Membrane Ruffling is a Mechanosensor of Extracellular Fluid Viscosity

Abstract

Cell behaviour is affected by the physical forces and mechanical properties of the cells and of their microenvironment. The viscosity of extracellular fluid — a component of the cellular microenvironment — can vary by orders of magnitude, but its effect on cell behaviour remains largely unexplored. Using bio-compatible polymers to increase the viscosity of the culture medium, we characterize how viscosity affects cell behaviour. We find that multiple types of adherent cells respond in an unexpected but similar manner to elevated viscosity. In a highly viscous medium, cells double their spread area, exhibit increased focal adhesion formation and turnover, generate significantly greater traction forces, and migrate nearly two times faster. We observe that when cells are immersed in regular medium, these viscosity-dependent responses require an actively ruffling lamellipodium — a dynamic membrane structure at the front of the cell. We present evidence that cells utilize membrane ruffling to sense changes in extracellular fluid viscosity and to trigger adaptive responses.

It has been extensively demonstrated that cellular behaviors are governed not just by biochemical signaling, but also by biophysical cues¹. For example, stiff substrates can steer stem cells towards osteogenic differentiation² and can prompt otherwise non-invasive tumor cells to invade³. Local topography can determine cell morphology⁴ and direct cell migration^5,6. Cells are capable of responding to externally applied forces, such as stretching and shearing, by rearranging their cytoskeleton and adjusting gene expression^7-10. One biophysical component of the cellular microenvironment that has received relatively little attention, however, is the viscosity of extracellular fluid (ECF). Cells in vivo are often immersed in fluids with viscosities orders of magnitude higher than that of culture medium^11-14 – secreted mucins make airway mucus and saliva viscous¹², excess fibrinogen can raise blood viscosity¹³, and hyaluronic acids contribute to the viscosity of interstitial fluids¹⁴. Moreover, abnormal ECF viscosity has been implicated in various diseases, including cystic fibrosis¹² and cancer^11,15-17. Despite the evidence suggesting a physiological role of ECF viscosity, there has been limited research addressing the subject, and no conclusive trend has been established in the few pioneering studies. In two studies, viscous fluid was found to decrease motility in non-adherent cells or cells moving through fluid in three dimensions^18,19. In another, there was no change in the motility of adherent keratocytes in viscous ECF²⁰. Recently, Gonzalez-Molina et al. observed that viscous ECF actually increased the migration speed of adherent cells²¹. This motivated us to investigate the effects of ECF viscosity on cell behavior at the subcellular scale to gain mechanistic insight into how cells sense and respond to viscosity. Counterintuitively, cells we examined in highly viscous medium moved nearly twice as fast as cells in regular medium. This speed increase was preceded by striking morphological and functional changes – cells in viscous medium increased their spread area and became flatter, displayed explosive remodeling of focal adhesions (FA), and generated stronger traction forces. We found that the response was not binary; rather, cells were able to tailor their response to the degree of viscosity. After characterizing these behaviors and examining cytoskeletal dynamics, we identified membrane ruffling at the lamellipodia as the viscosity sensor that regulates downstream changes in morphology and motility.

Elevated viscosity enhances single cell motility

To simulate viscous ECF, we added high molecular weight hydroxypropyl-methylcellulose (MC) to culture medium at concentrations ranging from 0.25% to 2%. MC is an inert thickening agent that has been used as a supplement in stem cell culture medium²² and in actin gliding assays²³ to increase fluid viscosity. We characterized these MC-supplemented media using a rheometer (Supplemental Figure 1a). The highest concentration used in this study, 2% MC, increases viscosity by three orders of magnitude (Supplemental Figure 1b), creating a solution with a consistency similar to that of honey. We then examined the motility of MDA-MB-231 breast cancer cells in regular versus viscous medium. Surprisingly, cells in viscous media exhibited speed increases of up to 2-fold (Figure 1a, Supplemental Video 1). The increases were graded, with higher viscosities corresponding to faster speeds until a plateau was reached. We then inspected cell speed during the transition from regular to viscous medium and found that speed began increasing instantaneously upon the addition of viscous medium before reaching a new steady state after approximately 30 minutes (Figure 1b). Despite the increase in speed, cell movement remained random, and the directional persistence of cell migration was not significantly altered by viscous medium (Supplemental Figure 1c, 1d). Cells immersed in viscous medium continued to move faster for at least 24 hours (Figure 1c, 1d), indicating a prolonged effect of elevated viscosity on cellular behavior.

