Mechanochemical feedback and control of endocytosis and membrane tension

Plasma membrane tension is an important factor that regulates many key cellular processes. Membrane trafficking is tightly coupled to membrane tension and can modulate the latter by addition or removal of the membrane. However, the cellular pathway(s) involved in these processes are poorly understood. Here we find that, among a number of endocytic processes operating simultaneously at the cell surface, a dynamin and clathrin-independent pathway, the CLIC/GEEC (CG) pathway, is rapidly and specifically upregulated upon reduction of tension. On the other hand, inhibition of the CG pathway results in lower membrane tension, while up regulation significantly enhances membrane tension. We find that vinculin, a well-studied mechanotransducer, mediates the tension-dependent regulation of the CG pathway. Vinculin negatively regulates a key CG pathway regulator, GBF1, at the plasma membrane in a tension dependent manner. Thus, the CG pathway operates in a negative feedback loop with membrane tension which leads to a homeostatic regulation of membrane tension.


Introduction 40
Living cells sense and use force for multiple functions like development 1 , 41 differentiation 2 , gene expression 3 , migration 4 and cancer progression 5 . Cells respond 42 to changes in tension, passively by creating membrane invaginations/ blebs 6-8 and 43 actively, by modulating cytoskeletal-membrane connections, mechanosensitive 44 channels and membrane trafficking 4,9,10 . Membrane trafficking through endo-exocytic 45 processes can respond and modulate the membrane tension 10 . While exocytosis acts to 46 reduce plasma membrane tension as a consequence of increasing net membrane area, 47 endocytosis could function to reduce membrane area and enhance membrane tension. 48 Membrane tension has long been shown to affect the endocytic process. Decreasing 49 tension upon stimulated secretion or by addition of amphiphilic compounds increases 50 endocytosis 11 12 . On the other hand, an increase in tension upon hypotonic shock 11 or 51 as evinced during mitosis 12 , results in a decrease in endocytosis. Increase in 52 membrane tension on spreading is also compensated via an increase in exocytosis 53 from endocytic recycling compartment providing extra membrane 13 . Together these 54 observations suggest that endocytosis responds to changes in membrane tension or 55 changes in membrane area. However, the specific endocytic mechanisms involved in 56 these responses have not been elucidated. The well-studied dynamin-dependent, 57 clathrin-mediated endocytic (CME) pathway is relatively unaffected by changes in 58 tension 14 while caveolae help to passively buffer membrane tension 7 . 59 We had recently shown that upon relaxing the externally induced strain on cells, 60 tubule like membrane invaginations termed as 'reservoirs' are created 6 . This initial 61 response is a purely passive mechanical response of the plasma membrane and is also 62 observed in synthetic lipid membranes. Similar structures are also formed on recovery 63 from hypo-osmotic shock termed as 'vacuole like dilations' (VLDs). VLDs form 64 passively, similar to reservoirs, albeit different in shape. Subsequent to the formation 65 of either reservoirs or VLDs, cells engage in an active ATP-dependent response that 66 occurs efficiently only at physiological temperatures to restore their morphology and 67 membrane tension 6 . This indicates the deployment of specific active cellular 68 processes following the passive response (see the cartoon in Fig 1a). 69 Here we have explored the nature of such active responses. We have tested the 70 functioning of multiple endocytic pathways on modulation of membrane tension by 71 different approaches. In parallel, we have determined the effects of modulating 72 endocytic processes on membrane tension by utilizing optical or magnetic tweezers to 73 measure membrane tension. Subsequent to the passive membrane response we had 74 observed earlier 6 , we find that a clathrin, caveolin and dynamin-independent 75 endocytic mechanism, the CLIC/GEEC (CG) pathway, rapidly responds to changes in 76 membrane tension, acting to restore it to a specific set point. Perturbing the CG 77 pathway directly modulates membrane tension forming a negative feedback loop with 78 membrane tension to maintain homeostasis. A previously identified mechanical 79 transducer, vinculin, is involved in the homeostatic control of tension; in its absence 80 the CG pathway fails to respond to changes in membrane tension, thereby altering the 81 set point. 82

