Disrupting biological sensors of force promotes tissue regeneration in large organisms

Inhibiting FAK in large mammals allows tissue regeneration

To evaluate the effects of blocking mechanotransduction on tissue repair in large animals, we employed a pharmacologic inhibitor of FAK (FAKI) (VS-6062) in red duroc pigs. VS-6062 (formerly known as PF-562,271) has been well characterized as having high specificity for FAK over a wide panel of other kinases^22,23. We first measured local and systemic toxicity and observed that neither topical administration of FAKI solution nor subcutaneous implantation of FAKI hydrogels resulted in any adverse reactions in unwounded porcine skin (Supplementary Fig. 1a, b). Furthermore, porcine serum FAKI concentrations following local treatment were almost undetectable and less than 1% of the maximum tolerated human dose observed in a previous Phase 1 clinical trial (Supplementary Fig. 1c)²⁴.

FAKI was delivered to partial-thickness excisional wounds on the dorsum of red Duroc pigs in a controlled manner, using a biodegradable and biocompatible pullulan-collagen-based hydrogel scaffold (Fig. 1a-c; Supplementary Fig. 1d)¹¹. We found that wounds treated with FAKI hydrogels (W_HF) had fully healed at postoperative day (POD) 14 ± 2.3, more than 10 days earlier than wounds treated with standard dressings (W) or empty hydrogels (W_H) (****p < 0.0001) (Fig. 1d, e). Importantly, this pharmacologic blockade of mechanical signaling resulted in skin that had less scar formation with more hair upon gross inspection, at PODs 40, 60, and 90 (Fig. 1f; Supplementary Fig. 1e). Using a tissue cutometer, we demonstrated that FAKI-treated wounds were less firm and more elastic than untreated wounds and exhibited more similar biomechanical properties to unwounded skin (Fig. 1g).

Wound tissue treated with FAKI exhibited dramatic regrowth of hair follicles and subcutaneous glands (sweat and sebaceous) (Fig. 1h), as well as newly regenerated peri-follicular adipose tissue, demonstrated by immunofluorescent staining for the adipocyte marker perilipin A (Fig. 1i). In contrast, control wounds failed to regenerate secondary structures and instead exhibited increased collagen deposition, fibrosis (Fig. 1h), and an increased number of alpha smooth muscle actin (αSMA) expressing myofibroblasts (Fig. 1j), a key cell that is known to drive tissue fibrosis and scar formation^1,2.

We performed a detailed quantitative assessment of the collagen architecture of the wounds at POD90, using the software algorithms CT-FIRE, CurveAlign, and MatFiber, which have all been previously developed to analyze collagen fiber properties in histologic images^25,26,27. Control wounds were found to exhibit a significant disruption of dermal architecture across these metrics, with collagen fiber elongation and increased unidirectional alignment²⁸ (Fig. 2a). Wounds with pharmacological blockade of FAK, by contrast, healed with a basket weave-like collagen fiber network similar to unwounded skin across a wide range of metrics (Fig. 2a). Specifically, FAK-inhibited wounds and unwounded skin both demonstrated decreased alignment, fiber length, angle kurtosis, and box density, while also both having an increased number of shorter collagen fibers (*p < 0.05) (Fig. 2b). Utilizing all 24 collagen structural parameters from these analyses (Supplementary Fig. 2), we performed a principal component analysis (PCA) and found that the first principal component (PC1) distinctly separated the fibrotic wounds (W, W_H) from regenerative healing (W_HF, UW) (Fig. 2c). We can interpret PC1 as an axis of fibrosis-regeneration that quantifies the significant collagen structural differences between fibrosis and regeneration across these 24 parameters.