Figure 1.

Elevated viscosity enhances single cell motility and induces graded, reversible cell spreading. a) After 2 hours incubation in media of different viscosities, MDA-MB-231 cells showed a graded increase in cell motility. Error bars: SD; p values were calculated with respect to DMEM, where p ≥ 0.05 (n.s.), < 0.001 (***), < 0.0001; N = 110 (DMEM), 136 (0.25% MC), 136 (0.5% MC), 172 (0.75% MC), 129 (1% MC) 92 (1.5% MC), 84 (2% MC). b) Upon changing medium to 1% MC, cells sped up and reached a new steady-state speed within 30 minutes; medium changed at 180 minutes (indicated by an arrow). Envelopes: SD; N = 43 (DMEM), 47 (1% MC). c,d) Viscous media made with different polymers increased cell motility, but non-viscous dextran medium with similar osmolarity to other viscous media did not for c) MDA-MB-231 cells or d) NIH 3T3 cells; measurements were performed after overnight incubation. Error bars: SD; p values were calculated with respect to DMEM, where p ≥ 0.05 (n.s.), < 0.0001 (****); for c) N = 293 (DMEM), 393 (1% MC), 176 (1% PEO), 75 (1% dextran), for d) N = 89 (DMEM), 120 (1% MC), 102 (1% PEO), 114 (1% dextran). e) Cells started spreading immediately upon the addition of 1% MC at 05:30. Left: eGFP-F-tractin, center: interference reflection microscopy (IRM), right: brightfield (BF); timestamp: mm:ss; scale: 20 μm. f) The increase in cell spread area plateaued approximately 20 minutes after medium change, with the final degree of spreading depending on the degree of viscosity. Envelopes: SEM; N = 19 (0.5% MC), 25 (0.75% MC), 16 (1% MC) 15 (1.5% MC), 20 (2% MC). g) Viscosity-induced increases in spread area were reversed by replacing 1% MC with DMEM. Envelope: SD; N = 19. h) Increases in spread area and speed in response to viscosity changes were correlated; R² = 0.84.

We next sought to verify that enhanced cell motility in MC-medium was due to elevated viscosity, rather than some chemical property of MC or the slightly increased osmolarity of the medium. We measured the speed of cells immersed in medium supplemented with 1% poly(ethylene oxide) (PEO), which increased the viscosity to 160 cP, similar to that of 0.75% MC. We observed a 1.4-fold migration speed increase in cells in 1% PEO, suggesting that the cell speed increase is not polymer-specific (Figure 1c). We also measured cell speed in medium supplemented with 1% dextran, where the osmolarity was similar to 1% MC but the viscosity was comparable to regular medium. Cells in 1% dextran did not exhibit any speed increase (Figure 1c). These results show that the cell speed increase can be attributed to elevated viscosity alone – it is not a result of changes in chemical composition or osmolarity of the medium.

To determine whether the effect of viscosity on cell motility was similar across cell types, we measured the speed of NIH 3T3 fibroblasts in viscous medium. We found the same trend of sustained speed increases in 1% MC and 1% PEO with no change in motility in 1% dextran (Figure 1d). We saw a similar speed increase after treating RAW 264.7 macrophages with 1% MC (Supplemental Figure 1e), and found that elevated viscosity induced faster migration in primary fibroblasts collected from patients with wound healing disorders as well (Supplemental Figure 1f). Gonzalez-Molina and colleagues also observed viscosity-induced speed increases in collective migration using hepatic cell lines²¹. These observations suggest the existence of a common mechanism across cell types allowing cells to respond to changes in viscosity. The fact that cells were able to move faster in increasingly viscous medium was counterintuitive – increased resistance should have led to a decrease rather than increase in cell speed. Indeed, a model of crawling cells assuming no adaptive changes in cytoskeletal dynamics predicted that increasing viscosity would slow down cell motility²⁴. This led us to hypothesize that cells were able to actively sense and respond to changes in ECF viscosity.

