Shifts in keratin isoform expression activate motility signals during wound healing

Summary

Keratin intermediate filaments confer structural stability to epithelial tissues, but the reason this simple mechanical function requires a protein family with 54 isoforms is not understood. During skin wound healing, a shift in keratin isoform expression alters the composition of keratin filaments. If and how this change modulates cellular functions that support epidermal remodeling remains unclear. We report an unexpected effect of keratin isoform variation on kinase signal transduction. Increased expression of wound-associated keratin 6A, but not of steady-state keratin 5, potentiated keratinocyte migration and wound closure without compromising mechanical stability by activating myosin motors to increase contractile force generation. These results substantially expand the functional repertoire of intermediate filaments from their canonical role as mechanical scaffolds to include roles as isoform-tuned signaling scaffolds that organize signal transduction cascades in space and time to influence epithelial cell state.

eTOC blurb

Nanes et al. explore why the seemingly simple mechanical function of keratin intermediate filaments requires a protein family with 54 isoforms. They find that skin wound healing triggers a change in keratin isoform expression not to alter force resistance, but to organize signals activating myosin and promoting cell motility.

Graphical Abstract

Introduction

Effective tissue barrier formation depends on keratin intermediate filaments¹. Joined through desmosomes, keratin filaments form strong trans-cellular networks, which function as mechanical scaffolds². Fifty-four different keratin isoforms are expressed in an intricate pattern depending on tissue, anatomic site, and differentiation state³. Keratin filaments are particularly important in the skin, where their disruption causes epidermal fragility in a number of diseases⁴, and specific keratin isoform mixtures mark epidermal layers and associated cell states (Figure 1A) ¹. However, epidermal stability must be balanced with plasticity¹. Too much stability relative to plasticity can limit remodeling, resulting, for example, in non-healing wounds⁵. Too much plasticity relative to stability can result in uncontrolled or abnormal remodeling, a feature of numerous skin diseases ranging from cancer to psoriasis^6,7. The molecular mechanisms underlying the balance between epidermal stability and plasticity largely remain to be uncovered.

Figure 1. Epidermal remodeling triggers a keratin isoform switch.

A. Main keratin isoforms expressed in steady-state and actively remodeling epidermis.

B. Top: A skin excision specimen includes healing edges from the prior biopsy wound. Middle: Hematoxylin and eosin (H&E)-staining. Dotted outlines, epidermal migration fronts. Epidermal layers: B, basal; S, spinous (suprabasal); C, cornified. Bottom: Immunofluorescence labeling of wound-associated K6 and K17, and steady-state basal epidermal K5.

C. Intermediate filament expression in acute (Top, GSE97615) and chronic (Bottom, GSE80178) human skin wounds and intact skin controls.

D-E. Immunofluorescence of steady-state K5 and wound-associated K6 (D) or K17 (E) at keratinocyte monolayer wound edges.

F. Keratin transcript levels measured by qRT-PCR from intact or wounded keratinocyte monolayers (see Methods). n = 3 monolayers per group.

G-H. Stratified epidermal cultures. G, H&E staining after different growth periods. H, Immunofluorescence labeling of basal epidermal keratin K5 and differentiated epidermal keratin K10.

I. Top: Schematic of the epidermal culture wound model (see Methods). Middle: H&E staining after 72 hours of migration. Bottom: Immunofluorescence of wound-associated K6 and K17, and steady-state basal K5. Dashed vertical lines, image re-alignment correcting for folds in the tissue section. See also Figure S1B.

During wound healing, the required balance between epidermal stability and plasticity shifts, as keratinocytes change shape, migrate into the wound area, and reform the epidermal architecture¹. These tissue-level processes are accompanied at the cellular scale by increased expression of five wound-associated keratin isoforms: keratin 6A (K6A), K6B, K6C, K16, and K17 (Figure 1A) ^8–10. Expression of wound-associated keratins also increases in other active remodeling epidermal states, including psoriasis^11,12, squamous cell carcinoma¹³, hypertrophic scars¹⁴, and following tissue expander placement¹⁵. While expression of wound-associated keratin isoforms is a reliable marker of an active remodeling state, determining whether these keratin isoforms have specific cellular functions that support wound healing has been experimentally and conceptually challenging. Depleting or mutating individual keratins frequently disrupts keratin filament architecture and the more general mechanical scaffold function¹⁶. Partial redundancy between keratin isoforms further complicates the interpretation of monogenic models¹⁷. Wound-associated keratin knock-out mice often demonstrate prominent defects in epidermal integrity and keratinocyte cell-cell adhesion but lack generalized wound-healing defects^18–21. Similarly, disease-causing mutations in human wound-associated keratin genes broadly disrupt epidermal stability where the affected isoform is expressed^22,23. Thus, while wound-associated keratins contribute to the canonical intermediate filament mechanical scaffold, it is less clear if additional functions of these isoforms specifically promote wound closure.

In order to disentangle the specific contributions of wound-associated keratins to epidermal remodeling from the general function of keratin mechanical scaffolds, we took a gain-of-function rather than a loss-of-function approach. We found that increased expression of wound-associated K6A, but not steady-state K5, potentiated keratinocyte migration without compromising cell-cell adhesion. Rather than altering the mechanical response to external force, K6A expression increased cellular force generation via a phosphorylation switch activating actomyosin contractility. Thus, by wound-induced keratin isoform switching, epidermal cells maintain their mechanical barrier function while tuning a signal transduction cascade to promote critical functions for the wound healing process. These results highlight the complex interplay between the mechanical and signaling roles of the intermediate filament cytoskeleton, demonstrating that shifts in keratin isoform composition not only mark changes in epithelial cell state, but functionally contribute to epithelial state determination.

Results

An epidermal culture model of wound-induced keratin switching

Keratin expression change following wounding has been observed in animal models and humans^8–10. We reproduced this phenomenon in a human skin excision specimen (Figure 1B) and in a focused reanalysis of transcriptomics data from acute²⁴ and chronic wounds²⁵ (Figure 1C). However, wound-induced keratin switching has not previously been observed in cultured keratinocytes. Furthermore, cultured keratinocytes often express wound-associated keratins at baseline²⁶, possibly reflecting the increased proliferative rate of keratinocytes in culture²⁷. This has complicated efforts to understand the implications of keratin switching at the cellular and subcellular levels.

Consistent with prior reports^28,29, we found that an hTERT and Cdk-4 immortalized human keratinocyte cell line grown in submerged monolayer culture expressed steady-state basal level keratins K5 and K14 and wound-associated K6 and K17 (Figure S1A). Immunofluorescence of monolayer scratch wounds did not show any change in expression of K6 or K17 relative to K5 at the wound edge (Figure 1D–E). Similarly, bulk qRT-PCR did not show any change in wound-associated keratin transcript levels in scratched keratinocyte monolayers compared to unwounded controls (Figure 1F).

While keratin expression remained stable in monolayer culture, keratinocytes at an air-liquid interface formed a multilevel structure undergoing differentiation, with increased expression of K10 in the upper layers (Figure 1G–H), closely resembling the stratified squamous architecture of the epidermis³⁰. We leveraged these stratified epidermal cultures to create a three-dimensional wound model. Cultures were grown within a barrier mold, followed by barrier removal to allow keratinocyte migration (Figure 1I). In stark contrast to monolayer scratch wounds, wound-associated K6 and K17 expression was notably increased at the migrating edge of stratified cultures (Figures 1I and S1B). Hence, these data established a physiologically realistic epidermal culture model that allowed us to investigate the cellular and subcellular functions of wound-associated keratins.

Wound-associated K6A supports epidermal migration without compromising mechanical stability

We sought an interventional strategy for modulating relative levels of wound-associated versus steady-state keratins. Prior attempts to knock down or delete keratin genes have resulted in prominent loss of cell-cell adhesion and decreased mechanical integrity^31–35, a finding we reproduced with even heterozygous deletion of KRT6 isoforms (Figure S1C–D). To circumvent such global effects, we took a gain-of-function approach and co-expressed fluorescently tagged K6A and K5, the closest basal level paralog to K6A with 80% amino acid sequence identity, in our keratinocyte cell line. We then used fluorescence activated cell sorting to isolate cell populations with different relative keratin expression (Figure 2A).

Figure 2. Wound-associated K6A supports epidermal migration without compromising mechanical stability.

A. Diagram of the gain-of-function keratin expression model. Cultured keratinocytes were transduced with lentivirus to express tagged keratin, then divided by fluorescence activated cell sorting based on keratin expression levels.

B. Western blot analysis of endogenous (K5, K6) and exogenously expressed (K5-G, K6-R) keratins in parental (wt), K5^high, and K6A^high cell lines (see also Figure S1E). Intermediate bands likely represent degradation products of the tagged constructs. Right: Ratio of K5-G or K6-R to K5 or K6. n = 4 samples per group.

C. Proliferation of K5^high and K6A^high cells in culture measured by uptake of 5-ethynyl-2’-deoxyuridine (EdU). n = 6 samples per group.

D. Thin-section transmission electron microscopy images of typical desmosomes with keratin intermediate filament insertions.

E. Mechanical stability of keratinocyte monolayers measured using a fragmentation assay (see Methods). n = 12 monolayers per group.

F. Paired epidermal cultures were grown from K5^high or K6A^high cells. Following barrier removal and 12-hours of migration, the remaining gap was measured in H&E-stained sections. n = 8–9 epidermal culture pairs per group.

G-J. Paired epidermal cultures grown from K5^high and K6A^high cells on opposite sides of each barrier. Cultures were imaged live following barrier removal (G; see also Video S1). Red and green lines, farthest migration extent of the K6A^high and K5^high halves at each time point. H, Farthest distance migrated by the K6A^high half minus farthest distance migrated by the K5^high half for each culture pair. I, Average migration speed for each culture half over 20-hours. J, Time needed for the K5^high and K6A^high halves of each culture pair to reach distance checkpoints. Dots below the diagonal represent culture pairs where the K6A^high half reached the checkpoint first; dots above the diagonal represent culture pairs where the K5^high half reached the checkpoint first. n = 16 epidermal culture pairs.

We created K5^high/K6A^low (K5^high) and K5^low/K6A^high (K6A^high) cell lines with moderately increased (approximately 35%) levels of the keratin marked “high” (Figures 2B and S1E), reasoning that a small change in keratin expression could avoid filament network disruption observed with higher levels of overexpression. ^36,37 Because cells maintain an equimolar balance between type-I and type-II keratin isoforms, which polymerize as obligate heterodimers, we expect that modestly increasing K5 or K6A protein levels, both type-II keratins, may have resulted in small compensatory shifts of endogenous keratins, distributed over all expressed isoforms. ³ However, we confirmed that endogenous keratin protein and transcript levels were similar in K5^high and K6A^high cells (Figure S1F–G), indicating that comparison of these cell lines reflects directly engineered keratin expression changes, rather than indirect compensation. Importantly, increasing K5 or K6A protein levels did not affect cell proliferation (Figure 2C), cell junction morphology and number (Figures 2D and S1H–I), or mechanical integrity (Figure 2E).

