Type III Collagen Regulates Matrix Architecture and Mechanosensing during Wound Healing

Abstract

Postnatal cutaneous wound healing is characterized by development of a collagen-rich scar lacking the architecture and functional integrity of unwounded tissue. Directing cell behaviors to efficiently heal wounds while minimizing scar formation remains a major wound management goal. In this study, we demonstrate type III collagen (COL3) as a critical regulator of re-epithelialization and scar formation during healing of COL3-enriched, regenerative (Acomys), scar-permissive (CD-1 Mus and wild-type Col3^B6/B6 mice) and COL3-deficient, scar-promoting (Col3^F/F, a murine conditional knockdown model) cutaneous wound models. We define a scar-permissive fibrillar collagen architecture signature characterized by elongated and anisotropically aligned collagen fibers that is dose-dependently suppressed by COL3. Furthermore, loss of COL3 alters how cells interpret their microenvironment—their mechanoperception—such that COL3-deficient cells display mechanically active phenotypes in the absence of increased microenvironmental stiffness through the upregulation and engagement of the profibrotic integrin α11. Further understanding COL3’s role in regulating matrix architecture and mechanoresponses may inform clinical strategies that harness proregenerative mechanisms.

INTRODUCTION

The ideal response to injury is regeneration, with complete restoration of normal tissue architecture and function. The healing of cutaneous injuries in most postnatal mammals, including humans, occurs by repair rather than regeneration, with an ensuing fibroproliferative response that promotes both re-epithelialization and scar formation. For more than 40 million Americans each year (Levinson, 2013; Sen et al, 2009), pathologic wound responses result in clinical outcomes at the extremes of the healing spectrum—persistent, nonhealing wounds on 1 end and excessive scar formation on the other. Both chronic wounds and pathologic scars lead to functional impairment, discomfort, and psychosocial morbidity, with an annual United States healthcare cost exceeding $40 billion (Sen et al, 2009). The ability of the midgestational human fetus and certain species of adult mammals (Adzick and Lorenz, 1994; Clark et al, 1998; Gawriluk et al, 2016) to regenerate skin after injury suggests that healing in people could be improved by a more thorough understanding of mechanisms promoting regenerative responses. Identifying such mechanisms that promote efficient wound healing while minimizing scarring would be highly impactful for fundamental science and, importantly, would also support the development of novel therapies to improve the lives of millions of people worldwide.

Tissue repair is a complex, orchestrated event involving a variety of cells, soluble factors, and extracellular matrix (ECM) proteins and their modifiers, which is characterized by dynamic reciprocity, a process whereby cells and the ECM influence and respond to each other over time (Schultz et al, 2011). Notably, alterations in collagen architecture, density, type, and resulting tissue stiffness provide topographical, biochemical, and biomechanical cues to cells after injury (Wietecha et al, 2020). Collagens have a dual role in skin wound repair. Although collagen deposition is required for efficient re-epithelialization and restoration of the dermis during cutaneous wound repair, the replacement of normal skin structures with collagen results in scar tissue, which is functionally inferior to uninjured skin. Furthermore, excessive collagen deposition results in pathologic scar formation. Scars are typically characterized as having an anisotropic alignment of collagen fibers oriented parallel to the neoepidermis, contrary to the architecture of unwounded dermis where collagen fibers are oriented bi-isotropically, in a basket-weave pattern. However, the characteristics of a proregenerative collagen matrix capable of recapitulating the integrity and function of uninjured skin remain ill defined.

Type III collagen (COL3), the second most abundant collagen in the body, plays a critical role in fetal development as well as in postnatal tissue maintenance and repair (Cooper et al, 2010; Liu et al, 1997). Although a regulatory role for COL3 in cutaneous wound healing has long been proposed on the basis of its induction early in healing (Hurme et al, 1991; Merkel et al, 1988), our laboratory first revealed a specific role for COL3 in regulating cutaneous repair and matrix remodeling/scar formation through its ability to attenuate myofibroblast activation and persistence (Volk et al, 2011). Furthermore, increased COL3 has been observed in 2 mammalian models of scar-free cutaneous wound healing—the midgestational fetus (Cuttle et al, 2005) and the African Spiny Mouse (Acomys spp) (Brant et al, 2015; Harn et al, 2021; Seifert et al, 2012)—although the spatiotemporal effects and mechanisms through which COL3 contributes to regenerative responses remain ill defined in these models and in other tissues. In this study, we investigate the mechanistic roles of COL3 in promoting the formation of a fibrillar collagen network that both supports efficient re-epithelialization and suppresses scar formation. Given COL3’s role in regulating and remodeling matrix architecture, we examined the evolution of collagen architecture in regenerative (Acomys cahirinus) and scar-permissive (Mus, Col3^B6/B6, and Col3^F/F) healing microenvironments and used second harmonic generation (SHG) microscopy to define key collagen microarchitectural features that direct a switch between scarless and scarring wounds. Furthermore, our work reveals that COL3 promotes the efficiency and quality of cutaneous wound repair through its ability to accelerate re-epithelialization and limit scar-promoting fibroblast activation and collagen organization. Our data also suggest that COL3 regulates fibroblast mechanoperception in the absence of increased granulation tissue (GT) stiffness in an α₁₁ integrin–dependent response, providing unique insight into how COL3 might orchestrate dynamic reciprocity between stromal cells and collagens in repair and regeneration of cutaneous and noncutaneous tissues and the tumor microenvironment (Brisson et al, 2015; Miedel et al, 2015; Volk et al, 2014, 2011). Defining such mechanisms has important implications for fine tuning COL3-directed therapies that may be harnessed to more effectively treat diseases encompassing the wound healing–fibrosis–cancer triad.

RESULTS

Cutaneous wounds in regenerative mammals recapitulate collagen architecture of unwounded dermis

To investigate the evolution of a collagen architecture capable of supporting a regenerative/scarless response after cutaneous injury, we examined collagen fiber characteristics of the GT in a mammalian model of regeneration, Acomys, 7, 14, and 35 days post-wounding (DPW) compared with that in the common laboratory mouse, Mus. We hypothesized that a proregenerative matrix would maintain collagen micro-architectural characteristics similar to that found in unwounded dermis, that is, collagen fibers arranged in a bi-isotropic, basketweave pattern instead of anisotropically aligned collagen fibers parallel to the neoepidermis exhibited in scars. Unwounded dermis and postinjury GT were imaged with brightfield (H&E) and SHG imaging (Figure 1a and b [for Mus and Acomys, respectively]). Consistent with previous reports (Brant et al, 2016; Harn et al, 2021; Seifert et al, 2012), healing of skin wounds in Mus results in the formation of a prominent scar by 35 DPW (Figure 1a), in contrast to wounds in Acomys, which heal scarlessly with the formation of new adnexal structures within the injury site (Figure 1b). SHG imaging of unwounded dermis reveals the canonical basketweave collagen architecture in dermis of both species (Figure 1a and b [for Mus and Acomys, respectively]) with no significant interspecies differences observed in collagen fiber characteristics (length, width, number of fibers/field of view) (Figure 1c–e) prior to injury. However, significant differences in GT collagen architecture were noted during healing between the scarless healing process in Acomys and the reparative response in Mus. Notably, there was a relative paucity of visualized fibers during the early healing stage (7 DPW) (Supplementary Figure S1) compared with that in unwounded skin and later time points (14 and 35 DPW) (Figure 1), precluding direct comparison of fiber characteristics between these stages; however, interspecies comparison 7 days after wounding reveals that fiber density was significantly reduced in GT of Acomys relative to that in Mus (Supplementary Figure S1g). Fourteen DPW, Acomys GT was composed of shorter and less dense (Figure 1c and e, respectively) SHG-detected collagen fibers than that of Mus. Although collagen fiber width increased between 14 and 35 DPW in both species (Figure 1d), length and density remained unchanged in the GT of Mus. Thirty-five DPW, no significant differences in these fiber properties existed between the 2 species, and both exhibit fiber length, width, and abundance similar to those of unwounded dermis.

Figure 1. Evolution of collagen architecture in wounds of regenerative Acomys and scarring Mus species after cutaneous injury.

(a, b) Representative H&E-stained sections and SHG images of (a) Mus and (b) Acomys unwounded dermis and cutaneous wounds evaluated 7^†, 14, and 35 DPW. (c–e) SHG-detected collagen architecture was assessed by comparing (c) fiber length, (d) fiber width, and (e) number of fibers 14 and 35 DPW. (f–h) Collagen fiber alignment of unwounded dermis and of 14 and 35 DPW granulation tissue was assessed by determining the (f) aspect ratio of the FFT of each region and (g) using OrientationJ to compare the distribution and angle orientation of fibers. (h) Fibers aligned within ±15° of the modal fiber orientation angle was compared between unwounded and healing dermis of both species. Large opaque dots represent the average values from n = 4 mice per time point from n = 5 regions (transparent dots) within each sample. *P < .05, **P < .01, ***P < .001, and ****P < .0001 through 2-way ANOVA with Tukey’s multiple comparisons. Biologically relevant comparisons (intraspecies between time points and interspecies at each time point) are shown, and all statistical comparisons can be found in Supplementary Tables S3–7. Statistical significance bars are color coordinated to highlight intraspecies comparisons (same color as data) or interspecies comparisons (shown in red). Bars = 1 mm or 20 μm (insets). ^†Note: 7 DPW SHG images (red boxes) were taken at a higher 2-photon laser intensity to visualize collagen architecture and therefore were not included in quantitative comparisons with unwounded and 14 and 35 DPW samples (7 DPW analysis can be found in Supplementary Figure S1). DPW, days post-wounding; FFT, fast Fourier transform; SHG, second harmonic generation; Unw., unwounded.