Cells spread rapidly in response to elevated viscosity

Because the cell speed increase began right after the change to viscous medium (Figure 1b), we recorded timelapse images of cells during the transition from regular to viscous medium and examined parameters known to affect cell motility, including spread area, FA dynamics, actin flow, and traction forces²⁵. First, we used interference reflection microscopy (IRM) to examine the spread area of MDA-MB-231 cells on the substrate, and found that the area began increasing immediately following medium change (Figure 1e, Supplemental Video 2). This increase in cell spread area continued for approximately 20 minutes before plateauing (Figure 1f). Inspection of timelapse 3D images revealed that as cells spread more, they also became flatter (Supplemental Video 3, Supplemental Figure 2). The same trend was observed in HEK 293 cells (Supplemental Video 4), further supporting the notion of a viscosity-sensing mechanism common across cell types.

Examining cell spreading at various viscosities revealed that, like cell speed, the increases in spread area were graded, with higher viscosity inducing more spreading. The most viscous medium, 2% MC, caused an increase in average cell spread area of more than 2-fold (Figure 1f). The instantaneous onset of cell spreading and the viscosity-dependent maximum speed and spread area implied a sensing mechanism in response to increases in viscosity. To test whether cells were equally able to sense a reduction in viscosity, we gently diluted and removed 1% MC before replacing it with regular medium. We observed an immediate decrease in spread area that continued for 20 minutes before leveling off (Figure 1g). Complete removal of the viscous medium was technically not possible, and the residual MC likely rendered the replenished medium slightly viscous, which could explain the small discrepancy between initial and final spread area. Plotting the viscosity-dependent speed increases against spread area increases, we found a strong linear relationship (Figure 1h), which is in agreement with a previously proposed model predicting that larger spread area and cell flattening increase speed²⁶.

Membrane ruffling is essential for the response to viscosity

We noticed that in the regions of the cell that spread in response to elevated viscosity, membrane ruffles were suppressed; inversely, in regions where no ruffling had been observed in regular medium, no spreading was detected after changing to viscous medium (Figure 2a; Supplemental Video 5). Membrane ruffling and lamellipodial protrusion are hallmarks of mesenchymal cell motility, and cycles of protrusion and retraction and ruffling are thought to facilitate haptotaxis and durotaxis²⁷. To measure ruffling quantitatively, we transfected MDA-MB-231 cells with eGFP-F-tractin to label F-actin, and acquired z-stack images in time lapse. Peripheral membrane ruffles, which occur when protruding lamellipodia detach from the substrate, are actin-rich structures that move upward in the z-axis and then inwards. Because the membrane ruffles spanned multiple focal planes, we segmented the ruffles by projecting the eGFP-F-tractin fluorescence into a single composite image, where we thresholded and integrated the signal to represent the extent of ruffling. We observed that membrane ruffles were suppressed to a greater degree with increasingly viscous medium (Figure 2b). Like the increase in spread area, the change in ruffling was reversible, and it recovered within approximately 10 minutes of replacing 1% MC with regular medium (Figure 2d, Supplemental Video 6). Plotting both the normalized cell spread area and degree of membrane ruffling over time, we found that the two exhibited an inverse relationship (Figure 2c). A steep decrease in ruffling preceded a more gradual increase in spread area in 1% MC.

Figure 2.

Membrane ruffling is critical for the cellular responses to viscosity changes, and is inversely correlated with viscosity-dependent cell spreading. a) MDA-MB-231 cells are shown just before [i] and 1 minute after [ii] the addition of 2% MC. Note that only actively ruffling areas (red arrowheads) and not areas devoid of ruffles (white arrowheads) spread in response to viscous medium. Scale: 20 μm. b) Viscous medium added at 15 minutes suppressed membrane ruffling, and higher viscosity led to more complete and more rapid suppression. Representative images of eGFP-F-tractin and segmented ruffles (red) are shown in the inset. Envelopes: SD; N = 12 (0.5% MC), 11 (1% MC), 14 (2% MC). c) Membrane ruffling and cell spreading exhibit an inverse relationship, and a decrease in membrane ruffling immediately precedes an increase in spread area. Envelopes: SEM; N = 8. d) A representative image of an MDA-MB-231 cell shows that membrane ruffling was suppressed in 2% MC, but recovered after the viscous medium was washed out. Cell contour traced in red; scale: 20 μm. e,f) Fish keratocytes, typically devoid of ruffles even in regular culture medium, did not exhibit increases in e) speed or f) spread area when immersed in viscous medium. Error bars: SD; p ≥ 0.05 (n.s.); e) N = 28 (RPMI), 23 (1% MC), f) N = 17. g,h) MDA-MB-231 cells treated with WGA to suppress ruffling did not exhibit increases in g) spread area or h) speed in 1% MC. g) Envelopes: SD; N = 8. h) Error bars: SD; p ≥ 0.05 (n.s.); N = 53 (DMEM + WGA), 59 (1% MC + WGA).

