Disease-associated keratin mutations reduce traction forces and compromise adhesion and collective migration

Cells sense and respond to surrounding mechanical cues by activating mechanotransduction pathways, which govern fundamental processes, including development, tissue repair and homeostasis (DuFort et al., 2011; Eyckmans et al., 2011; Evans et al., 2013; Iskratsch et al., 2014; Wang et al., 2015; Hsu et al., 2018). Tensile forces on the actin cytoskeleton and intermediate filaments (IFs) promote formation and maturation of adhesive structures. Adaptation to their mechanical environment requires cells to balance outside and inside forces by exerting appropriate forces against adhesion sites. Basal skin keratinocytes are under tension at the cell–extracellular matrix (ECM) and at cell–cell contact sites, and rely on keratin IFs to provide essential structural support by connecting to the ECM via hemidesmosomes and to neighboring cells via desmosomes (Block et al., 2005; Hatzfeld et al., 2017; Etienne-Manneville, 2018). Mutations in keratin or in desmosomal and hemidesmosomal protein-coding genes cause cell and tissue fragility and compromise mechanical resilience, resulting in skin blistering such as that observed in epidermolysis bullosa simplex (EBS) (Coulombe and Lee, 2012; Homberg and Magin, 2014). Mutation of keratin 14 (K14; also known as KRT14) at residue K14_R125 (corresponding to the mouse K14_R131 residue) is the most frequent mutation associated with EBS (www.interfil.org/; Szeverenyi et al., 2008) and causes extensive keratin aggregation and adhesion defects (Homberg et al., 2015; Seltmann et al., 2015). Although these data establish that disrupted keratins cause EBS through compromising mechanical resilience in keratinocytes, the overall contribution of keratins to mechanotransduction remains unclear.

Rho-family GTPases play key roles in actin cytoskeletal reorganization and focal adhesion (FA) formation (Chrzanowska-Wodnicka and Burridge, 1996; Ishizaki et al., 1997; Jaffe and Hall, 2005). In response to mechanical stimuli, increased RhoA activity adapts cells to their mechanical environment. In this setting, keratin 18 (K18; also known as KRT18) mediates tensional force-induced RhoA activation and stress fiber reinforcement (Fujiwara et al., 2016), suggesting that keratins are important regulators of RhoA activity. Changes in local microenvironment, particularly force distribution, may affect keratin-mediated signaling pathways through changing keratin network conformation and/or properties, following posttranslational modifications (Snider and Omary, 2014; Loschke et al., 2015). Moreover, keratins act by scaffolding signaling proteins such as Src, PKC, TRADD, 14-3-3, and Fas receptors (Gilbert et al., 2001; Inada et al., 2001; Kröger et al., 2013; Rotty and Coulombe, 2012; Sanghvi-Shah and Weber, 2017; Mariani et al., 2020).

Here, we investigated the involvement of keratins in mechanotransduction of basal epidermal keratinocytes. We compared mouse keratinocytes stably expressing wild-type (WT) K14 or the K14_R131P mutant in cells lacking all type I keratin genes (KtyI^−/−) (Homberg et al., 2015). We report that intact keratin networks are critical for traction force generation, focal adhesion formation, stiffness-dependent cell adhesion, and directed collective migration. Notably, the EBS-associated K14_R125/131 mutation significantly reduces RhoA activity, thereby compromising cell adhesion. Moreover, Src–FAK signaling pathways are involved in keratin-mediated signal transduction. Our data provide new insights into the involvement of keratin IFs in cellular mechanotransduction and the pathophysiology of EBS.

We have previously shown that mutation of keratin 14 (K14; also known as KRT14) at residue K14_R125 decreases keratinocyte stiffness (Homberg et al., 2015). Because cell mechanics is highly related to cellular mechanotransduction, we hypothesized that KtyI^−/− cells expressing wild-type K14 (K14-WT) and KtyI^−/− cells expressing K14_R131P (K14_R131P) differ in their ability to generate traction forces. Traction forces were measured using silicone elastomer substrates, which allow visualization and quantification of traction forces following wrinkle formation on substrates (Ichikawa et al., 2017; Fig. 1A,B). Wrinkle length and the percentage of cells forming wrinkles were significantly decreased in K14_R131P compared to K14-WT, indicating a lower ability to generate traction forces in the former (Fig. 1C,D). These results suggest that intact keratin networks assume key roles in traction force generation, resulting from mechanotransduction at cell–ECM adhesion sites.