We then proceeded to test the migration potential of K5^high and K6A^high keratinocytes. In three-dimensional epidermal migration assays, K6A^high epidermal cultures migrated farther than K5^high epidermal cultures over a 12-hour period (Figure 2F). We further took advantage of the bright fluorescence signals from the tagged keratins to enable live imaging of paired K5^high and K6A^high epidermal cultures following removal of a separating barrier (Figure 2G). While the partially translucent polycarbonate filter limited image quality, the fluorescence signal was sufficient to identify and track the migration edges using custom software (Figure 2G and Video S1) ³⁸. We observed that K6A^high epidermal cultures moved faster into the gap than their K5^high counterparts, and K6A^high cultures maintained their migration advantage over a 20-hour migration period (Figure 2H–J). Therefore, increased expression of wound-associated K6A offers a significant benefit to wound closure in three-dimensional epidermal cultures.

Wound-associated K6A supports migration in monolayers and individual cells

In order to perform high-resolution imaging of potential differences in cytoskeletal organization and dynamics between K5^high and K6A^high cells, we turned to a monolayer wound healing assay. In contrast to the three-dimensional epidermal wound model, we were initially unable to identify a clear difference in migration speed between K5^high and K6A^high monolayers (Figure S2A–D). Since the monolayer migration assay had significant wound-to-wound variability (Figure S2B), we wondered whether subtle but systematic differences in migration behavior between individual K6A^high and K5^high cells were missed in the coarser population level analysis.

To test this possibility, we created a mosaic population of keratinocytes including K5^high, K6A^high, K5^high/K6A^high, and K5^low/K6A^low cells and performed live imaging of the monolayer wound healing response (Figure 3A–B and Video S2). We applied particle image velocimetry to track local migration throughout the monolayer (Figure 3B, bottom; Video S2, right) ³⁸ and segmented the monolayers into different keratin expression regions (Figure 3B, top; Video S2, left), permitting comparison of migration speeds between keratin expression regions within the same wound and controlling for between-wound variability.

Figure 3. Wound-associated K6A supports keratinocyte migration in monolayers and single cells.

A. Diagram of a mosaic monolayer containing K5^high (green), K6A^high (red), K5^high/K6A^high (K5^hi6^hi, yellow), and K5^low/K6A^low (K5^lo6^lo, grey) cells during migration.

B. Top: Keratin expression regions. Bottom: Local migration speeds calculated using a computer vision pipeline (see Methods). White lines, 200-μm band tracking the wound edge. See also Video S2.

C. Average local migration speed 10-μm to 200-μm from the wound edge aligned relative to the global peak in migration speed. n = 36 scratch wounds.

D. Relative local migration speed within each keratin expression region (compared to average across all regions) at different time points relative to the peak in migration speed. n = 36 scratch wounds.

E-F. Pairwise comparison of local migration speed between keratin expression regions at the peak in migration speed (E) and 2-hours later (F). Each point represents one scratch wound. n = 36 scratch wounds.

G-I. Movement of sparsely seeded, individual K5^high and K6A^high keratinocytes. G, Aligned cell trajectories. Rings, quantiles of maximum distance from the origin. See also Video S3. H-I, Mean track velocity (H) and mean squared displacement rate (I) in K5^high and K6A^high cells. n = 7–10 movies with 144–232 tracks per group.

All wounds followed a similar pattern, with the overall migration speed rapidly increasing to a peak followed by slow deceleration, though the height and timing of the peak varied between experimental repeats (Figure 3C). Around the time of the overall migration speed peak, K6A^high regions exhibited a small, but consistent, migration advantage (Figure 3D). In the large majority (31 of 36) of wounds, K6A^high regions had higher speeds than K5^high regions (Figure 3E). K6A^high regions also had higher speeds than K5^high/K6A^high and K5^low/K6A^low regions, and K5^high/K6A^high regions had higher speeds than K5^high regions, indicating a dose-dependent effect (Figure 3E). Interestingly, the K6A^high migration advantage was transient. By 2 hours after the migration peak, K5^high region speeds began to approach those of the K6A^high regions (Figure 3F). Consistent with these results, K6A^high cells were modestly enriched near the wound edge of a mosaic population monolayer fixed after 18-hours of migration (Figure S2E).

Given that K5^high and K6A^high keratinocytes retain similar cell-cell adhesion properties (Figure 2E), we asked whether the K6A^high migration advantage was solely an emergent property of collective migration or if it could also be detected in individual keratinocytes. To address this question, we seeded K5^high and K6A^high cells at low density and tracked their movement over 12-hours using live imaging (Figure 3G and Video S3). Much like in the monolayer migration assays, K6A^high cells displayed a small advantage in migration speed and displacement, which peaked 4 to 8-hours after seeding (Figure 3G–I). Also similar to monolayer migration, the migration advantage was temporary, with the differences between K5^high and K6A^high cells becoming less apparent by 8 to 12-hours after seeding (Figure 3G–I). Thus, wound-associated K6A transiently potentiates keratinocyte migration at least in part through a cell autonomous effect.

Wound-associated K6A alters keratin filament dynamics

To address how the relative levels of keratin isoforms affect migration, we examined possible differences in keratin filament organization between K5^high and K6A^high cells. We imaged keratin filaments over time and used a computer vision pipeline to segment the filament networks (Figures 4A–B, S3A, and Video S4) ³⁹. In snapshots, the K5^high and K6A^high filament networks were similar. K5^high filament networks appeared to contain a somewhat higher number of very short filaments (Figure S3B), and an associated slight increase in overall filament density (Figure S3C), compared to K6A^high networks, possibly reflecting a slight difference in the imaging properties of the two fluorescent tags. Filament curvature was indistinguishable between K5^high and K6A^high cells (Figure S3D–F). However, when analyzed over time, K5^high filaments displayed larger intracellular displacements than K6A^high filaments (Figure 4A–B).

Figure 4. Wound-associated K6A increases cellular force generation by activating myosin.

A. Keratin filaments in K5^high and K6A^high cells. Right, pseudo-color overlay of filament images captured 3 minutes apart.

B. Filament networks delineated by a computer vision pipeline (see Methods). As in (A), overlay of networks 3 minutes apart highlights filament movement. See also Video S4.

C. Schematic of filament dynamics score calculation (see Methods).

D. Smoothed maps of local filament dynamics scores over different time intervals.

E. Comparison of filament dynamics scores between K5^high and K6A^high cells. Each data point represents spatially averaged filament dynamics scores for a cell. n = 110–139 cells in 24–25 movies per group.

F. Comparison of filament dynamics scores between keratinocytes expressing a chimeric keratin containing only the head domain of K6A joined to the remainder of K5 (K6A^h5^rt) with K5^high cells. n = 44–47 cells in 28 movies per group. See also Figure S4D.

G. Traction force microscopy (TFM) of K5^high and K6A^high cells. Left, reconstructed traction vectors (arrows) on a heatmap of force magnitude and a watermark of keratin filament TIRF images. Rings, visible borders and expanded boundary for strain energy calculation (see also Figure S4F). Right, strain energy density. Outlined dots, connected between groups, mean values of experimental replicates in separate batches of TFM substrates. n = 51–58 cells per group.

H. TFM of cells expressing a chimeric keratin containing the head domain of K6A joined to the remainder of K5 (K6A^h5^rt) or full-length K5. See also Figure S4H. n = 40 cells per group.

I. Diagram of the proximity ligation assay (PLA).

J. PLA detecting activated myosin by colocalization of phosphorylated myosin regulatory light chain (pRLC) and non-muscle myosin heavy chain (MYH9). Left, annotated images of K5^high (green outline) and K6A^high (red outline) cells. White diamonds, PLA detections (see Methods). Heatmap, detection density. Right, PLA detection density in K5^high and K6A^high cells. Outlined dots, connected between groups, mean values of experimental replicates. n = 66–80 cells per group in 37 total images.

K-L. PLA detecting activated myosin near keratin filaments by colocalization of pRLC and K5 (K) or K17 (L). K, n = 94–115 cells per group in 20 images. L, 99–108 cells per group in 23 images.

M-N. PLA detecting total myosin near keratin filaments by colocalization of total myosin regulatory light chain (RLC) and K5 (M) or K17 (N). M, n = 117–119 cells per group in 19 images. N, 121–129 cells per group in 21 images.

O. Schematic diagram of the proposed model. K6A-containing filaments preferentially recruit regulatory kinases, reducing the dimensionality of kinase–RLC interaction to increase myosin activation. Question marks indicate that the identity of the kinase and the mechanism of kinase–intermediate filament association remain to be determined.

To quantify differences in filament network dynamics, we created a metric reflecting changes in position and orientation of filaments between timepoints at one-minute intervals³⁹. Each point on the filament network at baseline is scored by the distance and difference in orientation of the nearest filament in the next considered timepoint (Figure 4C). We then spatially smoothed the local scores to generate continuous filament dynamics score maps and averaged over individual cells (Figure 4D). Comparing K5^high to K6A^high cells, we found that K6A systematically decreased intracellular dynamics of keratin filaments (Figure 4E). We confirmed that the difference in filament dynamics between K5^high and K6A^high cells is unlikely the consequence of small differences in the number of very short filaments or filament density (Figure S3G–H).

We further extended this approach using two-color live imaging of K5^high, K6A^high, and K5^high/K6A^high cells, allowing us to segment and calculate filament dynamics scores for K5 and K6A simultaneously (Figure S3I–O and Video S4). We found a dose-dependent effect of K6A on filament dynamics, with the highest dynamics scores for K5^high cells, the lowest dynamics for K6A^high cells, and intermediate dynamics for K5^high/K6A^high cells (Figure S3M–N). Furthermore, we could compare filament segmentations and dynamics scores between the K5 and K6A channels in K5^high/K6A^high cells. As expected, given the known propensity for individual keratins to intermix within filament networks⁴⁰, K5 and K6A network segmentations overlapped considerably (Figure S3L and Video S4). Additionally, K5 and K6A dynamics scores for K5^high/K6A^high cells were highly correlated, without evidence of bias between detection channels (Figure S3O), confirming the robustness of the filament segmentation algorithm and network dynamics score. Taken together, these results establish a direct relationship between the mixture of keratin isoforms assembled into filaments and the dynamics of the networks that those filaments form.

The filament dynamics score is a composite measure of both intrinsic properties of the filaments, such as stiffness, and extrinsic interactions between the filament network and other cellular components, such as the contractile actomyosin cytoskeleton. We reasoned that if a difference in contractility was responsible for the difference between K5^high and K6A^high filament dynamics, inhibiting myosin motors would reduce that effect. Indeed, we found that the specific myosin II inhibitor Blebbistatin completely eliminated the difference between K5^high and K6A^high filament dynamics (Figure S4A), while two upstream ROCK inhibitors had smaller effects (Figure S4B). While insufficient on their own to define a mechanism, these results establish a link between keratin filament dynamics and the actomyosin system.

