been reported for each individual TSP and show distinct tissue phenotypes and altered pathological susceptibilities that relate to altered cell-ECM interactions and signaling, as well as effects on ECM properties (1923). Mice lacking Thbs1, the gene that encodes TSP1, have increased vascular density in many tissues (24, 25), delayed wound healing, altered inflammatory responses, reduced abundance of active transforming growth factor–β (TGF-β) under various challenges, and increased susceptibility to tumor progression due to loss of the normal antiangiogenic and p53-regulatory activities of TSP1 (2628). An emerging coincident phenotype reported for Thbs1−/−, Thbs2−/−, Thbs4−/−, and Thbs5−/− mice is the disorganization of collagen fibril packing and altered fibril organization, manifested as increased cross-sectional areas of collagen fibrils in various tissues of young, healthy mice and correlated in several instances with mechanical weakening of the tissue (2123, 29, 30). These phenotypes are reminiscent of tissue-specific alterations to collagen fibrils observed in the dermis or tendons upon knockout of small leucine-rich proteoglycans (SLRPs) such as decorin or biglycan (31, 32). Similarly, mice lacking LOX also show impaired collagen fibril organization (33). LOX is a vital, copper-dependent enzyme with roles in ECM integrity, vessel wall mechanotransduction, and tumor progression that initiates extracellular cross-linking of collagen or elastin (34).

Investigations of the molecular basis for the TSP-null phenotypes have emphasized indirect regulatory mechanisms. For example, the activities of matrix metalloprotease-2 (MMP-2) and MMP-9 are altered in Thbs1−/− and Thbs2−/− mice, and TSP1 and TSP2 bind to and inhibit the activities of these proteases in vitro (3537). Cell type–specific effects on collagen transcription have also been noted: Thbs1−/− vascular cells have increased expression and secretion of collagen α1(I) and α2(I) chains (38), whereas the N-terminal domain of TSP1 increased collagen production and ECM assembly in vitro and in a sponge implant in vivo (39). Because TSP1 binds to and activates latent TGF-β, reduced TGF-β signaling in tissues of Thbs1−/− mice also affects ECM properties (28, 40). There remains a major gap in knowledge on the mechanisms of direct interaction of TSPs with fibrillar collagens and the importance of these in cell-ECM physiology. In vitro biochemical and electron microscopy studies have shown that several TSPs bind fibrillar collagens I, II, III, and V, and TSP5 has also been shown to bind to collagen IX (41, 42). However, many of these studies relied on supraphysiological Zn2+ concentrations to demonstrate the protein-protein interactions (43).

In view of the potential physiological importance of TSP-collagen interactions in ECM reorganization during aging, tumor progression, tissue fibrosis, and other disease states, we have focused on TSP1 and set out to elucidate the molecular basis of interactions between TSP1 and fibrillar collagen. We quantified the status of fibrillar collagens and LOX in the skin of Thbs1−/− mice and examined the binding of TSP1 to native and denatured collagens, recombinant collagens and collagen-derived peptides, and a library of synthetic triple-helical peptides designated the “Collagen Toolkit” (44). The Toolkit peptides are homotrimeric, which simplifies their synthesis. However, because there is much sequence homology between collagens I, II, and III, observations made using Toolkits II and III can often be applied directly to collagen I. We applied these findings to the physiological context of human dermal fibroblasts (HDF) to examine the functional and regulatory roles of these interactions. Our data reveal that TSP1 participates in both intracellular and extracellular associations with fibrillar collagens that affect collagen cross-linking and fibroblast phenotype.

RESULTS

Abundance of collagen I and LOX is altered in the skin of Thbs1−/− mice

Collagen fibrils in the dermis of young, healthy Thbs1 knockout (Thbs1−/−) mice are characterized by irregular packing and increased cross-sectional areas (29). To investigate the underlying molecular basis, we examined 8-week-old, male C57BL/6 and C57BL/6Thbs1−/− mice (38). We confirmed the Thbs1−/− status of the latter mice by immunoblotting extracts from spleen, a tissue that normally has high abundance of TSP1 (Fig. 1A). We extracted collagens from the dermis of wild-type and Thbs1−/− mice with acetic acid (HAc), which releases fibrillar collagen molecules polymerized by aldimine cross-links, and then measured the hydroxyproline content of these extracts to quantify collagen content. The HAc extracts from Thbs1−/− samples contained less hydroxyproline, an indication of altered collagen cross-linking, than samples from wild-type mice (Fig. 1B). Analysis by reducing SDS–polyacrylamide gel electrophoresis (PAGE) and subsequent quantification revealed a significant decrease in collagen α2(I) monomers and altered proportions of cross-linked β1,2 dimers in Thbs1−/− samples compared to controls (Fig. 1, C to E). Because both of these features are indicative of altered collagen cross-linking, we examined the status of the major collagen cross-linking enzyme LOX by immunoblotting of tissue extracts. The Thbs1−/− samples had significantly lower abundance of both proLOX, the catalytically inactive LOX proprotein, and mLOX, the catalytically active form (Fig. 1, F to H). The Thbs1−/− samples also tended to show an increased ratio of mLOX to proLOX (Fig. 1I). Collectively, these data implicated both TSP1-collagen I and TSP1-LOX interactions as possible mechanisms underlying the perturbations to collagen fibrils observed in Thbs1−/− mice.

Fig. 1 Fibrillar collagens and LOX in skin samples from wild-type and Thbs1−/− mice.

(A) Immunoblots showing TSP1 in spleen samples from wild-type (WT) and Thbs1−/− mice (four mice of each genotype, numbered 1 to 4). Ponceau staining of the gel used for immunoblotting is shown as a loading control. (B) Hydroxyproline content of HAc extracts from WT and Thbs1−/− skin samples. n = 4 for each genotype. (C) SDS-PAGE analysis of acid-extracted collagens from the skin of WT and Thbs1−/− mice under reducing conditions. The samples were equalized for hydroxyproline content. Bands corresponding to various forms of collagen trimers, dimers, and monomers are noted. n = 4 for each genotype. (D) Quantification of collagen α2(I) in WT and Thbs1−/− samples. n = 4 for each genotype. (E) Proportions of different collagen forms extracted from the skin of WT and Thbs1−/− mice, given as mean ± SD. n = 4 animals per genotype. (F) Immunoblot showing proLOX and mLOX in total skin extracts from WT and Thbs1−/− mice. Ponceau staining of the gel used for immunoblotting is shown as a loading control. n = 4 for each genotype. (G and H) Quantification of proLOX (G) and mLOX (H) from immunoblots, normalized to the major loading control band. n = 4 for each genotype. a.u., arbitrary units. (I) mLOX/proLOX ratio from the quantitations in (G) and (H) (P = 0.057). In each scatter plot, the bar indicates the mean. Data were analyzed by Mann-Whitney U test. Each dot in (B) and (D) represents the value from an individual animal. Each dot in (G) to (I) represents the mean of four separate measurements from each animal.

