Versican binds collagen via its G3 domain and regulates the organization and mechanics of collagenous matrices

Abstract

Type I collagen is the most abundant structural protein in the body and, with other fibrillar collagens, forms the fibrous network of the extracellular matrix. Another group of extracellular matrix polymers, the glycosaminoglycans, and glycosaminoglycan-modified proteoglycans, play important roles in regulating collagen behaviors and contribute to the compositional, structural, and mechanical complexity of the extracellular matrix. While the binding between collagen and small leucine-rich proteoglycans has been studied in detail, the interactions between collagen and the large bottlebrush proteoglycan versican are not well understood. Here, we report that versican binds collagen directly and regulates collagen structure and mechanics. Versican colocalizes with collagen fibers in vivo and binds to collagen via its C-terminal G3 domain (a non-GAG-modified domain present in all known versican isoforms) in vitro; it promotes the deposition of a highly aligned collagen-rich matrix by fibroblasts. Versican also shows an unexpected effect on the rheology of collagen gels in vitro, causing decreased stiffness and attenuated shear strain stiffening, and the cleavage of versican in the liver results in reduced tissue compression stiffening. Thus, versican is an important collagen-binding partner and plays a role in modulating collagen organization and mechanics.

Keywords: collagen Ligands Collection; glycosaminoglycan; hyaluronic acid; liver rheology; versikine.

Figure 1

Versican co-localizes with collagen fibers in mouse extrahepatic bile duct and collagen gels.A, schematic showing the different isoforms of versican. The antibody used for IEM binds to the β GAG domain in V0 and V1 (shown in green). B and C, representative IEM images showing localization of versican (black dots, indicated by teal arrows) on collagen fibers (blue arrows) in postnatal day 3 (B) and adult (C) mouse bile ducts (N = 2). D, quantification of the number of gold particles, indicating versican, per total fiber area (manually selected and measured in ImageJ); p = 4.1e-14. E and F, representative IEM images showing the localization of versican in plugs of co-gelled collagen (blue arrows) and versican (black dots indicated by teal arrows) (E) or collagen alone (F). Scale bar = 200 nm. Data represent mean ± SD and were analyzed by Mann–Whitney U test; ∗∗∗∗p < 0.0001.

Figure 2

Versican interacts with collagen via its G3 domain.A, Plates were coated with versican (Ver, red) or the V3 isoform (V3, blue) at 0.05, 0.1, 0.25, 0.5, and 1 μg/ml. The absorbance of collagen, added at 2.5 μg/ml and detected via a biotin-conjugated antibody, was measured colorimetrically. B, plates were coated with 0.25 μg/ml versican (red), V3 (blue) or versican after chondroitinase ABC digestion (Ver-ChABC, orange). The binding of increasing concentrations of collagen (0.1, 0.5, 1.0, 2.5, and 5 μg/ml) was assayed. C, plates were coated with 0.25 μg/ml recombinant G1 (pink) or G3 (teal) and the binding of increasing concentrations of collagen (0.1, 0.5, 1.0, 2.5, and 5 μg/ml) was assayed. D, plates were coated with 0.25 μg/ml versican (red), V3 (blue), decorin (Dec, cyan), lumican (Lum, orange dashed line) or aggrecan (Agg, purple), and the binding of increasing concentrations of collagen (0.1, 0.5, 1.0, 2.5, and 5 μg/ml) was measured. E, plates were coated with 0.25 μg/ml V3 to which was added collagen (1 μg/ml) mixed with increasing concentrations of HA (0.1, 0.5, 1, 5, and 10 ng/ml). F, plates were coated with 0.25 μg/ml V3 to which was added HA (10 ng/ml) mixed with increasing concentrations of collagen (0.1, 0.5, 1.0, 2.5, and 5 μg/ml). Three independent experiments were carried out for each panel in the figure and all data are shown. AU represents Absorbance Unit.

