Actomyosin-dependent cell contractility orchestrates Zika virus infection Free

Zika virus (ZIKV) has gained notoriety in recent years as there are no targeted therapies or vaccines available to date (Cao-Lormeau et al., 2016; Pierson and Diamond, 2018; Pierson and Graham, 2016; Poland et al., 2019; Rasmussen et al., 2016). ZIKV is a positive-sense, single-stranded RNA virus displaying a sophisticated life cycle that occurs entirely in the cytoplasm of the infected cell (Musso and Gubler, 2016; Sager et al., 2018). To coordinate distinct steps, from entry, replication to assembly and egress, ZIKV exploits the dynamic properties of cellular cytoskeleton, and therefore it provides an interesting system to study how cytoskeletal-based biophysical properties orchestrate its infection (Zhang et al., 2019). Of the most importance, the necessity and function of cytoskeleton-regulated cell contractility in ZIKV infection has remained elusive.

The ability of host cells to maintain morphological integrity, adhere to and exert forces on the environment largely depends on their contractility (Sweeney and Hammers, 2018). Actomyosin bundles facilitate generation of contractility via sliding of bipolar myosin II filaments along actin filaments (Pollard and Cooper, 2009). Proteins regulating actomyosin formation are thus the interior determinants for cell contractility. On the other hand, the stiffness of the extracellular matrix (ECM) is the exterior determinant for cell contractility (Gerardo et al., 2019; Patwardhan et al., 2021; Roberts et al., 2022), the underlying mechanisms of which mainly are mediated via actomyosin bundles (Park et al., 2020). Despite the importance of cell contractility, the molecular machinery controlling the host mechanics upon ZIKV infection has not been extensively examined.

In this study, we demonstrate the positive correlation between host cell contractility and ZIKV infection efficacy, and potentiate insight into development of actomyosin-dependent contractility and its regulatory molecules as novel and effective anti-ZIKV strategy.

To study whether the mechanobiological attributes of host cells respond to ZIKV infection, we examined cell contractility by performing traction force microscopy with human osteosarcoma cells (U2OS), which express abundant mechanosensing cytoskeletal actomyosin bundles and are highly susceptible to viruses (Hackett et al., 2019; Jiu et al., 2015; Rausch et al., 2017) (Fig. 1A,B). Cells were infected with the ZIKV Asian lineage strain, SZ01, with low multiplicity of infection (MOI 0.1) and measured at 12- and 24-h post infection (hpi), respectively. Intriguingly, cells exerted stronger forces to the substratum upon ZIKV infection at 12 hpi, leading to an increase of the strain energy, a parameter indicating the integrated measure of cell traction (Fig. 1A,B). At 24 hpi, however, the increased cell contractility with ZIKV infection mildly dropped. Concurrently, by using the gel deformation assay, it was revealed that there is similar increase in cell contractility at 12 hpi, which did not substantially change any further by 24 hpi (Fig. 1C,D). With a time-course of infection experiment, viral RNA replication progression was confirmed by quantitative RT-PCR (Fig. S1A), which indicated that 12 hpi is a time point when viral genome starts rapid replication, whereas 24 hpi is a time point when viral genome replication reaches maximum speed. Collectively, these results suggest that ZIKV infection modulates host cells contractility maximumly at the early stage of viral RNA replication.

Actin network and corresponding actomyosin bundles could be responsible for the cell contractility (Nishikawa et al., 2017; Svitkina, 2018; Zhou et al., 2009). This was first verified by phalloidin staining to visualize actin filaments in ZIKV-infected cells. The width of actin filaments were subsequently calculated by ImageJ, confirming that there was a significant increase in levels of thick bundles (Fig. 1E,F). The small GTPase RhoA regulates myosin II activity and the activities of several actin-binding proteins to promote contractility of actomyosin bundles and their assembly (Guilluy et al., 2011a,b; Lessey et al., 2012). We thus hypothesized that levels of active RhoA might be regulated by ZIKV infection. By using G-LISA, a small GTPase activation assay, we discovered that ZIKV infection significantly increased the level of active GTP-bound RhoA (Fig. 1G). Consistent with this, similarly elevated active RhoA level was also observed in VeroE6, another high susceptible cell for ZIKV infection (Fig. S1B). Active RhoA could promote myosin II activity by elevating the phosphorylation level of the regulatory myosin light chain (MLC, herein referring to MYL2). Accordingly, MLC showed increased phosphorylation at Ser18 and Ser19 (Fig. 1H; Fig. S1C), key sites for its activation, without any influence on the MLC mRNA level (Fig. S1D), indicative of high actomyosin contraction and consistent with the upregulated cellular force upon ZIKV infection.