TSP1 inhibits proLOX cleavage and binds to multiple fibrillar collagens in vitro

In view of the relative increase in mLOX relative to proLOX in Thbs1−/− skin, we examined in vitro whether TSP1 affects the processing of proLOX to mLOX by the metallopeptidase BMP-1 (5). In in vitro proLOX cleavage assays, under conditions of partial proLOX cleavage, TSP1 inhibited BMP-1 activity in a concentration-dependent manner, but bovine serum albumin (BSA) did not (Fig. 2, A and B). With regard to the above data and the complex processing of fibrillar procollagen that generates the mature collagen molecule (Fig. 2C), we next tested the binding of native TSP1 isolated from human platelets to different types and forms of fibrillar collagens in vitro by solid-phase binding assays. We tested TSP1 binding to denatured bovine collagen I; pepsin-extracted bovine collagens I and II from the skin and trachea, respectively; or pepsin-extracted human collagen III from placenta and native collagen I fibrils isolated from equine tendon. Pepsin cleaves collagen molecules at the inner ends of the N- and C-telopeptides, thus leaving the triple-helical domains in isolation (Fig. 2C). Synthetic triple-helical Gly-Pro-Pro peptide (GPP10) (45) and BSA were included as negative controls, and all assays included [Ca2+] and [Zn2+] in the physiological range. TSP1 bound equally well to pepsin-extracted collagens I, II, or III (Fig. 2D). However, as investigated with regard to collagen I, the binding of TSP1 was related to the molecular form of collagen because we detected no specific TSP1 binding to denatured collagen I, moderate binding to pepsin-extracted collagen I, and strong binding to native, cross-linked collagen I fibrils that also include the N- and C-telopeptides (Fig. 2D). We also tested binding of TSP1 to recombinant trimers of the collagen C-propeptide domain (C-Pro as shown in Fig. 2C), which drives triple-helix assembly in fibrillar procollagens and also prevents inappropriate fibril formation intracellularly (2, 4). We observed concentration-dependent binding of TSP1 to recombinant homotrimers of procollagen α1(I) C-propeptides (CPI) but not to recombinant homotrimers of procollagen α1(III) C-propeptides (CPIII), which share 66% sequence identity with CPI (Fig. 2E). Binding of TSP1 to CPI depended on the presence of calcium, which is required for homotrimerization of the C-propeptides (Fig. 2F) (4, 46). Comparison of the concentration dependence of TSP1 binding to CPI, pepsin-digested collagen I, or a recombinant, homotrimeric “mini” procollagen I (rProCOL1A1; this protein includes the N-propeptide, N-telopeptide, the 33 most N-terminal and 33 most C-terminal Gly-Xaa-Yaa repeats, C-telopeptide, and C-propeptide) demonstrated enhanced binding of TSP1 to the latter protein (Fig. 2G). Thus, TSP1 binding to procollagen I involved the CPI domain and at least one other binding site within the triple-helical region of this fibrillar collagen.

Fig. 2 TSP1 binds to proLOX and fibrillar collagens I, II, and III in vitro.

(A) Representative immunoblots of assays for cleavage of recombinant human proLOX by BMP-1 in the absence or presence of increasing concentrations of TSP1, as indicated. BSA was used as a negative control. n = 3 independent experiments. (B) Quantification of the concentration-dependent inhibition of BMP-1–mediated proLOX cleavage by TSP1. Each data point represents the mean from three independent experiments, and the error bars indicate SEM. (C) Schematic diagram of the domains of a fibrillar procollagen molecule. (D) Binding of TSP1 to immobilized collagens in solid-phase binding assays. Collagen I (Col I) (denatured), Col I, Col II, and Col III samples were derived from pepsin-digested collagen preparations. GPP10 and BSA were included as negative controls. One-way analysis of variance (ANOVA) and Bonferroni’s multiple comparison test were performed against GPP10. n = 4 independent experiments. (E) Concentration dependence of TSP1 binding to the indicated immobilized proteins (ligands). The ligands (CPI or CPIII) were immobilized and incubated with or without TSP1 (T). A mouse antibody recognizing TSP1 (C9) and goat anti-mouse antibody (GAM) were included in each reaction to quantify binding. n = 4 independent experiments. (F) Solid-phase binding assays testing TSP1 (T) binding to immobilized CPI or BSA in the presence of either Ca2+ or EDTA and quantified with the antibodies C9 and GAM. n = 4 independent experiments. One-way ANOVA and Bonferroni’s multiple comparison test were performed against BSA. (G) Concentration dependence of TSP1 (T) binding to the indicated immobilized proteins (ligands), rhProCOL1A1, Col I, and CPI. Antibodies C9 and GAM were used to quantify binding. n = 4 independent experiments. Dotted lines in (E) and (G) indicate negative control assays without TSP1. Data points in (D) to (G) indicate the mean, and error bars indicate the SEM.

TSP1 binds to specific peptides within the triple-helical domain of collagen II

To identify the putative binding motif(s) for TSP1 within fibrillar collagen triple-helical domains, we screened purified human TSP1 against the Collagen Toolkit II, a library of triple-helical peptides that spans the entire triple-helical domain of homotrimeric collagen II (44), by a solid-phase binding assay similar to an enzyme-linked immunosorbent assay. Specific Toolkit peptides were identified to bind TSP1 (Fig. 3A). We retested these candidate peptides in multiple independent experiments that included the additional negative control of adding the primary and secondary antibodies in the absence of TSP1 (Fig. 3B). Peptides II-5, II-44, and II-52 were confirmed to bind TSP1 in a specific and statistically significant manner (Fig. 3B) compared to GPP10. We also screened Collagen Toolkit III, a library of triple-helical peptides that spans the entire triple-helical domain of homotrimeric collagen III (fig. S1A) (45), and identified TSP1 binding to peptides III-5, III-52, and III-53 greater than binding to GPP10, but this binding was lower than to the equivalent collagen II peptides (fig. S1B). We tested whether binding of TSP1 to the collagen II peptides was affected by chelation of Ca2+ ions because the three-dimensional structure of the C-terminal region of TSPs depends on bound Ca2+ ions (4749). We found that TSP1 binding to peptide II-44 above background was abolished, and binding to II-52 was greatly reduced in the absence of cations. Binding to peptide II-5 was highly variable in the presence of EDTA (Fig. 3C). Similar results were obtained with the Toolkit III peptides (fig. S1C).

Fig. 3 Identification of specific TSP1-binding peptides within the triple-helical region of collagen II.

(A) Quantification of the binding to TSP1 to the 56 peptides (numbered for every fifth peptide as II1, etc.) of the Collagen Toolkit II in the presence of Ca2+ and Zn2+. The horizontal line represents the background binding to GPP10. Each bar represents the mean of duplicate samples. (B) Binding of TSP1 (T) to specific TSP1-binding peptides as identified from the initial hits in Toolkit II in the presence of Ca2+ and Zn2+. Binding was quantified by indirect colorimetric assay using the antibodies C9 and GAM. n = 4 independent experiments. (C) Quantification of the binding of TSP1 (T) to the indicated peptides in the presence of the Ca2+ chelator EDTA. n = 4 independent experiments. (D) Binding of proLOX to immobilized TSP1 and the indicated proteins and peptides derived from fibrillar collagens. n = 4 independent experiments. In all panels, GPP10 and BSA were included as negative controls. Pepsin-digested collagens I and III were used as positive controls in (B) and (C). In (B) to (D), one-way ANOVA and Bonferroni’s multiple comparison tests were performed against GPP10. Each bar indicates the mean, and error bars indicate the SEM.

In view of a previous report that LOX binds to collagen fibrils (50), we also tested for binding of proLOX to TSP1, pepsin-digested collagen molecules, C-propeptides, rProCOL1A1, collagen I fibrils, or TSP1-binding collagen triple-helical peptides. Under conditions in which we readily detected proLOX binding to TSP1, no specific binding of proLOX to pepsin-digested collagen molecules or TSP1-binding collagen triple-helical peptides was detected (Fig. 3D). Notably, proLOX bound to CPI and rProCOL1A1 but not to CPIII, indicating the presence of a specific LOX-binding site within the CPI domain of procollagen I (Fig. 3D).