RESULTS: 83
A rapid endocytic response to changes in membrane tension 84 Active cellular processes are involved in resorbing the 'reservoirs' or 'VLDs' formed 85 following a strain relaxation 6 . To determine whether endocytosis could be one such 86 active process, we monitored the extent of endocytosis by providing a timed pulse of 87 a fluid-phase marker, fluorescent-dextran (F-Dex), during and immediately after the 88 stretch-relax procedure (using a custom built stretch device 6 shown in Fig. 1b). 89 Compared to cells at steady state, there was a dramatic increase in fluid-phase 90 endocytosis immediately after relaxation of the strain (Fig. 1c) while uptake was 91 markedly reduced during the application of the strain ( Supplementary Fig. 1a). This 92 increase in endocytosis was transient and disappeared as early as 90 seconds after 93 strain relaxation (Fig. 1c). This also corresponds to the time scale of resorption of 94 reservoirs by an active process observed earlier 6 . By rapidly upregulating endocytosis, 95 cells thus respond to a net decrease in tension in a fast, transient fashion returning 96 swiftly to a steady state. 97 Earlier studies indicated that exocytosis helped add membrane rapidly in response to 98 increased membrane tension during cell spreading 15 . On deadhering, cells round off 99 decreasing their surface area while on replating, cells spread by adding membrane. 100 Thus it is likely that endocytic pathways could help retrieve membrane on deadhering 101 due to decrease in net membrane tension 8,16 . We reasoned that if an endocytic process 102 is responding to the release of strain during the deadhering process, it would be 103 upregulated during the detachment process. To monitor the extent of endocytosis we 104 followed the uptake of a timed pulse of F-Dex during and immediately after the 105 detachment (3 minutes) and compared it to that measured in the spread state ( Fig.1d  106 schematic). Our results showed that the net fluid-phase uptake underwent a rapid 107 increase while the cells were de-adhering (3 minutes), but subsided back to the steady 108 state level once it was de-adhered and held in suspension (Fig. 1d). Recycling of the 109 endocytic material is similar between steady state and deadhering (Supplementary 110 Fig. 1b). This indicates that there is no block in the recycling rate during deadhering 111 and increase in uptake on deadhering is due to a transient increase in endocytic 112 potential. 113 To further consolidate our findings, we used an alternate method to alter membrane 114 tension. We shifted cells from hypotonic to isotonic medium, which made passive 115 invaginations similar to reservoirs called VLD's 6 . This method also results in an 116 enhancement of fluid-phase endocytosis ( Supplementary Fig. 1c), consistent with the 117 results obtained by the other two methods of altering membrane tension. Together 118 these results suggest that reduction of membrane tension via a number of different 119 methods triggered a fast and transient endocytic response on the time scale of 120 seconds. 121

Membrane tension and the response of multiple endocytic pathways 122
To ascertain which of the multiple endocytic pathways respond to changes in tension, 123 we examined cargo previously shown to be endocytosed via these distinct pathways.