Manipulating mechanical forces modulates fibrotic behavior

To understand how disruption of mechanotransduction may be relevant to human healing, we investigated the behavior of human fibroblasts in response to changes in tissue force. We employed a three-dimensional (3D) culture system that permits the precise manipulation of mechanical strain (and therefore stress) applied to cells (Fig. 3a-c)²⁶. Fibroblasts were isolated from human tissue samples, seeded within 3D collagen scaffolds, and either restrained (no strain, NS) or subjected to 10% strain (Strain, S) or 10% strain with FAK inhibition (Strain + FAKI, S + FAKI). We have previously demonstrated that fibroblasts cultured in this system display physiologic morphology and actin/stress fiber machinery matching fibroblasts in mechanically stressed in vivo scar environments^26,29. To further demonstrate that our culture system accurately recreates the in vivo wound environment, we also observed increases in αSMA+ myofibroblasts induced by mechanical strain (Fig. 3d), matching observations seen in vivo (Fig. 1j). When cultured in a uniaxial strain environment²⁶, these fibroblasts demonstrated elongated, unidirectional cellular alignment similar to the highly aligned fibrotic scar observed in human and porcine scars^28,30 (Fig. 3e), while fibroblasts blocked from sensing mechanical forces demonstrated a multi-directional organization similar to native skin architecture²⁸ (Fig. 2a, b).

Fibroblasts play a critical role in reorganizing extracellular matrix (ECM) by depositing and remodeling collagen to develop long, aligned fibers³¹. Using a contraction assay, we observed that fibroblasts lost their ability to remodel the surrounding ECM environment upon FAKI treatment (Fig. 3f)¹³. This attenuation of remodeling was also correlated with a decrease in YAP expression, a downstream transcription factor of the FAK pathway (Fig. 3g)^32,33. These data demonstrate that our 3D human culture system provides a physiological mechanical environment that induces fibroblast phenotypes consistent with previous findings on αSMA and YAP mechanotransduction expression during collagen remodeling and deposition^11,13,32. During the process of wound healing, increased mechanical forces trigger activation of FAK and increase integrin-ECM connections, which in turn promote stabilization of the f-actin cytoskeleton, αSMA expression, and cellular tension³⁴. αSMA stabilization promotes expression of transnuclear proteins, such as YAP, which in turn translocates into the nucleus to activate a cascade of profibrotic and mechanotransduction signaling³³. Since our 3D culture system accurately recreates the in vivo mechanical environment by inducing fibrotic myofibroblast phenotypes, we then used the system to examine specific cellular signaling pathways governing regeneration or fibrosis in human fibroblasts.

Mechanotransduction inhibition induces regenerative programs

We have previously shown that targeting mechanotransduction with FAKI in a murine model of hypertrophic scar (HTS) disrupts the FAK-ERK-MCP-1 pathway and reduces the expression of those specific fibrotic and inflammatory signals¹³. However, these previous small animal studies were limited in scope and unable to fully examine the plethora of cellular signaling pathways, which are altered in response to mechanical stress disruption. Recent advances in single-cell transcriptomics have increased our ability to explore heterogeneous cellular responses, such as those associated with modulation of mechanical forces³⁵.

To examine the molecular drivers of fibrosis and regeneration in human fibroblasts, we used single-cell RNA sequencing (scRNA-seq) in combination with our 3D collagen scaffold system (Fig. 4a). Fibroblasts were isolated from a variety of anatomic locations, and each patient’s fibroblasts were separately subjected to the previously mentioned NS, S, and S + FAKI conditions. After 48 h, the collagen scaffolds were enzymatically digested and scRNA-seq was performed on the cells using the 10× Genomics Chromium platform³⁵. Data for individual cells were subjected to Louvain-based clustering and embedded into a two-dimensional UMAP space in a manner blinded to the phenotype of origin³⁶. Overall, we found that mechanical strain shifted fibroblast transcriptional programs away from the NS cell states (Fig. 4b). Conversely, FAKI treatment pushed the cells to a new transcriptional metastate, and this shift was present even when analyzing each set of patient’s cells individually (Fig. 4b, top three plots). These findings highlight the robustness of FAKI treatment across three separate tissue donor sites collected on three different days.