The conserved KGHR site, involved in cross-linking of collagen molecules, is a minimum motif for TSP1 binding

With the exception of peptide II-44, the Collagen Toolkit peptides that bound to TSP1 each contain a common amino acid sequence motif, KGHR (44). Therefore, we explored whether TSP1 binding depended on the KGHR motif, first by testing TSP1 binding to shorter derivatives of peptide II-52 that include the KGHR motif. TSP1 bound equally well to peptide II-52a, which contains four guest GXY triplets, peptides II-52b and II-52c, which contain three GXY triplets, and peptide II-52d, which consists of two GXY triplets, thus establishing GLKGHR as a minimum triple-helical region sufficient for TSP1 binding (Fig. 4A). The reduced binding of these short peptides relative to the longer peptide II-52 may reflect conformational differences in these very short regions of native sequence or a possible positive contribution of a flanking sequence, as observed for the interaction between matrix metalloproteinase 13 and collagen II (51). Next, we tested versions of the GLKGHR peptide in which each residue, except for the glycine residues that are required to maintain the triple-helical conformation, was mutated to alanine. Binding to TSP1 was maintained when leucine was replaced with alanine and was abolished when any one of the lysine, histidine, or arginine residues was replaced with alanine (Fig. 4A). Thus, an intact cluster of positively charged residues was essential for TSP1 binding. Although not examined further in the context of this study, we noted that peptide II-44 contains the motif RGER and speculate that the closely spaced arginine residues may explain the binding of TSP1 to this otherwise unrelated peptide.

Fig. 4 KGHR is a minimal and conserved motif for TSP1 binding.

(A) Binding of TSP1 (T) to derivatives of TSP1-binding peptides from the Collagen Toolkit II, truncated or mutated as indicated (guest peptides). Binding was quantified by indirect colorimetric assay using the antibodies C9 and GAM. Data are from five independent experiments. (B) Alignment showing the KGHR motifs (red text) in the N-terminal and C-terminal triple-helical regions of the indicated collagen chains of Homo sapiens (Hs). The numbers 87 and 930 refer to the amino acid position in the triple-helical domain of collagen α1(I). Black shading, identical residues; gray shading, conservative substitutions; no shading, no conservation. (C) Concentration-dependent inhibition of TSP1 binding to collagen I fibrils by Collagen Toolkit peptide II-52. Each data point represents the mean, and bars represent the SEM of three independent experiments. (D to F) Merged phase-contrast and immunofluorescence images of TSP1 binding to collagen I fibrils isolated from tendons (D); TSP1 binding to collagen I fibrils in the presence of the indicated Collagen Toolkit II peptides (E) or in the presence of peptides of the indicated sequence (F). Images are representative of four independent experiments. Scale bar, 15 μm. (G) Quantification of immunofluorescence staining for TSP1 bound to collagen fibrils, calculated as total fluorescence intensity within fibrils normalized on pixel count within fibrils. Each dot represents the value from an individual experiment, each bar indicates the mean, and the error bars indicate the SEM from four independent experiments. Data were analyzed by Kruskal-Wallis test and pairwise tests against BSA with Holm’s correction.

The KGHR motif occurs at only two sites within fibrillar collagens. The N- and C-terminal KGHR sites are centered at or close to Gly88 and Gly931, respectively, of the 1014-residue consensus triple-helical domain. In the fibrillar collagen chains of humans, the N-terminal motif is fully conserved in all collagens, except for COL1A2, COL5A3, COL24A1, and COL27A1, which present variant KGIR, KGQR, KGLK, or KGHK motifs, respectively (Fig. 4B). At the C-terminal site, the motif is conserved in most fibrillar collagens yet is shifted by one GXY triplet and replaced by KGHN in COL1A2 and is disrupted in COL24A1 and COL27A1 (Fig. 4B). KGHR motifs are also substantially conserved in various metazoans (52).

KGHR peptides inhibit the binding of TSP1 to native collagen I fibrils

The KGHR motif has a known role in the molecular cross-linking of mature collagen molecules as a target of reactive lysine aldehyde residues in the N- and C-terminal telopeptides that are generated by oxidative deamination by LOX (6). To determine whether the TSP1-KGHR interaction identified here is physiologically relevant for TSP1 binding to native fibrillar collagen, we tested whether KGHR-containing, triple-helical peptides can inhibit binding of TSP1 to native, cross-linked collagen I fibrils isolated from equine tendons. We found that the non–TSP1-binding peptide II-8 (Fig. 3) had minimal inhibitory activity at all concentrations tested, whereas the KGHR-containing peptide II-52 showed concentration-dependent inhibition with an approximate IC50 (mean inhibitory concentration) of 12.5 μM (Fig. 4C). By examining the TSP1 binding by indirect immunofluorescence, we found that, under control conditions, TSP1 bound along the length of the fibrils in a punctate pattern (Fig. 4D). In comparison to fibrils incubated with BSA, the TSP1 antibody reactivity depended on the addition of TSP1 and was sensitive to the concentration of TSP1 (Fig. 4D). Preincubation of TSP1 with peptide II-8 before their application to collagen fibrils did not affect binding of TSP1 to fibrils (Fig. 4E), whereas preincubation of TSP1 with peptide II-52 substantially reduced TSP1 binding to fibrils (Fig. 4E). We found that an intact KGHR motif was necessary and sufficient for inhibition of TSP1 binding, as established by comparing the activities of GLKGHR, to which TSP1 binds specifically, or GLAGHR, which is not bound by TSP1 (Fig. 4A). Only GLKGHR inhibited binding of TSP1 to fibrils effectively (Fig. 4F). These observations of TSP1 binding were substantiated by quantitative image analysis of fibrils from multiple experiments (Fig. 4G).

TSP1 associates with fibrillar collagens and LOX in human dermal fibroblasts

The above in vitro data and the strong binding of TSP1 to rProCOL1A1 indicated a complex, collagen-regulatory mechanism of action for TSP1. Considering that procollagen processing and collagen fibril assembly depend on both intracellular and extracellular events, we turned to human dermal fibroblasts (HDFs) to identify the cellular sites where TSP1, fibrillar collagens, and LOX could potentially interact. Immunoblotting of cell extracts, conditioned media (CM), and isolated ECM from 24- to 96-hour cultures of HDFs detected time-dependent abundance of TSP1 in whole-cell extracts (fig. S2A), CM, and isolated ECM (fig. S2B). Collagen I was detected in cell extracts and CM throughout the time course of the experiments and in ECM at 96 hours. The CPI fragment was present in CM but not in ECM (fig. S2, A and B). Immunofluorescence using the LOX antibody, which detects both proLOX and mLOX, showed that proLOX predominated over mLOX in cell extracts, whereas mLOX and a subfragment thereof were detected transiently in CM. No mLOX was detected in ECM (fig. S2, A and B). By indirect immunofluorescence and confocal microscopy of HDFs between 48 and 96 hours after plating, when both intracellular and extracellular TSP1 was present, a major site of colocalization of TSP1 and fibrillar collagens was in intracellular vesicular structures (Fig. 5A, as shown for 72-hour time point only). The association changed dynamically with time, increasing between 48 and 72 hours (Fig. 5B, colocalization per cell quantified by Pearson correlation). Partial, heterogeneous colocalization of extracellular TSP1 with nascent collagen meshworks or protofibrils was also detected in nonpermeabilized samples (Fig. 5, A and C). At 72 and 96 hours, cellular LOX (corresponding mostly to proLOX; fig. S2A) colocalized with TSP1 in a subset of intracellular vesicles and colocalized, in part, with fibrillar collagens in intracellular vesicles (Fig. 5, A and B). However, extracellular mLOX did not colocalize with collagen fibrils or TSP1 (Fig. 5, A and C).

Fig. 5 Localization of TSP1, fibrillar collagen, and LOX in human dermal fibroblast cultures.