We then postulated that determining the domain of the keratin protein responsible for the dynamics difference between K5^high and K6A^high filaments might further untangle this effect. Keratin proteins are composed of a central alpha helical rod domain flanked by intrinsically disordered head and tail domains^41,42. We created chimeric keratin constructs composed of the head and tail domains of K5 or K6A fused to the rod domain from the opposite keratin (K5^h6^r5^t and K6^h5^r6^t). Upon expressing each of these chimeras in cultured keratinocytes, we found that K6^h5^r6^t decreased filament dynamics compared to K5^h6^r5^t (Figure S4C). We further created chimeric keratins containing only the head or tail domain of K6A fused to the remainder of K5 (K6A^h5^rt and K5^hr6A^t) and determined that the K6A head domain alone was sufficient to decrease filament dynamics (Figures 4F and S4D), while the K6A tail domain had no effect (Figure S4E). Thus, the intrinsically disordered K6A head domain, rather than the structural rod domain, is responsible for decreasing keratin filament dynamics. While we cannot formally exclude contributions of the head domain to the mechanical properties of filaments, this result, combined with the similarity of K5^high and K6A^high filament networks in snapshots (Figure S3B–F), suggests that the changes in filament dynamics between K5^high and K6A^high cells are more likely related to changes in cytoskeletal force generation than changes in filament stiffness.

Wound-associated K6A increases cellular force generation by activating myosin

To directly test whether changing keratin filament composition affects cellular force generation, we compared the ability of K5^high and K6A^high cells to deform an elastic substrate using traction force microscopy⁴³. We found that K6A^high cells generated increased traction forces and strain energy compared to K5^high cells (Figures 4G and S4F). Using alternate cell lines expressing short-tagged keratins, we confirmed that fluorescent tags were not responsible for this effect (Figure S4G). Furthermore, expressing the K6A^h5^rt chimera also increased force generation (Figures 4H and S4H), indicating that, as with filament dynamics, traction force generation was modulated by differences in keratin head domains. Importantly, the increased strain energy, reflecting the total work of contraction executed by the cell, cannot be explained by changes in material properties such as filament elasticity or differences in organization of cell-matrix adhesions. The latter would in fact be expected as a downstream effect secondary to increased traction stress. ^44,45 This finding directly links shifts in the abundance of wound-associated K6A to cellular force generation.

Keratin filaments are not known to directly generate force. Rather, contractile forces are generated by myosin motor proteins linked to the actin cytoskeleton⁴⁶. Non-muscle myosin activation occurs through phosphorylation of myosin regulatory light chain (pRLC) ^47,48. To directly compare pRLC between K5^high and K6A^high cells, we employed a proximity ligation assay (Figure 4I) ⁴⁹. We probed for colocalization of myosin heavy chain (MYH9) and pRLC in a mosaic culture of K5^high and K6A^high keratinocytes. Consistent with the finding of increased contractility of K6A^high cells, we measured increased colocalization of pRLC with MYH9, indicating increased myosin activation, in K6A^high cells compared to K5^high cells (Figure 4J).

We further probed for colocalization of pRLC with keratin filaments using either antibodies against K5 or K17. In both cases, we detected increased colocalizations in K6A^high cells compared to K5^high cells (Figure 4K–L), suggesting that myosin motors in the neighborhood of keratin filaments are more frequently activated when the relative amount of K6A is increased. Importantly, the colocalization between total regulatory light chain and K5 or K17 was similar or decreased in K6A^high cells compared to K5^high cells (Figure 4M–N), underscoring that increased pRLC near keratin filaments must reflect a signaling phenomenon leading to myosin activation, not merely a differential physical coupling between keratins and the actomyosin machinery. Altogether, these results show that shifting keratin filament isoform composition can modulate signaling circuits without compromising mechanical scaffold function.

Discussion

Our data reveal an unexpected dual role for keratin filaments as isoform-tuned signaling scaffolds in addition to their canonical role as mechanical scaffolds. Increasing expression of wound-associated K6A supports epidermal remodeling not by altering the cell’s mechanical capacity for force resistance, but by organizing molecular signals that activate force generation to drive migration. This mechanism thus establishes a functional coupling between the cellular machinery of force resistance and force generation precisely in the context where the stability–plasticity balance must be finely tuned. During wound healing, the epidermis remains subject to mechanical stress, both from the usual external insults and directly resulting from tissue remodeling, so widespread disassembly of the keratin mechanical scaffold to increase tissue plasticity is an untenable strategy. Shifting the keratin isoform mixture therefore represents an elegant mechanism to maintain the mechanical capacity for force resistance while modulating non-mechanical functions through cellular signals. Dependence on the intrinsically disordered head domain, which varies considerably between keratin isoforms, rather than on the highly conserved structural rod domain, further supports a non-mechanical function for keratin isoform switching.

The data presented here do not define precisely how different keratin isoforms may organize cellular signals to modulate myosin activation. Based on our data, we postulate that K6A-containing filaments exhibit higher affinity for components of a kinase signaling cascade, reducing the dimensionality of kinase–RLC encounters from the 3D volume of the cytoplasm to a confined 1D scaffold, amplifying the chances for productive biochemical reactions between enzyme and substrate (Figure 4O). The identity of the signaling components and the molecular mechanisms involved in their differential binding to different keratin isoforms to generate this antenna effect for motility signals remain open questions which we will examine in future work.

Whereas signaling functions of keratins have long been postulated, ^50–52 how such signaling might be deconvolved from the critical mechanical scaffold role has not previously been explored. Most knock-out models and filament-disrupting mutants are clearly confounded by mechanical defects^18,53–61. In one notable example, and in seeming disagreement with the data presented here, keratinocytes from a Krt6a/b-null mouse migrated faster than wild-type controls in a Src- and myosin-dependent process^18,31–33. Yet these studies also noted broad disruption of cell-cell adhesion, ^18,31–33 and two other mouse models with varying degrees of K6 depletion but apparently preserved mechanical integrity demonstrated unchanged or decreased keratinocyte migratory ability^17,21. Based on our ability to induce gain-of-function conditions in a 3D epidermal model, we arrive at the plausible and unifying interpretation that increased keratinocyte migration in the Krt6a/b-null mouse results from general loss of keratin mechanical scaffold function and cell-cell adhesion, rather than isoform-specific effects of K6 per se. A key finding of our study, uncovered only by preserving the keratin filament network and mechanical scaffold, is that isoform-tuned signaling is separable from broad mechanical effects.

In a seeming paradox, we observed that elevated K6A expression increases force generation yet decreases intracellular movement of keratin filaments. This may simply reflect impairment of keratin filament movement by heightened cellular rigidity or subtly increased formation of actomyosin stress fibers^62,63. However, an intriguing alternative hypothesis is that keratin isoform-specific interactions with other cellular components might hinder K6A-enriched keratin filament movement under a given force. Expansive additional investigation will be needed to evaluate these possibilities.

Different keratin isoform mixtures serve as reliable markers of fundamental epithelial states, distinguishing simple versus stratified epithelia, stem cell versus differentiated cell compartments, and, as we have focused our attention in this study, stable versus actively remodeling tissues³. Our data indicate that these isoforms not only mark different epithelial states, but in fact establish those states through organization of signaling pathways or transcriptional networks. Thus, this study sets the stage for investigations of the diversity of intermediate filament interactions as a new regulatory motif in the adaptive control of cell function.

Limitations of the study

Although we detected no difference in mechanical integrity between K5^high and K6A^high cells, it remains possible that different keratin isoforms may influence the kinetics of subunit exchange or the formation of desmosomes or hemidesmosomes, including in potentially subtle ways via feedback loops related to actomyosin contractility. Dedicated studies will be needed to address these questions. Additionally, the full mechanistic link between actomyosin activation and increased wound closure is likely complex and is not comprehensively elucidated by our observation of increased force. Given that K6A^high cells possessed a transient migration advantage in two-dimensional culture but a persistent advantage in three-dimensional epidermal culture, we speculate that environmental cues significantly modulate the translation of changing force generation into cellular translocation. For example, while elevated activation of contractility may initially enhance cell speed, higher forces promote the maturation of focal adhesions, which may slow migration at later times. Future studies will explore these complex feedback mechanisms.

STAR Methods

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Gaudenz Danuser (gaudenz.danuser@utsouthwestern.edu).

Materials availability

Requests for reagents generated in this study will be fulfilled by the lead contact upon reasonable request.

Data and code availability

• Raw numerical data have been deposited at Zenodo and are publicly available as of the date of publication. DOIs are listed in the key resources table. All other data reported in this paper will be shared by the lead contact upon request.