We next explored whether there was a direct, causal relationship between the reduction of membrane ruffles and the increases in cell spreading and speed in viscous medium. To test this, we used fish keratocytes, which are well-characterized, highly motile cells with negligible membrane ruffling. Keratocytes treated with 1% MC did not exhibit any changes in spread area (Figure 2f) or cell speed (Figure 2e), agreeing with previous studies^20,28. We observed that a very small population (< 2%) of atypical keratocytes did exhibit ruffling and were stationary; following the addition of 1% MC, ruffling in these cells decreased and they showed large increases in spread area and motility (Supplemental Video 7). To further confirm that membrane ruffles were critical for viscosity-dependent cell spreading and speed increases, we treated MDA-MB-231 cells with wheat germ agglutinin (WGA), which has been shown to inhibit membrane ruffling²⁹. WGA treatment effectively inhibited membrane ruffling in regular medium, and subsequent addition of 1% MC did not induce any changes in spread area (Figure 2g) or speed (Figure 2h). These results demonstrated that membrane ruffling was critical for the increases in cell spread area and cell speed in response to viscosity.

Viscosity enhances focal adhesion formation and turnover

We next sought to identify the link between the changes in membrane ruffling and spread area at high viscosity by examining the role of integrin, a key molecule in cell adhesion and migration, also required in ruffle formation. Without extracellular matrix (ECM) proteins absorbed onto the substrate surface to facilitate integrin-ECM binding, cells did not display significant ruffling, and did not increase spread area in response to elevated viscosity (Figure 3a), implying that ruffling and the response to viscosity were mediated by integrin. To further confirm that integrin engagement was critical for viscosity-dependent spreading, we coated glass-bottom dishes with poly-L-lysine so that cell-substrate adhesion was mediated via electrostatic interactions³⁰ rather than integrin engagement. As expected, MDA-MB-231 cells adhered to the poly-L-lysine-coated surfaces exhibited minimal ruffling and did not increase spread area upon the addition of 1% MC (Supplemental Figure 3a).

Figure 3.

Viscosity-dependent spreading is mediated by integrin. a) In the absence of ECM proteins for integrin binding, cells did not spread in response to 1% MC, which was added at 20 minutes. Envelopes: SEM; N = 16 (1% MC), 21 (1% MC, no ECM). b) A representative image of MDA-MB-231 cells shows extensive FA growth and rapid FA assembly and disassembly after the addition of 1% MC at 11 minutes. Timestamp: mm:ss; scale: 40 μm. c) The difference in intensity of talin signal between subsequent frames shows significant redistribution of talin immediately following the addition of 1% MC, suggesting extensive nascent adhesion growth. Envelope: SD; N = 9. d) FA assembly and disassembly rates increased significantly in 1% MC. Error bars: SD; p < 0.0001 (****); FA data were collected from 9 cells and pooled, with N = 196 (assembly, DMEM), 479 (assembly, 1% MC), 251 (disassembly, DMEM), 308 (disassembly, 1% MC). e) Integrated FA area per cell over time shows a decrease followed by partial recovery in adhesion area after the addition of 1% MC at 10 minutes. Envelope: SD; N = 20 cells. f) Average individual FA area decreased by 30% in 1% MC. g) The total number of FAs per cell per frame increased slightly in 1% MC. g,f) Error bars: SD; p < 0.01 (**), < 0.0001 (****); FA data were collected from 20 cells for 20 frames and pooled, with N = 364 (DMEM), 725 (1% MC).

