How Epithelial Cells Reorient in Response to Cyclic Stretching


 Many types of adherent cells are known to reorient upon uniaxial cyclic stretching perpendicularly to the direction of stretching to facilitate such important events as wound healing, angiogenesis, and morphogenesis. While this phenomenon has been documented for decades, the underlying mechanism remains poorly understood. Using an on-stage stretching device that allowed programmable stretching with synchronized imaging, we found that the reorientation of NRK epithelial cells took place primarily during the relaxation phase when cells underwent rapid global retraction followed by extension transverse to the direction of stretching. Inhibition of myosin II caused cells to orient along the direction of stretching, whereas disassembly of microtubules enhanced transverse reorientation. Our results indicate distinct roles of stretching and relaxation in cell reorientation and implicate a role of myosin II-dependent contraction via a microtubule-modulated mechanism. The importance of relaxation phase also explains the difference between the responses to cyclic and static stretching.


23
Increasing attention has been paid over past decades to the effect of external mechanical 24 signals, such as stretching forces and substrate stiffness, on cellular behavior, including growth, 25 shape, adhesion, polarity, and migration 1-4 . In particular, cyclic stretching takes place ubiquitously 26 in the body, driven for example by beating of the heart, inflation of the lungs, or peristalsis of the 27 gut. Many types of cells, including smooth muscle cells, endothelial cells, and epithelial cells, are 28 exposed to cyclic stretching to affect both physiological and pathological events. For instance, 29 elevated cyclic stretching was proposed to induce aortic valve calcification 5 , while proper 30 stretching activities promote the differentiation of embryonic stem cells into muscle cells 6 . 31 As was first discovered four decades ago, cyclic stretching caused fibroblasts to reorient 32 perpendicularly to the direction of stretching 7 . Subsequent studies confirmed that many different 33 types of adherent cells, such as endothelial and epithelial cells, can realign similarly when exposed 34 to uniaxial cyclic stretching [8][9][10][11][12] . It is generally believed that this stretch-induced realignment is of 35 importance to such events as wound healing, angiogenesis, and morphogenesis 13-16 . The 36 intriguing response, accompanied by reorganization of actomyosin contractility and realignment 37 of the actin and microtubule cytoskeleton 17-22 , prompted various hypotheses based, for example, 38 on the maintenance of tensional homeostasis or dissipation of stored elastic energy 9,23,24 . In 39 addition to the contraction of actin cytoskeleton, microtubules may also play a regulatory role by 40 stabilizing the cell shape and polarity 18,19,25 . 41 Stretching-induced transverse reorientation is sensitive to the amplitude, frequency, and 42 waveform of stretching 9,10 . The response requires a minimal strain of around 3% 26 , above which 43 the extent of reorientation shows a correlation with the magnitude of strain. Interestingly, static 44 uniaxial stretching was found to cause cells to spread and migrate toward the direction of 45 stretching 27,28 , in contrast to the transverse reorientation induced by cyclic stretching. Fibroblasts 46 reorientation requires a minimal frequency of 0.01 Hz and saturates at 1 Hz, while smooth muscle 47 cells show optimal realignment responses at 0.5 Hz 9,29 . Triangular, square, or asymmetric 48 waveforms induce different extents of cell reorientation and stress fiber redistribution 10,30 , which 49 suggests sensitivities to the slopes of stretching and/or duration of stretching/relaxation. 50 The above observations suggest that, to unveil the mechanism of cyclic stretching-induced cell 51 orientation, it is important to understand potentially differential responses during the stretching 52 and relaxation phases. To this end, we have developed an on-stage cell stretcher based on a 53 motorized microscope stage and an elastic, patternable polyacrylamide substrate. The system 54 was programmed to allow synchronized image recording and cyclic stretching. Differences 55 between consecutive images then revealed changes in cell shape during different stages of cyclic 56 stretching and led us to a working model for transverse cell reorientation. 57