Given that scar-permissive wounds heal with increased anisotropic alignment of collagen fibers, we assessed SHG-detected fiber alignment in Acomys and Mus throughout the course of healing by investigating the aspect ratio of the fast Fourier transform of each region (Figure 1f). Despite the convergence of fiber length, width, and number of fibers 35 DPW in both species, the overall alignment of fibers was significantly reduced in Acomys relative to that in Mus wounds at that time (Figure 1f–h). To test the hypothesis that Acomys can regenerate a bi-isotropic matrix similar to unwounded dermis, we analyzed SHG images from wounds in both species 7, 14, and 35 DPW with OrientationJ to compare the distribution and angle orientation of fibers (Figure 1g and Supplementary Figure S1f) (distributions normalized with respect to the modal angle at 0°). As expected with the basket-weave pattern of unwounded dermis in both species, Mus and Acomys skin exhibited a bimodal distribution of fibers (blue lines), with ~20% of the fibers within ±15° of the modal fiber angle orientation (Figure 1g and h). During scar formation, there was an increasing frequency of fibers aligned in 1 direction in Mus (Figure 1g, leftmost graph), in contrast to Acomys fibers, which initially exhibited no observable increase in a preferred fiber orientation (Supplementary Figure S1f) and then transitioned to a bimodal distribution 35 DPW (Figure 1g, right graph). Thirty-five DPW, Mus wounds were observed to have ~50% of fibers within the scar area within ±15° of the modal angle, compared with those in Acomys in which significantly fewer (~20%) fibers were found within ±15° of the modal angle (Figure 1h). Consistent with the association of collagen fiber alignment and propensity for scar formation, Mus fibers exhibit a positive linear correlation between DPW and the percentage of fibers within ±15° of the modal angle (R² = 0.7920), whereas no correlation was observed in Acomys (R² = 0.02145) (slopes of linear regression are significantly different between Mus and Acomys, P < .0001).

COL3 expression is enriched and sustained throughout healing in the matrices of regenerative wounds

Induction of COL3 during the early stages of cutaneous healing is well-established, as are its relative higher expression in models of scarless healing such as Acomys (Brant et al, 2015; Harn et al, 2021; Seifert et al, 2012) and the midgestational mammalian fetus (Cuttle et al, 2005). Although COL3 expression is transiently increased in early stages of GT formation in multiple species (Clore et al, 1979; Hurme et al, 1991; Merkel et al, 1988), its temporal expression pattern throughout healing in Acomys is poorly defined. To test the hypothesis that the changes in matrix architecture seen in Acomys wounds noted earlier are associated with increased and sustained COL3 expression throughout healing, compared with its early but transient induction in Mus wounds, COL3 content in wounds from both species was assessed by immunofluorescence (Figure 2a–d). In addition, SHG imaging of fibrillar collagen density was performed to determine COL3 content relative to overall fibrillar collagen content in wounds over time. Baseline COL3 immunoreactivity and SHG-integrated density in unwounded dermis were not significantly different between the 2 species (Figure 2e and f). As expected, an induction of COL3 was noted 7 DPW in both species compared with that in their respective unwounded tissues (Figure 2e). As previously observed in postnatal mammalian healing (Hurme et al, 1991; Merkel et al, 1988), COL3 content of GT decreased significantly (to preinjury levels) between 7 DPW and 14 DPW in Mus (no significant differences in preinjury, levels at 14 and 35 DPW), whereas overall fibrillar collagen content begins to accumulate. In contrast, Acomys wound COL3 density continues to increase between 7 DPW and 14 DPW, at which time Acomys wounds exhibit significantly higher COL3 integrated density than time-matched Mus wounds. Thirty-five DPW, COL3 content in Acomys wounds was decreased compared with that at 14 DPW but remained elevated relative to that in Mus wounds 35 DPW. Although overall SHG signal associated with total fibrillar collagen increased progressively over the course of healing, no intraspecies differences in overall SHG signal in GT was detected (Figure 2f). Comparison of the ratio of COL3:SHG integrated density revealed an increase in Acomys wounds compared with that in Mus wounds 7 and 14 DPW, with normalization of this ratio between species by 35 DPW. Together, our data demonstrate that generation of a scar-permissive matrix is associated with lower COL3 expression, particularly in the context of overall fibrillar collagen, during proliferative and early remodeling stages of wound repair.

Figure 2. Acomys wounds have increased and sustained COL3 expression during healing compared with Mus wounds.

(a–d) Representative microscopy images with SHG (cyan) and COL3 (magenta) immunofluorescence for (a) unwounded dermis and granulation tissue at (b) 7, (c) 14, and (d) 35 DPW (bars = 20 μm). (e) COL3-integrated density, (f) SHG-integrated density, and (g) the ratio of COL3:SHG intensity in unwounded Acomys and Mus skin and 7, 14, and 35 DPW wounds. *P < .05, ***P < .001, and ****P < .0001 through 2-way ANOVA with Tukey’s multiple comparisons. Biologically relevant comparisons are shown in figure; full ANOVA results and P-values can be found in Supplementary Tables S8–10. n = 3–4 mice per species and time point. DPW, days post-wounding; SHG, second harmonic generation.

COL3 deficiency promotes assembly of aligned collagen matrices

Having identified the characteristics of a proregenerative collagen matrix in vivo, including a bi-isotropic architecture and enhanced COL3 expression, we next asked whether COL3-deficient (Col3^−/−) fibroblasts would produce a more aligned, scar-permissive matrix in vitro compared with COL3-expressing (wild-type, Col3^+/+) fibroblasts. Murine dermal embryonic fibroblasts, isolated from 5 independent pairs of Col3^+/+ and Col3^−/− embryonic littermates, were cultured on physiologically relevant and hyperphysiologically stiff silicone substrata for 7 days, and their fibroblast-derived matrices (FDMs) were subsequently imaged with SHG (Figure 3a–d). We observed that both increased stiffness and COL3 deficiency promoted the production of FDMs with characteristics (Figure 3e–k) associated with the scar-permissive matrices observed in vivo (Figure 1). Col3^+/+ fibroblasts cultured on stiff substrata (E ~ 1 MPa [Palchesko et al, 2012]) generated FDMs with longer (Figure 3e) and more aligned fibers (Figure 3i–k) than those cultured on physiologic (soft, E ~ 5 kPa [Palchesko et al, 2012]) substrata. Col3^−/− fibroblasts cultured on soft silicone assembled a denser matrix (Figure 3g and h) comprised of longer (Figure 3e), wider (Figure 3f), and more aligned (Figure 3j) fibers than Col3^+/+ fibroblasts under similar conditions. We also compared Col1a1, Col1a2, and Col3a1 expressions between COL3-producing and COL3-deficient fibroblast–ECM units. Although we observed the expected lack of Col3a1 expression in Col3^−/− fibroblasts, we did not observe a difference in Col1a1 or Col1a2 expression between genotypes and stiffness under these culture conditions (Supplementary Figure S2), suggesting the importance of COL3-dependent remodeling on matrix architectural differences, as seen with SHG, and/or the dependency on longer time courses seen under in vivo conditions to result in increased scarring associated with COL3 loss. These data indicate that fibroblasts cultured in scar-permissive conditions—COL3 deficiency and stiffer microenvironments—assemble collagen matrices with fiber microarchitectural features reminiscent of developing scar tissue, whereas fibroblasts cultured in a softer, COL3-containing microenvironment assemble a matrix with scar-restrictive characteristics.

Figure 3. Col3^−/− FDMs exhibit scar-permissive characteristics.

(a–d) Representative SHG images of FDMs from (a, b) Col3^+/+ and (c, d) Col3^−/− fibroblasts grown on (a, c) soft and (b, d) stiff silicone substrates. (e–h) CT FIRE analysis reveals that Col3^−/− fibroblasts assemble matrices with (e) longer, (f) wider, (g) more abundant, and (h) more matrix than a Col3^+/+ FDMs in a soft microenvironment. (i–k) Col3^+/+ FDMs on a stiff microenvironment and Col3^−/− FDMs assemble a more aligned matrix assessed by (i) the alignment of the FFT and (j, k) distribution of fiber orientation. *P < .05, **P < .01, and ***P < .001 through 2-way ANOVA with Tukey’s multiple comparisons. Bars = 50 μm. COL3, type III collagen; FDMs, fibroblast-derived matrices; FFT, fast Fourier transform; SHG, second harmonic generation.

Loss of COL3 impairs early wound healing and accelerates the assembly of a scar-permissive collagen architecture

Given our findings of the proregenerative properties of a COL3-enriched matrix in vivo and the observation that COL3-deficient fibroblasts assemble an aligned, scar-permissive matrix in vitro, we next used our recently developed COL3-deficient murine model to investigate the roles of COL3 in wound healing in vivo. We generated a ROSA-Cre-Col3a1 conditional knockdown mouse model (Col3a1^{flox/flox-ERT2/ERT2}, referred to as Col3^F/F in this study) (Supplementary Figure S3) on a C57BL/6J wild-type background (Col3^B6/B6, wild-type control). Depletion of COL3 at the time of wounding was achieved through 3 daily tamoxifen injections starting 6 days before wounding (D-6, Figure 4a) and confirmed at both the RNA (Figure 4b) and protein levels (pro and mature COL3 protein) (Supplementary Figure S3). Wound closure was observed throughout a 14-day period, whereas wound histomorphometrics and GT collagen architecture were analyzed in histological sections by H&E and SHG imaging 7, 14, 28, and 35 days after wounding, respectively. Observation of gross wound area demonstrated that COL3 loss delayed wound closure in tamoxifen-treated Col3^F/F mice compared with that in Col3^B6/B6 littermates 3, 7, and 10 DPW (Figure 4c). Differences in wound closure at these early time points are consistent with the early expression of COL3 by stromal cells in the provisional matrix (Clore et al, 1979), although differences in compliance of surrounding skin cannot be ruled out as a contributing factor. Consistent with this finding, histomorphometry established that the mean neoepidermal gap was greater in Col3^F/F wounds than in Col3^B6/B6 mice 7 DPW (Figure 4f and Supplementary Figure S4). In support of a direct effect of COL3 on keratinocyte behavior, human keratinocytes closed an in vitro wound gap faster and exhibited less meandering on COL3-coated substrate than on type I collagen (COL1)–coated substrates (Supplementary Figure S4c, d, e, and f, respectively). The ability of COL3 to influence re-epithelialization in vivo is supported by the observed increased localization of COL3 beneath the neoepidermis in both Mus and Acomys wounds 14 DPW (Supplementary Figure S4g and h). Previously, we reported that COL3-haploinsufficient mice heal with impaired re-epithelialization and increased scar volume after cutaneous wounding relative to wild-type littermates (Volk et al, 2011). Consistent with these data, we found that COL3 loss in Col3^F/F mice is associated with an increased GT/scar area at 35 DPW compared with that in Col3^B6/B6 mice (Figure 4g).