• All original code has been deposited at Zenodo and is publicly available as of the date of publication. DOIs are listed in the key resources table.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Key resources table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Mouse anti-keratin 1	Novus	Cat#NB100-2756
Rabbit anti-keratin 5	Abcam	Cat#ab52635
Rabbit anti-keratin 5	Cell Signaling Technology	Cat#25807
Rabbit anti-keratin 6	Abcam	Cat#ab93279
Rabbit anti-keratin 10	Abcam	Cat#ab76318
Mouse anti-keratin 14	EMD Millipore	Cat#MAB3232
Mouse anti-keratin 16	Invitrogen	Cat#LL025
Rabbit polyclonal anti-keratin 17	Abcam	Cat#ab53707
Mouse anti-beta-actin	Sigma	Cat#A1978
Rabbit polyclonal anti-non-muscle myosin heavy chain IIA	Invitrogen	Cat#PA5-17025
Mouse anti-phospho-myosin regulatory light chain	Cell Signaling Technology	Cat#3675
HRP, goat anti- mouse IgG	Invitrogen	Cat#31430
HRP, goat anti- rabbit IgG	Invitrogen	Cat#31460
Alexa Fluor 488, goat anti- mouse IgG	Invitrogen	Cat#A-11209
Alexa Fluor 488, goat anti- rabbit IgG	Invitrogen	Cat#A-11009
Alexa Fluor 568, goat anti- mouse IgG	Invitrogen	Cat#A-11004
Alexa Fluor 568, goat anti- rabbit IgG	Invitrogen	Cat#A-11036
Alexa Fluor 647, goat anti- rabbit IgG	Invitrogen	Cat#A-21244
Bacterial and virus strains
Ad5-CMV-Cre	Baylor College of Medicine Gene Vector Core	https://www.bcm.edu/research/atc-core-labs/gene-vector-core/services-and-fees/in-stock-adenoviral-vectors
Biological samples
Skin excision specimen	Fresh and Archived and Skin Tissue Repository (FASTeR) in the Department of Dermatology at UT Southwestern	https://www.utsouthwestern.edu/education/medical-school/departments/dermatology/
Chemicals, peptides, and recombinant proteins
Dispase	Stemcell Technologies	Cat#07913
Bovine type I collagen	Advanced Biomatrix	Cat#5005
Agarose (for epidermal culture histology)	Lonza	Cat#50000
Blebbistatin	Sigma	Cat#B0560
Y27632	Selleckchem	Cat#S1049
GSK 269962A	Selleckchem	Cat#S7687
DMSO	Fisher	Cat#BP231
QGel 920	CHT	Cat#QGEL-920
(3-aminopropyl)triethoxysilane (APTES)	Sigma	Cat#440140
Dark Red Carboxylate-Modified Microspheres	ThermoFisher	Cat#F8789
1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC)	ThermoFisher	Cat#22980
L-ascorbic acid	Sigma	Cat#A92902
anhydrous copper(II) sulfate	Acros	Cat#42287
Cy5-azide	Sigma	Cat#777323
5-ethynyl-2’-deoxyuridine (EdU)	ThermoFisher	Cat#A10044
Critical commercial assays
RNeasy Mini Kit	Qiagen	Cat#74104
SuperScript IV First-Strand Synthesis System	Invitrogen	Cat#18091050
iTaq Universal SYBER Green Supermix	BioRad	Cat#1725120
Duolink In-Situ PLA Probe Anti-Mouse MINUS	Sigma	Cat#DUO92004
Duolink In-Situ PLA Probe Anti-Rabbit PLUS	Sigma	Cat#DUO92002
Duolink In-Situ PLA Detection Reagents FarRed	Sigma	Cat#DUO92013
Q5 Site-Directed Mutagenesis Kit	New England Biolabs	Cat#E0554S
Lipofectamine LTX with PLUS Reagent	Invitrogen	Cat#A12621
Deposited data
Raw quantitative data	This paper	DOI 10.5281/zenodo.11509646
Punch biopsy wound RNAseq data	Iglesias-Bartolome et rfal. 24	GEO GSE97615
Foot ulcer microarray data	Ramirez et al. 25	GEO GSE80178
Experimental models: Cell lines
Human keratinocyte Ker-CT cells	Laboratory of Jerry Shay (UT Southwestern) 28,29; ATCC	CRL-4048
HEK293 cells	ATCC	CRL-1573
Oligonucleotides
Primers and synthetic DNA for cloning, see Table S1	This paper	N/A
qRT-PCR primers, see Table S1	This paper	N/A
Recombinant DNA
pLVX-IRES-Puro	Clontech	Cat#632183
pLVX-IRES-Neo	Clontech	Cat#632181
pBabe-RFP1-KRT5-hygro	Wang et al. 97; Addgene	Cat#58493
pDONR22-KRT6A	DNASU	Clone HsCD00039474
pX333	Maddalo et al. 75; Addgene	Cat#64073
pMA-tia1l	Lackner et al. 74	N/A
pLVX-KRT5-mNG-IRES-Puro	This paper	N/A
pLVX-KRT6A-mRb-IRES-Neo	This paper	N/A
pLVX-K5h6Ar5t-mNG-IRES-Puro	This paper	N/A
pLVX-K6Ah5r6At-mRb-IRES-Puro	This paper	N/A
pLVX-K6h5r5t-mNG	This paper	N/A
pLVX-K5h5r6t-mRb	This paper	N/A
pLVX-K5-FLAG-IRES-mNeonGreen	This paper	N/A
pLVX-K6A-FLAG-IRES-mNeonGreen	This paper	N/A
Software and algorithms
MATLAB	The Mathworks Inc.	https://www.mathworks.com/products/matlab.html
MonolayerKymographs package	Zaritsky et al. 38	https://github.com/DanuserLab/MonolayerKymographs
u-delineate package	Gan et al. 39	https://github.com/DanuserLab/u-delineate
u-inferforce package	Han et al. 43	https://github.com/DanuserLab/u-inferforce
Fiji distribution of ImageJ	Schindelin et al. 98	https://fiji.sc/
Fiji Image Stitching Plugin	Preibisch et al. 80	https://github.com/fiji/Stitching
TrackMate plugin for ImageJ	Tinevez et al. 82	https://github.com/trackmate-sc/TrackMate
Slide Set plugin for ImageJ	Nanes93	https://github.com/bnanes/slideset
Python	Python Software Foundation	https://www.python.org/
R	R Project for Statistical Computing	https://www.r-project.org/
oligo package	Carvalho et al. 95	https://doi.org/doM0.18129/B9.bioc.oligo
Custom MATLAB and Python scripts	This paper	DOI 10.5281/zenodo.11507378; https://github.com/DanuserLab/nanes_2024_dev-cell_krt-signaling
Other
Keratinocyte SFM	Gibco	Cat#17005042
DMEM/F-12	Gibco	Cat#11320033
DMEM, high glucose, pyruvate	Gibco	Cat#11995065
Polycarbonate Cell Culture Inserts, 0.4pm pore size, 3.14cm2 area	Nunc, ThermoFisher	Cat#140640
2-well silicone culture inserts	Ibidi	Cat#80209

Experimental model and study participant details

Cell lines

Human keratinocyte Ker-CT cells (ATCC CRL-4048) were a kind gift from Dr. Jerry Shay (UT Southwestern Medical Center). This cell line was developed from male human foreskin keratinocytes through expression of HTERT and CDK4. ^28,29 Unless otherwise specified, keratinocytes were cultured on tissue-culture treated plastic coated with bovine type I collagen (PureCol, Advanced Biomatrix #5005) by pre-incubating dishes in a 100 μg/mL collagen solution for 30 minutes. Cells were maintained in Keratinocyte serum-free medium (K-SFM; Gibco #17005042) with the included 5 ng/mL human recombinant epidermal growth factor and 50 mg/mL bovine pituitary extract supplements according to the manufacturer’s instructions in a humidified incubator at 37 °C and 5% CO₂. K-SFM contains a low calcium concentration (about 0.12mM according to the manufacturer) which supports keratinocyte growth without inducing differentiation, but which also limits desmosome formation⁶⁴. To induce desmosome formation and support differentiation for certain experiments, we developed a keratinocyte differentiation medium based on a simplification of the classic Rheinwald and Green formula^65,66. This medium was composed of fully-supplemented K-SFM mixed in equal parts with DMEM/F12 (Gibco #11320033) and additionally supplemented with 2% fetal bovine serum (FBS; Sigma F0926). Based on an estimated calcium concentration of 3.75mM in FBS⁶⁷, the calcium concentration of the differentiation medium was 0.65mM. As indicated below, 0.5mM additional calcium chloride, yielding a final concentration of 1.15mM, was added for some experiments. HEK293 cells (ATCC CRL-1573) used for generation of lentiviral particles were maintained in DMEM (Gibco #11995–065) supplemented with 10% FBS and antibiotic–antimycotic reagent (Gibco #15240–062) in a humidified incubator at 37 °C and 5% CO₂. This polyploid cell line was isolated from a female human embryonic kidney. All cell lines were periodically tested for mycoplasma using a PCR-based Genlantis Mycoscope Detection Kit (MY01100). Cell lines were not authenticated.

Human skin wound tissue sections

We obtained anonymized H&E stained and unstained tissue sections from a skin excision specimen leftover following routine clinical care from the Fresh and Archived and Skin Tissue Repository (FASTeR) in the Department of Dermatology at UT Southwestern. As this was a de-identified archived sample, patient demographic information is not available. This study was approved by the UT Southwestern Medical Center Institutional Review Board.

Method details

Plasmids

Primer and synthetic DNA sequences (Integrated DNA Technologies) used to assemble the following constructs are provided in Table S1.

Keratin sequences were cloned by polymerase chain reaction (PCR) from pBabe-RFP1-KRT5-hygro (Addgene #58493) and pDONR22-KRT6A (DNASU clone HsCD00039474) and inserted into pLVX-IRES-Puro or pLVX-IRES-Neo lentiviral expression vectors (Clontech) with C-terminal fluorescent tags using seamless cloning (HiFi DNA Assembly Master Mix, New England Biolabs). The pLVX-KRT5-mNG-IRES-Puro construct includes an mNeonGreen (Allele Biotechnology) ⁶⁸ sequence in-frame with the KRT5 sequence separated by a short peptide linker (DPAFLY). Similarly, the pLVX-KRT6A-mRb-IRES-Neo construct includes an mRuby2 (Addgene #40260) ⁶⁹ sequence in-frame with the KRT6A sequence separated by the same linker.

Keratin head/tail domain chimera constructs were assembled using seamless cloning into pLVX-IRES-Puro, with keratin domains cloned from the above pLVX-KRT5-mNG-IRES-Puro and pLVX-KRT6A-mRb-IRES-Neo constructs by PCR. Keratin domains were defined based on NCBI Reference Sequence annotations (NP_000415.2, NP_005545.1). Specifically, the K5 head and tail domains included amino acids 1–167 and 478–590 respectively; the K6A head and tail domains included amino acids 1–162 and 473–564 respectively.

For proximity biotin labeling experiments, the sequence of miniTurbo, a small promiscuous biotin ligase⁷⁰, was synthesized with a C-terminal Flag tag and flanking XbaI and BamHI restriction sites. This fragment was then inserted into pLVX-IRES-Puro using the XbaI and BamHI sites. Finally, the KRT5 and KRT6A sequences were amplified by PCR from the above pLVX-KRT5-mNG-IRES-Puro and pLVX-KRT6A-mRb-IRES-Neo constructs and inserted in-frame and N-terminal to the biotin ligase with a short peptide linker (DPAFSR) using seamless cloning. To create a marker construct for these experiments, EGFP from pEGFP-C1 (Clontech) was inserted into pLVX-IRES-Puro using the SnaBI and BamHI restriction sites.

To create short-tagged keratin constructs without linked fluorescent proteins, the Q5 Site-Directed Mutagenesis Kit (NEB) was used to delete the miniTurbo sequence from the above pLVX-KRT5-miniTurbo-FLAG-IRES-Puro and pLVX-KRT6A-miniTurbo-FLAG-IRES-Puro constructs, leaving the FLAG joined in-frame to the C-terminal end of the keratin sequence. Seamless cloning was then used to replace the puromycin resistance cassette after the internal ribosome entry sequence with an mNeonGreen sequence, allowing for expression of a non-linked fluorescent marker from the same plasmid.

Cell line creation

Lentiviral particles were generated using the pLVX system (Clontech) with packaging vectors psPAX2 and pMD2.G (Addgene plasmids #12260 and #12259). HEK293 cells were transfected with expression and packing plasmids following standard calcium phosphate or polyethylenimine (Polysciences #23966) protocols. Supernatant was collected two days after transfection, filtered through 0.45-μm mixed cellulose esters membrane syringe filters (Fisher Scientific #09–720-005) and incubated on target cells overnight. After multiple days of culture and cell expansion, cells expressing the relevant fluorescent marker were collected on a FACS Aria II SORP flow cytometer equipped with 405 nm, 488 nm, 561 nm, and 633 nm lasers (Beckton Dickinson) and returned to culture.

Western blotting

Cells were grown to confluence and switched to differentiation medium 24 hours prior to sample collection. Because keratin filaments are highly insoluble, a high-ionic-strength urea buffer was used in place of a standard lysis buffer⁷¹. The 6.5 M urea lysis buffer also contained 50 mM tris, 1 mM EGTA, 2 mM DTT, 50 mM sodium fluoride, 1 mM sodium vanadate, and Halt Protease Inhibitor Cocktail (ThermoFisher #78430), and was adjusted to pH 7.5. After washing twice with PBS, cells were collected into urea lysis buffer using a cell scraper, then incubated at 4 °C for 20 minutes on a shaker. The lysates were then sonicated using a Fisherbrand Model 505 Sonic Dismembrator (ThermoFisher) twice for 10 seconds at 20% power with a 10 second intervening rest. Any remaining insoluble material, which was typically negligible, was pelleted by centrifugation at 16.1×10³ g for 20 minutes at 4 °C, and the supernatant was transferred to a clean tube. Protein concentrations were measured using a BCA assay (ThermoFisher #23228). Laemmli sample buffer (BioRad #161–0747) with 2-mercaptoethanol (ThermoFisher #BP176) was added to the samples, which were then heated to 95 °C for 5 minutes.