58
Reorientation Response of NRK-52E Epithelial Cells to Cyclic Stretching 59 NRK-52E cells, a rat kidney epithelial cell line that exhibits high circularity and large spreading 60 area at steady state, were used for studying shape responses to cyclic stretching. A novel stretcher 61 for real-time imaging was developed using a microscope with a motorized stage. Cells were 62 cultured on an elastic polyacrylamide (PAA) substrate 400 um in thickness. One end of the 63 substrate was attached to the supporting coverslip underneath, which moved with the motorized 64 stage. The other end was anchored with a rod, which pushed against a handle attached to the 65 surface ( Fig. 1a and 1b). A custom computer program allowed synchronized image acquisition at 66 various times relative to the stretching cycle ( Fig. 1a and 1b). 67 We used PAA gels with a Young's modulus of 20-30 kPa (prepared with 10% acrylamide and 68 0.3% bis-acrylamide), which is in the physiological range of tissue stiffness, as the substrate. 69 Micropatterns on the gel surface was used to indicate the applied strain ( Fig. 1c; Supplementary 70 Fig. 1f, 1i, and 1k). The gel showed a nearly ideal elastic behavior when stretched by up to 20% at 71 0.5 Hz in a square waveform (Fig. 1d). Upon relaxation from 45 min of continuous cyclic stretching 72 with 15% strain, the residual strain was <0.72% along the direction of stretching (referred to as 73 axial) and <0.55% along a perpendicular direction (referred to as transverse) ( Supplementary Fig.  74 2). The substrate remained intact following >72h of up to 18% of continuous cyclic stretching at 75 0.5Hz. In addition, transverse strain remained below 16% of axial strain, suggesting that the strain 76 was predominantly uniaxial (Fig. 1d). 77 To avoid the complications of cell-cell mechanical interactions, we have limited the analysis to 78 isolated cells. With 15% of cyclic strain at 0.4-0.5 Hz, cells started to show shape change as soon 79 as 5 min and became oriented in a transverse direction within 30min ( Fig. 2a and 2b). By 80 measuring the length along axial or transverse direction, we found that axial length decreased 81 rapidly during the first 30 min, while transverse length steadily increased over 90 min (Fig. 2c). 82 Consistent with the changes in length, spreading area decreased during the first 30 min then 83 gradually recovered over the following 60 min (Fig. 2d). These observations indicated that 84 stretching-induced reorientation involved a rapid axial shortening phase and a steady transverse 85 elongation phase. 86

Distinct Shape Responses during Stretching and Relaxation Phases 87
The above observations suggested that cellular responses to cyclic stretching may involve 88 distinct protrusive/retractive responses in a spatially/temporally dependent manner, while the 89 reorientation may represent a cumulative result of incremental shape changes. We, therefore, 90 applied difference imaging as a sensitive means for detecting local changes in cell area (Fig. 3a). 91 We first examined the response to a single pulse of stretching for 10s by generating difference 92 images before and after stretching, during stretching, and during post-stretching relaxation (Fig.  93 3b). To compare the activities along axial and transverse directions, we further divided each cell 94 into two axial and two transverse quadrants (Fig. 3a). 95 Difference images taken during relaxation showed predominantly transverse protrusions ( Fig.  96 3b), which were most prominent during the first 10 s post stretching then decreased rapidly ( Fig.  97 3b and 3c). In contrast, difference images during stretching showed only baseline activities similar 98 to those of unstretched cells, while difference images immediately before and after stretching 99 showed prominently retractions along both axial and transverse directions (Fig. 3c), suggesting 100 that retraction took place immediately after the release of stretching. Decreasing the duration of 101 stretching from 10 s to 1 s did not significantly decrease the extent of retraction but reduced the 102 subsequent transverse protrusions and the difference in protrusive activities between transverse 103 and axial directions (Fig. 3c). 104 We then examined the response to stretching after applying cyclic stretching for up to 45 min 105 at 0.5 Hz (Fig. 4). Interestingly, while protrusive activities of transverse quadrants were 106 maintained following multiple cycles of stretching (Fig. 4, upper left blue bars), protrusions of axial 107 quadrants decreased progressively (Fig. 4, upper right orange bars). In contrast, similar retraction 108 responses were observed along axial and transverse quadrants immediately after the relaxation 109 of stretching, showing strong retraction after a single cycle of stretching but much weaker 110 retraction after additional cycles (Fig. 4). Together, these results explained cyclic stretching-111 induced reorientation as a cumulative consequence of persistent transverse protrusions coupled 112 with diminishing axial protrusions. 113