124
A number of endocytic pathways function concurrently at the cell surface [17][18][19][20] . In 125 addition to the well characterized CME pathway, there are pathways that are 126 independent of clathrin but utilize dynamin for vesicle pinching 19,21 . Additionally, 127 there are clathrin and dynamin independent pathways which function in a number of 128 cell lines 22-24 , but not in all 25 . The CLIC/GEEC (clathrin independent carrier/ GPI-129 anchored protein enriched early endosomal compartment) pathway is a clathrin and 130 dynamin-independent pathway, responsible for the internalization of a major fraction 131 of the fluid-phase and several GPI-anchored proteins (GPI-AP) 22,24 , and other plasma 132 membrane proteins such as CD44 26 . Therefore, we used these specific cargoes to test 133 the response of different endocytic pathways while membrane tension was altered. 134 The endocytic uptake of the transferrin receptor (TfR), a marker of CME, did not 135 increase in the cells which exhibited a transient rise in the fluid-phase after a 136 hypotonic shock (Fig 2a) or detachment (Fig. 2b) as visualized using two color 137 fluorescence microscopy. However, uptake of the folate receptor, a GPI-AP that is 138 internalized via the CG pathway 27,28 , exhibited a considerable increase (Fig. 2c). This 139 indicated that clathrin-independent endocytosis rather than CME might be involved in 140 the fast response to a decrease in membrane tension. 141 There are endocytic pathways which utilize dynamin independent of clathrin 142 function 19,22 . Therefore we tested whether the increase in fluid-phase uptake requires 143 dynamin function. We used a conditional triple knock out cell line that removes 144 Dynamin 1, 2 and 3 from the genome 29 , thereby abolishing all the dynamin-mediated 145 endocytic pathways ( Supplementary Fig. 2a). The dynamin triple knockout mouse 146 embryonic fibroblasts (MEFs) shows higher steady state fluid-phase endocytosis 29 . 147 However, cells lacking all forms of dynamin also transiently increased their fluid-148 phase endocytosis upon both stretch-relax cycles to the same extent as wild type 149 (WT)-MEFs (Fig. 3a) and hypotonic/isotonic media changes ( Supplementary Fig. 2b). 150 Thus, neither CME nor dynamin-dependent endocytic pathways appear to respond to 151 an acute reduction in membrane tension. 152 A caveolin-dependent endocytic process is important to retrieve specialized 153 membrane on deadhering 30 , and a caveolae-mediated passive mechanism is reported 154 to buffer the increase in membrane tension and prevent cell lysis triggered by the 155 flattening of caveolae 7 . To test if caveolin-dependent endocytic mechanisms could be 156 important for this rapid endocytic up-regulation, caveolin null MEFs were subjected 157 to the stretch-relax protocol. These cells exhibited a transient increase in fluid-phase 158 uptake similar to their WT controls (Fig. 3a). In addition, caveolin-null cells also 159 exhibit a fast transient upregulation of fluid-phase endocytosis during de-adhering as 160 well ( Supplementary Fig. 2c). 161 We next examined the morphology of the endocytic carriers formed by reduction of 162 membrane tension induced by deadhering using electron microscopy (EM). For this, 163 we utilized Cholera Toxin bound HRP (CTxBHRP), which marks the internalized 164 plasma membrane. We used a procedure in which the surface remnant peroxidase 165 reaction product is quenched with ascorbic acid, revealing only the internalized 166 CTxBHRP labeled membrane 26 . After 5 minutes post-deadhering the major endocytic 167 structures labeled had the typical morphology of CG carriers (or CLICs) comprising 168 structures with tubular and ring-shaped morphology (arrows, Supplementary Fig. 2d). 169 Morphologically-identical structures were also observed in WT MEFs at steady 170 state 31 and in Cav1 -/-MEFs (arrows, Supplementary Fig. 2d) consistent with the 171 observation of fast fluid-phase uptake in Cav1 -/cells via CG ( Fig. 3a and  172 Supplementary Fig. 2c). At this time point, surface-connected caveolae (containing no 173 peroxidase-reaction product) persist in the Cav-expressing WT cells (arrowheads, 174 Supplementary Fig. 2d), consistent with the possibility that the caveolar pathway does 175 not play a significant role in transiently modulating endocytosis at these early times of 176 deadhering. 177 Together, these experiments indicated that the clathrin, dynamin or caveolin 178 dependent endocytic mechanisms do not exhibit a rapid respond to a reduction in 179 membrane tension. This is in contrast to fluid-phase or GPI-anchored protein uptake 180 which is endocytosed via the CG pathway. CG-mediated endocytosis is a high 181 capacity pathway capable of internalizing the equivalent of the entire plasma 182 membrane area in 12 minutes 26 , and of recycling a large fraction of endocytosed 183 material 32 . This pathway is also implicated in the delivery of membrane on cell 184 spreading in response to increases in membrane tension, thus helping to maintain 185 membrane homeostasis 9,13 . 186