UMAP-based clustering identified nine transcriptionally distinct fibroblast clusters (clusters 0–8) in the pooled dataset (Fig. 4c). Unstrained fibroblasts were found to aggregate together as a relatively homogeneous group, representing the majority of cells in our putative cluster 0. These cells, defined primarily by consistent expression of fibroblast housekeeping genes such as RND3, likely represent the native fibroblast steady-state in our system (Fig. 4c, d)³⁷. By contrast, mechanical strain strongly altered transcriptomic profiles and considerably increased fibroblast transcriptional heterogeneity. Strained fibroblasts differentiated among five different heterogeneous clusters, delineated by clusters 2, 4, 5, 7, and 8 (Fig. 4b, c). Overall, all of these strained fibroblasts were defined by collagen production (COL1A1, COL3A1), myofibroblast differentiation (POSTN, ACTA2 [encoding αSMA], PDGFRA, RUNX1, ZEB2)^34,38,39, and genes related to mechanotransduction PTK2 (encoding FAK) and its downstream effector YAP1 (Fig. 4f, Supplementary Fig. 3a-c), recapitulating our findings identifying αSMA and YAP upregulation both in vivo (Fig. 1g) and in vitro (Fig. 3d, g). We also observed overall upregulation of EN1 (encoding engrailed-1) and PRRX1, which our group has previously identified as hallmarks of a profibrotic fibroblast lineage in small animal models (Fig. 4f; Supplementary Fig. 3a-c)^40,41.

Next, we used Genetrail3, a computational pipeline for over-representation analysis (ORA) of specific genesets on a single-cell level, to further investigate differential regulation of cellular signaling pathways among human fibroblast groups⁴². We found that mechanically strained fibroblasts exhibited a significant induction of genesets for focal adhesion (WP306), MAPK6/MAPK4 signaling (WP3307), positive regulation of actin filament binding assembly (GO-BP:0032233), TGF-beta signaling pathway (KEGG:M2642), and EGF/EGFR signaling pathway (WP437) (Fig. 4g; Supplementary Fig. 3d), with an enrichment of the genes EGFR, MAPK1, ROCK1, and RAC1 (Supplementary Fig. 3a, c)^{43,44,45,46,47}. Furthermore, strained fibroblasts showed a significant enrichment of activated immune response pathways (GO-BP:0002253) (Fig. 4g) and positive regulation of the immune response (GO-BO:0002218) (Supplementary Fig. 4b), corresponding to upregulation of inflammatory genes such as CCL2 (Supplementary Fig. 3a). We have previously shown that MCP-1 (protein form of CCL2) is an inflammatory cytokine upregulated by fibroblasts in response to mechanical activation during murine fibrosis¹³. These findings further demonstrate that mechanical strain promotes inflammation by upregulating fibroblast MCP-1 expression, contributing to fibro-proliferation that attracts inflammatory cells to the wound site and further aggravates inflammation and fibrosis.

The strained fibroblast clusters also expressed considerable heterogeneity, representing several differentiation fates that each contribute to the development of fibrosis. Cluster 2 cells upregulated CKS2, which increases cellular proliferation and metabolic activity⁴⁸, and demonstrated an enrichment for cell division pathways (GO-BP:0051301), signifying a highly proliferative state (Supplementary Fig. 4b). Especially mechanosensitive clusters 4, 5, and 7 demonstrated upregulation of pathways for the alteration of the YAP1/ECM axis (WP3967) (Supplementary Fig. 4b), corresponding with increased activation of myofibroblast differentiation gene POSTN (Fig. 4f). Clusters 5 and 8 represented especially fibrotic clusters, with an enrichment for fibrosis pathways (WP3624) (Supplementary Fig. 4b) and upregulation of HES1 and SOX4 (Fig. 4d-e; Supplementary Fig. 3a-c). HES1 is a downstream effector of NOTCH3 and has been related to a variety of human fibrotic diseases⁴⁹, and SOX4 expression has been tightly linked to various profibrotic factors, such as TGFB, Wnt, and NOTCH^50,51,52. Clusters 2 and 5 also both demonstrated positive regulation of the immune response (GO-BP:0002218), indicating proinflammatory phenotypes. Within each cluster, individual cells also each exhibited heterogeneity when expressing key genes (Supplementary Fig. 4a). From this analysis, cluster 2 could represent an early proliferative and proinflammatory fibroblast state, clusters 4 and 7 could represent mechanoresponsive fibroblast states that drive differentiation into myofibroblasts, and clusters 5 and 8 could represent fibroblasts found in chronic, late-stage fibrotic conditions.