To understand the molecular mechanisms of keratin network-mediated mechanotransduction, we focused on RhoA signaling, which is critical in actomyosin regulation that underlies traction force generation. Examination of active RhoA in K14_R131P compared to that in K14-WT revealed significantly lower levels in the former (Fig. 1E). The finding that active RhoA was also strongly diminished in human EBS keratinocytes expressing K14_R125C (K14_R125C) compared to normal human keratinocytes (NHK) (Fig. S2C), suggested that EBS-associated K14 mutations downregulate RhoA activity.

Traction force generation depends on proper FA formation and actin reorganization. To understand mechanisms underlying reduced traction force generation in K14_R131P compared to K14-WT, we analyzed actin and FA organization. We confirmed the expression of K14 and keratin 5 (K5; also known as KRT5) in both cell lines (Fig. S2G). K14-WT are characterized by well-organized keratin networks evenly distributed throughout the cell. By contrast, K14_R131P formed extensive aggregates throughout the cytoplasm instead of filaments (Fig. 2A; Fig. S3C; Homberg et al., 2015). Whereas most K14-WT formed pronounced peripheral actin filament bundles, K14_R131P possessed actin-rich protrusions, but lacked peripheral actin bundles and showed smaller cell size compared to K14-WT (Fig. 2A–C, Water). Vinculin staining revealed large FAs exclusively at the cell periphery in K14-WT, whereas peripheral FA were significantly reduced in K14_R131P (Fig. 2A,D, Water). Careful quantification of the FA area, based on vinculin fluorescence signals (Fig. S4), revealed a decrease in the ratio of FA to cell area in K14_R131P compared to K14-WT (Fig. 2E). Immunoblotting confirmed unaltered vinculin levels in both cell lines, indicating that the K14_R131P mutation did not affect vinculin levels but suppressed its localization to FAs (Fig. S2H). Immunostaining showed that RhoA localized diffusely throughout the cytoplasm and at the periphery in K14-WT, with some perinuclear enrichment, as previously reported (Kranenburg et al., 1997; Michaelson et al., 2001; Yonemura et al., 2004). By contrast, in K14_R131P, RhoA was distributed throughout the cytoplasm, but was diminished at the periphery (Fig. S2I). These results suggest that intact keratin networks are important for the localization of RhoA to the cell periphery to promote actin reorganization and peripheral FA formation.

The reduced RhoA activity in human EBS keratinocytes expressing K14_R125C, compared to NHK (Fig. S2C), raises the question of whether these alterations might be relevant for the pathophysiology of EBS. In agreement with previous results, K14_R125C showed aggregated keratins in the cell periphery, in addition to thick perinuclear keratin bundles, whereas NHK displayed an extensive keratin cytoskeleton (Fig. S2A; D'Alessandro et al., 2002; Russell et al., 2004). Based on vinculin staining, the ratio of FA to cell area was significantly lower in K14_R125C compared to NHK (Fig. S2B), suggesting that FA formation is downregulated in K14_R125C. Thus, K14_R125C mutations cause similar alterations in EBS keratinocytes and engineered mouse keratinocytes.

Depletion of keratin 6 (K6) in keratinocytes, and of vimentin in fibroblasts and carcinoma cells, reinforces actin cytoskeletal structures and increases cellular traction forces (Wong and Coulombe, 2003; Jiu et al., 2017). Consistent with these reports, KtyI^−/− keratinocytes showed intense vinculin staining, and the ratio of FA area to cell area was increased compared to wild-type (WT) keratinocytes (Fig. S2D,E). In addition, the activity of RhoA was significantly upregulated in KtyI^−/− (Fig. S2F). Considering the suppression of FA in K14_R131P (Fig. 2A,D), our data demonstrate that absence of keratins and expression of EBS-associated keratin mutants have opposite effects. How disease-associated keratin aggregates disturb the machinery controlling actin reorganization and FA formation requires future studies. The current data establish that intact keratin networks mediate actin reorganization and FA formation.