Figure 3

Potential versican binding sites on collagen identified using the CLC II.A and B, Binding between CLC II peptides and 10 μg/ml recombinant V3 (A) or G3 (B) was tested using a solid-phase binding assay. Empty wells in the CLC-coated plate contained full-length collagen as a positive control and data were normalized to the positive control (see dashed red lines). Three independent experiments were carried out and all data are shown. C, binding motifs (in color and shaded) identified by the alignment of versican-binding CLC peptides (where O represents hydroxyproline). Note that peptide II-5 has two instances of the potential motif.

Figure 4

Versican and its V3 isoform upregulated the deposition of collagen-rich matrices by fibroblasts and led to increased fiber alignment.A–C, representative SHG imaging of the fibroblast-deposited matrices: control plate, with vitronectin coating (A), versican-coated plate (B), V3-coated plate (C). D, quantification of the intensity of the SHG signals, normalized to the value for the control group (p < 1.0e-15 for Control vs. Ver and Control vs. V3; p = 0.0313 for Ver vs. V3). E, the distribution of collagen fiber orientation was analyzed using OrientationJ. The data were normalized to the dominant angle of each SHG image. For each condition, SD is represented by dotted lines. The statistical significance of differences between conditions is shown in Table S3. F–H, representative confocal imaging of immunostaining for fibronectin in fibroblast-deposited matrices (red - fibronectin, blue - DAPI): vitronectin coated plate as a control (F), versican-coated plate (G), V3-coated plate (H). I, quantification of the intensity of fibronectin staining (p = 0.0426 for Control vs. V3). J, quantification of cell numbers after 7 days in culture. Four independent experiments were carried out with two technical repeats for each coating condition per experiment and 9 SHG images were obtained and analyzed for each technical repeat. Scale bar = 100 μm. Data represent mean ± SD; D was analyzed using the Kruskal-Wallis test with Dunn’s multiple comparisons, I and J were analyzed using one-way ANOVA with Tukey’s multiple comparisons, E was analyzed using two-way ANOVA with repeated measurements with Tukey’s multiple comparisons; ∗p < 0.05 and ∗∗∗∗p < 0.0001.