Samples were run on 4–20% Mini-PROTEAN pre-cast gels (Biorad) and transferred to nitrocellulose membranes (ThermoFisher #88018). Once membranes were blocked with 5% bovine serum albumin (BSA; Equitech-Bio BAH65–0500) in Tris buffered saline (TBS) with 0.1% Tween 20 (TBST; ThermoFisher BP337) for 1 hour at room temperature, they were incubated in primary antibodies overnight at 4 °C. Membranes were washed with TBST and probed with secondary antibodies conjugated with horseradish peroxidase. Once they were washed again, bands were detected through enhanced chemiluminescence using freshly prepared substrate solution based on 100 mM Tris (ThermoFisher BP152) adjusted to pH 8.5 with hydrochloric acid (ThermoFisher A144S-212), 1.2 mM luminol (Sigma A8511), 2.5 mM p-coumaric acid (Sigma C9008) and 0.04% hydrogen peroxide (w/v, Alpha Aesar L14000) on a Syngene G:BOX imager equipped with a Synoptics 4.2-megapixel camera and controlled with Syngene Genesys software.

Primary antibodies used for Western blotting include Mouse anti-keratin 1 (Novus LHK1) at 1:250; Rabbit anti-keratin 5 (Abcam ab52635), targeting an N-terminal epitope, at 1:1,000; Rabbit anti-keratin 5 (Cell Signaling Technology #25807), targeting a C-terminal epitope, at 1:1,000; Rabbit anti-keratin 6 (Abcam ab93279), targeting an epitope conserved in K6A, K6B, and K6C, at 1:2,000; Rabbit anti-keratin 10 (Abcam ab76318) at 1:1,000; Mouse anti-keratin 14 (EMD Millipore MAB3232) at 1:200; Mouse anti-keratin 16 (Invitrogen LL025) at 1:500; Rabbit polyclonal anti-keratin 17 (Abcam ab53707) at 1:2,000; and Mouse anti-beta-actin (Sigma A1978) at 1:5,000. Secondary antibodies used for Western blotting at 1:1,000 dilution include HRP-conjugated goat anti- mouse IgG (Invitrogan #31430) and HRP-conjugated goat anti- rabbit IgG (Invitrogan #31460).

Quantitative reverse transcription PCR

Keratinocytes were grown to confluence in 6-well tissue culture plates. For comparison between cell lines (Figure S1G), medium was switched from K-SFM to differentiation medium 24 hours before sample collection. For scratch wound experiments (Figure 1F), culture medium was switched from K-SFM to differentiation medium, then treatment samples were scratched with a 1000 mL pipette tip in a cross-hatch pattern. The scratch procedure was repeated every other day for 10 days. Samples were collected one day following the last scratch procedure. RNA was isolated from the samples using RNeasy spin column kits (Qiagen #74004) and reverse transcription was performed using SuperScript IV First-Strand Synthesis System with oligo(dT)₂₀ primers (Invitrogen #18091050). Real-time PCR using a C1000 thermal cycler with a CFX96 optical reaction module (Bio-Rad), iTaq Universal SYBER Green Supermix (Bio-Rad), and transcript-specific primers (Table S1) was performed to quantify transcript levels. The 2^−ΔΔCT method⁷² was used to compare each individual keratin to the mean level of all keratins measured.

Keratinocyte monolayer fragmentation assay

We adopted the commonly used dispase-based fragmentation assay to verify the ability of our keratinocyte cell lines to form stable monolayer barriers⁷³. Keratinocytes were seeded into collagen-coated 12-well tissue culture plates and grown to confluence. Upon reaching confluence, culture medium was changed from K-SFM to differentiation medium with additional 0.5 mM calcium chloride. After an additional 48-hour incubation, the monolayers were washed twice with PBS then incubated in dispase solution (5 units/mL in HBSS, Stemcell Technologies #07913) for 80 minutes at 37 °C to detach the monolayers from the plate. The detached monolayers were then exposed to mechanical shear stress by slow passage through a 1000μL pipette tip 10 times. Each plate was then imaged twice with an Apple iPhone 13 mini with gentle rocking between the image pairs to displace the monolayer fragments. Monolayer fragments were manually annotated and counted. If a different number of fragments was counted in each of the image pairs for a given well, the higher value was used.

KRT6A/B/C cluster heterozygous deletion

The human genome contains three type-II wound-associated keratin genes, KRT6A, KRT6B, and KRT6C, located in series with no other intervening genes in a 48 kilobase region on chromosome 12. We adopted a two-stage strategy of loxP cassette insertion followed by Cre recombination to attempt deletion of the entire genomic cluster. First, we used CRISPR-Cas9 with homology directed repair to insert selection cassettes at sites flanking the keratin gene cluster. Each cassette contained a fluorescent protein expression sequence (either mNeonGreen or mCherry) and a loxP site, designed such that if the cassettes were correctly positioned and oriented, Cre-mediated recombination would remove the keratin gene cluster along with the fluorescent markers. Each cassette was synthesized and inserted into a tia1l self-targeting carrier plasmid⁷⁴ by seamless cloning. Guides were synthesized and inserted into the pX333 system (Addgene #64073), which expresses tandem sgRNAs as well as Cas9⁷⁵. The three plasmids were transfected together into keratinocytes using Lipofectamine LTX with PLUS reagent (Invitrogen #A12621) according to the manufacturer’s protocol. After 7 days in culture to allow cassette insertion and cell expansion, cells expressing both fluorescent markers (about 1% of the total population) were collected by FACS and returned to culture. After an additional 20 days in culture to ensure stable integration of the cassettes, cells were again sorted for expression of both fluorescence markers. To finally excise the keratin cluster, we transduced the cells with Ad5-CMV-Cre adenoviral vector (purchased from Baylor College of Medicine Gene Vector Core). After 7 days of culture, cells lacking both fluorescent markers, indicating successful deletion of the keratin cluster, were collected by FACS, and clonal lines were created by limiting dilution. We were unable to isolate clones completely lacking K6. However, we did isolate a clone with K6 protein decreased by approximately 50% (Figure S1D). We isolated genomic DNA from this clone using a Qiagen DNEasy kit and used PCR to amplify a region from the type-II wound-associated keratin cluster as well as a region formed by the predicted deletion. Both bands were present, indicating a heterozygous clone (Figure S1D). This cell line was used as a positive control for the keratinocyte monolayer fragmentation assay (Figure S1C).

Electron microscopy of keratinocytes

For transmission electron microscopy of keratinocyte monolayers en face, glow-discharged carbon-film gold EM grids (EMS CF200-Au) were coated with bovine type I collagen (PureCol, Advanced Biomatrix #5005) by adding one drop of a 100 μg/mL collagen solution to each grid and incubating at 37 °C for 30 minutes. The collagen solution was then removed by absorption onto a strip of filter paper. Keratinocyte cell suspensions were prepared with 10⁴ cells/mL in differentiation medium, and one drop of cell suspension was added to each grid. After 24 hours, the grids were processed as follows⁷⁶: grids were washed by dipping into 4 droplets of cytoskeleton buffer (CB; 10 mM MES, pH 6.1 with 150 mM NaCl, 5 mM EGTA, 5 mM glucose, and 5 mM MgCl₂) ⁷⁷, after wicking off the last droplet, the grids were lightly fixed and extracted in CB containing 0.25% glutaraldehyde and 0.5% Triton-X 100 for 1 minute. Grids were washed by dipping into 1 droplet of CB. The grids were then fixed for 10 minutes in CB containing 2% glutaraldehyde. After washing in 3 water droplets for 2 minutes each, the grids were stained with 1% uranyl acetate for 2 minutes; washed twice in water droplets for 2 minutes each, then air dried. Imaging was done on a JEM-1400 Plus transmission electron microscope equipped with a LaB6 source operated at 120 kV using an AMT-BioSprint 16M CCD camera. Cell junction widths were measured using ImageJ by manually drawing line regions of interest across the thickest portion of electron-dense intercellular bridges parallel to the cell-cell border.

For transmission electron microscopy of thin transverse sections of keratinocyte monolayers, cells were cultured in 35mm glass-bottom dishes (Mattek P35G-1.5–14-C) until confluent. The cells were washed once with PBS and twice with 100 mM sodium cacodylate buffer, then fixed in 2.5% glutaraldehyde in 100 mM cacodylate buffer. Following fixation, samples were rinsed five times in 100 mM sodium cacodylate, then post-fixed in 1% osmium tetroxide with 1.5% K₃[Fe(CN)₆] in 100 mM sodium cacodylate for 1 h at room temperature. Cells were rinsed with water and stained en bloc with 0.5% aqueous uranyl acetate in 25% methanol overnight at 4 °C. After five rinses with water, specimens were stained with 0.02 M lead nitrate in 0.03 M L-aspartate for 30 minutes. Samples were dehydrated with increasing concentration of ethanol, infiltrated with Hard Embed-812 resin and polymerized at 70 °C for 48 hours. Embed-812 discs were removed from the plastic housing by submerging the dish in liquid nitrogen. Paired disks of the same sample were then assembled with a drop of Hard Embed 812 resin at the center and polymerized for 24 hours. Blocks were sectioned with a diamond knife (Diatome) on a Leica Ultracut UCT(7) ultramicrotome (Leica Microsystems) and collected onto copper grids, post stained with 2% uranyl acetate in water and lead citrate. Images were acquired on a JEOL JEM-1400 Plus transmission electron microscope equipped with a LaB6 source operated at 120 kV using an AMT-BioSprint 16M CCD camera.

Live imaging of monolayer migration

Sorted or mixed populations of keratinocytes were prepared depending on the experiment and seeded into collagen-coated 6-well tissue culture plates. Cells were grown to confluence, then the culture medium was changed from K-SFM to differentiation medium. After 24 hours, a linear scratch was created in each monolayer using a 1000 μL pipette tip. Monolayers were washed three times with PBS, which was then replaced with fresh differentiation medium. Samples were imaged using an inverted phase-contrast and epifluorescence Nikon ECLIPSE Ti microscope equipped with a motorized and programmable stage, a Nikon Perfect Focus System, a Hamamatsu Orca-Flash4.0 scientific CMOS camera, a SOLA solid state white-light excitation system, an OKO lab custom-built environmental chamber with temperature control and CO2 stage incubator, and Nikon Elements acquisition software. The environment was maintained at 37 °C with 5% CO₂ during the imaging period. Beginning 30 minutes after wounding, phase-contrast and fluorescence images were acquired every 6 minutes using a Plan Fluor 10× 0.3NA objective.