Because membrane ruffling is a result of retrograde forces overcoming the strength of cell-substrate adhesion at the lamellipodia and pulling the leading edge rearward^27,31,32, we proceeded to investigate whether FAs, which facilitate cell-substrate adhesion, were affected by elevated viscosity. We recorded timelapse images of MDA-MB-231 cells transfected with mEmerald-talin, a protein recruited early during the formation of nascent adhesions³³. Inspecting the changes in mEmerald-talin intensity between subsequent frames, we observed immediate redistribution of talin to the expanding lamellipodia following the addition of 1% MC (Figures 3b & 3c, Supplemental Video 8), suggesting extensive nascent adhesion formation. We also observed significant changes in FA assembly and disassembly rates (Figure 3d). The assembly and disassembly rates were both approximately 25% higher in 1% MC than in the regular medium. The average adhesion size decreased by 30% in viscous medium (Figure 3f), while the total number of adhesions per cell increased slightly (Figure 3g), leading to an overall reduction in total FA area per cell (Figure 3e). Efficient FA remodeling and turnover are necessary for cell motility, and faster FA turnover has been shown to enhance cell motility^34,35. Furthermore, FA size governs cell speed, with optimal FA size leading to faster migration³⁶. Our results suggest that viscosity-dependent cell spreading is mediated by integrin and that the cell speed increase in viscous medium is a product of increased FA turnover and tuning of FA size.

Actin, but not myosin, is required for spreading

FA-mediated cell motility is driven by actin polymerization. Actin filaments continuously grow towards the leading edge, which both drives protrusion and causes filaments to undergo retrograde flow, the rate of which decreases when actin filaments are coupled to the substrate through FAs³⁷. The rapid cell spreading in viscous medium implied a decrease in the rate of retrograde flow and an increase in actin-driven protrusion. We first tested whether actin polymerization was involved by treating MDA-MB-231 cells with 100 nM latrunculin A (latA), a drug that inhibits F-actin polymerization, before immersing them in 1% MC. Viscosity-induced spreading in latA-treated cells was significantly attenuated (Figures 4a & 4b, Supplemental Video 9), confirming that actin polymerization was necessary. Examining kymographs of the leading edge in cells transfected with eGFP-F-tractin, we found a 10-fold reduction in retrograde flow speed for cells in 1% MC (Figures 4c & 4d), indicating coupling of actin to FAs to slow retrograde flow and drive protrusion.

Figure 4.

Actin polymerization is required for the cellular response to elevated viscosity, but myosin II contractility is not. a) Representative timelapse images show that latA-treated MDA-MB-231 cells did not exhibit significant increases in spread area after changing medium to 1% MC, unlike untreated cells in Figure 2a. Timestamp: mm:ss; scale: 20 μm. b) Viscosity-induced spreading was significantly attenuated after latA treatment. Envelope: SEM; N = 16 (1% MC), 15 (1% MC, latA). c) Representative kymographs show that retrograde flow was slower in 1% MC. eGFP-F-tractin scale: 20 μm; kymograph scale: t = 60 sec, x = 2 μm. d) Analysis of kymographs revealed the retrograde flow decreased 10-fold in 1% MC. Error bars: SD; p < 0.0001 (****); N = 37 (DMEM), 27 (1% MC). e) Blebbistatin-treated cells still spread robustly in 1% MC. Envelopes: SEM; N = 16 (1% MC), 17 (1% MC, blebbistatin).

To evaluate the role of myosin contractility in viscosity-induced spreading, we treated cells with the myosin II inhibitor blebbistatin (20 μM). After the addition of 1% MC, the blebbistatin-treated cells showed a trend very similar to untreated cells, with rapid spreading ultimately leading to a 1.7-fold increase in cell spread area (Figure 4e). These results demonstrated that cell spreading in viscous medium was independent of myosin contractility, and instead proceeded through reinforcement of cell-substrate adhesions and engagement of the molecular clutch, which slowed down retrograde flow and led to actin-driven protrusion at the leading edge. This agrees with a recent study in which lamellipodial spreading was found to be dependent on nascent adhesion stabilization and independent of myosin³⁸. As viscous medium suppresses membrane ruffling at the cell edge, prolonged contact between the lamellipodia and substrate could allow for integrin-ECM engagement and nascent adhesion formation, a critical early step in the formation of FAs. Indeed, it has been shown that temporal maintenance of contact between lamellipodial protrusions and the substrate facilitates nascent adhesion formation and growth³².