Figure 4. COL3 loss impairs wound healing and accentuates assembly of a scar-permissive collagen architecture.

(a) Timeline depicting tamoxifen-induced knockdown of Col3a1 and wound surgery experiment. (b) Knockdown of Col3a1, evidenced by qPCR, is observed in Col3^F/F (Col3a1 knockdown) mice after 3 tamoxifen injections and not in treated Col3^B6/B6 (wild-type) mice; knockdown in Col3^F/F is sustained through the longest observed time point in the study. n = 7–9 mice per genotype and treatment. (c) Representative gross wound photos and quantitated wound closure of wounds in both Col3^F/F and Col3^B6/B6. (d, e) Representative H&E and SHG images of wounds in (d) Col3^B6/B6 and (e) Col3^F/F mice over the course of healing. (f, g) Histomorphometric analysis of (f) epidermal gap (*P < .05 through unpaired t-test) and (g) granulation tissue area. (h, i) CT-FIRE quantitation of SHG-detected collagen (h) fiber length and (i) straightness; no differences between genotype were observed in fiber width or number (Supplementary Figure S4). (j, k) OrientationJ analysis of SHG-detected collagen fibers in Col3^B6/B6 and Col3^F/F granulation tissue comparing (j) the number of fibers within 15° of the modal orientation angle with (k) corresponding rose plots depicting the distribution of fibers. Note that scales on rose plots are different for each genotype; corresponding inset contains rose plots of each genotype at the same scale. (l) Linear regression models of collagen fiber alignment over time have increased positive correlation with decreasing COL3 expression during the later stages of healing. *P < .05, **P < .01, ***P < .001, and ****P < .0001 through 2-way ANOVA with Tukey’s multiple comparisons. Biologically relevant comparisons are shown in figure; full ANOVA results and P-values can be found in Supplementary Tables S9–13. Statistical significance bars are color coordinated to highlight intragenotype comparisons (same color as data) or intergenotype comparisons (shown in red). Bars = 1 mm or 20 μm (insets). COL3, type III collagen; DPW, day post-wounding; SHG, second harmonic generation.

Focusing on the collagen organization in both unwounded skin and wound GT, we found, similar to Mus and Acomys, that collagen within the unwounded dermis of both Col3^B6/B6 and Col3^F/F exhibited a canonical basket-weave pattern (Figure 4d and e, respectively). The evolution of GT collagen architecture in Col3^B6/B6 wounds was similar to that seen in Mus wounds (Figure 1), with the formation of a progressively anisotropically aligned collagen matrix (Figure 4d). Notably, collagen architecture in Col3^F/F GT revealed an accentuated scar-permissive matrix with longer fibers than that in Col3^B6/B6 wounds (Figure 4e). Seven DPW, Col3^B6/B6 and Col3^F/F mice had shorter, thinner, and fewer fibers than each respective unwounded dermis, although no differences were detected between genotypes at this time point or 14 DPW (Figure 4h and i and Supplementary Figure S5). Both 28 and 35 DPW, Col3^F/F wounds had significantly longer collagen fibers than Col3^B6/B6 wounds (Figure 4h), with a progressive increase in fiber length in Col3^F/F wounds between 28 and 35 DPW. No significant difference was observed in fiber width or density between genotypes (Supplementary Figure S5), with the width of the fibers in the GT being similar to that of the unwounded skin it replaced by 35 DPW. Fiber straightness was increased in GT compared with that in unwounded skin and significantly increased in response to COL3 loss (Col3^F/F vs Col3^B6/B6 wounds) 35 DPW (Figure 4i).

Given our in vitro observations that Col3^−/− fibroblasts generate a more aligned matrix (Figure 3j and k), we compared the collagen fiber distribution of Col3^B6/B6 wounds with that of Col3^F/F wounds. We observed that Col3^F/F wounds had significantly more fibers within 15° of the mode angle orientation than Col3^B6/B6 wounds at each time point and throughout the course of healing (Figure 4j and k). Furthermore, 35 DPW, wound matrices of both genotypes were more aligned than those from 7 and 14 DPW, respectively. Linear regression models of collagen fiber alignment over time revealed an increased positive correlation with decreasing COL3 expression (Figure 4l); slopes between linear regressions of Col3^B6/B6 (R² = 0.4952) and Col3^F/F (R² = 0.7121) were significantly different (P = .0028). In addition, slopes of linear regressions were significantly different between COL3-enriched Acomys wounds and wildtype and COL3-deficient mice (P = .0017 and P < .0001 compared with those of Col3^B6/B6 and Col3^F/F, respectively). Thus, our data reveal a role for COL3 in regulating matrix architecture during healing by regulating collagen fiber length and orientation.

COL3 deficiency promotes myofibroblast activation and alters cellular mechanoperception

Myofibroblasts play a key role in the production and remodeling of collagen during wound healing. Activation of recruited fibroblasts into myofibroblasts occurs in response to a variety of mechanisms, including but not limited to key biomechanical and biochemical regulators such as increased matrix stiffness and TGFβ signaling. Our previous work revealed that COL3 loss increases myofibroblast activation in culture and within GT in a dose-dependent manner (Volk et al, 2011). Although we established that the N-propeptide of COL3 plays a role in regulating TGFβ bioavailability (Brisson et al, 2022), such that loss of COL3 increases TGFβ signaling in vitro and in vivo, the observed impact on matrix architecture suggests that COL3 may also directly regulate cellular mechanotransduction in healing tissues. To minimize potential confounding effects of the matrix architecture on the mechanical microenvironment, we first assessed the impact of COL3 loss on the phenotype of fibroblasts cultured on substrates of known stiffness (physiologic and hyper-physiologic) in short-term (72-hour) cultures to minimize accumulated matrix. Embryo-paired Col3^+/+ and Col3^−/− fibroblasts were cultured on soft silicone (E ~ 5 kPa) and glass (E ~ 1 GPa) to compare propensity for myofibroblast activation, as evidenced by the incorporation of α-smooth muscle actin (αSMA) into F-actin stress fibers (Figure 5a and b). As expected, an increase in substrate stiffness promoted myofibroblast activation. Interestingly, a loss of COL3 also resulted in increased αSMA incorporation into stress fibers, even in the absence of increased stiffness. Consistent with our previous data showing that Col3^+/+ fibroblasts contract 3-dimensional collagen lattices less efficiently than COL3-deficient fibroblasts (Col3^+/+ < Col3^+/− < Col3^−/−) (Volk et al, 2011), functional myofibroblast activation was also observed through increased contractility in Col3^−/− fibroblasts by more wrinkling of a deformable silicone substrate than in Col3^+/+ fibroblasts on both COL1- and COL3-coated silicone (Figure 5c and d). In addition, Col3^+/+ fibroblasts exhibited less overall wrinkling on COL3-coated silicone than on COL1-coated silicone.

Figure 5. COL3 deficiency promotes myofibroblast activation and alters cellular mechanoperception.

(a) Representative embryonic littermate-paired Col3^+/+ and Col3^−/− fibroblasts cultured on soft silicone (E ~ 5 kPa) and glass (E ~ 1 GPa) substrates for 72 hours were stained for αSMA and F-actin. (b) Quantitative analysis of colocalized αSMA and F-actin pixels (canonical indicator of myofibroblast activation); n = 4 fibroblast pairs. (c) Brightfield images of silicone wrinkling by Col3^+/+ and Col3^−/− fibroblasts on COL1- or COL3-coated substrata and (d) corresponding quantitated total wrinkled area from n = 4 fibroblast pairs. (e) Representative images of YAP immunolocalization in paired Col3^+/+ and Col3^−/− fibroblasts cultured on soft silicone and glass with (f) quantitative analysis of the percentage of immunopositive nuclear YAP relative to the total cellular YAP signal from the averages of n = 3 fibroblast pairs. *P < .05, **P < .01, and ***P < .001 through 2-way ANOVA with Tukey’s multiple comparisons. Bars = 20 μm. αSMA, α-smooth muscle actin; COL1, type I collagen; COL3, type III collagen.