Analysis of migration images was based on a previously established computer vision pipeline for migration velocity estimation and wound edge detection³⁸. In brief, velocity measurements were computed from the phase-contrast images using a particle image velocimetry (PIV)-based method with a 15-μm square patch size. The pipeline includes a correction for possible microscope stage drift. Monolayer contours were then determined by segmentation of the velocity map, based on the intuition that while the cells are moving, there should be no motion in the empty wound area. The algorithm imposes two additional constraints to increase robustness of the segmentation. First, the monolayer is assumed to be a continuous region. Second, the monolayer contour is assumed to move monotonically in one direction. The second condition means that the pipeline will fail if the monolayer edge retracts. Monolayers where this occurred were excluded from the analysis. While the pipeline is robust to the presence of stationary debris or irregularities in the surface of the wound area, floating debris may interfere with velocity estimation and identification of the monolayer contour. Therefore, images with floating debris were also excluded from the analysis.

Downstream analysis steps were performed using custom MATLAB scripts. For uniform cell populations, two measures of migration were calculated. First, local migration speed, as estimated by particle image velocimetry, was averaged over an area between 10-μm and 200-μm from the monolayer contour. Because the monolayer contour as estimated from 15-μm PIV patches is necessarily imprecise, the 10-μm closest to the monolayer contour was excluded. Average local migration speed followed a consistent pattern, quickly increasing to a peak speed, then gradually declining over the course of the experiment (Figures S2B and 3C). However, the timing of this peak varied between individual scratch wounds. For more robust comparisons, migration speeds traces were aligned in time based on the time of peak migration speed for each wound. Second, area closed was calculated based on advancement of the monolayer contour. This measure was not adjusted for differences in timing of peak migration speed.

For mixed cell populations, monolayers were additionally segmented into regions with different keratin expression patterns. To accomplish this, each of the fluorescence images corresponding to tagged keratins were segmented using the Rosin thresholding method⁷⁸. These masks were then used to define high and low expression regions for each keratin. Average local migration speeds were calculated separately for each keratin expression region in the 10-μm to 200-μm band from the monolayer contour, as well as for the band as a whole. As with the analysis for uniform cell populations, migration speed traces were aligned in time based on the time of peak migration speed. Since area closed is an overall measure that does not allow comparisons between different keratin expression regions, it was not considered in analysis of the mixed population monolayers.

Fixed imaging of monolayer migration

Sorted or mixed populations of keratinocytes were prepared and seeded onto collagen-coated 18 mm-diameter #1.5 coverslips in tissue culture dishes. After cells were grown to confluence, the culture medium was switched from K-SFM to differentiation medium. After 4 hours, a line scratch was created in each coverslip using a 1000 μL pipette tip. The cultures were then returned to the incubator for 24 hours to allow migration to occur. After that period, coverslips were removed from the incubator, washed 3 times with PBS, incubated in 4% paraformaldehyde (Electron Microscopy Sciences #15713) in PBS for 10 minutes, incubated in 0.1% Triton X-100 in PBS for 8 minutes, and washed 3 more times with PBS. For experiments involving immunofluorescence labeling, samples were incubated in primary antibody diluted in PBS for 30 minutes at 37 °C, washed 3 times with PBS, incubated in fluorophore-conjugated secondary antibody diluted in PBS for 30 minutes at 37 °C, and washed 3 more times with PBS. Finally, coverslips were mounted onto glass slides using Fluoromount-G with DAPI (ThermoFisher 00–4959-52).

Primary antibodies used include Rabbit anti-keratin 5 (Abcam ab52635) at 1:200; Rabbit anti-keratin 6 (Abcam ab93279), epitope conserved in K6A, K6B, and K6C at 1:50; and Rabbit polyclonal anti-keratin 17 (Abcam ab53707) at 1:1,000. Secondary antibodies used at 1:500 dilution include Alexa Fluor 488, goat anti- mouse IgG (Invitrogen A-11209); Alexa Fluor 488, goat anti- rabbit IgG (Invitrogen A-11008); Alexa Fluor 568, goat anti- mouse IgG (Invitrogen A-11004); Alexa Fluor 568, goat anti- rabbit IgG (Invitrogen A-11036); and Alexa Fluor 647, goat anti- rabbit IgG (Invitrogen A-21244).

For analysis of relative enrichment of cells from a mixed population with different keratin expression levels at the monolayer edge (Figure S2E), slides were imaged using an inverted epifluorescence Nikon ECLIPSE Ti microscope equipped with a Zyla sCMOS camera (Andor), SOLA solid state white-light excitation system, and μManager acquisition software⁷⁹. Partially overlapping images were acquired using a Plan Fluor 40× 1.3NA objective, then stitched together pairwise using the Fiji Image Stitching Plugin⁸⁰. Analysis was performed on the composite stitched images. Monolayer edges were manually segmented using ImageJ. Individual cells were segmented using Cellpose version 1.0 with the pretrained “cyto” model⁸¹. A normalized sum of the tagged keratin fluorescence signals was used as input for the cytoplasm channel, and DAPI signal was used as input for the nuclear channel. Segmentations were manually reviewed, and any image areas with grossly incorrect cell segmentations were excluded from analysis. A custom Python script was used for downstream analysis. The median tagged keratin fluorescence signal was calculated for each cell. Cells were annotated as expressing high or low levels of each keratin using a linear discriminator function developed based on thresholds set between bimodal peaks in the fluorescence signal distributions. The fraction of cells expressing a high level of each keratin was then compared between the first 100-μm from the monolayer edge and the next 400-μm.

For immunofluorescence imaging of endogenous keratin expression in migrating monolayers (Figure 1D–E), slides were imaged using an inverted epifluorescence Nikon ECLIPSE Ti microscope equipped with a Zyla sCMOS camera (Andor), SOLA solid state white-light excitation system, and μManager acquisition software⁷⁹. Images were acquired using Plan Apo 20× 0.75NA and Plan Apo 40× 1.3NA objectives.

Live imaging of single-cell migration

Keratinocytes were sparsely seeded onto collagen-coated #1.5 glass-bottom culture dishes (Cellvis D35–20-1.5-N). Cells were imaged using an inverted phase-contrast and epifluorescence Nikon ECLIPSE Ti microscope equipped with a motorized and programmable stage, a Nikon Perfect Focus System, a Hamamatsu Orca-Flash4.0 scientific CMOS camera, a SOLA solid state white-light excitation system, an OKO lab custom-built environmental chamber with temperature control and CO2 stage incubator, and Nikon Elements acquisition software. The environment was maintained at 37 °C with 5% CO₂ during the imaging period. Beginning 22-minutes, 35-minutes, or 6-hours after seeding, fluorescence images were acquired every 3 minutes for 6 to 18-hours using a Plan Fluor 10× 0.3NA objective.

Images were analyzed using the TrackMate plugin for Fiji⁸². Cells were identified using the Laplacian of Gaussian detector, and tracks were identified using the sparse linear assignment problem tracker⁸³. Downstream analysis was performed using custom R scripts. Two complementary measures of migration speed were considered. First, mean velocity of a track was defined as the length of the track divided by the track duration. Second, mean squared displacement (MSD) rate was defined as the square of track displacement (linear distance from start to end positions) divided by the track duration, averaged over the set of tracks. For freely diffusing particles, MSD increases linearly with time at a rate proportional to the diffusion coefficient⁸⁴. Since sparsely seeded cells migrating in the absence of a directional que do not fully resemble freely diffusing particles, we do not estimate a diffusion coefficient, and instead use the MSD rate as an approximation to aggregate information from multiple tracks of potentially different lengths.

Stratified epidermal migration model

We adapted an epidermal culture model from existing protocols^85,86. While epidermal equivalent cultures lack the fibroblast-containing dermal equivalent present in skin equivalent cultures ⁶⁶, their simplified design allowed us to prepare them in removable barrier molds to model epidermal migration. To prepare epidermal cultures, polycarbonate filters with 0.4μm pore size (3.14cm² area, Nunc #140640) were placed in larger tissue culture dishes. Two silicone inserts, each with two chambers separated by a 500μm gap (Ibidi #80209), were then pressed firmly into place on each filter using sterile forceps.

On day 0, a cell suspension of 2.5×10⁵ keratinocytes in 70μL of K-SFM was added into each chamber, and additional K-SFM was added to the culture dish to reach the level of the filter. Throughout the procedure, medium in the culture dish was refreshed every two to three days, and we verified that culture medium did not leak to the top surface of the filter outside of the silicone inserts. On day 3, culture medium was aspirated from above the filter within the silicone inserts, effectively lifting the epidermal culture to an air-liquid interface. At the same time, medium in the culture dish was changed from K-SFM to differentiation medium. On day 7, an additional 0.5 mM calcium chloride was added to the differentiation medium in the culture dish.

For migration experiments, the silicone inserts were removed on day 9 by grasping the insert and filter housing with sterile forceps and gently lifting the insert while holding the filter in place. The filter was then returned to the culture dish, maintaining epidermal cultures at an air-liquid interface during the migration period. This time point, 6 days after exposure of the culture to an air-liquid interface, corresponded with development of clear basal and suprabasal keratinocyte layers and an early cornified layer, but precedes the formation of a thick cornified layer, which was more easily disrupted during removal of the silicone inserts (Figure 1G). Cultures were either imaged live during migration or fixed at the appropriate timepoint.

Fixed imaging of epidermal cultures

Prior to fixation, medium was aspirated from the tissue culture dish and the top and bottom surfaces of the filters were washed twice with PBS, taking care not to dislodge the epidermal cultures from the top surface of the filters. The filters were then removed from the culture dishes and incubated overnight in 4% paraformaldehyde solution at 4 °C (Electron Microscopy Sciences #15713). In preparation for paraffin embedding and sectioning, the filters were cut out from their plastic housing using a #15 scalpel blade (GF Health Products 2976#15), bisected perpendicularly to the migration edge, and embedded in a 2.5% agarose gel (Lonza SeaKem LE Agarose #50000). Samples were submitted to the UT Southwestern Histo Pathology Core Facility for paraffin embedding, sectioning, and preparation of hematoxylin and eosin (H&E) stained and unstained slides.

In preparation for immunofluorescence staining, unstained paraffin-embedded tissue sections on glass slides were deparaffinized by 3 successive 5-minute incubations in xylene followed by rehydration in 3 successive 3-minute incubations in 100% ethanol and 2 successive 2-minute incubations in deionized water. For antigen retrieval, slides were placed in Tris-EDTA buffer (10 mM Tris base, 1 mM EDTA, 0.05% Tween 20, pH 9), heated to 95 °C in a microwave oven, and incubated until the buffer cooled to room temperature, approximately 45 minutes. Slides were rinsed three times in Hanks Balanced Salt Solution (HBSS) for 5 minutes each, and a PAP pen was used to circle the tissue sections. Slides were then incubated in blocking solution [5% goat serum (Abcam ab7481), 5% bovine serum albumin (Equitech-Bio BAH65–0500), and 0.5% Triton X-100 (Sigma X100) in PBS] for 1 hour. Primary antibodies were diluted in PBS and incubated on the tissue sections overnight at 4 °C in a humidified chamber. Following primary antibody incubation, slides were washed 3 times with HBSS for 10 minutes each. Secondary antibodies were diluted in PBS and incubated on the tissue sections for 60 minutes. Finally, the slides were washed 3 times with HBSS for 10 minutes each, and a small amount of Fluoromount-G with DAPI (ThermoFisher 00–4959-52) was used to affix coverslips to the slides.