Functional Roles of Myosin II and Microtubules in Cyclic Stretching-Induced Cell Reorientation 114
Previous studies have implicated actomyosin contractility in cyclic stretching-induced cell 115 reorientation 17,19 . By inhibiting myosin II with blebbistatin, we found that cells became highly 116 branched while showing a weak orientation along the axial direction ( Fig. 5a and 5b). In addition, 117 both axial shortening and transverse elongation were suppressed as compared to untreated cells 118 ( Fig. 5c), suggesting that actomyosin contractility is directly or indirectly required for both 119 activities. 120 Since microtubules have been suggested to control cell shape and polarity 25,31,32 , we tested the 121 effect of microtubule disassembly on the responses to cyclic stretching. As shown in Fig. 6a and 122 6b, treatment with nocodazole caused a visibly higher degree of reorientation than controls. 123 Moreover, both axial shortening and transverse elongation reached a greater extent than control 124 cells (Fig. 6c), suggesting that microtubules play a role in tempering the response. Further analyses 125 of the responses during the first 10 s of relaxation after various periods of cyclic stretching 126 revealed that axial protrusion was inhibited transiently after 15min of stretching (Fig. 6d), while 127 transverse protrusion showed an increase after 45 min (Fig. 6d). These effects of nocodazole were 128 consistent with the kinetics of length change as shown in Fig. 6c and implicated a microtubule-129 mediated mechanism that affects protrusive activities in a location-dependent manner. 130