We then took advantage of 3D structural illumination microscopy (3D-SIM) super-resolution imaging experiments, where the myosin II motor domain was visualized with an antibody against the regulatory light chain (RLC), and the myosin II tail domain with an antibody against its C-terminal region. The assembly of contractile actomyosin bundles involves registered alignment of myosin II filaments and their subsequent fusion into large stacks (Jiu et al., 2019). Importantly, the myosin II filament stacks were thicker and more filaments were well aligned in ZIKV-infected cells compared to uninfected Mock cells (Fig. 1I,J). Quantification of the 3D-SIM data from a region covering 5 µm from the leading edge of the cell demonstrated that the myosin II filament density as well as the average lateral stack diameter, were mildly but significantly increased in ZIKV-infected cells (Fig. 1K,J). To explore how ZIKV infection leads to the activation of RhoA, we exogenously expressed viral structural protein envelop (Env), non-structural 1 (NS1) and non-structural 4B (NS4B) proteins fused with EGFP, respectively, and measured the active RhoA levels. Interestingly, only Env expression is able to activate RhoA (Fig. 1M), indicating Env is the key underlying insight for the RhoA-actomyosin-dependent contractility. Together, these results demonstrate that there are increased contractility and enrichment of actomyosin bundles in host cells upon ZIKV infection (Fig. 1N).

To examine whether cell contractility, in turn, regulates ZIKV infection, we applied blebbistatin, a myosin II ATPase inhibitor, into the host cells. Traction forces were calculated upon 5 µM blebbistatin treatment (a common working concentration used in most study), and reduced apparently (Fig. 2A,B). Decreased phosphorylation level of the MLC by blebbistatin was subsequently verified by western blot (Fig. 2C; Fig. S2A), which support the traction force data. To further confirm that blebbistatin leads to unorganized actomyosin bundles, which are responsible for cell contraction, phalloidin staining was used to visualize actin network (Fig. 2D). Based on above measurements, we concluded that treatment with the commonly used 5 µM of blebbistatin is sufficient to decrease host cell contractility. To further narrow down the reasonable blebbistatin concentration, we checked the cell toxicity by both bright-field imaging to observe the morphology of the cell and the BrdU assay to quantify the cell viability (Fig. S2B,C), which indicates that 1 µM and 5 µM of blebbistatin application are apparently show no side effect to cells, whereas 10 µM treatment started to deform cell shapes and slightly decreased cell viability.

Given that the cell contractility increase was visualized from 12 hpi onwards, we speculated that viral invasion was not affected in blebbistatin-treated contractility-losing cells. To confirm this, we used two assays to examine the viral genome at early (binding and entry) steps with blebbistatin treatment (Fig. 2E), respectively. Cells were infected by ZIKV for 1 h at 4°C, then cell lysates were collected directly for measurement of RNA copies by qRT-PCR, or incubated for one more hour at 37°C before qRT-PCR (Fig. 2E). The results confirmed that ZIKV entry was not impaired by blebbistatin (Fig. 2F).

We next wanted to assess whether cell contractility affected ZIKV infection after invasion. To address this, we subsequently examined the infection at 24 hpi, the fast replication period during the viral life cycle, and applied blebbistatin at 0.5 h before infection (Fig. 2G) or 1 h after infection (Fig. 2I), respectively. Three readouts were quantified to affirm infection defects, including the infection proportion by immunofluorescence, the viral genome replication by quantitative RT-PCR and the viral production by titer assay. The percentage of virally infected cells at 24 hpi in both control and drug application modes were similar, with a 100% success rate (Fig. 2H,J), regardless of time point of drug application. It is apparent that treatment with blebbistatin caused a significant reduction of both viral RNA copies and virus production, in a concentration-dependent manner (Fig. 2H,J; Fig. S2D). It is notable that the extent of the defects in the case of blebbistatin treatment 0.5 h before infection resembles that for blebbistatin treatment 1 h after infection. Together, these data indicate that host cell contractility indeed regulates ZIKV infection, particularly in the after-entry time slot.