(A) Dual indirect immunofluorescence staining of the indicated pairs of proteins in permeabilized or nonpermeabilized HDFs cultured for 72 hours. The regions shown at higher magnification in the insets are boxed by dotted lines in the main panels. Nuclei were stained with 4′,6-diamidino-2-phenylindole (DAPI; blue). Images are representative of four independent experiments. (B) Quantification by Pearson correlation (per cell) of colocalization of the indicated pairs of proteins in permeabilized (P) HDF cultures at the indicated times. C, collagen I; T, TSP1; L, LOX. n = 4 independent experiments with at least 36 cells analyzed per condition per experiment. Data were analyzed by Kruskal-Wallis test and Dunn’s comparison. (C) Quantification by Pearson correlation (per field) of colocalization of the indicated pairs of proteins in nonpermeabilized (NP) HDF cultures at the indicated times. n = 4 independent experiments with at least six fields analyzed per condition per experiment. (D) Dual indirect immunofluorescence staining of the indicated pairs of proteins in permeabilized and nonpermeabilized HDFs after 10 days of culture. Images are representative of four independent experiments. The regions shown at higher magnification in the insets are boxed by dotted lines in the main panels. (E) Quantification by Pearson correlation (per field) of colocalization of the indicated pairs of proteins in HDFs after 10 days of culture. n = 4 independent experiments with at least six fields analyzed per condition per experiment. (F) In situ proximity ligation assay for the indicated pairs of proteins in HDFs cultured for 72 hours. Images are representative of four independent experiments. (G to L) Quantification of in situ proximity ligation signals for the indicated pairs of proteins over time in permeabilized (P) or nonpermeabilized (NP) HDFs. 2Ab, secondary antibodies only. Each data point represents the mean from one experiment, and the bars indicate the mean. n = 4 (G to J) and 3 (K to L) independent experiments. At least 36 cells were analyzed per condition for each experiment. (A), (D), and (F) show results for HDFs at passage 7 and are representative of four experiments on HDFs between passages 4 and 8. Scale bars, 50 μm (main images) and 10 μm (insets).

Because cross-linking of collagen fibrils by mLOX takes place over extended time periods (53), HDFs were also examined after 10 days of culture (Fig. 5D). At this time, we observed that the network of fibrillar collagen surrounding the cells was extensive and limited access for antibody staining, as indicated by the limited collagen staining seen with staining of nonpermeabilized cells and the detergent-resistant fibrillar network that was revealed by staining after permeabilization (Fig. 5D). In addition to amorphous and punctate deposits of TSP1, extracellular TSP1 colocalized, in part, with collagen fibrils (Fig. 5, D and E). We did not detect colocalization of TSP1 and LOX, or LOX and fibrillar collagens, in either permeabilized or nonpermeabilized cells (Fig. 5, D and E). Thus, TSP1-LOX and collagen-LOX associations in HDFs were limited to intracellular sites, where proLOX predominates, and appeared to correlate with early-stage cultures, whereas TSP1 and collagen colocalized in intracellular vesicles and extracellularly over time.

To determine definitively whether fibrillar collagen or LOX associated with TSP1 in HDFs, we carried out in situ proximity ligation assays. This method uses antibodies to detect protein-protein associations in situ within a radius of ~40 nm (54). In HDFs cultured for 48 to 96 hours, TSP1 associated intracellularly with fibrillar collagens, and this colocalization was limited to vesicular structures (Fig. 5, F and G, and fig. S3A). We found that extracellular association of TSP1 and fibrillar collagens was restricted to discrete, extracellular patches and fibril-like structures and was maintained over time, reflecting associations within the ECM (Fig. 5, F and H, and fig. S3B). The intracellular association of TSP1 with LOX was also apparent over time (Fig. 5, F and I, and fig. S3A). Extracellularly, in line with the immunofluorescence results, a minor LOX-TSP1 association was detected above background (Fig. 5, F and J) at times when mLOX was the predominant form of LOX and located extracellularly (fig. S2B). In addition, in line with the immunofluorescence data, intracellular association of LOX with fibrillar collagen in permeabilized HDFs was minor (fig. S3A), whereas no extracellular association was detected (Fig. 5, F, K, and L, and fig. S3B). No signal was detected with single antibodies alone or antibodies to vimentin and TSP1 as a pair of proteins predicted not to associate in HDFs (fig. S3C). Overall, these experiments in HDFs established that the major sites of colocalization of TSP1 with fibrillar collagen or LOX in HDFs were intracellular and that TSP1 also associated, over extended periods of time, with extracellular collagen fibrils.

The association of TSP1 and collagen I does not depend on LOX activity

Given the colocalization of TSP1 with LOX in cells and the binding of TSP1 to the KGHR motifs, which are targets of reactive lysine aldehydes in collagen cross-linking, we determined whether inhibiting LOX activity with β-aminopropionitrile (βAPN) affected the ability of TSP1 to bind to collagen. We found that βAPN treatment altered the ratios of collagen and TSP1 in the deoxycholate-soluble and -insoluble fractions of HDFs and of soluble TSP1 in CM, whereas cellular LOX abundance was unchanged (Fig. 6, A and B). As expected, βAPN resulted in fewer extracellular collagen fibrils (Fig. 6C). However, we established by both immunofluorescence (Fig. 6D) and in situ proximity ligation (Fig. 6, E and F) that TSP1 remained associated with the residual collagen fibrils. Thus, the interaction of TSP1 with fibrillar collagen did not depend on collagen cross-linking by LOX.

Fig. 6 Effects of LOX inhibition or KGHR-containing peptides on the localization of TSP1 and fibrillar collagen in human dermal fibroblast cultures.

(A) HDFs were cultured for 10 days in the absence or presence of the LOX inhibitor βAPN. The soluble and insoluble fractions of cell extracts and CM from each treatment group were analyzed by immunoblotting for collagen I α(l), TSP1, and proLOX. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) is a loading control (CTRL). Molecular markers are indicated to the left of the blots in kilodaltons. (B) Quantification of the indicated proteins from the immunoblots in (A). Protein abundances were normalized to GAPDH and expressed as a ratio versus control cells. Each bar represents the mean of three independent experiments, and the error bars indicate SEM. Mann-Whitney U tests were performed for each pair (not significant). (C) Indirect immunofluorescence images showing TSP1 and collagen I in HDFs cultured for 10 days in the absence or presence of βAPN. Nuclei were stained with DAPI (blue). Images are representative of four independent experiments. The regions shown at higher magnification in the insets are boxed by dotted lines in the main panels. (D) Quantitation by Pearson correlation (per field) of colocalization of TSP1 and collagen I under the indicated conditions. n = 4 independent experiments with at least 36 cells analyzed per condition per experiment. (E) In situ proximity ligation assays in permeabilized HDFs after 10 days culture in the absence or presence of βAPN. Images are representative of four independent experiments. (F) Quantitation of proximity ligation signals under the indicated conditions. n = 4 independent experiments with at least 36 cells analyzed per condition per experiment. (G) Immunofluorescence showing collagen and TSP1 in HDFs cultured and treated for 48 hours with the indicated peptides, then fixed, and stained by indirect immunofluorescence (IF) or in situ proximity ligation (Prox). In the top row, the regions shown at higher magnification in the insets are boxed by dotted lines in the main panels. Each row is representative of four independent experiments. (H) Quantification by Pearson correlation of immunofluorescence colocalizations as in (G). n = 4 independent experiments with at least 36 cells analyzed per condition per experiment. (I) Quantification of proximity ligation signals for conditions as in (G). n = 4 independent experiments with at least 36 cells analyzed per condition per experiment. Data in (H) and (I) were analyzed by Kruskal-Wallis test and pairwise tests against HAc with Holm’s correction. Scale bars, 50 μm (main images) and 10 μm (insets).