To examine the significance of RhoA levels in cell adhesion and morphology, we restored endogenous Rho activity using the Rho activator CN03. CN03 treatment increased phosphorylation levels of myosin light chain-2 (Myl2), as shown by immunoblot and immunofluorescence analysis (Fig. 1F,G). CN03 treatment did not significantly affect actin organization or the area and localization of FAs in K14-WT (Fig. 2A,C–E), possibly because Rho activity in these cells is already high in the experimental conditions. However, CN03 treatment increased the ratio of FA area to cell area in K14_R131P (Fig. 2E); at the same time, the peripheral distribution of actin bundles and of FAs remained largely unaltered (Fig. 2C,D). This suggests that global activation of Rho, in the presence of aggregated keratins, promotes FA maturation but is insufficient to restore actin reorganization and FA localization. We hypothesize that intact keratin networks contribute to localized RhoA activation in order to regulate actin reorganization and to promote peripheral FA formation.

The cytolinker protein plectin plays important roles in cytoskeletal network reorganization (Wiche, 1998). For example, plectin mediates the functional interplay between vimentin IFs and actomyosin arcs, and plectin depletion suppresses the dynamics of these two cytoskeletons (Jiu et al., 2015). Because actin structures were different between K14-WT and K14_R131P, we asked whether K14_R131P-induced keratin aggregation might mediate actin reorganization via plectin. In both K14-WT and K14_R131P, plectin was distributed throughout the cytoplasm and remained unaltered in both cell lines transfected with control siRNA (Fig. S3C). Treatment with a targeting siRNA effectively suppressed plectin mRNA and protein expression (Fig. S3A,B); however, it did not affect distribution or size of aggregated keratins in K14_R131P. While we cannot exclude an involvement of plectin in cell contractility and adhesion, formation and localization of K14_R131P aggregates is independent of plectin.

Matrix stiffness controls proliferation, survival and migration (Engler et al., 2006; Yeh et al., 2017; Martino et al., 2018). To further elucidate the involvement of keratin organization in mechanotransduction, we examined cell adhesion formation on soft collagen substrates (Fig. 3A,B). K14-WT attached and spread in a single layer on a gel, similar to the morphology observed on a collagen-coated glass surface, whereas the percentage of K14_R131P that attached and spread on a gel was significantly decreased (Fig. 3B–D, Water). In addition, K14-WT showed vinculin signals enriched at the cell periphery, whereas in K14_R131P, vinculin lacked an obvious enrichment at the plasma membrane (Fig. 3B, Water). To examine the involvement of Rho signaling in keratin-mediated cell adhesion on a soft substrate, endogenous Rho activity was restored using CN03 treatment. Treatment of K14_R131P with CN03 significantly increased the percentage of spread cells with peripheral FAs (Fig. 3D,E).

To substantiate the involvement of keratins in mechanotransduction, we turned to YAP/TAZ (also known as YAP1 and WWTR1, respectively), which are important regulators of cell–matrix adhesion-mediated signaling by acting as mechanosensing switches. Stiffness-dependent cell spreading is associated with YAP/TAZ nuclear translocation to exert their co-transcriptional activity (Dupont et al., 2011; Nardone et al., 2017; Wang et al., 2020). Immunofluorescence analysis revealed that YAP mainly localized in the nucleus of K14-WT, whereas it remained in the cytoplasm of K14_R131P (Fig. 3F,G). Considering that cells actively sense the rigidity of the surrounding ECM by exerting traction forces to determine cell functions, including adhesion formation (Chen et al., 2015; Jansen et al., 2015), our results raise the possibility that an intact keratin network contributes to substrate stiffness sensing and peripheral FA formation, involving Rho signaling.

FAK (also known as PTK2) is a key regulator of FAs. Because size and distribution of FAs were different between K14-WT and K14_R131P, we focused on FAK signaling. The Y397 of FAK is the main site of autophosphorylation induced by integrin binding to the ECM. Phosphorylation of Y397 generates an SH2 binding site for Src family kinases and induces phosphorylation of other residues, including Y925 (Westhoff et al., 2004; Deramaudt et al., 2011). Analysis of FAK phosphorylation showed lower phosphorylation levels of FAK-Y925 in K14_R131P compared to control cells, whereas FAK-Y397 remained unaltered (Fig. 4A–C). Analysis of Src, which phosphorylates FAK-Y925 (Brunton et al., 2005), revealed significantly lower levels in K14_R131P compared to levels in K14-WT (Fig. 4D,E). To substantiate the involvement of Src, we used the inhibitor PP1 to block Src phosphorylation and activation. PP1 indeed inhibited Src phosphorylation and significantly decreased FAK-Y925 phosphorylation levels (Fig. 4F–H). Thus, an intact keratin network is required for Src activation and Src-mediated phosphorylation of FAK-Y925. To correlate FAK activity to the state of FA, we monitored its localization. Although FAK localized along the entire basal surface and concentrated at the cell periphery in K14-WT, the peripheral localization was not obvious in K14_R131P (Fig. 4I,J). This suggests that the K14_R131P mutation affects focal adhesion formation, accompanied by altered localization and activation of FAK.