Suppression of mechanical signaling by FAKI in strained fibroblasts abrogated nearly all of these transcriptomic signatures, effectively blocking fibroblast differentiation into profibrotic and proinflammatory myofibroblast clusters. Instead, we found that disruption of FAK signaling shifted fibroblasts toward a more homogeneous metastate (Fig. 4b, c). We observed that blocking mechanical signaling strongly reduced the transcription of collagen encoding genes, such as COL1A1 and COL3A1 (Fig. 4e, f), and strongly induced the expression of the matrix-metalloproteinase (MMP) genes MMP1, MMP3, and MMP10 (Fig. 4e, f, Supplementary Fig. 3a-c). MMPs reduce fibrosis across a wide range of disease models through collagen degradation^53,54,55, and also promote cellular migration and re-epithelialization^53,56. Finally, FAK inhibition induced the expression of the antifibrotic gene STC1 (stanniocalcin-1), which has also been shown to promote wound healing and re-epithelialization^57,58.

We observed a significant reduction of the MAPK-ERK mechanotransduction pathway but preservation of AKT after blocking mechanical signaling with FAK inhibition (Fig. 4e, f, Supplementary Fig. 3a-c), consistent with prior findings^11,13. Differential expression analysis revealed that FAK-inhibited fibroblasts upregulated EGR1 (encoding early growth response protein 1) and MFGE8 (encoding milk fat globule-EGF factor 8 protein or lactadherin) expression, which are both mediated by AKT signaling. Recent studies have found that the AKT-EGR1 pathway has been specifically associated with regenerative phenotypes that regulate tumor suppression^59,60,61, while MFGE8 mitigates scar by tagging collagen molecules for phagocytosis⁶². MFGE8 also promotes regeneration by positively regulating vascular endothelial growth factor (VEGF) expression and angiogenesis through AKT phosphorylation⁶³. FAK inhibition consistently promoted these regenerative transcriptional profiles, even when analyzing each set of patient cells individually (Supplementary Fig. 5a, b). Furthermore, we found that even cells experiencing a basal level of physiologic strain responded to FAK inhibition (No Strain + FAKI group) with similar shifts in gene expression by decreasing profibrotic signaling and increasing regenerative transcription (Supplementary Fig. 3e).

Genetrail3 analysis of FAK-inhibited fibroblasts demonstrated a significant induction of genesets for ECM disassembly (GO-BP:0022617), activation of MMPs (WP2769), and collagen catabolic process (GO-BP:0030574), consistent with an antifibrotic phenotype (Fig. 4g). Moreover, FAK inhibition strongly induced beneficial transcriptional genesets for cellular detoxification (GO-BP:1990748), epithelial cell migration (GO-BP:0010634), cellular homeostasis (GO-BP:0019725), cell redox homeostasis (GO-BP:0045454), and collagen catabolic processes (GO-BP:0030574) (Fig. 4g; Supplementary Fig. 3d; Supplementary Fig. 4b). Cluster 3 fibroblasts showed a specific enrichment for pathways related to tissue development (GO-BP:0009888) and adipogenesis (WP236), demonstrating that this cluster of cells could potentially have the highest regenerative potential (Supplementary Fig. 4b). Based on these findings, we postulated that FAK inhibition promotes collagen degradation and reduces profibrotic fibroblast phenotypes by inhibiting a wide range of mechanotransduction pathways, such as MAPK-ERK and YAP, while preserving the AKT pathway to induce regenerative phenotypes through EGR1 and MFGE8.