Peptide II-52 inhibits the association between collagen and TSP1 in the ECM of fibroblasts

To investigate the relevance of KGHR-dependent binding of TSP1 to fibrillar collagens in vitro to the associations of TSP1 with collagen in a cellular context, we treated HDFs for 48 or 72 hours with selected collagen triple-helical peptides at concentrations that effectively competed with collagen I for binding to TSP1 in vitro. By immunofluorescence of nonpermeabilized, control HDF cultures, we observed that nascent networks of collagen fibrils started to form after 48 hours (Fig. 6G), as expected (Fig. 5A). The colocalization of fibrils and TSP1 was retained in HDFs treated with peptide II-2, or the mutants GLAGHR or GLKGAR, none of which bind to TSP1 (Fig. 6, G and H). However, we found that collagen fibrils were more fragmentary and aggregated after treatment with either peptide II-52 or GLKGHR, and TSP1-collagen colocalization was decreased (Fig. 6, G and H). Peptide II-52 was confirmed to decrease the extracellular association of TSP1 and fibrillar collagens by in situ proximity ligation (Fig. 6, G and I), whereas the short GLKGHR peptide had a minor effect, possibly explained by its lower TSP1-binding activity. Thus, peptides that include an intact KGHR motif compete with cell-derived, collagenous ECM binding to TSP1, and reduced TSP1-binding affects the organization of the collagen ECM.

Peptide II-52 increases myofibroblast differentiation through a TGF-β receptor 1–dependent mechanism

Because KGHR-containing peptides perturbed collagen-TSP1 interactions in the ECM of fibroblasts and cell-ECM interactions are important regulators of myofibroblast differentiation (1), we examined possible physiological effects of KGHR-containing peptides on fibroblast differentiation. We first established conditions for the induction of two myofibroblast markers—fibronectin containing the alternatively spliced domain EDA (EDA-FN) and α-smooth muscle actin (αSMA)—by TGF-β1, a known inducer of myofibroblast differentiation in HDFs and measured the abundance of TSP1, LOX, and β-catenin after TGF-β1 treatment (fig. S4, A to G). To determine whether TSP1-binding peptides influence the differentiation of myofibroblasts, we measured the abundance of αSMA in cells treated for 96 hours with the TSP1-binding peptide II-52 or the non–TSP1-binding peptide II-8. II-52 specifically increased the number of αSMA-positive cells, although to a lesser extent than did TGF-β1, whereas peptide II-8 had no effect on the number of αSMA-positive cells (Fig. 7, A and B). Neither peptide altered HDF proliferation over time relative to vehicle treatment (Fig. 7C). The KGHR motif was important for the induction of αSMA by II-52 because only the wild-type GLKGHR peptide increased the percentage of αSMA-positive cells, whereas the peptides bearing the alanine substitutions in the motif did not (Fig. 7, D and E). The GLKGHR peptide was less efficient than peptide II-52 at inducing αSMA, likely because of the reduced TSP1-binding activity of GLKGHR compared to II-52. None of the peptides altered HDF cell numbers at 96 hours (Fig. 7F).

Fig. 7 KGHR-containing, TSP1-binding collagen triple-helical peptides increase myofibroblast differentiation through a TGF-βR1–dependent mechanism.

(A) Merged immunofluorescence images of HDFs cultured for 96 hours in the presence of HAc, TGF-β1, collagen peptide II-8, or collagen peptide II-52 and stained for αSMA (green) and nuclei (DAPI; blue). Images are representative of three independent experiments. (B) Quantification of the percentage of αSMA-positive cells under the indicated conditions from three independent experiments. (C) Effects of the various treatments on cell numbers over time; n = 3 independent experiments. (D) Merged immunofluorescence images showing αSMA (green) and DNA (blue) in HDFs cultured for 96 hours in the presence of HAc, TGF-β1, or the indicated peptides. Images are representative of three independent experiments. (E) Quantification of αSMA-positive cells per field after treatment with the indicated peptides. n = 3 independent experiments. (F) Effects of the indicated peptides on cell numbers at 96 hours. n = 3 independent experiments. (G) Merged immunofluorescence images showing αSMA (green) and DNA (blue) in HDFs cultured for 96 hours in the presence of the inhibitor of TGF-βR1 kinase activity SB-431542 (SB) or the β-catenin signaling inhibitor PNU-74654 (PNU). Images are representative of three independent experiments. (H) Quantification of αSMA-positive cells after 96 hours in the presence of the indicated inhibitors. n = 3 independent experiments. (I) Quantification of bioactive and total (determined after acid activation) active TGF-β1 in media from HDFs cultured under the indicated conditions. UT, untreated. n = 3 independent experiments with duplicate samples per condition per experiment. In all graphs, each data point indicates the mean, and error bars indicate the SEM. In (B), (C), (E), (F), and (H), at least 30 cells were scored per condition per experiment. Data were analyzed by Dunnett’s multiple comparison test (one-way ANOVA) against HAc (B) or by Kruskal-Wallis test and pairwise tests against HAc (E and F) or against dimethyl sulfoxide (DMSO) (H). Scale bars, 50 μm.

We also investigated whether the signaling process by which II-52 increases the number of αSMA-positive cells is related to that of TGF-β signaling pathways, which are the most widely studied inducers of myofibroblast differentiation. We treated HDFs with an inhibitor of TGF-β receptor 1 (TGF-βR1) kinase activity, SB-431542 (55), or an inhibitor of β-catenin signaling, PNU-74654 (56), because β-catenin signaling can also promote αSMA expression. First, we established effective concentrations of these compounds to inhibit the TGF-β1–induced increases in EDA-FN and αSMA protein abundance in HDFs (fig. S4, A and B). Neither inhibitor affected the abundance of TSP1, but SB-431542 abolished the TGF-β1–induced increase in proLOX and mLOX abundance (fig. S4, A, C, and D to G). SB-431542 significantly reduced the induction of αSMA by II-52 and by TGF-β1 (Fig. 7, G and H), whereas PNU-74654 had a small effect on the percentage of αSMA-positive cells only for TGF-β1–treated cells (Fig. 7, G and H). Thus, promotion of myofibroblast differentiation by II-52 is mediated by TGF-βR1. To investigate whether this process is related to the known role of extracellular soluble TSP1 as an activator of the latent TGF-β1 complex (28, 40), we quantified bioactive and total (that is, after exogenous acid activation) TGF-β1 in media from HDF cultures treated with peptide II-8 or II-52 for 96 hours. Bioactive TGF-β1 was negligible under all conditions, and the addition of acid to the medium to activate TGF-β1 did not indicate any differences in the abundance of total TGF-β1 between the conditions (Fig. 7I). Thus, the activity of II-52 in perturbing extracellular TSP1-collagen interactions also promotes myofibroblastic differentiation by a mechanism that does not relate to altered abundance of extracellular bioactive TGF-β1.

DISCUSSION

The organization and cross-linking of fibrillar collagens are fundamental to tissue integrity. In many tissue settings, TSPs coincide with collagen-rich ECM, which suggests associations that might be highly relevant to normal tissue physiology and fibrotic pathologies. Here, we elucidate that TSP1, the prototypic TSP family member, undergoes both intracellular and extracellular molecular interactions with fibrillar collagens. We show the latter to be important for control of myofibroblast phenotype. In vivo, in the dermis of Thbs1−/− mice, we observed altered cross-linking of fibrillar collagens, reduced abundance of proLOX and mLOX, and an increased ratio of mLOX to proLOX. We established in vitro that TSP1 bound to proLOX, inhibited the cleavage of proLOX to mLOX by BMP-1, and bound to multiple sites on procollagen I molecules (the C-propeptide domain and the KGHR motifs that are conserved in the triple-helical domains of collagens II and III). In cultured fibroblasts, TSP1 colocalized with collagen I and, to a lesser extent, LOX in intracellular vesicles and also associated with extracellular collagen fibrils by binding the KGHR motif. The latter association promoted collagenous ECM organization and likely inhibits myofibroblastic differentiation. Treatment of cells with exogenous, KGHR-containing peptides promoted myofibroblast differentiation through a TGF-βR1–dependent process without altering the extracellular abundance of bioactive TGF-β1. These data establish that TSP1 coordinates posttranslational collagen I processing and assembly through multiple mechanisms that act at different stages of collagen fibril production, both intracellular and extracellular.