Keratinocytes often collectively migrate during fundamental biological processes, such as wound healing and morphogenesis. The dynamics of collective migration are determined by the force balance between protrusive and adhesive forces (Trepat et al., 2009; George et al., 2017). IFs are involved in collective cell migration by regulating cell adhesion. For example, knockdown of GFAP, vimentin, and nestin restricts collective migration of astrocytes by compromising dynamics of adherens junctions and FAs (de Pascalis et al., 2018), and knockout of K6 in keratinocytes disrupts desmosomal adhesion and increases FA turnover, which accelerates collective migration (Wang et al., 2018). We investigated the impact of keratin network disruption on collective migration in a wound healing assay (Fig. S1A). Time-lapse imaging showed that leader cells appeared and migrated in both K14-WT and K14_R131P collectives (Fig. S1B; Movies 1, 2). Whereas leader cells in K14-WT group migrated along relatively straight tracks oriented almost vertically to the wound edge, K14_R131P leaders showed little directionality during migration (Fig. S1B,D), and the directional persistence was significantly lower than that of K14-WT (Fig. S1C,F). The average migration velocity of K14_R131P was significantly faster than that of K14-WT (Fig. S1E). These results suggest that the migration velocity and the coordinated movement of cell groups depends on an intact keratin network. The activity of Rho GTPases plays a crucial role in directed collective migration by regulating protrusions and retractions through cytoskeletal reorganization (Lawson and Ridley, 2018). Impaired directed collective migration and reduced RhoA activity in K14_R131P is consistent with defects in directed migration in RhoA-null keratinocytes (Fig. 1E; Jackson et al., 2011). This suggests that disturbed migration results from defective RhoA activation, in addition to weakened desmosome adhesion in K14_R131P (Homberg et al., 2015).

It remains unknown how keratins contribute to the regulation of RhoA signaling. p190RhoGEF (also known as ARHGEF28) interacts with FAK and is activated by mechanotransduction (Zhai et al., 2003). The interaction of Solo (also known as ARHGEF40) with keratin 18 is important for tensional force-induced RhoA activation and traction force generation (Fujiwara et al., 2016, 2019). We hypothesize that keratin networks participate in regulation of RhoA by affecting activities and/or localization of RhoGEFs such as p190RhoGEF and Solo, contributing to traction force generation through actin reorganization and FA formation. Activation of FAs can further activate RhoA, thus traction force generation drives a positive feedback loop. Considering that FA formation is downregulated in K14_R131P, unlike in keratin knockout cells, aggregated keratins possibly inhibit RhoA activation, resulting in force imbalance. Our observations suggest that defects in force sensing and mechanotransduction associated with keratin mutations contribute to EBS.

Type I collagen (354236) was purchased from Corning (Corning, NY). Rho activator II (CN03) was purchased from Cytoskeleton (Denver, CO) and used at 4 μg/ml for 4 h. PP1 was purchased from Biomol (Hamburg, Germany) and used at 10 μM for 1 h. Alexa Fluor^®-647-labeled phalloidin (A22287) was purchased from Thermo Fisher Scientific (Waltham, MA). An antibody against plectin is from Dr H. Herrmann [Division of Cell Biology, German Cancer Research Center (DKFZ), 69120 Heidelberg, Germany]. Other antibodies were purchased as follows: anti-keratin14 (ab181595, Abcam, Cambridge, UK), anti-vinculin (V9131, Sigma Aldrich, St Louis, MO), anti-RhoA (sc-418, Santa Cruz Biotechnology, Dallas, TX), anti-GAPDH (sc-25778, Santa Cruz Biotechnology or G7895, Sigma Aldrich), anti-YAP (sc101199, Santa Cruz Biotechnology), anti-phospho FAK (Tyr397) (3283S, Cell Signaling Technology, Danvers, MA), anti-phospho FAK (Tyr925) (3284S, Cell Signaling Technology), anti-FAK (3285S, Cell Signaling Technology), anti-phospho Src (Tyr416) (6943S, Cell Signaling Technology), anti-Src (2108S, Cell Signaling Technology), anti-phospho-Myl2 (Thr18/Ser19) (3674S, Cell Signaling Technology) and anti-Myl2 (3672S, Cell Signaling Technology). Secondary antibodies used for immunoblotting and immunofluorescence analysis were purchased from Dianova (Hamburg, Germany).