In our previous studies, Myo18B plays a critical role in maintaining higher-order myosin II stack structures for generation of mature contractile actomyosin bundles (Jiu et al., 2019). Unc45a promotes generation of contractile actomyosin bundles through synchronized myosin II folding and filament-assembly activities (Lehtimäki et al., 2017). It appears that these two genes are critical for cells to establish and maintain contractility via regulating actomyosin bundle integrity, we thus decided to use Myo18B and Unc45a gene knockout cells, respectively, to check whether abolishing contractility by gene modification would affect ZIKV infection, similar to what was seen with the drug treatment approach. Western blots with respective antibody probes were used to verify the knockout efficiency (Fig. 3A; Fig. S3A,B). Both knockouts appear to have no apparent defects on cell viability (Fig. S3C,D). Cells plated on coverslips display a large variation in their morphologies, it is not feasible to obtain quantitative information about subcellular localizations of actomyosin bundles. To overcome this obstacle, we plated wild-type and knockout cells on crossbow-shaped fibronectin-coated micropatterns, where they display regular shape. Thus, we are able to witness the loose unconsolidated actin network as well as smaller focal adhesions, a machinery to exert force to the substrate, in both gene knockout cells (Fig. 3B; Fig. S3E–G). Aligning with these micropattern data, we observed that cells indeed exerted weaker forces to the substratum by traction force microscopy in both Myo18B and Unc45a depletion cells (Fig. 3C,D).

To further dissect the role of cell contractility in ZIKV infection, we infected both Myo18B and Unc45a depletion cells with a relatively low MOI (0.1) and moderately high MOI (1) of ZIKV, and observed the infection progression at both 12 hpi and 24 hpi. Intriguingly, the infected proportion could be distinguished (Fig. 3E). Moreover, it is apparent that Myo18B and Unc45a knockout caused significant reduction of both viral RNA copies and virus production, in a MOI- and hpi-dependent manner (Fig. 3F,G). The localization of ZIKV Env has been shown to represent the viral replication factory (Zhang et al., 2022). We thus stained the Env upon ZIKV infection to visualize the replication factory in Myo18B and Unc45a knockout cells, which displays slightly dispersive distribution compared to wild-type cells, indicating a structural lose of viral replication machinery (Fig. 3H). Moreover, we observed a mild decrease in the intensity of Env per cell, indicating the less effective replication (Fig. 3I). Together, these results suggest that less-contractile cells indeed lead to insufficient ZIKV infection, perhaps via manipulating the condensation and efficiency of replication factory.

Except for when modulated through either drug application or gene editing, cells grown on compliant matrix (i.e. 50 kPa) failed to assemble thick contractile actomyosin bundles and thus had compromised contractility. To verify this, we fabricated compliant substrates (collagen-coated polyacrylamide gels with elastic moduli of 50 kPa) and observed host cells grown on top of it. We confirmed that there are lower levels of phosphorylated MLC and a looser actin network when cells cultured on compliant matrix (Fig. 4A,B; Fig. S3H).

To investigate whether having less cell contractility by culturing on a substate with a low extracellular stiffness is also able to restrain ZIKV infection, we applied the infection in cells cultured on the compliant 50 kPa substrate. Infection proportion was saturated in either case (Fig. 4C). Consistent with previous data, cells on compliant ECM had suppressed numbers of viral RNA copies and virus production (Fig. 4D,E). Taken together, our results suggest that less-contractile cells, no matter whether this is caused by internal or external reasons, indeed lead to lower ZIKV infection, through either drug application or gene editing.

In this study, we provide evidence that not only does ZIKV infection lead to a greater stiffness for host cells, but also host cell contractility in turn contributes to ZIKV infection efficacy (Fig. 4F). Owing to the fact that real infection in vivo occurs within a compliant ECM microenvironment compared to the stiff plastic- or glass-bottomed Petri dish in the laboratory, our study highlights the importance of host cell mechanical property during viral infection process, and implies the synthetic progression for potential solid tumorigenesis upon infection.