The absence of TSP family members leads to disorganization of collagen fibrils in tissues of mice (4, 17, 2527). By use of binding assays in vitro and demonstrations of protein localizations and colocalizations in cultured cells, we were able to identify specific interactions of TSP1 and then ascribe cellular contexts to these interactions. With regard to proLOX, we identified that TSP1 inhibited BMP-1–mediated cleavage of proLOX in vitro and associated in HDFs only with intracellular LOX. Thus, BMP-1 activity on proLOX may rise in the TSP1-null context, resulting in altered spatiotemporal aspects of proLOX production and thereby affecting collagen cross-linking. Other studies have reported cardioprotective actions of intracellular TSP4 during ER stress within the ER of cardiomyocytes, but the status of LOX or collagen ECM organization was not examined (57).

We also investigated whether fibrillar collagens are both binding partners and substrates of LOX. We demonstrated, in vitro, specific binding of proLOX to CPI and rProCOL1A1 but not to CPIII, mature collagens, or KGHR-containing collagen triple-helical peptides. These data suggest an interaction of proLOX with the CPI region of procollagen I before secretion. Consistent with this, collagen and LOX colocalized in HDFs; however, in the absence of antibodies specific to proLOX or mLOX, we could not determine whether this interaction is specific to proLOX. Free, extracellular CPI is present physiologically, has various functions (58, 59), and was detected in early-stage HDF cultures (fig. S2); however, mLOX only appeared in CM after 48 hours. Our studies of nonpermeabilized HDFs at 10 days did not detect any collagen-LOX association above background. This is in line with the transient, dynamic nature of LOX enzymatic interactions with collagen telopeptide lysine residues.

With regard to binding of TSP1 to fibrillar collagens, we show that TSP1 exhibits dual direct binding to fibrillar intracellular procollagen I (containing the CPI domain and KGHR sites) and to mature, extracellular collagens I, II, and III (via KGHR sites). Given the increased collagen cross-linking in the Thbs1−/− context, it is relevant to consider that both mechanisms may limit the rate of homeostatic collagen cross-linking. TSP1 binding to the CPI domain could shield intracellular procollagen molecules from BMP-1 cleavage until secretion, or from active LOX, or from both. TSP1 binding to the KGHR motif could protect KGHR motifs from cross-linking by lysine aldehyde attack. We established that TSP1 bound to synthetic KGHR-containing peptides and to native cross-linked collagen I fibrils and that KGHR-containing peptides inhibited TSP1 binding to native fibrils. Hence, the cross-linking of lysine residues does not appear to prevent TSP1 binding at this site. The enhanced TSP1 binding to collagen I fibrils relative to pepsin-digested collagen I suggests that the spatial presentation of KGHR sites may influence extracellular TSP1 binding, perhaps through increased avidity.

Because TSP1-binding collagen triple-helical peptides disrupted TSP1 binding to native collagen I fibrils (Fig. 4), we were able to explore the functional significance of collagen-TSP1 associations in HDFs. We focused on the early time points at which we had detected collagen-TSP1 or collagen-LOX associations (as in Fig. 5). KGHR-containing peptides markedly perturbed nascent collagen fibrils, resulting in aggregated collagen “patches,” greater variability in fibril size and shape, and reduced extracellular association of TSP1 and collagen I. KGHR-containing peptides also had specific effects on myofibroblast differentiation (Fig. 7). The induction of αSMA by disruption of the TSP1-collagen interaction (using the TSP1-binding peptide II-52) required TGF-βR1 kinase activity, as established using specific inhibitors of either TGF-βR1 or β-catenin signaling. Although extracellular TSP1 is known to activate latent TGF-β1 (28), the application of II-52 did not increase the amount of extracellular bioactive TGF-β1 or total active TGF-β1. These results are not definitive, and the mechanism will require further investigation to determine whether the TSP1-collagen interaction inhibits signaling through TGF-βR1, for example, by controlling the abundance or activity of TGF-βR1.

Known molecular partners of fibrillar collagens include FACITs (fibril-associated collagens with interrupted triple helices) (60), which align on the surface of the growing fiber and may restrict further recruitment of fibrillar collagen molecule, and the SLRPs. Different classes of SLRPs bind to different sites on the surface of collagen fibrils to modulate protofibril nucleation and the subsequent axial and lateral growth of fibrils (61). In common with TSP knockout mice, mice null for individual SLRP family members are viable but typically show tissue-specific alterations to collagen fibril structures and sizes. For example, mice lacking the SLRP fibromodulin (Fmod) have disorganized and smaller collagen fibrils in tendons and increased C-telopeptide cross-linking of collagen I (6264). Fmod is proposed to facilitate extracellular collagen cross-linking because it binds to KGHR motifs in fibrillar collagens and binds and activates LOX without affecting LOX maturation (65, 66). TSP1 and Fmod are the only ECM proteins identified to date that bind to the KGHR motif of fibrillar collagens, although binding of PEDF (pigment epithelium–derived factor) to collagen I depends on the arginine residue in this motif (67). Thus, TSP1 and Fmod bind the triple helix of fibrillar collagens at a site distinct from the motifs recognized by the major collagen receptors and other collagen-binding proteins (6870).

Overall, the collective action of many ECM proteins on fibrillar collagens emphasizes the evolution of a complex network of mechanisms to control collagen fibril assembly and the activity of LOX. The central conceptual advance of this study is the identification of an additional form of molecular control of fibrillar collagens. Our data reveal actions of TSP1 on fibrillar collagen, exemplified here by collagen I, at several steps during the molecular processing and assembly of collagen fibrils. Detailed studies will be needed to determine whether these activities influence procollagen trafficking, mask KGHR cross-linking sites, or control the spatiotemporal positioning of proLOX or its access to collagen molecules before secretion. The mechanism of action of TSP1 is unique in regulation of collagen homeostasis because it acts on both intracellular procollagen and proLOX, as well as on extracellular fibrillar collagen.

MATERIALS AND METHODS

Materials and cells

Normal HDFs from foreskins of healthy juveniles (C-12300, PromoCell) were cultured in fibroblast growth medium (C-23010, PromoCell) with l-ascorbic acid (50 μg/ml) and used for experiments between passages 3 and 8. Primary antibodies were used for detection of αSMA, collagen I, EDA-FN, TSP1, vimentin, collagen I, CPI (LF41) (71), and LOX. Full details for each antibody are given in table S1. Secondary antibodies were as follows: horseradish peroxidase (HRP)–conjugated antibody recognizing mouse immunoglobulin G (IgG) or rabbit IgG; fluorescein isothiocyanate (FITC)–conjugated antibody for mouse IgG; and tetramethyl rhodamine isothiocyanate (TRITC)–conjugated antibody for rabbit IgG (table S2). Chemicals used and supplier information are listed in table S3. The following inhibitors were used: βAPN, SB431542, and PNU-74654 (full details are in table S4). Purified and recombinant proteins used are listed in table S5. Buffer compositions are given in table S6.