Cells at elevated viscosity generate stronger forces

The rates of FA assembly and disassembly increased at elevated viscosity, and both processes are force-dependent^7,39,40. To examine whether faster FA turnover in response to increased viscosity was a result of increased force generation, we performed continuum traction force microscopy⁴¹. We found that total traction forces were approximately 2-fold higher when cells were immersed in 1% MC, and maximum traction stresses increased significantly as well (Figures 5a & 5c, Supplemental Figure 3b). We next examined traction forces transmitted to the substrate via integrin at the single-molecule level using tension gauge tether (TGT)⁴². Timelapses of cells on TGT-coated surfaces showed a 2.4-fold increase in TGT ruptures following the addition of 1% MC (Figures 5d & 5e). This result demonstrated extensive engagement of integrin molecules transmitting forces greater than the 54 pN rupture threshold and suggested dynamic and extensive coupling between integrin and the actin cytoskeleton.

Figure 5.

Cells generate stronger forces at high viscosity independent of myosin II activity. a,b) Representative images of MDA-MB-231 cells show that traction forces increased in 1% MC regardless of whether or not myosin II activity was inhibited by blebbistatin. c) Quantification shows that total traction forces increased approximately by 2 fold in 1% MC. Error bars: SD; N = 18 for each condition. d) Single-molecule force measurement using TGT revealed extensive engagement of integrin molecules transmitting forces > 54 pN as cells spread in response to elevated viscosity. Color indicates the time of TGT rupture; scale: 20 μm. e) TGT ruptures over time increased by 2.4-fold after the addition of 1% MC. Error bars: SD; p < 0.001 (***); N = 8. f) Inhibition of myosin II contractility by blebbistatin did not inhibit the cell speed increase in 1% MC. Error bars: SD; p ≥ 0.05 (n.s.), < 0.0001 (****); N = 293 (DMEM), 393 (1% MC), 47 (DMEM, blebbistatin), 39 (1% MC, blebbistatin). g) The phase diagram generated by the model after inputting experimental measurements shows that, given the observations that in viscous medium 1) FA assembly/disassembly increase, 2) protrusion increases, and 3) FA area decreases, cortical tension is predicted to decrease as spread area increases upon elevated viscosity. h) Maximal and average drag forces on a ruffling membrane in DMEM and in 1% MC were estimated using finite element analysis, and the results showed that drag forces increased by two orders of magnitude in 1% MC. i) We propose that membrane ruffling is a mechanosensor of ECF viscosity.

Inhibition of myosin II contractility by blebbistatin decreased cellular forces but did not abolish the increase in traction force induced by 1% MC (Figures 5b & 5c, Supplemental Figures 3b & 3c). This agrees with a previous finding that traction force magnitude correlates with changes in cell spread area⁴³, as both blebbistatin-treated and untreated cells showed increases in spread area and traction forces in viscous medium. In agreement with the findings that blebbistatin did not prevent increases in spread area or traction force generation in viscous medium, blebbistatin-treated cells still increased speed by 1.5 fold in 1% MC, an increase comparable to that of the control group (Figure 5f).

Interestingly, observing cells labeled with mApple-myosinIIA, we found that a lamella-wide distribution of myosin filaments emerged as cells spread in 1% MC (Supplemental Figure 3d). Some of myosin filaments formed structures known as “myosin stacks”⁴⁴, which were absent in the cells prior to the addition of the 1% MC. So, although myosin II contractility was dispensable for the response to viscosity, these data indicate that myosin could play a structural role in viscosity-induced rearrangement of the actin cytoskeleton.