ZIKV replicates their RNA in endoplasmic reticulum (ER)-enclosed cytoplasmic sites (Brand et al., 2017; Mohd Ropidi et al., 2020; Oyarzún-Arrau et al., 2020). Several studies have pointed out that actomyosin contractility is vital for many viral infections. For instance, the fusion of Sendai virus with host cells is markedly reduced when using blebbistatin or siRNA targeted to non-muscle myosin II, showing that contractility signaling pathway might provide a barrier for host cells to against viral infection (Das et al., 2015). After fusion, intracellular trafficking of viral components are essential for the initial of infection and subsequent viral life cycle. It is also reported that this process could occur in an actomyosin-dependent manner during fig mosaic virus infection (Ishikawa et al., 2015). According to these results, the mechanism of how actomyosin affects ZIKV infection appears quite similar to that shown in previous studies. Our results show that disrupting cell contractility and myosin activity by either chemical treatment or gene editing does not compromise ZIKV invasion. Hence, we hypothesize that actomyosin contractility mainly affects post-entry steps of ZIKV infection, such as intracellular trafficking or replication. We also speculate that the potential consequences of increased contractility upon ZIKV infection could facilitate the collection of growing replication sites around the nucleus, as has been studied in the context of Vaccinia virus infection (Schramm et al., 2006).

In addition, ZIKV infection enhances monocyte adhesion and transmigration (Ayala-Nunez et al., 2019), which could be one of the reasons for highly efficient virus spreading. We thus propose that the actomyosin cytoskeletal rearrangements induced by ZIKV are a prerequisite for the observed cell contractility and migration activities that apparently contribute to the infectious life cycle of ZIKV.

Human osteosarcoma (U2OS) cells and African green monkey kidney epithelial cells (VeroE6) cells were cultured at 37°C with 5% CO₂ in Dulbecco's modified Eagle's medium (DMEM; Biological Industries) supplemented with 10% fetal bovine serum (FBS) (Gibco), and 1% penicillin and streptomycin (Gibco). C6/36 cells were cultured in minimum essential medium (MEM) (Gibco) supplemented with 10% FBS and 2% non-essential amino acids (Solarbio, N1250-100) at 28°C in 5% CO₂. ZIKV strain SZ01 was used in this study (GeneBank: KU866423.2). Virus stocks were prepared by virus amplification in C6/36 cells at a multiplicity of infection (MOI) of 0.1. Virus-containing supernatant medium was harvested from day 4 post infection and stored at −80°C. For lentivirus production, the pLKO.1 shRNA plasmid (Addgene #10878) was transfected into HEK293T cells together with psPAX2 packaging plasmid (Addgene #12260) and pMD2.G envelop plasmid (Addgene #12259) by using FuGENE HD (Promega). Supernatants were collected 48 h post-transfection, filtered through a 0.45 μm filter to remove the cells debris and stored at −80°C. The myosin ATPase activity inhibitor blebbistatin (Sigma-Aldrich, 182515, Saint Louis, MO, USA) was solubilized in DMSO and used at the concentration of 1, 5 and 10 μM for the time indicated in the Fig. 2. DMSO was used as control treatment.

Collagen was added into gel mix (10× Dulbecco's PBS, 0.23% 1 M NaOH and H₂O) to generate the 2% collagen gel. Cells were collected in a tube and centrifuged at 100 g for 5 min at 4°C. The cells were resuspended with 2% collagen gel at the density of 10⁶/ml. The cell mix were seeded into a 48-well plate (NEST) and incubated in a cell incubator for 30 min. An appropriate volume of medium was added into the wells. The gels were imaged using a camera (Nikon, Tokyo, Japan) and the gel area was visualized by staining with Crystal Violet dye and the change in area was calculated with ImageJ.

The width of thick actin bundles in U2OS cells with/without ZIKV infection were quantified with ridge detection plugin from Fiji ImageJ. The parameters used for quantifying were: line width 29.0, high contrast 230, low contrast 87, sigma 8.87, low threshold 0.0, and upper threshold 0.17. Please note that the detection of thick bundles identified some dorsal stress fibers, whereas the defects supporting the decreased contractile bundles are still significant. We manually outlined the dorsal stress fibers in cells on Petri dishes and the ventral stress fibers in cells on micropattern chips, and subsequently measured the lengths for quantification. All analyses were undertaken in an unbiased manner by a researcher who was not aware of the experimental conditions.