Sample preparation and collagen extraction from mouse skin

Collagen was extracted from the skin of 8-week-old wild-type C57BL/6 male mice (n = 4; JAX 000664, the Jackson Laboratory) and B6.129S2-Thbs1tm1Hyn/J male mice (n = 4; JAX 006141, the Jackson Laboratory). Mutant mice were backcrossed to C57BL/6J for eight generations before establishment of the stock at the Jackson Laboratory (19). Skin (16 cm2) was dissected from each mouse and shaved, any underlying adipose was removed, and samples were cut into 3-mm-wide strips. Two grams of skin per mouse was incubated with phosphate-buffered saline (PBS) containing protease inhibitor cocktail (2 mM N-ethylmaleimide, 10 mM leupeptin, and 20 mM pepstatin A) under rotation at 4°C overnight, and extracts were collected as supernatants after centrifugation at 14,000g for 30 min at 4°C. Pellets were resuspended in 3 ml of 0.5 M HAc containing protease inhibitor cocktail and incubated under rotation for 24 hours at 4°C. Supernatants were collected after centrifugation at 14,000g for 30 min at 4°C to obtain the HAc-soluble fraction, and Trizma base was added to bring the pH to 7, followed by hydroxyproline quantitation. Aliquots of acid-soluble collagens from each mouse (corresponding to 250 μmol hydroxyproline) were boiled in reducing SDS-PAGE sample buffer before analysis by SDS-PAGE.

Hydroxyproline quantitation

Acid-extracted collagens were brought to a concentration of 600 μg/ml and subjected to hydrolysis with 6 M HCl for 20 hours at 95°C. Samples were centrifuged at 14,000g for 10min, and supernatants were diluted in deionized H2O to a final concentration of 4 M HCl. Hydroxyproline assay kit was used according to the manufacturer’s procedures (QuickZyme Biosciences).

ECM proteins and collagen peptides

The following proteins were used (details, catalog numbers, and suppliers of proteins are listed in table S5): recombinant human TGF-β1; purified human TSP1; recombinant human proLOX; rProCOL1A1; denatured bovine collagen I; pepsin-digested collagens I, II, and III; native collagen I fibrils; and CPI and CPIII homotrimers (4, 72). Collagens were dissolved overnight in 0.5 M HAc to final concentrations of 1 or 5 mg/ml. Collagen Toolkit II and derivative peptides were synthesized as C-terminal amides on TentaGel R RAM resin using Fmoc (9-fluorenyl methoxycarbonyl)/tBu (tert-butyl) method (44) and using CEM Liberty or Liberty Blue automated microwave peptide synthesizers. Each guest sequence (27 residues for Toolkits or shorter sequences containing KGHR and its derivatives, as indicated) is placed between GPC(GPP)5– and –(GPP)5GPC host sequences to stabilize the triple-helical conformation. GPP10, a negative control peptide (GPC-[GPP]10GPC-NH2), represents the combined host sequence. All peptides were dissolved in 10 mM HAc at a final concentration of 5 mg/ml.

Solid-phase binding assays

Collagen triple-helical peptides at 10 μg/ml in 10 mM HAc were coated onto wells of 96-well plates (Immulon 2HB, 3455, Thermo Fisher Scientific) overnight at 4°C. Collagen fibrils, pepsin-digested collagens, or denatured collagen I was adsorbed at a saturating concentration of 10 μg/ml in 10 mM HAc. In other experiments, wells were coated overnight at 4°C with TSP1 diluted in tris-buffered saline (TBS) containing 2 mM CaCl2 or with CPI or CPIII diluted in TBS containing 0.5 mM CaCl2, each to a final concentration of 24 nM or as indicated in individual figure panels. rProCOL1A1 was diluted in 10 mM HAc to a final concentration of 24 nM. All of the following steps were at room temperature (RT). Wells were blocked with BSA Fraction V (50 μg/ml). Test proteins in solution and primary or secondary antibodies were each diluted in incubation buffer (IB; table S6), containing either 2 mM EDTA or 2 mM CaCl2 and 15 µM ZnSO4. Antibodies were diluted as given in tables S1 and S2. All steps were performed for 1 hour, and each was followed by three washes in IB. Final concentrations of proteins in solution were 8 nM TSP1 or 24 nM rProCOL1A1, CPI, CPIII, or proLOX . For colorimetric detection, 100 μl of a 1:1 mixture of 3,3′,5,5′-tetramethylbenzidine and 0.02% H2O2 in citric acid buffer were added. Reactions were stopped with 100 μl of 2 M H2SO4, and absorbance was measured at 450 nm in a 96-well plate reader (M2/SpectraMax, Molecular Devices). As a negative control to confirm the specificity of each primary antibody, 8 nM BSA was added in IB instead of the test protein. For competition assays, TSP1 was preincubated with increasing amounts of Toolkit peptides II-8 and II-52 or selected derivatives (II-52d and II-52dK3A) for 1 hour before adding to wells coated with collagen I fibrils, and assays were developed as described above.

ProLOX cleavage

Recombinant proLOX (50 nM) was incubated without or with 0.18 nM recombinant human BMP-1 in the presence of either 225 nM BSA or TSP1 at concentrations of 25, 50, 110, or 225 nM at 37°C for 15 min. Reactions were ended by addition of an equal volume of SDS-PAGE sample buffer, and samples were resolved on 10% polyacrylamide gels under reducing conditions and immunoblotted for LOX or TSP1. ProLOX cleavage was quantified by measuring the pixel intensities of proLOX and mLOX bands using ImageJ software. Percentage of LOX cleavage was then calculated as follows: % of LOX cleavage = [(mLOX/proLOX)lane/(mLOX/proLOX) from BMP1 + BSA condition] × 100.

Fluorescence microscopy

For binding of TSP1 to fibrillar collagen, glass coverslips were coated with 60 μl of native collagen I fibrils (2 μg/ml in 0.01 M HAc) overnight at 4°C. In some experiments, TSP1 (8 nM in IB) was incubated with selected collagen triple-helical peptides for 1 hour at RT. Peptides used were as follows: II-8, II-52, II-52d (GLKGHR peptide), or II-52dK3A (GLAGHR peptide); each at 45 μM in 0.01 M HAc or the equivalent volume of 0.01 M HAc as a negative control. All ligands and antibodies were diluted in IB containing 2 mM CaCl2 and 15 μM ZnSO4. In parallel, collagen-coated coverslips were blocked with 5% BSA and then TSP1, without or with peptide, added for 1 hour, followed by anti-TSP1 for 1 hour, and then FITC-conjugated antibody to mouse IgG for 1 hour. All steps were followed by three washes in IB. Coverslips were washed and mounted in Vectashield (Vector Laboratories) and examined under a Leica DMI6000 inverted epifluorescence microscope with a HCX PL APO 100× 1.40–numerical aperture (NA) oil objective. XY images were captured with a DFC365FX Leica monochrome charge-coupled device camera run by Leica Application Suite X software (v3.0.2). TSP1 bound to collagen fibrils was quantified with Volocity 6.3 software by (i) selecting areas in phase-contrast images that corresponded to collagen fibrils by finding objects with intensity values between 0 (lower) and 60 (upper) and by (ii) finding objects bigger than 4 μm2 under the green fluorescence channel within the above fibril areas. The total fluorescence intensity in each area was then normalized over the pixel counts for the collagen fibril area.