Primary antibodies used include Rabbit anti-keratin 5 (Abcam ab52635) at 1:200; Rabbit anti-keratin 6 (Abcam ab93279) at 1:50; and Rabbit polyclonal anti-keratin 17 (Abcam ab53707) at 1:100. Secondary antibodies include the same set of secondary antibodies listed for immunofluorescence above.

H&E slides were imaged using an Olympus BX53 upright microscope with brightfield illumination, an Olympus DP21 2.01 megapixel color CCD camera, and Plan N 4× 0.1NA, 10× 0.25NA, 20× 0.4NA, and 40× 0.65NA objectives. For migration assays, the gap width between epidermal culture pairs was measured using ImageJ by manually drawing polyline regions of interest along the filter in the gap. Immunofluorescence slides were imaged using an inverted epifluorescence Nikon ECLIPSE Ti microscope equipped with a Zyla sCMOS camera (Andor), SOLA solid state white-light excitation system, and μManager acquisition software⁷⁹. Partially overlapping images were acquired using a Plan Fluor 40× 1.3NA objective, then stitched together pairwise using the Fiji Image Stitching Plugin⁸⁰.

Live imaging of epidermal culture migration

Epidermal cultures were imaged using an inverted phase-contrast and epifluorescence Nikon ECLIPSE Ti microscope equipped with a motorized and programmable stage, a Nikon Perfect Focus System, a Hamamatsu Orca-Flash4.0 scientific CMOS camera, a SOLA solid state white-light excitation system, an OKO lab custom-built environmental chamber with temperature control and CO2 stage incubator, and Nikon Elements acquisition software. With epidermal cultures on filters sitting in tissue culture dishes placed on the stage of the inverted microscope, images were acquired from below, through the semi-translucent filter (Figure 2H). The environment was maintained at 37 °C with 5% CO₂ during the imaging period. Beginning 15 minutes after barrier removal, fluorescence images were acquired every 12 minutes using a Plan Fluor 10× 0.3NA objective.

Although resolution was limited by imaging through the filters, bright fluorescence signals from the tagged keratin constructs allowed us to identify cell locations. To analyze the resulting fluorescence images, we modified the monolayer migration analysis pipeline described above³⁸ as follows: First, all images were smoothed using a Gaussian filter with a 3.25-μm (5 pixel) standard deviation to decrease the influence of the filter mesh pattern on the PIV calculations. Second, since the initial texture-based segmentation seed frequently failed with fluorescence images instead of phase-contrast images, we used Otsu thresholding as an alternative⁸⁷. Finally, because interpretation of the PIV-based local velocity estimates was complicated by potential differences in cell motion at different vertical levels within the three-dimensional epidermal culture, migration distance and speed calculations were based only on the distance travelled by the monolayer contour.

Keratin filament network analysis

Keratinocyte cell lines expressing fluorescently tagged keratins were seeded onto collagen-coated #1.5 glass-bottom culture dishes (Cellvis D35–20-1.5-N). For knockdown experiments, cells were transduced with lentivirus to express targeting shRNA (Table S1) or a scrambled control 72 hours before analysis. Once cells grew to discrete colonies, but before reaching confluence, the culture medium was switched from K-SFM to differentiation medium. After 8 hours, cells were imaged using an inverted phase-contrast and epifluorescence Nikon ECLIPSE Ti microscope equipped with a motorized and programmable stage, a Nikon Perfect Focus System, a Hamamatsu Orca-Flash4.0 scientific CMOS camera, a SOLA solid state white-light excitation system, an OKO lab custom-built environmental chamber with temperature control and CO2 stage incubator, and Nikon Elements acquisition software. The environment was maintained at 37 °C with 5% CO₂ during the imaging period. Images were acquired every 1 minute for 4 minutes using a Plan Apo 60× 1.4NA objective, generating an effective pixel size in object space of 108 nm. For applicable experiments, 30 μM Blebbistatin (Sigma B0560), 5 μM Y27632 (Selleckchem S1049), 1 μM GSK 269962A (Selleckchem S7687), or DMSO (Fisher BP231) control were added 1 hour (Blebbistatin) or 2 hours (other inhibitors) prior to imaging.

Images were analyzed using the previously published u-delineate package written in MATLAB³⁹. In brief, the pipeline segments the filament network using a multi-scale steerable filter and iterative graph matching to connect fragments into complete filaments. The pipeline also calculates the local structural similarity of two filament networks based on the distance and orientation of individual filament pixels in the networks. Each pixel in the first network is matched to the closest pixel in the second network within a defined radius, and vice versa. A radius of 2.2 μm was used for all studies here. For each matched pair, the difference in position and filament orientation are calculated and used to create distance and angle maps, which are spatially smoothed by a Gaussian filter, combined, and scaled so the final similarity score decreases monotonically from 1 to 0 as the distance increases to the search radius or the angle increases to π/3. The dynamics score is defined as 1 minus the combined similarity score. The final dynamics score is 0 for identical networks and increases to 1 with increasing local differences in filament position or orientation.

To measure the effect of keratin filament composition on filament network dynamics within cells, the keratin filament images at baseline were compared to images 1, 2, 3, or 4 minutes later (Figure 4A–B). Because this analysis cannot reliably distinguish movement of filaments within cells from movement of the cells themselves, all image sequences were pre-screened to ensure that only stationary cells were included in the analysis. Cells were manually segmented using ImageJ, and the u-delineate package was run to segment the keratin filament networks and calculate dynamics score maps (Figure 4C–D). The dynamics score map was then averaged over each cell area mask, excluding regions without a filament match. As a check, a quality score was computed for each cell, defined as the fraction of the cell mask area with a filament match, which was typically higher than 0.9. Image sequences of cells with low quality scores were reviewed by visual inspection. They tended to show failed filament segmentation due to low signal-to-noise ratio, debris, or cell movement, which was missed in the initial screen; these cells were excluded from subsequent analysis. For the filament dynamics analysis of keratinocytes expressing different levels of multiple keratins, the quality score on each channel also served as a robust marker of high versus low expression of each keratin, with clear bimodal distributions (Figure S3J).

In addition to the dynamics score, the u-delineate package was used to calculate additional keratin filament network properties from the baseline keratin filament segmentation. Filament density was defined as the fraction of cell area pixels segmented as filaments. Filament straightness was defined as the ratio of the linear start to end distance to the filament length. Filament curvature was defined as the mean curvature of a polynomial fit at each pixel along the filament. The median filament length, straightness, and curvature were calculated for each cell. An alternate cell summary measure of filament curvature was also calculated as the median curvature of each pixel in all filaments in the cell.

Traction force microscopy

Silicone gel substrates for traction force microscopy (TFM) were prepared in #1.0 glass-bottom 35-mm tissue culture dishes (Cellvis D35–14-1.0-N). Prepolymers for high-refractive index (n = 1.49) 8-kPa gels were prepared by mixing components A and B of QGel 920 (Quantum Silicones) in a 1:1.15 ratio as previously described^43,88. The prepolymers were spread onto the coverglass bottoms of the dishes by spinning at 1,250 r.p.m for 30 seconds in a spin-coater (Laurell Technologies WS-650Mz-23NPPB), then cured at 100 °C for 2 hours, producing gels approximately 35-μm thick. The gels were then treated with 5% (3-aminopropyl)triethoxysilane (APTES) in ethanol to functionalize their surfaces. To allow for visualization and tracking of substrate deformation, 40-nm carboxylate far-red fluorescent beads were covalently linked to the gel surfaces by incubating the gels under a suspension of the beads (1:5,000 dilution from the 5% stock suspension) in 40-mM HEPES, pH 8, with 0.01% 1-ethyl-3-(3-dimethylaminopropyl) carbodiimide (EDC) as a catalyst. To facilitate cell adhesion during cell seeding, the substrates were coated with fibronectin by incubation with 50-μg/ml of fibronectin and 1 mg/ml EDC in PBS for 15 minutes at room temperature. The coated dishes were washed three times with PBS and filled with cell culture medium before cell seeding.

To measure keratinocyte traction force generation, 4×10⁴ cells were seeded onto a substrate in differentiation medium, then returned to the incubator for 4.5 hours to allow attachment to occur. For knockdown experiments, cells were transduced with lentivirus to express targeting shRNA or a scrambled control 72 hours before seeding onto TFM substrates. Following attachment, cells were imaged using a GE DeltaVision OMX SR inverted microscope equipped with a motorized and programmable stage, three PCO sCMOS cameras, solid state lasers (405nm, 488nm, 568nm, and 640nm), and environment control. Images of the substrate beads and keratin tags were acquired in Ring-TIRF mode with a U Plan Apo 60× 1.49NA objective (generating an effective pixel size in object space of 80 nm), with imaging performed at 37 °C. Following initial imaging, during which cell positions were recorded, a prewarmed solution of dilute bleach was added to remove cells from the substrate, thereby allowing repeat imaging of the same locations with the substrate in its relaxed, strain-free state.

Image analysis was performed using the previously published u-inferforce package written in MATLAB^43,89. Briefly, the substrate deformation field was determined by tracking the displacement of each bead between the deformed and relaxed images using cross-correlation analysis. The displacement field, after outlier removal and stage drift correction, was then used to estimate the traction field over areas of interest surrounding each cell. Traction fields were estimated using the boundary element method⁴³. While more computationally intensive than the more commonly used Fourier transform traction cytometry method, the boundary element method allows L1-norm rather than L2-norm regularization, which resolves force variation at shorter length scales⁴³. A regularization parameter of 0.05 was selected based on L-curve analysis of a subset of images⁴³. This parameter value was applied to all images analyzed to permit comparison of the resulting force fields. Cell area was segmented based on the tagged keratin signal. Since this signal did not always clearly extend to the cell boundary (for example, Figure 5A), masks were dilated by 1.6 μm (20 pixels), although we verified that the precise amount of dilation did not meaningfully affect the results (Figure S4C). Strain energy, which represents the mechanical work of the cell on the elastic substrate, was calculated as ½ × (displacement · traction), integrated over the segmented cell area⁹⁰.

Proximity ligation assay

Mixed populations of keratinocytes were prepared and seeded onto 8-chamber glass slides with removable wells (Lab-Tek, ThermoFisher #154534). Cells were grown to approximately 50% confluence, then culture medium was switched from K-SFM to differentiation medium for 24 hours. For knockdown experiments, cells were transduced with lentivirus to express targeting shRNA (Table S1) or a scrambled control 72 hours before processing. Slides were removed from the incubator and washed three times with PBS. For experiments using antibodies against keratins, the slides were incubated in methanol for 2 minutes. For other experiments, the slides were incubated in 4% paraformaldehyde in PBS for 10 minutes, then incubated in 0.1% Triton X-100 in PBS for 8 minutes. Slides were then washed twice with PBS and processed using Duolink Proximity Ligation Assay (PLA) reagents (In-Situ PLA Probe Anti-Mouse MINUS, Sigma DUO92004; In-Situ PLA Probe Anti-Rabbit PLUS, Sigma DUO92002; In-Situ Detection Reagents FarRed, Sigma DUO92013) according to the manufacturer’s in situ fluorescence protocol. Briefly, slides were incubated in the provided Blocking Solution for 60 minutes, incubated in primary antibodies diluted in the provided Antibody Dilutant solution for 30 minutes, incubated in PLA probe solution for 60 minutes, incubated in the ligation solution for 30 minutes, and incubated in the amplification solution for 100 minutes, with washes in the provided buffers between each step. Finally, a #1.5 cover glass was affixed to the slide using Duolink In-Situ Mounting Medium with DAPI (Sigma DUO82040).