Mechano-modulation of two opposing fibroblast trajectories

Traditional differential expression analysis (Fig. 4) only provides a snapshot of mRNA expression. We therefore employed RNA velocity analysis using the scVelo package to explore the comparative abundance of spliced and unspliced pre-mRNA transcripts in fibroblast clusters⁶⁴. scVelo uses a dynamical likelihood-based model, which identifies velocity states and transcriptional dynamics of each individual cell in an unbiased manner (Supplementary Fig. 6a)^64,65. Two opposing trajectories of fibroblast differentiation were identified in both the pooled and individual patient datasets, triggered either by the activation of mechanotransduction pathways or the disruption of mechanical signaling by FAK inhibition (Fig. 5a, Supplementary Fig. 7a). We found that FAK inhibition strongly increased the transcriptional activity of mechanically activated fibroblasts, resulting in a higher proportion of unspliced pre-mRNA, which accounted for 60% of all mRNA transcripts versus 30% in control and strained cells (Fig. 5b). To quantify the relationship between fibroblast clusters resulting from either mechanical activation or disruption of mechanotransduction, we applied partition-based graph abstraction (PAGA) informed by velocity-inferred directionality to quantify the relationship between fibroblast clusters resulting from either mechanical activation or disruption of mechanotransduction (Fig. 5c)⁶⁶. The fibroblasts of the control group (cluster 0) were identified as the origin of the underlying Markov transition matrix, confirming their identity as root cells of fibroblast differentiation (Fig. 5d). Partition-based graph abstraction (PAGA) identified trajectory vectors pointing from the control fibroblasts toward either activated profibrotic clusters in response to mechanical strain (2, 4, 5, 7, 8) or regenerative clusters in response to mechanotransduction blockade (1, 3, 6) (Fig. 5c).

Using RNA velocity analysis, we identified several genes with differential proportions of unspliced to spliced mRNA among the treatment groups. Specifically, EGR1 showed a high proportion of unspliced pre-mRNA in FAK-inhibited cells, providing further evidence for its role as a driver gene of regenerative fibroblast clusters (Fig. 5e). Additionally, MDM2, a gene regulated by AKT expression that regulates apoptosis, showed similar kinetics and further demonstrates how AKT-reliant genes persist within the FAK-inhibited fibroblast subset (Supplementary Fig. 4c)⁶⁷. RNA velocity analysis also revealed an enrichment of pre-mRNA for PIEZO1 and CAPZA2 in FAK-inhibited cells. PIEZO1 has been recently implicated in regulating mechanotransduction in macrophages⁶⁸, and PIEZO1 activation in fibroblasts may also mitigate fibroblast migration into the wound and contribute to the antifibrotic effects observed in response to FAK inhibition⁶⁹. CAPZA2 is an f-actin capping complex that binds to the barbed ends of actin filaments, preventing further addition of actin monomers. This reduces subsequent actin polymerization and may limit the ability to form a functional contractile myofibroblast phenotype in FAK-inhibited fibroblasts⁷⁰. We also compared the relative differentiation states of individual cells based on the distribution of unique mRNA transcripts using CytoTRACE⁷¹. Mechanically stimulated fibroblasts appeared less differentiated compared to other cells due to a high number of uniquely expressed mRNA features, further demonstrating how strain initiates transcription of a wide range of unique profibrotic gene expression profiles in fibroblasts (Fig. 5f; Supplementary Fig. 7b).