To determine whether this phenotypic shift in fibroblast activation was associated with altered mechanotransduction, we next asked whether COL3 deficiency promotes the translocation of the mechanosensitive protein YAP into the nucleus (Dupont et al, 2011) by measuring the percentage of nuclear/total YAP ratios. As expected, both Col3^+/+ and Col3^−/− fibroblasts on a stiff substrate had a higher percentage of YAP colocalized in the nucleus compared to those on a soft substrate. Whereas Col3^+/+ fibroblasts on a soft substrate had a low percentage of nuclear YAP as expected, surprisingly, Col3^−/− fibroblasts on soft substrata showed a marked increase of translocated YAP into the nucleus, as evidenced by their increased percentage nuclear/total YAP, reaching levels comparable with those of cells on glass (Figure 5e and f). This phenomenon was also observed with the mechanosensitive transcriptional regulator TAZ (as detected by a WWTR1 antibody) (Supplementary Figure S6). To test the hypothesis that myofibroblast activation in COL3-deficient fibroblasts was in part due to altered mechanosensing, we drove Col3^+/+ and Col3^−/− myofibroblast activation by culturing on tissue culture plastic (E ~ 1 GPa) and treated with verteporfin (0.5 μM), a YAP inhibitor that prevents its nuclear accumulation (Supplementary Figure S7). Decreased αSMA/F-actin colocalization was observed in Col3^+/+ and Col3^−/− fibroblasts (Supplementary Figure S7a and b) when treated with verteporfin for 72 hours, concomitant with decreased nuclear YAP in both genotypes (Supplementary Figure S7c and d). We also compared the mRNA expression of Ccn2, a downstream target of activated YAP signaling, in Col3^+/+ and Col3^−/− fibroblasts. Although we did not observe significant differences in mRNA expression with 0.5 μM verteporfin (data not shown), we saw a decrease in Ccn2 in both wild-type and COL3-deficient fibroblasts treated with 1 μM verteporfin (Supplementary Figure S7e and f). These data thus support our hypothesis that COL3 deficiency drives increased myofibroblast activation through altered mechanosignaling.

We next asked whether the COL3-mediated myofibroblast activation and in vitro mechanoresponses were also operative in vivo in our Col3^F/F wound model. Similar to our previous observations in COL3-haploinsufficient (Col3^+/−) mice (Volk et al, 2011), we saw increased αSMA expression in wounds of Col3^F/F mice (Figure 6a–d). Seven DPW, αSMA expression was more highly concentrated at the wound edges in Col3^F/F wounds relative to that in Col3^B6/B6 wounds and compared with that found in the center of Col3^F/F wounds 7 DPW (Figure 6a and b). Fourteen DPW, αSMA+ cells were distributed throughout the GT in both Col3^F/F and Col3^B6/B6 mice; however, there was an increase in myofibroblast density in COL3-deficient wounds of Col3^F/F mice relative to those of Col3^B6/B6 littermates at each time point analyzed (14, 28, and 35 DPW) (Figure 6c and d). Given the well-established mechanism through which tissue stiffness regulates myofibroblast activation, we hypothesized that an overall increase in GT stiffness secondary to COL3-dependent changes in matrix architecture would provide the mechanistic explanation for the observed differences in myofibroblast density. Using atomic force microscopy (AFM)–based nanoindentation, we interrogated the stiffness of GT in Col3^F/F wounds compared with wounds in wild-type (Col3^B6/B6) mice throughout the course of healing. As expected, GT stiffness increased over time in wild-type wounds; however, GT in Col3^F/F wounds did not show a time-dependent stiffening. In fact, Col3^F/F wounds were significantly softer than Col3^B6/B6 wounds 28 and 35 DPW (Figure 6e). No difference was observed in the stiffness of unwounded dermis between genotypes (7.8 ± 3.0 kPa and 7.2 ± 2.3 kPa for Col3^B6/B6 and Col3^F/F, respectively, P = 0.4120 by unpaired t-test). To address the discordance between the decrease in stiffness and increase in activated myofibroblast density in the GT of Col3^F/F wounds compared with that of Col3^B6/B6 wounds, we investigated markers of mechanotransduction in GT stromal cells throughout healing. Surprisingly, despite a reduction in GT stiffness in Col3^F/F wounds compared with that in Col3^B6/B6 wounds at later time points, we observed an increase in YAP immunoreactivity associated with COL3 loss (28 and 35 DPW) (Figure 6f and g), suggesting that GT stromal cells exhibited positive mechanoresponses, despite the softer microenvironment. In addition, we observed increased expression of mechanosensitive markers RhoA in COL3-knockdown wounds 7, 28, and 35 DPW and phosphorylated focal adhesion kinase (pFAK) at 28 and 35 DPW (Figure 6h–k), providing additional support that COL3 loss drives cellular mechanoresponses in the face of diminished GT stiffness.

Figure 6. COL3 loss increases cellular mechanotransduction despite a decrease in granulation tissue stiffness.

(a, b) Representative immunofluorescence images of αSMA expression in Col3^B6/B6 and Col3^F/F wounds at 7 DPW. (a) Insets show differences in αSMA intensity between the edge and center of granulation tissue between genotypes, and (b) quantitated integrated density data show spatially different αSMA expression in wounds (7 DPW). (c, d) Representative insets of (c) αSMA immunoreactivity 35 DPW in Col3^B6/B6 and Col3^F/F wounds and (d) quantitated αSMA integrated density within wounded and unwounded dermis. (e) AFM assessment of granulation tissue stiffness in Col3^B6/B6 and Col3^F/F wounds. (f–j) Representative immunofluorescence images of Col3^B6/B6 and Col3^F/F wounds 35 DPW and quantified intensity data for mechanosensitive proteins (f, g) YAP, (h, i) pFAK, and (h, j) RhoA. n = 4–6 mice per genotype and time point. *P < .05, **P < .01, ***P < .001, and ****P < .0001 through 2-way ANOVA with Tukey’s multiple comparisons. Biologically relevant comparisons are shown in figure; full ANOVA results and P-values can be found in Supplementary Tables S16–18. Bars for all images = 1 mm for whole-tissue sections and 20 μm for all insets. αSMA, α-smooth muscle actin; AFM, atomic force microscopy; COL3, type III collagen; DPW, days post-wounding; pFAK, phosphorylated focal adhesion kinase.

Altered mechanoperception with loss of COL3 depends on integrin α11

After observing that Col3^F/F fibroblasts exhibited increased mechanotransduction on soft substrates in vitro and in spite of decreased GT stiffness in vivo, we next asked how COL3-dependent interactions between fibroblasts and their surrounding matrix might regulate myofibroblast activation. Given that fibroblasts isolated from patients with vascular Elhers-Danlos syndrome (EDS), who present with sequence variation in the COL3A1 gene, exhibit increased use of integrin αv and decreased expression of integrins α2β1 and α5β1 (Zoppi et al, 2004), we explored the possibility of integrin switching as a driver of fibroblast activation in response to COL3 loss. Integrin subunit α11 (murine gene name Itga11) is a major collagen receptor on stromal cells that controls myofibroblast activation as well as remodeling of the surrounding matrix by myofibroblasts (Carracedo et al, 2010; Gullberg et al, 1995; Popova et al, 2007; Velling et al, 1999). Knockdown of Itga11 in mice limits myofibroblasts and their associated scar deposition. We thus hypothesized that knockdown of Col3a1 in fibroblasts would promote incorporation of the integrin α11 subunit into focal adhesions and drive myofibroblast activation. Indeed, we observed increased integrin α11 expression in unwounded skin and wounds (14, 28, and 35 DPW) of Col3^F/F mice compared with that in Col3^B6/B6 mice (Figure 7a and b) and increased integrin α11 protein produced in vitro in Col3^−/− fibroblasts compared with that in embryonic littermate-paired Col3^+/+ fibroblasts (Figure 7c). These data suggest that the loss of COL3 drives integrin switching toward the utilization of proscarring integrin α11.

Figure 7. COL3 loss drives myofibroblast activation by promoting incorporation of profibrotic integrin α11 into focal adhesions.

(a) Representative immunoreactivity of integrin α11 in wounds of Col3^B6/B6 and Col3^F/F mice at 35 DPW. Bars = 1 mm and 20 μm for the tile scans and insets, respectively. (b) Quantitated integrin α11 expression for wounded and unwounded dermis, n = 3–5 mice per genotype. (c) Representative western blots and densitometry of integrin α11 expression in n = 5 embryo-paired Col3^+/+ and Col3^−/− fibroblasts (**P < .01, paired t-test). (d) Immunofluorescent images of paxillin (green) and integrin α11 (red) in fibroblasts cultured on soft silicone (E ~ 5 kPa) and glass (E ~ 1 GPa). Bars = 5 μm. (e) Focal adhesion size and (f) colocalization of integrin α11 into focal adhesions were quantitated in fibroblasts cultured on substates of different stiffnesses. (g–i) Representative immunofluorescent images of αSMA and F-actin (g) in fibroblasts treated with siRNA against Itga11 (si-Itga11) and controls (si-Control) cultured on soft silicone (E ~ 5 kPa, bars = 20 μm). (h) si-Itga11 treatment effectively knocks down Itga11 expression in fibroblasts of both genotypes. (i) Quantitative analysis of αSMA (green) and F-actin (red) assessing myofibroblast activation in si-Itga11– and si-Control–treated fibroblasts. *P < .05, **P < .01, and ***P < .001 through 2-way ANOVA with Tukey’s multiple comparisons. αSMA, α-smooth muscle actin; COL3, type III collagen; DPW, days post-wounding; si-Control, control-targeted small interfering RNA; si-Itga11, Itga11-targeted small interfering RNA; siRNA, small interfering RNA.

To confirm that COL3 deficiency promotes the incorporation of integrin α11 into focal adhesions to engage the microenvironment, Col3^+/+ and Col3^−/− fibroblasts were cultured in vitro on soft and stiff substrates, and focal adhesion composition in the fibroblasts was assessed after 72 hours of culture (Figure 7d). Col3^+/+ fibroblasts on soft silicone (E ~ 5 kPa) exhibited a decrease in average focal adhesion size (as detected by area of paxillin+ immunocytochemical staining) compared with Col3^+/+ fibroblasts on glass or Col3^−/− fibroblasts cultured on either soft silicone or glass (Figure 7e). There was also an increase in colocalization of integrin α11 with paxillin+ focal adhesions in Col3^+/+ fibroblasts on glass and Col3^−/− on soft silicone and glass compared with that in Col3^+/+ fibroblasts on soft silicone (Figure 7f). We thus saw that stiffness and COL3 loss drive the utilization of integrin α11 during myofibroblast activation.