Modeling cell spreading in response to viscosity

We developed a mathematical model to describe the dynamic changes to the cell upon increased viscosity, incorporating measurements of cell spread area, leading edge protrusion, FA area, FA assembly/disassembly rates, actin retrograde flow, and traction force. The following principles known to govern cytoskeleton mechanics (detailed further in Methods) were implemented in the model⁴⁵. First, tension resists protrusion at the leading edge, resulting in actin retrograde flow; second, forces exerted by the retrograde flow facilitate nascent adhesion formation and subsequent FA growth; third, FAs exhibit centripetal growth as a reaction-diffusion-convection process, driven by retrograde flow; and finally, strong integrin-ECM engagement promotes leading edge protrusion against cortical tension, where increasing tension causes membrane buckling and ruffling. Mapping our data to the parameter space, we found that increased FA assembly/disassembly rates as seen in 1% MC could lead to three possible scenarios immediately following an increase in viscosity: decreased protrusion (Δl < 0) and increased FA area (ΔFA > 0); Δl > 0 and ΔFA > 0; Δl > 0 and ΔFA < 0. Our model predicted that only if cortical tension decreased upon a viscosity increase would the experimental observation of Δl > 0 and ΔFA < 0 be possible (Figure 5g).

We tested this by measuring cortical tension using laser ablation, a technique in which tension stored in the cortex is reported by the distance and rate of actin arc recoil following ablation⁴⁶. Confirming the prediction of the model, cortical actin arcs recoiled only half as far after ablation during the 15 minutes following 1% MC addition, representing decreased cortical tension (Supplemental Figure 4, Supplemental Video 10). Notably, after 20 minutes in 1% MC, coinciding with the plateau in cell spreading, cortical tension returned to the level observed in DMEM. Our observation agreed with a previous report that cell spreading was associated with an initial drop followed by recovery in cortical tension as cells spread and available membrane reservoirs were depleted⁴⁷. Cross-examining the dynamics of ruffling and cortical tension in cells subjected to 1% MC, we found that ruffling began to decrease instantaneously, whereas cortical tension dropped significantly only after a 10-minute delay (Supplemental Figure 4c), indicating that ruffle suppression occurs prior to cortical tension reduction.

Additionally, the model predicted that, to achieve a 2-fold increase in traction force in 1% MC, there must be an increase in traction force associated with nascent adhesion formation in addition to the force transmitted through mature FAs. To test this prediction of increased nascent adhesion formation, we measured changes in nascent adhesion integrated intensity upon increased viscosity, and observed a rapid increase in nascent adhesion formation (Figure 3c, Supplemental Video 11) amounting to 40% increase in nascent adhesion formation in 1% MC compared to that in DMEM (Figure 3c), consistent with the predictions of the model.

Membrane ruffling acts as sensor of extracellular viscosity

Because the presence of membrane ruffling in regular medium dictated whether or not spread area and speed would increase in viscous medium, we reasoned that suppression of ruffling triggered these downstream responses. At elevated viscosity, the retrograde forces of contractility and membrane tension that drive ruffling, buckling the lamellipodium and moving it upward and rearward ^27,48,49, would be insufficient to overcome the hydraulic resistance imposed by the medium (details included in Methods). We performed simulations to evaluate the forces required to move the lamellipodium upward and rearward, and found that the drag force on a ruffling membrane increased by two orders of magnitude in 1% MC (Figure 5h, Supplemental Figure 5, Supplemental Videos 12 & 13). Upon the viscosity increase, the forces driving ruffling are counteracted by the drag force, which increases proportionally to viscosity. Because of the decrease in net force on the membrane, the membrane can no longer be displaced upward and rearward to form ruffles to the same extent as in regular medium. Accordingly, we observed an inverse relationship between viscosity and ruffling (Figure 2b), supporting the notion that higher hydraulic resistance suppresses ruffling more effectively. As a result, the membrane remains in close contact with the substrate, leading to increased nascent adhesion growth at the leading edge; indeed, it has been shown that temporal maintenance of lamellipodial protrusions stimulates nascent adhesion formation³². The membrane is then pushed forward more efficiently by actin polymerization as actin filaments are coupled to new adhesions, causing a reduction in retrograde flow and an increase in cell spreading. This increase in cell spread area leads to higher traction force generation, which in turn accelerates FA turnover and tunes FA size, driving faster cell migration.