The siRNA targeting the murine sequence of plectin was purchased from siTOOLs Biotech (Munich, Germany). siRNA was transfected using Lipofectamine^® RNAiMAX (Life Technologies). The siRNAs were used at a final concentration of 30 nM. Transfected cells were cultured for 48 h before being subjected to experiments. Knockdown efficiency of plectin1 mRNA expression was analyzed by reverse transcription PCR followed by quantitative PCR analysis using specific primers (plectin1 forward, 5′-CCCAGGATGAGAAGGAACAAC-3′; plectin1 reverse; 5′-GCACAAGAAAGCACACAGAAG-3′; GAPDH forward, 5′-GTGTTCCTACCCCCAATGTG-3′; GAPDH reverse, 5′-AGGAGACAACCTGGTCCTCA-3′).

Mouse keratinocytes were established and cultured as described previously (Kröger et al., 2013; Ramms et al., 2013; Homberg et al., 2015). For culturing cells on a plastic, a glass, or a silicone substrate, the surface was coated with 5 µg/cm² type I collagen. To prepare a type I collagen gel for analyzing cells cultured on a soft substrate, a collagen solution was neutralized by adding NaOH, NaHCO₃, and 10× PBS, and the concentration was adjusted to 3% with culture medium consequently to have an elastic modulus of 50–100 Pa (Valero et al., 2018; and according to the application notes of the manufacturer) and spread on 13-mm coverslips placed in a 24-well culture plate. After fabrication of the collagen by placing the plate at 37°C for at least 20 min, cells were seeded in the well and cultured for 48 h. Normal human keratinocytes (NHK) and K14_R131C cells were kindly provided by Prof. Dr C. Has (Uniklinik Freiburg, Freiburg im Breisgau, Germany) and cultured in keratinocyte serum-free medium supplemented with L-glutamine, EGF and BPE (#17005-075, Thermo Fisher Scientific).

The active form of RhoA was detected by pulldown assays using the Rho-binding domain (RBD) of rhotekin fused to GST, as described previously (Ren and Schwartz, 2000). Cells were cultured for 48–72 h in growth medium, and cell lysates were subjected to pulldown assays.

Total cell lysates were subjected to analysis of protein phosphorylation and protein expression analysis. Protein phosphorylation levels were calculated by correcting for the total amount of the proteins. Immunoblotting was performed under the following conditions: blocking with 5% skim milk in TBS-T (Tris-buffered saline containing 0.1% Tween^® 20) for 40 min at room temperature; primary antibody incubation in TBS-T containing 5% BSA overnight at 4°C; secondary antibody incubation in TBS-T containing 5% skim milk for 50 min at room temperature. Primary antibodies were used at the following dilutions: anti-keratin 14, 1:50,000; anti-vinculin, 1:1000; anti-RhoA, 1:60; anti-GAPDH, 1:2000; anti-phospho FAK (Tyr397), 1:2000; anti-phospho FAK (Tyr925), 1:2000; anti-FAK, 1:2000; anti-phospho Src, 1:2000; anti-Src, 1:2000; anti-phospho Myl2, 1:2000; anti-Myl2, 1:1000; anti-plectin, 1:2000.