To determine whether myofibroblast activation in Col3^−/− fibroblasts was dependent on signaling through α11 integrin, Col3^+/+ and Col3^−/− fibroblasts were cultured on soft silicone for 72 hours after transfection with either a control-targeted small interfering RNA (si-Control) or a pool of small interfering RNAs (siRNAs) directed against Itga11, and the degree of myofibroblast activation was assessed by colocalization of αSMA and F-actin (Figure 7g). RT-qPCR for Itga11 confirmed that si-Control–treated Col3^−/− fibroblasts showed significantly higher Itga11 mRNA levels than si-Control–treated Col3^+/+ fibroblasts (Figure 7h), whereas both Itga11-targeted siRNA–treated Col3^+/+ and Col3^−/− fibroblasts showed persistent Itga11 knockdown. si-Control and Itga11-targeted siRNA Col3^+/+ fibroblasts showed no significant difference in the number of colocalized αSMA/F-actin pixels (Figure 7i), whereas si-Control Col3^−/− fibroblasts had significantly greater colocalization of αSMA/F-actin pixels, consistent with an activated myofibroblast phenotype. In support of our hypothesis, Itga11-targeted siRNA Col3^−/− fibroblasts had significantly lower colocalized αSMA/F-actin pixels than si-Control Col3^−/− fibroblasts. These data suggest that loss of COL3 drives myofibroblast activation in an integrin α11–dependent manner and that Col3^−/− myofibroblasts can be reverted to a more quiescent fibroblast phenotype by targeting Itga11.

DISCUSSION

The early temporal induction of COL3 during wound repair and the relative increased COL3 expression observed in mammalian scar-free wound healing models, such as the midgestational fetus (Cuttle et al, 2005) and the African Spiny Mouse (Acomys) (Brant et al, 2015; Harn et al, 2021), suggest an association between COL3 and increased regenerative potential after injury. However, COL3 has also been implicated as a profibrotic mediator due to its accumulation in fibrotic and desmoplastic processes (Nurmenniemi et al, 2012; Wang et al, 2016; Yang et al, 2022). Our study provides mechanistic insight into how COL3 promotes tissue regeneration and attenuates scar formation, with data demonstrating COL3 as a critical regulator of re-epithelialization, fibroblast activation, and matrix architecture during cutaneous wound healing. We demonstrate that COL3 regulates the evolution of the fibrillar collagen architecture during GT formation such that COL3-enriched microenvironments (Acomys) promote the reformation of a bi-isotropic basket-weave collagen architecture, whereas COL3-deficient microenvironments promote the formation of an aligned collagen matrix that is associated with scarring (Col3^F/F > Col3^B6/B6 ≅ Mus>Acomys) in a dose-dependent manner. Using in vitro and in vivo models of COL3 knockdown, we show that COL3-deficient environments promote assembly of a scarpermissive matrix characterized by increased collagen fiber length and alignment and impair wound closure through effects on re-epithelialization. Our data further show that although the loss of COL3 does not increase the stiffness of GT during healing, it does activate cellular mechanotransduction pathways resulting in fibroblast activation through a profibrotic integrin α11–dependent mechanism. Collectively, our data reveal that COL3 plays a critical role early in healing to mediate wound closure and suggest that its effects on scar prevention is regulated by a feed-forward process of dynamic reciprocity between GT fibroblasts and their surrounding collagen matrix.

Defining drivers of a proregenerative matrix has critical implications for the development of new regenerative therapies to reduce scarring and improve healing of problematic wounds. Despite the wealth of information regarding the cellular and molecular events that orchestrate wound healing and scar formation, both chronic wounds and pathologic scars remain significant clinical problems (Sen et al, 2009). Models of mammalian scarless healing, including the midgestational fetus (Leung et al, 2012) and Acomys (Seifert et al, 2012), have been used to identify the mechanisms that promote efficient wound healing while minimizing scarring. Our study shows that COL3 expression is increased in Acomys wounds relative to that in scar-forming Mus, agreeing with previous studies that have observed this relationship at the levels of gene and protein expressions (Brant et al, 2015; Harn et al, 2021). We further contextualize this observation by showing that whereas scarring Mus wounds have an initial increase in COL3 expression (similar to that in most other mammals), Acomys wounds exhibit both an increase in magnitude of induction as well as sustained duration of expression throughout the later stages of healing (Figure 2). In association with this increased COL3 expression, we show that Acomys wounds have shorter and fewer collagen fibers in the early remodeling phase of healing (14 DPW) (Figure 1) than Mus wounds as well as Acomys unwounded dermis and healed wounds (35 DPW). In contrast to collagen fibers in Mus wounds, which become progressively more aligned throughout healing, Acomys wounds maintain a bi-isotropically aligned matrix similar to that of unwounded dermis throughout healing. Recapitulating this COL3-enriched, bi-isotropically arranged matrix in tissue engineering strategies may be optimal for promoting dermal regeneration. Indeed, 1 study demonstrated superior fibroblast and keratinocyte migration as well as angiogenesis with fibrillar scaffolds arranged in a basket-weave–like pattern compared with aligned matrices in vitro and in vivo (Sun et al, 2018). It should be noted that Acomys fibroblasts maintain their scar-restrictive phenotype under in vitro conditions, suggesting cell-intrinsic properties (Brewer et al, 2021; Dill et al, 2023; Stewart et al, 2018). Acomys fibroblasts have been reported to show no significant change in functional myofibroblast activation in response to increased stiffness (Stewart et al, 2018) or through the addition of recombinant human TGFβ1 in culture (Brewer et al, 2021). In addition, Brewer et al (2021) demonstrated that Acomys fibroblasts have different adaptations in Hippo–YAP signaling that contribute to their regenerative capability in an ear-punch wound model. Although not examined in the context of this study, the production of COL3 by cultured Acomys fibroblasts may regulate both TGFβ1 responses (Brisson et al, 2022) as well as mechanotransduction as reported in this study in murine fibroblasts. The absolute requirement of COL3 in scarless repair of Acomys has yet to be determined; however, our study demonstrates that COL3 dose-dependently regulates matrix architecture and cell behavior that potentiates the switch between regenerative and scar-permissive responses after cutaneous injury and suggests that it at least plays a contributing role in the scarless phenotype in Acomys.

As cutaneous healing progresses, collagen regulates the biomechanical properties of GT, with current paradigms supporting that GT stiffness increases over time and that this process can perpetuate a feed-forward increase in myofibroblast activation and scar tissue deposition (Hinz et al, 2019). Given that increased collagen fiber alignment proportionally increases the bulk stiffness of collagen constructs (Sapudom et al, 2023; Seo et al, 2020) and our previous work that showed increased collagen deposition and myofibroblast activation during healing in Col3a1^+/− mice (Volk et al, 2011), we initially hypothesized that there would be an increase GT stiffness in our Col3^F/F mice relative to that in Col3^B6/B6 mice. Interestingly, although we saw a progressive stiffening of GT through 35 DPW in wild-type wounds (Figure 6e), we did not observe an increase in matrix stiffness during healing in our Col3^F/F model despite significant increases in SHG-detected fibril alignment and greater myofibroblast density (Figures 4 and 6, respectively). Patients with vascular EDS and models of COL3 deficiency have been reported to exhibit vascular rupture and tissue failure, with noted mechanically inferior unwounded adult tissue (Cooper et al, 2010; Liu et al, 1997). In addition, unwounded and healing dermis from type V collagen–haploinsufficent mice, used to model EDS type I, demonstrated lower tensile strength in association with increased collagen fibril diameter, decreased fibril number, and other microarchitectural abnormalities (as detected by transmission electron microscopy) in Col5a1^+/− mice than in wild-type mice (Wenstrup et al, 2006). Given that both COL3 and type V collagen have roles in regulating COL1 fibrillogenesis, loss of these regulatory collagens is likely to alter and disrupt collagen fiber assembly, thus resulting in mechanically inferior tissue as observed in patients with EDS. Differences between bulk and local mechanical properties may further explain our findings. The architectural complexity of biological tissues can complicate the interpretation of mechanical measurements owing to tissue-specific isotropic variance (Rubiano et al, 2019; Stewart et al, 2017). Our AFM nanoindentation methodology quantifies the modulus from a force applied perpendicular to the direction of the fibers, using a micro-spherical tip (radius ≈ 5 μm), which characterizes the relative tissue microstiffness that cells locally respond to within their milieu. Given the increased anisotropy observed with collagen fibers perpendicular to the neoepidermis in Col3^F/F GT, it is possible that mesoscale mechanical properties of COL3-deficient wounds may be stronger with an axial load applied in the direction of the fibers or with rheological characterization. Even if no differences in bulk or local mechanical properties in COL3-deficient tissues were to be observed with additional mechanical characterization, differences in matrix architecture may be influencing the way that cells experience their microenvironment, their mechanoperception, and myofibroblast activation independently of microstiffness. Seo et al (2020) demonstrated that myofibroblast activation and contractility can be directed by collagen architecture within 3-dimensional collagen lattices, independent of bulk construct stiffness. Given the dynamic reciprocity of cells and the ECM that orchestrates wound healing, further understanding of interplay between ECM architecture, local and bulk biomechanical properties, and stromal cell phenotype may be critical in our understanding of how to engineer and promote tissue regeneration.