RhoA activity was determined with a RhoA G-LISA absorbance-based biochemical assay kit (Cytoskeleton Inc., Denver, CO, USA) following the manufacturer's instructions. Briefly, cells were lysed, and binding buffer was added to the cell lysate, followed by an incubation on a RhoA-GTP affinity plate coated with RhoA-GTP-binding protein. The plate was placed on an orbital plate shaker at 400 rpm for 30 min at 4°C. After washing, primary anti-RhoA antibodies (dilution 1:250) and secondary HRP-linked antibodies (dilution 1:62.5) were sequentially added, followed by an incubation on an orbital shaker at 400 rpm for 45 min at room temperature. Afterwards, HRP detection reagents were added, and the absorbance was recorded via a plate reader spectrophotometer Enspire (PerkinElmer Company, Waltham, MA, USA).

Immunofluorescence (IF) experiments were performed as previously described (Jiu et al., 2019). Briefly, cells were fixed with 4% PFA in PBS for 15 min at room temperature, washed three times with 0.2% BSA in Dulbecco's PBS, and permeabilized with 0.1% Triton X-100 in PBS for 5 min. Cells were blocked in 1× Dulbecco PBS supplemented with 0.2% BSA. The following primary antibodies were used for immunofluorescence: NMIIA rabbit polyclonal antibody binding to C-terminal tail residues 1948–1960 (1:1000 dilution; 909801, BioLegend); NMIIA RLC mouse polyclonal antibody (1:100 dilution; M4401, Sigma); phospho-myosin light chain 2 (Thr18/Ser19) rabbit polyclonal antibody (1:50 dilution; #3674, Cell Signaling, Beverly, MA, USA); vinculin mouse monoclonal antibody (1:100 dilution; V9131, Sigma). Both primary and secondary antibodies were applied onto cells and incubated at room temperature for 1 h. Alexa Fluor-conjugated phalloidin was added together with primary antibody solutions onto cells. All immunofluorescence data were obtained with a Leica DM6000B wide-field fluorescence microscope with a HCXPL APO 63×, NA 1.40 oil objective. 3D-SIM imaging and processing was performed on a GE Healthcare DeltaVision OMX equipped with a 60×1.42 NA Oil objective and sCMOS camera as described previously (Jiu et al., 2019). For micropattern experiments, the cells were plated on CYTOOchips™ prior to fixation as described previously (Jiu et al., 2019).

All cell lysates were prepared by washing the cells once with PBS and scraping them into lysis buffer (50 mM Tris-HCl pH 7.5 150 mM NaCl, 1 mM EDTA, 10% glycerol and 1% Triton X-100) supplemented with 1 mM PMSF, 10 mM DTT, 40 μg/ml DNase I and 1 μg/ml of leupeptin, pepstatin, and aprotinin (MERCK). All preparations were conducted at 4°C. Protein concentrations were determined with Bradford reagent (#500-0006, Bio-Rad, Richmond, California, USA). Then western blotting was performed as described previously (Jiu et al., 2019). The following antibodies were used in this assay: phospho-myosin light chain 2 (Thr18/Ser19) rabbit polyclonal antibody (dilution 1:500; #3674, Cell Signaling); myosin light chain mouse monoclonal antibody (dilution 1:1000; #M4401, Sigma-Aldrich); myosin-18B rabbit polyclonal antibody (1:50 dilution; LS-C403352, LSBio); Unc45a mouse polyclonal antibody (1:300 dilution; H00055898-B01P, Abnova); and GAPDH mouse polyclonal antibody (1:1000 dilution; G8795, Sigma-Aldrich). Original blot images are shown in Fig. S4.

Total cellular RNA was extracted by EZ-press RNA Purification Kit (EZBioscience, #B0004DP) according to the manufacturer's protocols. Total RNA was reverse transcribed by using Color Reverse Transcription Kit (EZBioscience, #A0010CGQ). Real-time RT-PCR was carried out by using 2× Color SYBR Green qPCR Master Mix (ROX2 plus) (EZBioscience, #A0012-R2) in QuantStudio 1 system (Thermo). All readings were normalized to the level of GAPDH. Primer sequences are: ZIKV forward 5′-CAACCACAGCAAGCGGAAG-3′ and reverse primer 5′-AAGTGATCCATGTGATCAGTTGATCC-3′; GAPDH forward 5′-GCATCCTGCACCACCAACTG-3′ and reverse primer 5′-GCCTGCTTCACCACCTTCTT-3′.