Our simulations showed the vertical drag force on a ruffling membrane with an area of 4x8 μm² to be approximately 1 pN in 1% MC, and the lateral drag force on a protruding membrane to be of the same order of magnitude, agreeing with previous reports⁵⁰. In advancing lamellipodia, cell-generated protrusive forces are more than sufficient to overcome this level of resistance, as the force generated by just a single polymerizing actin filament is >1 pN⁵¹, and actin-driven cell protrusions have been reported to generate forces on the order of ones to hundreds of nN^52,53. The force generated by polymerizing actin filaments in the 4x8 μm² model protrusion would therefore greatly exceed the sub-pN drag force. However, this force generation machinery is confined in very thin, sheet-like lamellipodia at the protruding edge of the cell, and these forces are oriented almost entirely in the xy-plane (Supplemental Figure 5d). Similarly, the retrograde forces that cause out of plane buckling and lead to ruffling are primarily oriented in the xy-plane, and the component of these forces oriented along the z-axis causing ruffling is therefore extremely small. It is also known that viscous drag will significantly affect buckling at extremely low Reynolds number, which applies to motile cells and cell-scale filaments and sheets in viscous fluids^54-57, and that the force required to cause buckling increases proportionally with the coefficient of viscous drag (see Methods and Supplemental Figure 5e). Therefore, while the increased drag force is easily overcome in the xy-plane by actin-driven protrusion, it is sufficient to hinder ruffling in the z-direction.

Our results suggested that ruffles act as a mechanosensor of ECF viscosity rather than as a passive membrane reservoir. To further confirm ruffles actively sense and respond to changes in viscosity, instead of passively storing and releasing membrane, we perturbed membrane availability by other means. We first treated cells with 80 μM dynasore to suppress membrane recycling via endocytosis, effectively rendering more membrane available to be incorporated into the cell surface^58,59. No spread area increase was observed, and cells continued to ruffle. Subsequent addition of 1% MC led to an increase in spread area similar to that in untreated cells (Supplemental Figure 6a). Likewise, we observed no spread area increase after treating cells with hypertonic medium (530 mOsm), and subsequent addition of hypertonic 1% MC increased spread area (Supplemental Figure 6b). We next treated cells with hypotonic medium (160 mOsm) to increase cell volume and thereby membrane tension. We observed a slight decrease in spread area as cells swelled (Supplemental Figure 6c), but following treatment with hypotonic 1% MC, cell spread area increased. Notably, when cells were treated with hypotonic 2% MC, instead of 1% MC, the increase in cell spread area was more rapid and extensive (Supplemental Figure 6c). In all cases, perturbations to membrane availability did not induce responses typical of treatment with viscous medium, but subsequent addition of viscous medium did, indicating that ruffles are not a passive membrane reservoir, but rather a sensor of viscosity.

To summarize, suppression of ruffling at high viscosity initiates the following: enhanced integrin-substrate engagement, increased nascent adhesion growth, reduced actin retrograde flow, increased cell spreading, greater traction force generation, faster FA turnover, and ultimately faster cell speed. In effect, membrane ruffling acts as a sensor of ECF viscosity (Figure 5i).

Conclusion

In this study, we characterized the responses of cells subjected to high viscosity, and found that membrane ruffling serves as a sensor of viscosity. We concluded that the cellular response to viscosity is mediated not by a molecular sensor, but by a biophysical mechanism – an actively ruffling membrane probing the viscosity of the ECF. Our hypothesis is supported by the observation that cells with negligible or inhibited ruffling did not respond to viscosity. Given the variability of ECF viscosity in vivo, membrane ruffling may have an important role in regulating cell motility and force generation, both of which are significant in disease progression⁶¹. Whether membrane ruffling serves as a viscosity sensor for cells embedded in 3D matrices has yet to be explored, but we note that membrane ruffling has been observed in fibroblasts in 3D collagen matrices⁶². Our study focused on the effects of elevated viscosity on FA-mediated cell behavior, and the effects on other modes of cell motility, such as amoeboid movement⁶³ or nuclear piston-based migration⁶⁴, are still unknown. It is also possible that elevated viscosity could lead to changes in gene expression at longer timescales, as this study explored the instantaneous biophysical responses to viscosity. Further work is necessary to comprehensively understand how viscosity affects cell behavior across different timescales, modes of motility, and environments, but it is clear that the viscosity of the extracellular fluid plays an important role in regulating cell behavior.