Although changes in collagen architecture and expression can influence stromal cell phenotype (Sapudom et al, 2023; Seo et al, 2020), our study shows that the absence of Col3a1 expression alters their mechanoperception, such that both in vitro and in vivo cells within the healing microenvironment exhibit a mechanically active phenotype even in the absence of increased stiffness (Figures 5 and 6). We observed increased nuclear YAP translocation and larger focal adhesion assemblies (both indicative of a positive mechanoresponse) in vitro with our Col3^−/− fibroblasts even on soft, physiologically relevant (~5 kPa) substrates, in contrast to that in wild-type fibroblasts (Figures 5 and 7), and increased immunoreactivity of mechanosensitive proteins YAP, pFAK, and RhoA in wounds of Col3^F/F mice relative to that in wild-type littermates (Figure 6). The treatment of wild-type and COL3-deficient cells with verteporfin, a YAP inhibitor, reversed the activated myofibroblast phenotype in hyperphysiologically stiff microenvironments (Supplementary Figure S7), demonstrating that myofibroblast activation can be recovered by preventing a positive mechanoresponse (preventing nuclear YAP signaling). Furthermore, we show that this altered mechanoperception is due to an integrin switching mechanism, in which COL3-deficient cells upregulate and incorporate the profibrotic integrin α11 (Figure 7) into their focal adhesions. Differential integrin utilization and altered matrix assembly (Zoppi et al, 2004) have been reported in EDS resulting from mutations in either COL3A1 (vascular EDS type IV) or COL5A1 (EDS type I). Although a previous study revealed that loss of COL3 expression in fibroblasts from patients with vascular EDS resulted in decreased expression of integrin dimers α2β1 and α5β1 while increasing the profibrotic αvβ3 (Sarrazy et al, 2014; Zoppi et al, 2004), our COL3-deficient models demonstrate that the knockdown of the Col3a1 drives myofibroblast activation in an integrin α11–dependent manner. Although the overall expression of integrin α11 protein is higher in COL3-deficient wounds, in vitro data also suggest that Col3^−/− fibroblasts form larger focal adhesion complexes utilizing integrin α11 than Col3^+/+ fibroblasts. The development and maturation of focal adhesions may drive a positive mechanoresponse and is likely to further influence matrix deposition and remodeling by fibroblasts in a feed-forward–dependent manner. A better understanding of the mechanisms regulating integrin α11 utilization in vivo may provide important targets for influencing both wound healing and tumor microenvironments.

Although this work demonstrates COL3’s effects on matrix architecture and mechanotransduction during healing, COL3 may be influencing other aspects critical in promoting tissue regeneration rather than repair. The ability of COL3 to regulate paracrine signaling within healing wounds (Brisson et al, 2022) may further influence cell activities and fate in the wound microenvironment. Our group has previously shown that the N-terminal cysteine-rich propeptide of COL3 can bind selectively and sequester to TGFβ to attenuate but not eliminate its signaling (Brisson et al, 2022). Our data indeed support the presence of procollagen 3 in wounds 7 DPW (Supplementary Figure S3c), consistent with previous reports that the N-propeptide is retained to a greater extent in tissues compared with that in other fibrillar collagens (Wu et al, 2010). Given that we have previously demonstrated that the human alpha 11 integrin promoter is regulated by TGFβ in both SMAD- and SP1-dependent manner (Lu et al, 2010), regulation of physiologic and pathophysiologic responses to injury by both the COL3 N-propeptide and triple helical domains may contribute to the responses described. In addition, the COL3-enriched microenvironment in healing wounds of both the midgestational fetus and Acomys support reformation of adnexal structures. Although it has been shown that Acomys wounds support wound-induced hair follicle neogenesis throughout injured skin (Harn et al, 2021), the role that COL3 enrichment plays in this process remains undefined. Given the importance of matrix architecture in directing cell activities and fate in healing environments, this work supports COL3 as a driver of proregenerative matrix architecture formation and provides important information for biomaterial designs that reflect the spatiotemporal architectural differences observed in tissue/organ-specific wound healing.

It is important to acknowledge that this study focused on mechanoresponsive mechanisms through which COL3 regulates myofibroblast activation and persistence during cutaneous wound healing. Our findings do not rule out a similar role on other fibroblast populations, defined in recent studies as a heterogeneous population either by anatomic or by functional phenotype, within the GT (Guerrero-Juarez et al, 2019; Philippeos et al, 2018; Shook et al, 2018). A recent study using a thermal injury model showed that earlier upregulation of Col3a1 after injury is associated with decreased αSMA+ myofibroblasts and decreased collagen alignment within the wound area, which suggests that COL3’s regulation of fibroblast phenotype during healing may be conserved among other dermal fibroblast populations (Kuan et al, 2024). Given the development of aberrant matrix alignment in normotrophic, hypertrophic, and keloidal scars (Verhaegen et al, 2009), defining a role for COL3 in suppressing a proscarring phenotype has important clinical implications. It is notable that scar matrix architecture shares features of alignment while also exhibiting diverse higher-order organization. Although hypertrophic scars present a highly aligned, anisotropic collagen matrix (similar to that observed in our Col3^F/F model [Kelf et al, 2012]), keloidal scars are characterized by whirls of aligned, compacted collagen (Da Costa et al, 2008). Data support the ability of delivered COL3 to reduce scar formation and improve healing in the skin (Dong et al, 2023; Nuutila et al, 2015; Xu et al, 2022) and other tissues (Fagerholm et al, 2010; Merrett et al, 2008; Pupkaite et al, 2020; Walimbe et al, 2019; Wei et al, 2024). In fact, recent work in preclinical models suggests that COL3 will improve outcomes in patients with hypertrophic scars (Lin-Hui et al, 2024). Follow-up studies will also be required to determine whether COL3 impacts other aspects of healing such as inflammatory cell phenotype and fate during wound healing because an increasing number of studies has recognized the importance of myeloid cell mechanosignaling in regenerative responses during wound repair (Babaniamansour et al, 2024; Chen et al, 2022a, 2022b; Tschumperlin et al, 2018). Finally, our data reveal a positive impact of COL3 on re-epithelialization. In this study, we observe impaired re-epithelialization in our Col3^F/F mice compared with that in wild-type littermates (Figure 4 and Supplementary Figure S4), which is consistent with our previous data showing improved re-epithelialization in Col3^+/+ mice compared with that in COL3-haploinsufficient (Col3^−/−) littermates (Volk et al, 2011). In addition, our in vitro studies support that this improved epithelialization is secondary to direct responses of keratinocytes to COL3, with keratinocytes exhibiting improved directed migration when cultured on a COL3 substrate compared with that on a COL1 substrate. Finally, our data revealing COL3 localization beneath the neoepidermis in both Mus and regenerative Acomys wounds (Supplementary Figure S4) and data in human skin revealing the concentration of COL3 in the papillary dermis (Iriyama et al, 2022) provide further support for the direct interaction of COL3 with keratinocytes in reformation and maintenance of the epidermis in vivo. Whether these effects on re-epithelialization are mediated in part by a mechanodependent process or by additional mechanisms is under current investigation. Collectively, these data supporting a positive impact of COL3 on re-epithelialization and this and other reports supporting improved healing of the skin and in other tissues, with reduced scarring, suggest that COL3-inspired tissue engineering strategies have the potential to improve efficiency and quality of healing in both impaired/chronic wound healing and pathologic scars.

Our results demonstrate that COL3 plays a critical role in promoting regenerative healing after cutaneous injury. Collagen fiber architecture in COL3-enriched Acomys wounds supported scarless healing and was characterized by an architecture akin to unwounded skin, whereas COL3-deficient wounds were delayed in wound closure and resulted in increased scar formation characterized by a hyperaligned collagen architecture. Despite a lack of increase in GT microscale stiffness (measured by AFM nanoindentation used in this study), loss of COL3 also enhanced cellular mechanoperception and promoted myofibroblast activation in an integrin α11-dependent manner. Although our work suggests that COL3 regulates efficiency of healing and scar formation in large part through its effects on fibroblasts, COL3 deficiency likely influences regenerative potential in other aspects of wound healing. Our Col3-knockdown model suggests that COL3 deficiency impairs re-epithelialization, a defining feature of healed cutaneous wounds. Additional understanding of the spatiotemporal effects of COL3 and matrix architecture outcomes during healing may help to improve our understanding of tissue regeneration potential in mammals and refine tissue engineering approaches.

MATERIALS AND METHODS

Mice and excisional wound surgeries

An Acomys cahirinus breeding colony was established at the University of Florida. Mus musculus mice of the CD-1 outbred strain were purchased from Charles River. All experiments were conducted on groups of adult mice (aged 6 months to 1 year) of mixed sexes. The experiments were conducted in accordance with the Institutional Review Board on Animal Care at the University of Florida. Full-thickness excisional wounds (8-mm diameter) were made through the mid-dorsal skin, as previously described (Brant et al, 2016). At various times after wounding, the mice were euthanized, and the wound tissue with surrounding unwounded skin was excised for histology.

Conventional COL3-deficient mice used for embryo donors for fibroblast harvest were generated in a colony established at the University of Pennsylvania from breeder pairs of Col3a1 heterozygous (Col3^+/−) mice originally purchased from Jackson Laboratories (Volk et al, 2011). Procedures conducted were approved by the University of Pennsylvania Institutional Animal Care and Use Committee and guidelines set forth in the National Institutes of Health Guide for the Care and Use of Laboratory Animals.

Col3a1:LONG-cKO.B6/J mice were generated at the CRISPR/Cas9 Targeting Core at the University of Pennsylvania Perelman School of Medicine. Mutated Col3a1 sequences were generated with long single-strand DNA oligos to add loxP sites around the target gene sequences (Supplementary Figure S2a). Founders were initially backcrossed onto a C57BL/6J wild-type strain and then subsequently backcrossed for >8 generations with B6;129-Gt(ROSA)26Sor^{tm1(cre/ERT2)Tyj}/J mice (Jackson Laboratory) to yield mice with homozygous floxed or wild-type mice with homozygous Cre-ERT2:Col3a1^{Flox/Flox-ERT2/ERT2} (Col3^F/F, COL3 knockdown) or Col3a1^{B6/B6-ERT2/ERT2} (Col3^B6/B6, wild type). Genotype of mice was confirmed through qPCR using the primers listed in Supplementary Table S1. For wounding studies, temporal targeted knockdown of Col3a1 in Col3^F/F mice was accomplished by 3 daily intraperitoneal injections of tamoxifen (200 mg/kg, Sigma-Aldrich, T5648) beginning 6 days before wounding; Col3^B6/B6 littermates receiving this regimen served as controls. Overall weight of mice was not affected by genotype or short-term Col3a1 depletion induced by tamoxifen treatment performed for these studies Supplementary Figure S3b). Col3a1 knockdown in Col3^F/F dermis was confirmed by qPCR analysis of dermis samples collected prior to tamoxifen injection, at the day of surgery (6 days after first tamoxifen injection), and 35 DPW (Figure 3a and b). Western blotting (Supplementary Figure S3c) and immunohistologic analysis (Supplementary Figure S3d) provided confirmation of COL3 protein loss in tissues after tamoxifen treatment. All Col3a1:LONG-cKO.B6/J mice for this study were generated in a colony established at the University of Pennsylvania, and procedures conducted were approved by the University of Pennsylvania Institutional Animal Care and Use Committee and guidelines set forth in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Mice were implanted with a microchip (Allflex FDX-B transponders) at weaning and were subsequently identified by the month/year of the microchip implant and the last 4 digits of the microchip number to allow for blinding of genotype during wounding, tissue processing, and wound analysis. All mice were aged 8–12 weeks at the time of wound surgery.