To further understand the transcriptional shifts observed in our single-cell data, we constructed pseudotime trajectories⁷². Using the locus of origin identified by RNA velocity analysis (Fig. 5c, d), we found significant pseudotemporal divergence between mechanically strained, untreated fibroblasts and FAK-inhibited fibroblasts (Fig. 5g center; Supplementary Fig. 8a, b). These findings demonstrate that mechanotransduction alters fibroblast programming, and that blocking mechano-signaling in strained fibroblasts alters the resulting cellular program in a way that is transcriptionally similar to the native fibroblast base state of control cells. Using pseudotime analysis, we identified a regenerative axis leading from the control cells (origin, O) to the FAK-inhibited cell clusters (regeneration, R), and a fibrotic axis leading in the opposite direction toward the mechanically strained, untreated fibroblast clusters (fibrosis, F) (Fig. 5g, center; Supplementary Fig. 4b). These diverging trajectories were also observed with RNA velocity analysis and PAGA (Fig. 5a, c), with vectors radiating out of cluster 0 into either regenerative or fibrotic clusters. Along the fibrotic axis, we observed a transcriptional increase in the previously identified genes COL1A1, PDGFRA, POSTN, RUNX1, and EN1, which are hallmarks of myofibroblast proliferation, mechanotransduction, and collagen production (Fig. 5g; Supplementary Fig. 8a, b). This trajectory was also characterized by an induction of genesets for cell metabolism and regulation of DNA binding (GO:0051101), demonstrating an increasingly active and proliferative process (Fig. 5g, bottom right). In contrast, the trajectory from control to the FAK-inhibited cells (regenerative axis) exhibits a preservation of the AKT pathway (shown by AKT1 expression), along with increased EGR1, MFGE8, MMP1, and PIEZO1 expression. In contrast, myofibroblast markers, such as PRRX1, POSTN, RUNX1, EN1, and PDGFRA, clearly decreased along this axis (Fig. 5g; Supplementary Fig. 8a, b). Pathways, such as mRNA processing (WP411), RNA splicing (GO:0000375), and regulation of mRNA processing (GO:0050684), decreased along the regenerative trajectory (Fig. 5g, bottom left), correlating with our RNA velocity analysis. In summary, we utilized advanced bioinformatic tools to identify two diverging transcriptional trajectories, either toward fibrosis or toward regeneration. Along these opposing trajectories, we observed stark differences in transcriptional profiles that either defined fibrotic or regenerative fibroblast phenotypes.

Porcine and human confirmation of diverging trajectories

In order to confirm these diverging proregenerative and profibrotic differentiation trajectories as well as correlate human and large animal phenotypes (Figs. 1, 2), we first performed immunofluorescent staining of tissue sections from porcine wounds for the key regenerative and fibrotic markers identified in our human fibroblasts. FAK inhibition significantly blocked YAP expression at POD7 after injury (**p < 0.01) (Fig. 6a, b), while also promoting expression of top regenerative markers EGR1, MFGE8, and MMP1 (*p < 0.05) (Fig. 6a, b). These observations persisted at POD90 for both the fibrotic marker YAP and the regenerative marker EGR1, highlighting the long-lasting effects of alterations in mechanotransduction upon fibroblast phenotypes (Supplementary Fig. 9a, b). Expression of MMP1 and MFGE8 normalized by POD90, demonstrating that key ECM remodeling occurs early and tapers off by late timepoints (Supplementary Fig. 9c, d).

We then repeated our human fibroblast experiments to confirm these key markers on both the mRNA and protein level using qPCR and western blotting (Fig. 6c, d; Supplementary Fig. 10a). First, we confirmed that mechanical strain indeed induced fibrotic YAP mRNA and protein expression (Fig. 6c, d; Supplementary Fig. 10a). Mechanically strained fibroblasts treated with YAP1 small interfering RNA (siRNA) promoted EGR1 expression and also decreased expression of COL3A1 and POSTN (Supplementary Fig. 10b). These experiments demonstrated that YAP indeed acts as a master regulator of profibrotic differentiation, and that silencing YAP expression could, like FAK inhibition, promote regenerative EGR1 expression while downregulating collagen and myofibroblast markers.

We also confirmed on the mRNA and protein levels that FAK inhibition increased EGR1, MFGE8, and MMP1 expression, while decreasing YAP and preserving AKT expression (Fig. 6c, d; Supplementary Fig. 10a). Silencing EGR1 using siRNA, however, negated the beneficial effects of FAK inhibition by instead promoting profibrotic YAP1 and COL3A1 expression (Supplementary Fig. 10c). These findings confirmed that disruption of mechanotransduction unlocks a regenerative differentiation path driven by the master transcription factor EGR1.

Overall, these data strongly indicate the presence of two distinct profibrotic and proregenerative differentiation paths dependent on mechanotransduction signaling that occur after injury in both large animals and humans. On both the mRNA and protein levels, we confirmed the interlinked behavior of both YAP and EGR1 signaling and showed how silencing the expression of each of these master transcriptional factors directly pushes fibroblasts toward the opposing trajectory. Taken together, our findings identify a critical role of mechanical signaling in wound healing and scar formation for large organisms and highlight that true tissue regeneration can occur by blocking mechanical signaling.