Primary antibodies used include Rabbit anti-keratin 5 (Abcam ab52635) at 1:200; Rabbit polyclonal anti-keratin 17 (Abcam ab53707) at 1:1,000; Rabbit polyclonal anti-non-muscle myosin heavy chain IIA (MYH9; Invitrogen PA5–17025) at 1:200; and Mouse anti-phospho-myosin regulatory light chain (Cell Signaling Technology #3675) at 1:100.

Slides were imaged using an inverted phase-contrast and epifluorescence Nikon ECLIPSE Ti microscope equipped with a motorized and programmable stage, a Nikon Perfect Focus System, a Hamamatsu Orca-Flash4.0 scientific CMOS camera, a SOLA solid state white-light excitation system, an OKO lab custom-built environmental chamber with temperature control and CO2 stage incubator, and Nikon Elements acquisition software. Images were acquired on the tagged keratin, PLA reaction, and DAPI channels using a Plan Apo 60× 1.4NA objective (generating an effective pixel size in object space of 110 nm). Image analysis was performed using custom Python and MATLAB scripts. Individual cells were segmented using Cellpose version 1.0 with the pretrained “cyto” model⁸¹. A normalized sum of the tagged keratin fluorescence signals was used as input for the cytoplasm channel, and DAPI signal was used as input for the nuclear channel. The median tagged keratin fluorescence signal was calculated for each cell. Cells were annotated as expressing high or low levels of each keratin using a linear discriminator function developed based on thresholds set between bimodal peaks in the fluorescence signal distributions. Segmentations and classifications were manually reviewed, and any incorrectly segmented or classified cells were excluded from analysis. Cells that were poorly adhered to the slide or out of focus were also excluded. Next, individual local intensity clusters in the scaled PLA image were detected using previously published MATLAB code combining wavelet denoising and multiscale products of wavelet coefficients^91,92. Finally, PLA detections were counted within each segmented cell mask, and the density of PLA detections was compared between cells with different keratin expression levels.

Analysis of cell-cell junction density

Cells were prepared and seeded onto collagen-coated 18-mm-diameter #1.5 coverslips in tissue culture dishes. After reaching 25% confluence, the culture medium was switched from K-SFM to differentiation medium. After 5 hours, coverslips were removed from the incubator, washed 3 times with PBS, incubated in 4% paraformaldehyde (Electron Microscopy Sciences #15713) in PBS for 10 minutes, incubated in 0.1% Triton X-100 in PBS for 8 minutes, and washed 3 more times with PBS. Coverslips were then incubated in either Alexa Fluor 488 phalloidin (ThermoFisher A12379) or Alexa Fluor 555 phalloidin (ThermoFisher A34055) diluted 1:400 in PBS for 30 minutes at 37 °C. Finally, coverslips were mounted onto glass slides using Fluoromount-G with DAPI (ThermoFisher 00–4959-52). Slides were imaged using an inverted Nikon Ti-Eclipse microscope equipped with a pco.edge sCMOS camera with 6.5-μm pixel size (PCO), an Andor Diskovery illuminator coupled to a Yokogawa CSU-X1 confocal spinning disk head with 100 nm pinholes, and an Apo TIRF 60× 1.49NA objective (Nikon) with an additional 1.8× tube lens (yielding a final magnification of 108×; Andor Technology). This setup generated an effective pixel size in object space of 120 nm. Actin and keratin filament networks were segmented using the u-delineate package as described above in the section “Keratin filament network segmentation and analysis” based on images of phalloidin and tagged keratins respectively³⁹. Approximate cell-cell boundaries were manually annotated using ImageJ. Adherens junctions and desmosomes, defined as segmented actin and keratin filaments crossing annotated cell-cell boundaries, were counted using the Slide Set plugin⁹³. Junction density was defined as the number of junctions detected along a cell-cell boundary divided by the length of the boundary. To account for potential inconsistency in the start and end positions of the manually annotated cell-cell boundaries, boundary length was trimmed by the outermost junctions. Junction density was then compared between cell lines with different keratin expression levels.

Cell proliferation assay

Cells were prepared and seeded into collagen-coated 12-well tissue culture plates, 4×10⁴ cells per well. For knockdown experiments, cells were transduced with lentivirus to express targeting shRNA or a scrambled control after 4 hours, then analyzed after 96 hours with regular changes of culture medium. H2B-mCherry3 co-expression from the shRNA constructs allowed us to confirm an infection efficiency close to 100% (see “Plasmids” above). For direct cell line comparisons without shRNA knockdown, cells were analyzed after 4 days in standard culture conditions. Following the growth period, cells were incubated in 5-ethynyl-2’-deoxyuridine (EdU; ThermoFisher A10044) dissolved in K-SFM (20 mM final concentration) for 60 minutes at 37 °C. Cells were then washed 3 times with PBS, incubated in 4% paraformaldehyde (Electron Microscopy Sciences #15713) in PBS for 10 minutes, incubated in 0.1% Triton X-100 in PBS for 8 minutes, and washed 3 more times with PBS. To detect EdU incorporation into nuclear DNA, cells were incubated in a reaction mixture containing 20 mg/mL L-ascorbic acid (Sigma A92902), 0.5 mg/mL anhydrous copper(II) sulfate (Acros #42287), and 1 μM Cy5-azide (Sigma #777323) in PBS for 30 minutes at 37 °C. Cells were washed 3 times with PBS, then incubated in DAPI (ThermoFisher D1306) diluted 1:5,000 in PBS for 15 minutes at room temperature to provide a nuclear counter-stain. After 3 final washes in PBS, cells were imaged using an inverted phase-contrast and epifluorescence Nikon ECLIPSE Ti microscope equipped with a motorized and programmable stage, a Nikon Perfect Focus System, a Hamamatsu Orca-Flash4.0 scientific CMOS camera, a SOLA solid state white-light excitation system, an OKO lab custom-built environmental chamber with temperature control and CO2 stage incubator, and Nikon Elements acquisition software. DAPI and Cy5 images were acquired using a Plan Fluor 10× 0.3NA objective, generating an effective pixel size in object space of 650 nm.

Images were segmented using Cellpose version 1.0 with the pretrained “cyto” model⁸¹. The DAPI signal was used as input for the “cytoplasm” channel, and no input was provided for the “nuclear” channel, producing cell nucleus masks. The median Cy5 signal was calculated for each mask, and cells were annotated as EdU-positive (Cy5-high, proliferating) or EdU-negative (Cy5-low, quiescent) based on comparison of the median Cy5 signal to a threshold set between bimodal peaks in the Cy5 signal distribution.

Imaging of clinical wound samples

Routine skin biopsies performed for diagnostic purposes result in epidermal wounds, which typically heal over 2 to 6 weeks. If a malignant lesion is identified in the biopsy, an excision may be recommended to ensure that the malignancy is completely removed. If an excision is performed before the biopsy wound has completely healed, the resulting specimen will incidentally contain the re-epithelializing biopsy wound edges. We obtained anonymized H&E stained and unstained tissue sections from an excision specimen in which the re-epithelizing biopsy wound edges were visible, but no residual carcinoma was present. Unstained slides were processed for immunofluorescence and imaged using the procedures describe above in “Fixation, staining, and imaging of epidermal cultures.” This study was approved by the UT Southwestern Medical Center Institutional Review Board.

Clinical sample transcriptomics analysis

Two previously published transcriptomics datasets comparing wounded to intact skin were reanalyzed for keratin expression using custom R scripts. Iglesias-Bartolome and colleagues collected serial punch biopsy specimens from the arm and buccal mucosa, extracted RNA from the samples, and performed RNA sequencing²⁴. Unwounded skin was represented by the initial samples, and wounded skin was represented by follow-up samples taken from the same locations. We downloaded the expression data (Gene Expression Omnibus GSE97615) for the arm skin samples and plotted keratin transcript levels (log-transformed RPKM). Statistical analysis was not performed. Ramirez and colleagues collected tissue samples from diabetic foot ulcers and non-ulcerated foot skin, extracted RNA from the samples, and profiled transcripts using Affymetrix GeneChip Human Gene 2.0 ST microarrays^25,94. We downloaded the expression data (Gene Expression Omnibus GSE80178) and used the “oligo” R package to import, annotate, and preprocess the microarray data, including background-correction and normalization using the RMA algorithm⁹⁵. Keratin expression intensity values for wound and control samples were plotted, and statistical analysis was not performed.

Quantification and statistical analysis

All statistical analysis was performed using R (R Foundation for Statistical Computing). The Kruskal-Wallis rank sum test was used to evaluate non-parametric scaled data, with Dunn’s method for multiple comparisons used for experiments involving more than two groups⁹⁶. Two-tailed P values less than 0.05 are considered to be significant and are shown in figures or indicated in legends.

Additional resources

Danuser Lab software webpage: https://github.com/DanuserLab

Supplementary Material

1

Table S1. Primers and synthetic DNA sequences, Related to STAR Methods.

2

Video S1. Wound-associated K6A supports migration in three dimensional epidermal cultures, related to Figure 2.

Paired epidermal cultures of opposite cell lines initially separated by a 500-μm gap and imaged live over 24.6 hours at a sampling rate of 12 minutes per frame. Playback at 12 frames per second. Green and red lines indicate the farthest extent of migration for each culture. See also Figure 2H–K.

3

Video S2. Mosaic monolayer migration assays reveal subtle differences between K5^high and K6A^high cells during migration, related to Figure 3.

A mixed population monolayer containing K5^high, K6A^high, K5^high/K6A^high, and K5^low/K6A^low cells were scratched to create a wound and imaged over 12.5 hours at a sampling rate of 6 minutes per frame. Playback at 30 frames per second. See also Figure 3A–F. Keratin expression regions in the monolayer were segmented based on the fluorescence signal from the tagged keratin constructs (left). Local migration speeds computed in blocks of 15-μm side-length by particle image velocimetry (right; see Methods).

4

Video S3. Wound-associated K6A supports a transient migration advantage in single cells, related to Figure 3.

Individual keratinocytes were sparsely seeded and allowed to migrate without exogenous directional cues. Imaged over 18 hours at a sampling rate of 3 minutes per frame. Playback at 24 frames per second. See also Figure 3G–I.

5

Video S4. Wound-associated K6A alters keratin filament dynamics, related to Figure 4.

Keratin filaments in K5^high, K6A^high (plays first), and K5^high/K6A^high (plays second) cells were imaged live over 24 minutes at a sampling rate of 30 seconds per frame. Playback at 24 frames per second. Filament networks were segmented using a computer vision pipeline (see Methods). Filament segmentations are overlayed on the tagged keratin fluorescence images. See also Figures 4A–E and S3I–O.

Highlights.

• During skin wound healing, the isoform composition of keratin filaments changes

• Wound-associated keratin isoforms organize cellular signals to activate myosin

• Keratin isoform-dependent signaling is separable from canonical mechanical function