Cells cultured on type I collagen-coated coverslips and on 3% type I collagen gel were fixed with 2% (w/v) paraformaldehyde in phosphate-buffered saline (PBS) at room temperature for 20 min and then permeabilized with 0.1% Triton X-100™ in PBS for 3 min. Exceptions to this were that cells were fixed with 10% TCA on ice for 15 min for staining endogenous RhoA and with cold methanol at −20°C for 3 min for double staining of plectin and K14. Cells were treated with 5% BSA in TBS for blocking and were incubated with primary antibodies for 90 min and secondary antibodies for 40 min at room temperature. Primary antibodies were used at the following dilutions: anti-keratin 14, 1:600; anti-vinculin, 1:600; anti-RhoA, 1:50; anti-YAP, 1:200; anti-FAK, 1:400; anti-plectin, 1:400; anti-phospho-Myl2, 1:100. Fluorescence images were obtained using an LSM780 confocal microscope (Carl Zeiss) equipped with a PL-Apo 40× oil objective lens (NA 1.3) for mouse keratinocytes and a PL-Apo 63× oil objective lens (NA 1.46) for human keratinocytes. Image quantifications of FAs in each cell were performed using ImageJ software (National Institutes of Health). We used maximum intensity projection images of 0.5 μm from the ventral surface for the analysis. Images were processed with the ‘Subtract Background’ command, and the ‘Threshold’ command was run to convert images to binary images before executing the ‘Analyze Particles’ command (particle size, >0.1 μm²). The region of interest (ROI) was set to the outline of each cell. To calculate the proportion of total FA area in a cell area, total FA area of each cell was divided by the cell area measured by tracing the outline of each cell.

Cell traction force was visualized using a wrinkle-generating method as described previously (Ichikawa et al., 2017), except that the silicone substrates CY 52-276 were mixed at a weight ratio of 1.2:1 and spread on coverslips. The coverslip was put on a 6-well culture plate and coated with type I collagen. Cells were seeded in a well at a density that enabled recognition of the circumference of each single cell after culturing for 48 h–72 h, and images in which the total area covered by the cells were the same were selected for image analysis. Phase-contrast images were obtained using an IncuCyte microscope (Sartorius, Göttingen, Germany) equipped with a 20× objective lens (NA 0.45) placed in a 5% CO₂ incubator. Cellular traction force generation was evaluated by manually tracing the wrinkles on the silicone surface surrounding the cell in the phase-contrast image, and the length of wrinkles was analyzed by using ImageJ software. The wrinkle length per cell was calculated by dividing the wrinkle length in a whole image by the number of cells in the image. The ratio of the mean total wrinkle length per cell of K14_R131P to that of K14-WT was calculated in each experiment.

For observation of collective cell migration, keratinocytes were seeded in a rounded rectangle-shaped silicone well (growth area 0.22 cm², Ibidi) placed on a glass-bottomed dish coated with type I collagen in a low calcium medium (0.05 mM CaCl₂) and incubated overnight. Then the culture media were replaced with a high calcium medium (2 mM CaCl₂) and incubated for an additional 24 h, and the silicone well was removed. After culture for 5 h in the high calcium medium, cells were subjected to a time-lapse analysis using a spinning disc confocal microscope (Carl Zeiss) equipped with a 10×/0.3 EC Plan-Neofluar objective lens. During image acquisition, the dish was placed in a CO₂ chamber set on the stage of the microscope, and the microscope was covered with a temperature hood to maintain the stage temperature at 32°C. Phase-contrast images were acquired every 15 min for at least 5 h. Nuclei of leader cells that did not detach from the cell group during 5 h time-lapse imaging were tracked using the manual tracking tool in ImageJ. The coordinate of each cell was calculated by correcting the angle of the wound edge as 0° and the start position as: (X₀, X₀)=(0,0), and the parameters for 5 h of migration were calculated with the following formulae, where (X_n, Y_n) indicates the coordinate of a cell at the timepoint n (min). Distance (μm): ⁠. Total travel distance (μm): ⁠. Mean velocity (μm/min): v=d_total/300. Directional persistence: p=d/d_total.

Data are expressed as the mean±s.d. of more than three independent experiments. Statistical analysis was performed using Prism 8 (Graphpad Software, La Jolla, CA). P-values were calculated using a one-way analysis of variance (ANOVA) followed by Tukey's test or a two-tailed paired t-test. The threshold for statistical significance was set at P<0.05.

We thank T. S. Matsui (Osaka University, Japan) for his help and support with wrinkle formation assay and C. Has (Uniklinik Freiburg, Germany) for providing K14_R125C.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.243956.reviewer-comments.pdf