Wounding survival surgeries were carried out under general anesthesia through inhaled isoflurane. All mice received preemptive analgesia (0.1 mg/kg buprenorphine SR, subcutaneously). The dorsum was shaved with electric clippers, depilated by Nair application, and scrubbed with 4% w/v chlorhexidine gluconate solution. Two full-thickness wounds through the panniculus carnosus were made on either side of the spine and caudal to the last rib using 6-mm punch biopsies (Integra). Wounds were then covered with a sterile Curad pad (Medline Industries), and a circumferential sterile occlusive dressing (Tegaderm, 3M) was applied. Murine wounds were photographed with an in-frame ruler 0, 3, 5, 7, 10, and 14 DPW to assess wound closure. Open wound area was measured using Adobe Photoshop and plotted as a function of time using the average of both wound areas for each mouse for statistical analysis (Volk et al, 2011). Wound samples were collected at the corresponding time points by collecting a margin of unwounded tissue surrounding the wound and prepared for corresponding analyses.

Histology and histomorphometric analysis

Wound samples were fixed in a glutaraldehyde-based fixative (Prefer, Anatech) overnight at 4 °C prior to subsequent processing for histological wound assessments: H&E morphometry, immunohistochemistry, and SHG imaging. Briefly, harvested wounds with surrounding unwounded skin were bisected, and paraffin-embedded sections (10 μm) were obtained from the 2 halves. Histomorphometry was performed on paraffin-embedded H&E sections that exhibited the greatest overall wound length, representing the central wound, were used to assess wound healing quantitatively. The epidermal gap and GT area, were measured 7 DPW; GT area was additionally measured 14, 28, and 35 DPW. As previously described (Volk et al, 2011), the epidermal gap was defined as the distance between the encroaching epidermal elements. GT area was defined as the cellular area from the lateral transition zone from the normal epidermis to the hypertrophic epidermis, superior to the level of the panniculus carnosus and inferior to the epithelial basement membrane or open wound surface. In addition, for AFM measurements, the wound edge region was defined as the area within ~250 μm on the medial side of the transition zone flanking both sides of the GT area excluding adnexal structures.

SHG imaging

SHG imaging of fibrillar collagen in FDMs, obtained from fibroblastic/ECM units, and unwounded and wounded murine dermis was performed on a Leica SP8 confocal/multiphoton microscope (Leica Microsystems) by tuning the Coherent Chameleon Ultra II Ti:Sapphire laser (Coherent) to 910 nm and collecting SHG signal on a nondescanned detector configured to capture wavelengths at 455 nm (×20 [1.0 numerical aperture] water immersion objective). Regions for analysis were taken within the GT area defined by histomorphometry using the representative H&E of the wound. Each image taken was 1024 × 1024 pixels (553.57 × 553.57 μm) at a scan speed of 400 Hz and a line average set to 3. Imaging parameters were kept identical between imaging sessions to allow for comparable image analysis quantification between groups. The autofluorescence was subtracted from the original SHG images as previously described (Case et al, 2017). Backward SHG images were used for characterizing fiber properties in CT FIRE (version 2.0) (Bredfeldt et al, 2014) (fiber length, width, density, and straightness) using the default parameters. Forward SHG images were used for analyzing fiber distributions with the ImageJ plugin OrientationJ as has been described (Franco-Barraza et al, 2020). Rose plots of fiber distribution frequencies were made in MATLAB R2022a (MathWorks).

Immunohistochemistry

Sections from fixed, paraffin-embedded tissues were mounted onto charged glass slides. After deparaffinization and rehydration, antigen retrieval for αSMA, YAP, pFAK, RhoA, and integrin α11 was performed by boiling in a citrate buffer (Biogenex, HK0809K) for 8 minutes. Antigen retrieval for COL1 and COL3 was done by incubating in proteinase K (20 μg/ml) at 37 °C for 20 minutes in Trisethylenediaminetetracetic acid. Tissue sections were blocked with 5% normal donkey serum in Odyssey Blocking Buffer (LI-COR) for 1 hour at room temperature (RT). Primary antibodies against mouse anti-αSMA (1:500, Sigma-Aldrich, A2457), mouse anti-YAP (1:500, Santa Cruz Biotechnology, sc-17140), rabbit anti-pFAK (1:100, Invitrogen, 44-624G; phosphorylation at residue Y397), mouse anti-RhoA (1:100, Santa Cruz Biotechnology, sc-418), sheep anti–integrin α11 (1:250, R&D Systems, AF6498), rabbit anti-COL1 (1:1000, Thermo Fisher Scientific, PA5-9513), and goat anti-COL3 (1:1000, Southern Biotech, 1330-01) were incubated at 4 °C overnight in Odyssey Blocking Buffer. Appropriate secondaries were diluted in Odyssey Blocking Buffer with a nuclear dye and incubated for 1 hour at RT: goat anti-rabbit F(ab⁺)-Alexa Fluor 488 (Thermo Fisher Scientific, A11070) for pFAK, donkey anti-mouse Alexa Fluor 488 (Invitrogen, A21202) for YAP, donkey anti-mouse Alexa Fluor+555 (Thermo Fisher Scientific, A32773) for RhoA and αSMA, donkey anti-sheep Alexa Fluor 633 (Thermo Fisher Scientific, A2100) for integrin α11, and donkey anti-goat Alexa Fluor 633 (Thermo Fisher Scientific, A21082) for COL3. Sytox Green (1:100,000, Invitrogen, S7020) or DRAQ5 (1:10,000, BioLegend, 424101) were used for nuclear visualization.

AFM-based nanoindentation

Wound samples for AFM measurements were excised and embedded in optical cutting temperature (unfixed) and flash frozen on dry ice. Samples were then sagittally bisected, placed on edge, and re-embedded and frozen in optical cutting temperature. Optical cutting temperature sections were cut at a thickness of 20 μm through Kawamoto’s film-assisted cryosectioning to minimize confounding residual substrate effects on the indentation measurements during AFM (Kawamoto and Kawamoto, 2021). The average indentation modulus for each wound sample was determined by interrogating at least 15 different locations within the GT and flanking unwounded dermis to account for spatial heterogeneity. AFM nanoindentation was performed on wound cryosections using spherical colloidal tips (R ≈ 5 mm, nominal k ≈ 0.03 N/m, HQ:CSC38/tipless/Cr-Au, Cantilever B, NanoAndMore) and a Dimension Icon AFM (Bruker Nano), following our established procedure (Li et al, 2017). At each location, an indentation force versus depth (F–D) curve was obtained at an indentation rate ≈ 10 mm/s up to ≈ 500 nm maximum indentation depth. The effective indentation modulus, E_ind, was calculated by fitting the entire loading portion of each F–D curve to the Hertz model (Batista et al, 2014), assuming the Poisson’s ratio v ≈ 0.48 for the skin (Li et al, 2012). Samples were immersed in 1 × PBS at RT for the duration of AFM indentations to simulate a physiologic environment.

Fibroblast cell culture

Mouse dermal embryonic fibroblasts (embryonic day 18.5) were harvested through explant cultures and genotyped as previously described (Brisson et al, 2022; Volk et al, 2011). All fibroblasts were used within 4 passages of isolation. Cells were cultured in a humidified incubator with 5% carbon dioxide/95% air with complete media: DMEM with Glutamax (Gibco) supplemented with 10% fetal bovine serum (Atlanta Biologicals), antibiotics (100 U/ml penicillin and 100 g/ml streptomycin, Gibco), and 100 μg/ml of L-ascorbic acid (Sigma-Aldrich, A8960). Cells were tested routinely to ensure the absence of mycoplasma contamination. Itga11-targeted siRNA experiments were performed through transfection with 2 nM ON-TARGETplus mouse Itga11 siRNA SMARTPool (catalog number L-041795-01-0005, Horizon Discovery). The SMARTPool siRNA contains a combination of 4 siRNA sequences targeting Itga11. Transfection was achieved using Lipofectamine RNAimax (Invitrogen). YAP inhibition was tested by treating fibroblasts with 0.5 μM verteporfin (Sigma-Aldrich, SML0534) or 0.1% DMSO (vehicle control) for 72 hours prior to fixation for immunocytochemistry or lysis for RNA collection (details are provided in corresponding sections below). Media with vehicle or verteporfin treatments was exchanged daily.

For stiffness-based experiments and silicone wrinkling assays, Sylgard 527 (E ~ 5 kPa, used for silicone wrinkling [Palchesko et al, 2012]) or Sylgard 184 (E ~ 1 MPa [Palchesko et al, 2012]) was fabricated as described previously in 35-mm Petri dishes (Brisson et al, 2022; Stewart et al, 2018). Substrata for fibroblast/ECM units were prepared following an established protocol (Franco-Barraza et al, 2020) prior to addition of 500,000 cells/35 mm dish and allowed to grow for 7 days. For silicone wrinkling assays, functionalized silicone was coated with 10 μg/ml human COL1 (Advanced Biomatrix, 5007) or COL3 (Advanced Biomatrix, 5021) and imaged through brightfield microscopy (Olympus BX51, ×10 objective) after 72 hours of culture.