For localization of collagen I, LOX, αSMA, and TSP1 in HDFs (3 × 104 cells per well) were plated onto 13-mm glass coverslips for 48, 72, or 96 hours at 37°C. In some experiments, cells were treated with 15 μM Collagen Toolkit II peptides for 48, 72, or 96 hours. In other experiments, HDFs were treated with 10 μM SB-431542, 10 μM PNU-74654, or the equivalent volume of DMSO for 30 min before addition of Collagen Toolkit II peptides or TGF-β1 (2 ng/ml) for 96 hours or treated with 2 mM βAPN for 10 days. For colocalization in nonpermeabilized cells, coverslips were fixed in 4% paraformaldehyde (PFA). For permeabilization, cells were either treated with methanol/acetone (1:1) or fixed in 2% PFA and permeabilized in 0.5% Triton X-100 in PBS. All of the following steps were carried out at RT in a humidified chamber, and each step was followed by three washes in PBS. Coverslips were blocked with 1% BSA for 30 min, and then cells were stained with the appropriate primary antibody(s) diluted in 2% BSA for 1 hour and 30 min, washed in PBS, and incubated for 1 hour with FITC-conjugated antibody to mouse IgG and then with TRITC-conjugated antibody to rabbit IgG, each diluted in PBS containing 5% BSA. Coverslips were washed and mounted as above. Cells were examined under an inverted Leica SP5-AOBS confocal laser-scanning microscope, with an HCX PL APO lambda blue 63× 1.4-NA oil objective. XY images were captured as Z stacks with a 0.25-μm Z step size, with photomultiplier tube detectors with a photocathode made of gallium-arsenide-phosphide (Leica) for collecting light emission. Images were captured with Leica Application Suite AF software (v2.7.3.9723). For colocalization studies, Pearson correlation was measured in Volocity 6.3 software by the method of Costes to set automatic thresholds (73). At least 36 cells were analyzed per condition per experiment. For nonpermeabilized cells, Pearson correlation was measured per field; a minimum of six fields was taken for each condition in each independent experiment. For permeabilized cells (48- to 96-hour cultures), the Pearson correlation was measured for each cell. In 10-day cultures, nuclei could not be counted accurately due to dense ECM and overlapping of cells; therefore, the Pearson correlation is reported per field.

For in situ proximity ligation, HDFs were plated onto 13-mm glass coverslips, fixed, and incubated with primary antibodies, as described above. In situ proximity ligation assays were performed using Duolink (Sigma-Aldrich) according to the manufacturer’s instructions. Briefly, secondary antibodies coupled to complementary DNA probes were added for 1 hour at 37°C, and then fluorescent detection was performed with Detection Reagents Green. Coverslips were examined under a Leica wide-field microscope (Leica DMI4000 B) and an HCX PL APO 63× 1.40-NA oil objective. XY images were captured with a DFC310 FX Leica digital camera run by Leica Application Suite X software (v4.5). Images were analyzed in ImageJ (Fiji). The intensity range was set between 37 and 255 intensity values. The total area of particles per field was measured, and depending on the experimental design, the number of nuclei per field was counted.

Preparation of cell lysates and supernatants

HDFs (4 × 105 cells) were cultured for 24, 48, 72, or 96 hours at 37°C. In some experiments, cells were treated with 10 μM SB-431542 or PNU-74654 or DMSO for 30 min before addition of TGF-β1 (2 ng/ml) for 96 hours. Cells were lysed in SDS-PAGE sample buffer to obtain total cell extract or with 2% deoxycholate buffer. CM was harvested and mixed 1:1 with SDS-PAGE sample buffer. In parallel, heparin-binding proteins were collected from CM by incubation with 25 μl of Affi-Gel heparin beads for 90 min with rotation at 4°C, followed by washing in PBS and resuspension of beads in SDS-PAGE sample buffer. For measurements of TGF-β1 activity, HDFs were plated as for immunofluorescence microscopy and CM harvested after 96 hours of culture, and centrifuged to remove cell debris, and 100 μl of aliquots were analyzed for active TGF-β1 by TGF-β1 Enzyme Immunoassay (Enzo, ADI-900-155), without or with acid activation of TGF-β1, according to manufacturer’s procedures.

SDS-PAGE and immunoblotting

Proteins were resolved on 10 or 7.5% polyacrylamide SDS-polyacrylamide gels under reducing conditions. Gels were stained with GelCode Blue Stain reagent. For immunoblotting, proteins were transferred to polyvinylidene difluoride membrane (ISEQ00010, Immobilon-P, Millipore) and blocked in immunoblot blocking buffer (BB) overnight at 4°C. Membranes were probed with primary antibodies diluted in BB for 90 min at RT under agitation. After three washes in BB, membranes were probed with appropriate HRP-conjugated secondary antibodies for 1 hour at RT, washed, and incubated with Amersham ECL, and signals were detected on x-ray film (28906836, GE Healthcare). Protein levels from Coomassie-stained gels and immunoblots were quantified with NIH ImageJ software (Fiji) by boxing each lane and generating histogram plots showing peak areas corresponding to each band along the lane.

Multiple sequence alignment

Collagen sequences were obtained from National Center for Biotechnology Information protein division. Multiple sequence alignments were prepared in MUSCLE 3.8 (74) at default parameters through the resources of EMBL/EBI (www.ebi.ac.uk/Tools/msa) and are presented in BoxShade 3.21 (www.ch.embnet.org/software/BOX_form.html).

Statistical analysis

All experiments were carried out at least three times independently, unless stated otherwise in the figure legend. Statistical tests were made in GraphPad Prism v5.01. Data were analyzed with D’Agostino-Pearson omnibus and Shapiro-Wilk tests to check for normal distribution. More than two groups of data were analyzed by one-way ANOVA with Bonferroni’s multiple comparison test. Where distribution was not normal, two groups of data were analyzed by nonparametric Mann-Whitney U test. More than two groups of data were analyzed by nonparametric Kruskal-Wallis test with Holm’s correction. In the figures, ***P ≤ 0.0005, **P ≤ 0.005, and *P ≤ 0.05.

Ethical standards

The experiments comply with the current laws of the country in which they were performed.

SUPPLEMENTARY MATERIALS

www.sciencesignaling.org/cgi/content/full/11/532/eaar2566/DC1

Fig. S1. Identification of specific TSP1-binding peptides within the triple-helical region of collagen III.

Fig. S2. Detection of collagen I, TSP1, FN, LOX and collagen I C-propeptide in cell fractions, CM, and ECM isolated from HDFs.

Fig. S3. Detection of the association of TSP1 with collagens I and II and LOX in HDFs by in situ proximity ligation.

Fig. S4. TGF-β1–mediated induction of αSMA in HDFs.

Table S1. Primary antibodies and the dilutions used in this study.

Table S2. Secondary antibodies and the dilutions used in this study.

Table S3. Chemicals used in this study.

Table S4. Chemical inhibitors used in this study.

Table S5. Proteins used in this study.

Table S6. Buffers used in this study.

REFERENCES AND NOTES

Acknowledgments: We thank R. Stone for the assistance with setting up Collagen Toolkit assays, D. Bihan for the peptide synthesis and participation in initial experiments, and the Wolfson Bioimaging Facility at University of Bristol for the confocal microscopy facilities. We thank L. W. Fisher (National Institute of Dental and Craniofacial Research, NIH) for the antibody LF41. We thank J. Lawler for discussions and J. Rougier for the advice on nonparametric statistical analyses. Funding: We acknowledge the support of Medical Research Council (MRC) UK grant K018043 to J.C.A. and British Heart Foundation program grants RG/09/003/27122 and RG/15/4/31268 and Wellcome Trust Biomedical Resource grant 094470/Z/10/Z to R.W.F. We acknowledge the MRC and the Wolfson Foundation for establishing the Wolfson Bioimaging Facility at the University of Bristol. Author contributions: J.C.A. and R.W.F. designed the study and supervised the research. D.J.S.H., S.R., R.W.F., and J.C.A. gave intellectual input. S.R. carried out the experiments with input and/or contribution of specific protein or peptide reagents by N.P., A.M.B., D.J.S.H., R.W.F., and J.C.A. Data analysis was carried out by S.R., D.J.S.H., R.W.F., and J.C.A. The manuscript was drafted by S.R., R.W.F., and J.C.A. All authors contributed to and approved the final version of the manuscript. Competing interests: The authors declare that they have no competing interests. Data and materials availability: All data required to evaluate the conclusions in the paper are present in the paper or the Supplementary Materials.
View Abstract

Navigate This Article