Figure 5

Different matrix proteoglycans have distinct effects on the mechanics of collagen networks.A, gelation times for collagen-proteoglycan co-gels, with rheological measurements taken during gelation. Col: 2.5 mg/ml pure collagen gel; Col-Ver: 2.5 mg/ml collagen with 0.167 mg/ml versican; Col-V3: 2.5 mg/ml collagen gel with 0.167 mg/ml V3; Col-Agg: 2.5 mg/ml collagen gel with 0.167 mg/ml aggrecan; Col-Dec: 2.5 mg/ml collagen gel with 0.167 mg/ml decorin (p = 2.4e-8 for Col vs. Col-Ver; p = 1.6e-9 for Col vs. Col-V3; p = 1.6e-7 for Col-Ver vs. Col-Agg; p = 2.5e-6 for Col-Ver vs. Col-Dec; p = 1.1e-8 for Col-V3 vs. Col-Agg; p = 1.9e-7 for Col-V3 vs. Col-Dec). B, the shear storage modulus (G′) for the gels in (A), measured after complete gelation (p = 1.6e-5 for Col vs. Col-Ver; p = 2.0e-5 for Col vs. Col-V3; p = 1.3e-5 for Col-Ver vs. Col-Agg; p = 0.0021 for Col-Ver vs. Col-Dec; p = 1.7e-5 for Col-V3 vs. Col-Agg; p = 0.0022 for Col-V3 vs. Col-Dec). C, G′ for collagen gelled with the addition of high molecular weight (1.5 MDa) HA or both HA and V3. Col-0.1HA: 2.5 mg/ml collagen gel containing 0.1 mg/ml HA; Col-0.1HA-V3: 2.5 mg/ml collagen gel containing 0.1 mg/ml HA and 0.167 mg/ml V3 (p = 0.0248 for Col vs. Col-0.1HA). D, G′ of different collagen-proteoglycan co-gels measured by shear rheometry under increasing strain stiffening after full gelation (see Table S4). E, G′ of different collagen-proteoglycan co-gels measured with the gap remaining at 1 mm and under compression to 10% (gap changed to 0.9 mm) (at non-compression, p = 0.0001 for Col vs. Col-Ver, p = 0.026 for Col vs. Col-V3, p = 0.0073 for Col-Agg vs. Col-V3, p = 2.3e-5 for Col-Agg vs. Col-Ver, p = 0.0237 for Col-Dec vs. Col-V3, p = 9.9e-5 for Col-Dec vs Col-Ver; at 10% compression, p = 0.0081 for Col-Agg vs. Col-Ver). F, G′ values at 10% compression for the different co-gels in (E) were normalized to the G′ in the non-compressed state (p = 0.0056 for Col vs. Col-Agg, p = 0.0004 for Col-Ver vs. Col-Agg, p = 0.0031 for Col-Agg vs. Col-V3). For measuring G′ during gelation in A and B, N = 17 for Col, N = 12 for Col-Ver, N = 11 for Col-V3, N = 15 for Col-Agg and N = 11 for Col-Dec; in C, N = 13 for Col, N = 12 for Col-0.1HA and N = 10 for Col-0.1HA-V3. These gels were tested with either strain sweep (N = 3 for Col, N = 3 for Col-Ver, N = 3 for Col-V3, N = 4 for Col-Agg and N = 3 for Col-Dec) or compression (N = 3 for Col, N = 3 for Col-Ver, N = 4 for Col-V3, N = 4 for Col-Agg and N = 3 for Col-Dec); each gel was subject to only one test. The dotted lines in (D) represent SD. Data represent mean ± SD; A and B were analyzed using one-way ANOVA with Tukey multiple comparison, C was analyzed by Kruskal-Wallis test with Dunn’s multiple comparison, D, E, and F were analyzed using two-way ANOVA with repeated measurements with Tukey’s multiple comparison; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 and ∗∗∗∗p < 0.0001.

Figure 6

ADAMTS5 and ChABC treatment of rat livers alters compression stiffening behavior.A, sulfated GAG quantification after perfusion with either enzyme, compared to control (Hank’s Balanced Salt Solution; HBSS) (p = 0.0123 for HBSS vs. ADAMTS5 and p = 0.0019 for HBSS vs. ChABC). B, representative confocal imaging of immunostaining using an antibody against DPEAAE, the epitope exposed by ADAMTS5 cleavage of versican. DPEAAE (green), DAPI (blue). C, G′ was measured under 0%, 10%, 15%, 20% and 25% compression (with HBSS perfusion as a control). There is a significant difference at 25% compression (p = 0.0171 for HBSS vs. ChABC at 25% compression). D, Young’s modulus (E) was calculated from normal force and gap changes and plotted at 5%, 12.5%, 17.5% and 22.5% compression (p = 0.003 for HBSS vs. ADAMTS5 and p = 1.8e-5 for HBSS vs. ChABC at 22.5% compression). E, G′ measured at increasing strain from 1% to 50% (data did not show statistical differences.). N = 3 for HBSS, N = 4 for ADAMTS5, and N = 4 for ChABC-perfused livers. Compression and strain sweep experiments were done on the same liver samples, as was the assay for sulfated GAGs (A). Scale bar = 200 μm. Data represent mean ± SD; A was analyzed using one-way ANOVA, C, D, and E using two-way ANOVA with repeated measurements; post hoc test by Tukey’s multiple comparisons; ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001 and ∗∗∗∗p < 0.0001.

Figure 7

Model of interactions between HA, collagen, and versican.A, collagen and HA binding sites on different versican isoforms. B, versican may serve as a linker between collagen fibers and HA chains in collagen fibrous networks. The median diameter of a collagen fibril is approximately 200 nm (53), and the size of isolated versican, measured by dynamic light scattering (data not shown), is 203.4 nm. In this schematic model, the collagen fibril diameter and length of versican are therefore represented as being roughly the same size.