131
Although cellular response to cyclic stretching has been studied for decades due to its 132 physiological importance, the transverse reorientation as observed for adherent cells seems 133 counterintuitive. A serious limitation has been the reliance on endpoint analysis in most studies, 134 such that little is known about the dynamic events responsible for the reorientation. In the 135 present study, we have developed a novel cell stretcher based on a motorized microscope stage 136 that allows the recording of live cells in synchrony with cycles of stretching and relaxation. The 137 study was further facilitated by the nearly ideal elastic property of PAA substrates, and by the use 138 of difference imaging for detecting minute changes in cell shape in response to stretching or 139 relaxation. 140 We found that the response to cyclic stretching took place via an early phase that lasted for 30-141 45 min, when the reorientation involved primarily net axial retraction, followed by a late phase 142 when the reorientation involved primarily net transverse extension (Fig. 2c). This late phase of 143 extension is similar to the lagged increase in traction forces along the transverse direction as 144 reported previously 33 . In addition, the response to each stretching cycle involved no significant 145 response during stretching, immediate retraction upon relaxation, and protrusion that lasted for 146 10-20s during the ensuing relaxation ( Fig. 3c and Fig. 4). Retractive and protrusive activities 147 showed distinct temporal and spatial patterns. Retraction was global but its magnitude decayed 148 after a few cycles. In contrast, protrusion was more localized and persistent along the transverse 149 direction, while axial protrusions diminished gradually over 30 min (Fig. 4). Together, these events 150 explained not only the reorientation but also the transient decrease in cell area during the first 30 151 min (Fig. 2e). 152 The key role of relaxation phase in cyclic-stretching-induced reorientation provided a simple 153 answer to the puzzling difference between the responses to cyclic and static stretching, where 154 the reorientation was axial and too slow to be captured during the 1-10 s of stretching in the 155 present cyclic regimen 1,27,28,34 . The requirement of relaxation for reorientation may also explain 156 the dependence of the responses on the waveform of stretching; triangular waves were found to 157 elicit much weaker responses than rectangular or trapezoid waves of a similar peak magnitude. 158 We suspect that while rectangular and trapezoid waves contain discrete periods of relaxation to 159 support transverse elongation, relaxation in triangular waves is gradual and total relaxation is too 160 brief to allow much shape change 10,30 . 161 The present relationship between retraction and protrusion may be similar to that in the 162 process of "retraction induced protrusion", where tail retraction was followed by frontal 163 protrusion at the opposite end 32,35,36 . Similarly, symmetry breaking that initiates polarized cell 164 migration typically starts with the formation of a retracting tail, followed by protrusion at the 165 front 37,38 . All these activities may represent a common mechanism of retraction at one end 166 triggering protrusion signals at a distal end. Supporting this hypothesis, we showed that the 167 inhibition of myosin II with blebbistatin suppressed not only axial retraction but also transverse 168 extension (Fig. 5), causing cells to orient along the axial direction possibly as a passive response 169 to stretching 17,19 . 170 Previous investigations of the role of microtubules in cell reorientation upon cyclic stretching 171 have yielded conflicting results 19,39-41 . As microtubules are required for establishing cell polarity 172 and directional cell migration 25,41,42 , we suspect that they may play a role in coordinating 173 differential transverse and axial extensions, such that their disassembly may eliminate the 174 difference and inhibit cell reorientation. Surprisingly, the disassembly of microtubules caused an 175 enhancement of cyclic stretching-induced reorientation through the enhancement of both axial 176 retraction and transverse extension (Fig. 6). A possible explanation is that, as microtubules are 177 known to suppress cell contractility 43 , disassembly of microtubules may promote stretching-178 induced retraction and downstream events. A second, non-mutually exclusive possibility is that 179 microtubules, as relatively rigid structures, may be aligned by stretching to generate anisotropic 180 resistance to retraction 41 . Disassembly of microtubules may decrease this resistance to facilitate 181 axial retraction. A third possibility is that retraction signals may be generated at axial ends then 182 transported via microtubules to cause global retractions. Inhibition of microtubules would then 183 cause retraction signals to accumulate at axial ends, enhancing axial retraction while allowing 184 more extensions elsewhere ( Supplementary Fig. 3). This hypothetical mechanism is therefore 185 complementary to the local-excitation/global-inhibition (LEGI) mechanism proposed for polarized 186 cell migration 44,45 , where microtubules were assumed to transport retraction signals away from 187 the front to create a stable tail while protrusion signals were localized and self-amplified at the 188 front 25 . 189 In summary, the present results suggest that cyclic stretching induces two complementary 190 events during the relaxation phase-an immediate retraction that decays after the first several 191 cycles of stretching, and a slower but more persistent protrusion along the transverse direction. 192 Reorientation represents a cumulative result of stepwise axial retractions and transverse 193 extension at each cycle (Fig. 7) Fig. 1g). 214 A casting chamber was then assembled with the activated coverslip, a handle made of a PDMS 215 block attached to a bind-silane treated coverslip ( Supplementary Fig. 1h), and a top coverslip 216 micropatterned with gelatin ( Supplementary Fig. 1f, 1i, and 1k)  (TEMED, Bio-Rad, Hercules, CA) was injected into the chamber and allowed to polymerize for 220 45min. The coverslip with PAA substrate was then attached with vacuum grease to an acrylic block 221 with a hole, to form a culture chamber with stretchable substrate (Fig. 1B). 222 Micropatterned coverslips were prepared as described previously 47 Fig. 1d-e). 228 Strain Analysis. Fluorescent latex beads (0.2µm diameter, red, polystyrene; Molecular Probes, 229 Carlsbad, CA) were added to the acrylamide solution for revealing the ECM micropattern. 230 Microcontact-printed squares 50x50 µm 2 in area were used for visualizing the strain of the PAA 231 gel in response to the stress induced by stage movement. Deformation of the square pattern was 232 imaged in response to stage movements calculated to stretch the gel by 5%, 10%, 15%, or 20%. 233 Fluorescent images were analyzed with ImageJ (National Institutes of Health, Bethesda, MD), and 234 the length and width of the square as a function of stretching distance was analyzed with linear 235 regression. Residual strain upon relaxation was determined by measuring the offset of local marks 236 before and after stretching. 237 Microscopy and Live-cell Imaging. The microscope was covered with a plastic enclosure to serve 238 as an incubator, wherein the temperature and CO2 concentration was maintained for cell viability. Analysis of Cell Orientation, Cell Length, and Protrusion Activities. Cell outline was drawn 246 manually using ImageJ for determining the spreading area. The outline was then fit with an ellipse, 247 and the orientation index was determined as cos2θ where θ is the long axis of the ellipse. Perfect 248 alignment parallel and perpendicular to the direction of stretching was indicated by an orientation 249 index of 1 and -1, respectively. Cell outline was also fitted with a rectangular to determine the cell 250 transverse/axial length as the width/length of the rectangle. Protrusion and retraction activities 251 were calculated using the difference of consecutive images, using a custom MATLAB program. 252 Areas of net protrusion or retraction were normalized against the average cell spreading area.  Comparison of control and nocodazole treated cells for reorientation and length change suggests 463 that the effects of nocodazole take place primarily after 30min of cyclic stretching, as an increase 464 in both transverse elongation and axial shortening (* p < 0.05; ** p < 0.01; n = 98, Mean ± SEM). 465 Similar to the analysis in Fig. 4, changes of peripheral area are measured in nocodazole-treated 466 and control cells. immediately upon the release of stretching (red arrows), followed by progressive extension during 475 the relaxation phase (blue arrows). The effects accumulate upon prolonged cyclic stretching, 476 which leads to the change in cell shape and orientation. 477

Before Stretching
Uniaxial Cyclic Stretching After Stretching

Retraction Protrusion
Contour before stretching Contour before release of stretching Contour upon release of stretching