Zika virus titers were determined by a plaque assay performed on Vero cells. Briefly, Vero cells were seeded into 24-well plates at a density of 1×10⁵ cells/well and washed with pre-warmed phosphate-buffered saline (PBS). Cells were then infected with serial 10-fold dilutions of virus supernatants for 2 h at 37°C with 5% CO₂. Inoculum was removed and replaced with DMEM containing 1% carboxymethylcellulose (CMC) (Sigma, #C5678) and 1.5% FBS. At 4 days post-infection, cells were washed with PBS and fixed with 4% PFA at room temperature for 1 h, followed by staining with Crystal Violet (Beyotime, C0121) for 10 min. After rinsing with water, the numbers of visible plaques were counted, and the virus titers were calculated as plaque forming units (PFU) per ml.

As previously, Myo18B and Unc45a knockout cells were generated by CRISPR/Cas9 methods (Jiu et al., 2019; Lehtimäki et al., 2017). For Myo18B, a guide sequence targeting exon 1 of human Myo18B and exon 7 of the human Unc45a were selected based on CRISPR Design Tool (Lehtimäki et al., 2017) with quality scores of 83 and 91, respectively. Oligonucleotides for cloning guide RNA into pSpCas9 (BB)-2A-GFP vector (Addgene #48138, deposited by Feng Zhang) were designed. The forward primer for Myo18B was 5′-CACCGCTCATCACGCCTCGCCCTGT-3′ and reverse primer 5′-AAACACAGGGCGAGGCGTGATGAGC-3′; the forward primer for Unc45a was 5′-CACCATGTCAAAGCACTCTACCGG-3′ and reverse primer 5′-AAACCCGGTAGAGTGCTTTGACATC-3′. Transfected cells were detached 24 h post-transfection, suspended into complete DMEM with 20 mM HEPES and sorted with a FACS Aria II (BD Biosciences), using on low intensity GFP-expression pass gating, as single cells onto 96-well plate supplemented DMEM containing 20% FBS and 10 mM HEPES. CRISPR clones were cultivated for 2 weeks prior selecting clones with no discernible Myo18B or Unc45a protein expression.

Cell proliferation and viability were assessed by BrdU assay, which were performed via a BrdU kit (Abcam, Cambridge, UK) in accordance with the manufacturer's instructions. In brief, cells were labeled with BrdU and fixed, followed by an incubation with an anti-BrdU antibody and HRP-linked IgG. Afterwards, HRP substrate was added, and the absorbance at 450 nm was determined by using an Enspire plate reader spectrophotometer (PerkinElmer).

Gels of polyacrylamide (PAA) of various stiffness were prepared by mixing 40% polyacrylamide and 2% bis-acrylamide solution, as described previously (Pelham and Wang, 1997). Substrate preparation protocols and modulus values were adopted from previously published work (Tse and Engler, 2010). Briefly, the gel solution for desired stiffness was mixed with ammonium persulfate (APS; 1:100) and tetramethylethylenediamine (TEMED; 1:1000) and placed between a hydrophobic glass (octadecyltrichlorosilane treated; 104817, Sigma-Aldrich) and the transparency sheet treated with 3-APTMS (A17714, Alfa Aesar, Ward Hill, MA, USA). Once polymerized, the hydrophobic plate was carefully removed. The gel was conjugated with sulfo-SANPAH [sulfosuccinimidyl 6 (4-azido-2- nitrophenyl-amino) hexanoate] and incubated overnight with rat tail type I collagen (25 µg/ml) (A1048301, Invitrogen, CA, USA) at 4°C.

Numbers of actin filaments length and mean actin filaments width were quantified by the rigid detection plugin of Fiji ImageJ. Focal adhesion lengths were manually quantified with Fiji ImageJ. Statistical analyses were performed with Excel (Microsoft, Redmond, WA, USA) and SigmaPlot (Systat Software Inc, San Jose, CA, USA). Sample sizes and the numbers of replications are included in the images. A Student's two-sample unpaired t-test or Mann–Whitney–Wilcoxon rank-sum test was used to assess the statistical difference. P<0.05 was considered to be significant.

We thank Aguang Dai and Jing Zhao in the core facility for technical imaging support.

The peer review history is available online at https://journals.biologists.com/jcs/lookup/doi/10.1242/jcs.261301.reviewer-comments.pdf