Keratinocyte migration studies

Human primary epidermal keratinocytes were obtained from multiple foreskin donors from the University of Pennsylvania Skin Biology and Diseases Resource Center (for scratch assays) or CellNTec (for live-cell migration) and cultured in CnT-PR CnT-Prime (CellNTec) media in a humidified chamber at 37 °C and 5% carbon dioxide. For in vitro re-epithelialization scratch assays, human primary epidermal keratinocytes were seeded at confluency and scratched with a p10 micropipette tip to generate a wound area. Images were then taken after scratch and at time t = 6, 12, 18, and 24 hours after scratch. Open (nonepithelialized) area was measured at each time point in ImageJ and compared with the initial open area for each experiment. For live-cell migration (meandering) experiments, 12-well plates were coated with human recombinant COL1 or COL3 (1 μg/cm²) prior to placement of stainless-steel cylinders in the wells to create a cell-free circular zones and subsequent plating of human primary epidermal keratinocytes at confluency around the cylinder. The cylinders were then removed, and wells were imaged overnight using a ×10 objective on a Leica DMI4000 Microscope with a Yokagawa CSU-X1 confocal spinning disk (PennVet Imaging Core) within an environmental chamber to maintain 37 °C and 5% carbon dioxide. Velocity software was used to track individual cell movement at the edge of the cell-free zone. A total of 20 cells were tracked per treatment (5 cells in 4 replicate wells). Meandering data (directed migration, total migration/path length of a cell track) were calculated per cell.

Immunocytochemistry

Cell fixation for αSMA/F-actin analysis was performed as previously described (Brisson et al, 2022; Stewart et al, 2018). Mouse anti--αSMA (1:500, Sigma-Aldrich, A2547) was incubated in 5% BSA (Thermo Fisher Scientific) overnight at 4 °C. Donkey anti-mouse Alexa Fluor 488 (1:500, Invitrogen, A21202), Alexa Fluor 594–conjugated phalloidin (1:500, Invitrogen, A12381), and DRAQ5 (1:10,000) were incubated for 1 hour at RT the next day. Cells for other immunocytochemical staining were fixed in 3% paraformaldehyde for 30 minutes, permeabilized in 0.1% Triton-X 100, and blocked in 5% BSA for 1 hour prior to incubation with the following antibodies: mouse anti-YAP (1:500, Santa Cruz Biotechnology, sc-17140), rabbit anti-paxillin (1:500, Abcam, ab32084), and/or sheep anti-Integrin α11 (1:500, R&D Systems, AF6498) at 4 °C overnight. Corresponding secondaries were added the next day for 1 hour at RT. Fluorescent imaging was performed with a Leica SP-8 confocal microscope. CellProfiler was used to quantify the amount of positive YAP immunostaining within the nucleus (identified by DRAQ5) relative to that of the total positive YAP immunostaining of the nucleus and cytoplasm. Colocalization analyses of paxillin and integrin α11–positive pixels as well as αSMA+/F-actin+ stress fibers were performed in MATLAB (MathWorks).

Western blotting

In vivo wound samples (7 DPW) from Col3^B6/B6 and Col3^F/F mice were flash frozen and homogenized in RIPA buffer (Cell Signaling Technology, 9806) with 1:100 protease inhibitor (Sigma-Aldrich, P-8340) and 1:100 phosphatase inhibitor cocktail (Sigma-Aldrich, P-5726). RIPA-solubilized fractions and the insoluble pellet, which was subsequently processed and solubilized with urea, were used to assess total COL3 (soluble and insoluble fractions) (Supplementary Figure S3c and d, respectively). Secondary extraction with a urea-based buffer was performed by resuspending the pellets left over from the RIPA extraction in 6 M guanidine hydrogen chloride (Sigma-Aldrich, G3272) with the same protease and phosphatase inhibitor cocktail listed earlier and shaken overnight at 4 °C. Guanidine-extracted samples were then ethanol precipitated with 9 equivalent volumes of ice-cold 100% ethanol, incubated at −20 °C overnight, and separated by microcentrifugation at maximum speed for 20 minutes at 4 °C; resulting pellets were then washed with 90% ethanol and microcentrifuged again for 5 minutes at 4 C. Ethanol was then aspirated, and resulting pellets were air dried for 20 minutes at RT and subsequently resuspended in 6 M urea (Fisher, U15-500). Protein concentrations of pellets for both extractions were measured using a BCA kit (Pierce, 23227). All lysates were resuspended in 4X LDS sample buffer (Invitrogen, NP0007), and 5% β-mercaptoethanol was added to the lysate prior to boiling at 95°C for 5 minutes. A total of 15 μg of each sample was loaded on a 4–12% bis-tris gel (Invitrogen, NP-0321) and run at 120 V for 3 hours in MOPS buffer (Invitrogen, NP-0001). Protein bands were transferred to polyvinylidene difluoride membrane at 35 V overnight, and membranes were subsequently stained with a Ponceau stain (Cell Signaling Technology, 59803S), following manufacturer’s instructions. Western blots for COL3 were blocked with 5% milk + 2% BSA in tris-buffered saline (TBS) + 0.2% Tween-20 for 1 hour at RT. COL3 antibody (Southern Biotech, 1330-01) was diluted in blocking solution at 1:500 and incubated overnight. Goat horse-radish peroxidase (Abcam, ab205718) was used at 1:100,000 dilution in blocking solution for 1 hour at RT. Blots were treated with Pierce Supersignal femto substrate (Pierce, 34094), following manufacturer instructions, and developed using autoradiography film in a Konica SRX101A film processor. For loading controls (GAPDH for RIPA-extracted fractions and β-actin for guanidine extracted), blots were blocked in 5% milk + 0.2% Tween in TBS, GAPDH antibody (Cell Signaling Technology, 5174, 1:2000 dilution) or β-actin (Cell Signaling Technology, 4967, 1:500 dilution) was incubated in 5% BSA + 0.2% Tween in TBS overnight. Rabbit horseradish peroxidase (Cell Signaling Technology, 7074) was used at 1:2000 dilution in 5% milk +0.2% Tween in TBS for 1 hour RT and developed using Pierce ECL Substrate (Pierce 32106), following manufacturer instructions.

Quantitative real-time PCR

Genotyping was confirmed through qPCR from 3-mm tail snips taken from pups at postnatal day 21. Samples were digested in 50 mM sodium hydroxide and mixed with Promega GoTaq Green (M7122) with corresponding forward and reverse primers prior to running PCR (Supplementary Table S1). Samples were then run on 2% Tris acetate EDTA agarose gels and imaged on a gel doc imager (Bio-Rad Laboratory). For Col3a1-knockdown confirmation, shaved 4-mm skin biopsies were flash frozen on dry ice prior to being ground with a mortar and pestle in liquid nitrogen. In vitro fibroblast RNA was collected through cell lysis in RLT buffer (Qiagen). RNA extraction for all samples was performed using a Qiagen RNeasy Kit, per the manufacturer’s instructions, and RNA concentration was determined through Nanodrop (Thermo Fisher Scientific). A total of 1 μg of RNA was used to generate cDNA using Superscript III (Invitrogen, 18080093) prior to being diluted 10 times with nuclease-free water. cDNA was then mixed with SYBR green master mix (Applied Biosystems, 4309155) with the primer sequences listed in Supplementary Table S2. Fold change was subsequently determined using delta delta CT method (fold change = 2^ΔΔCt) with gene expression normalized to Gapdh.

Statistics and data analysis

For all figures, opaque dots represent the average value from each mouse/wound or in vitro murine dermal embryonic fibroblast pair, whereas small, transparent dots represent the individual image analysis regions or technical replicates from each experiment. Overall averages for each mouse/wound/murine dermal embryonic fibroblast pair (opaque dots) were used for statistical analysis. All values and error bars represent mean ± SD except for rose plots in Figures 1g, 3k, and 4k and Supplementary Figure S1f, which depict mean (solid line) ± SEM (shaded area). Unpaired or paired 2-tailed student’s t-tests (Figures 4g and 6b, respectively) were used in comparing 2 sample groups. Two-way ANOVAs were employed for comparing 3 or more groups, followed by Tukey’s multiple comparisons. For 2-way ANOVAs performed in Figures 1, 2, 4, 6, and 7b, biologically relevant comparisons (intraspecies between different times and interspecies at a given time point) are shown for presentation purposes, and full 2-way ANOVA results with all multiple comparison P-values can be found in corresponding supplementary tables listed in the figure caption. Figures 3, 5, and 7e–i have all statistically significant differences contained within the figure. Statistical significance bars are color coordinated to highlight intraspecies/intragenotype comparisons (same color as data) or interspecies/intergenotype comparisons (shown in red). All data and statistics were performed and graphed in GraphPad Prism 9. P < .05 was considered statistically significant. Full statistical analyses are presented in Supplementary Tables S3–18.

Supplementary Material

Supplementary material is linked to the online version of the paper at www.jidonline.org, and at https://doi.org/10.1016/j.jid.2024.08.013.

Abbreviations

αSMA

α-smooth muscle actin

AFM

atomic force microscopy

COL1

type I collagen

COL3

type III collagen

DPW

days post-wounding

ECM

extracellular matrix

EDS

Elhers-Danlos syndrome

FDMs

fibroblast-derived matrices

GT

granulation tissue

pFAK

phosphorylated focal adhesion kinase

RT

room temperature

SHG

second harmonic generation

si-Control

control-targeted small interfering RNA

siRNA

small interfering RNA

DATA AVAILABILITY STATEMENT

All data for this study are contained within the manuscript and supplementary figures and tables or are available by reasonable request from the corresponding author. No large datasets were generated or analyzed as part of this research.