CLIC/GEEC (CG) pathway responds to membrane tension 187
Since the CG cargo responded to changes in tension, we explored this finding in 188 further detail. CG pathway is regulated by small GTPase's, ARF1, its GEF GBF1, and 189 CDC42 at the plasma membrane 28,33,34  inhibitor of GBF1, which also decreases fluid-phase endocytosis in cells at steady 195 state but does not affect CME ( Supplementary Fig 3c). Inhibiting GBF1 prevents the 196 increase in fluid-phase endocytosis observed upon stretch-relax ( Fig. 3b) or 197 deadhering ( Supplementary Fig. 3d). Similar to the decrease in fluid-phase on 198 increasing tension during stretch ( Supplementary Fig. 1a), CD44, a CG pathway 199 specific cargo, shows reduced endocytosis during hypotonic shock (Supplementary 200 Fig. 3e). 201 To further confirm that this response is due to CG endocytosis, we assessed the effect 202 of the stretch-relax protocol on cells that lack CG endocytosis. HeLa cells have been 203 shown to lack a robust CG endocytic pathway 25,27 . While the molecular basis for this 204 defect is not understood, we find that fluid-phase endocytosis in HeLa cells is not 205 susceptible to GBF1 inhibition by LG186 ( Supplementary Fig. 4a). In addition, these 206 cells do not show an obvious recruitment of GBF1 to the plasma membrane in the 207 form of punctae as observed in cells exhibiting constitutive CG endocytosis such as 208 6 CHO cells as reported earlier 34 ( Supplementary Fig 4b/4c). Correspondingly, these 209 cells did not exhibit a rapid increase in fluid-phase endocytosis on a hypotonic to 210 isotonic shift ( Supplementary Fig. 4d). 211 These experiments, combined with the up-regulated endocytosis of a CG specific 212 cargo (GPI-AP) (Fig. 2c), suggest that the CG endocytic pathway is specifically 213 involved in the rapid, transient response to changes in membrane tension. 214 Passive and active membrane response to changes in membrane tension 215 As mentioned above, upon a rapid reduction in membrane tension, cells form passive 216 structures such as reservoirs and VLDs similar to the response of an artificial 217 membrane. Reservoirs are formed upon strain relaxation in the membrane after 218 stretching cells, whereas VLDs are formed by water expelled by the cell after a hypo-219 to-isotonic-shock recovery 6 . Both reservoirs and VLDs are reabsorbed and disappear 220 within a couple of minutes, coincidental with an increase in endocytosis. This led us 221 to test if inhibiting the CG pathway could have a measurable impact on the rate of 222 disappearance of such passive structures. We find that the CG pathway exhibits 223 exquisite temperature sensitivity and is barely functional at room temperature (RT), 224 and is not efficient even at 30 o C in comparison to CME in CHO cells (Fig. 4a).

225
Correspondingly, the reservoir resorption in CHO cells was impaired after lowering 226 temperature (Fig. 4b). In addition, inhibition of the CG pathway in CHO cells by 227 inhibiting GBF1 reduced the rate of reservoir reabsorption at 37 °C (Fig. 4a). In 228 contrast, HeLa cells lacking a characteristic CG pathway did not show any difference 229 in the rate of disappearance of reservoirs upon inhibiting GBF1 and was much less 230 affected by lowering of temperature than CHO cells ( Supplementary Fig. 4e). Thus, 231 perturbing CG endocytosis affects the kinetics of resorption of reservoirs. 232 We next examined if passively generated structures could help initiate endocytosis at 233 the sites of their formation. Since the disappearance of each reservoir is gradual and 234 not as a single step process 6 ( Fig. 4b), this indicates that reservoirs are not likely to be 235 pinched off directly as endosomes. Further, we do not observe endosomes form at the 236 site of the reservoirs ( Supplementary Fig. 5a). To test this, we took advantage of our 237 earlier observation that cells plated on polyacrylamide gels do not form VLDs upon 238 hypotonic to isotonic shifts 6 . Whereas the lack of generation of VLDs was confirmed 239 in our cells grown on polyacrylamide ( Supplementary Fig. 5b), the cells still showed 240 an increase in endocytosis similar to when plated on glass, upon exposure to hypo-to-241 isotonic-shock procedure ( Supplementary Fig. 5c). Together, these data suggest that 242 CG endocytosis occurs subsequent to the passive responses of the membrane but the 243 passive invagination formation is not necessary to form CG endosomes. However, the 244 transient increase in CG endocytosis following the passive response helps swiftly 245 resorb the excess membrane helping to restore the membrane morphology. 246

Role of the CG pathway in setting membrane tension 247
Since the CG endocytic pathway responded to changes in membrane tension we 248 hypothesized that it might be involved in the setting of steady state membrane tension 249 as well. To explore this hypothesis, we directly measured tether forces by pulling 250 membrane tethers using optical tweezers 37 . The force experienced by membrane 251 tethers provides a way to measure the effective membrane tension 38 (Fig 5a). We 252 found that acutely inhibiting the CG pathway by inhibiting GBF1 drastically reduced 253 the tether forces in a resting cell (Fig. 5b). To further assess this, we applied 0.5 nN 254 force pulses using a magnetic tweezer device to ConcanavalinA (ConA)-coated 255 magnetic beads attached to the cell membrane ( Supplementary Fig 5d). Consistent 256 with optical tweezers results, resistance to force (stiffness) was reduced in GBF1 257 inhibited cells (Supplementary Fig 5e). That this effect was due to a reduction in 258 membrane tension and not any effects on the cytoskeleton, was corroborated by the 259 lack of a change in the measured stiffness of fibronectin-coated beads attached to cells 260 via integrin-fibronectin adhesions with and without GBF1 inhibition (Supplementary 261 Fig 5f). 262 We next examined tether forces in cells wherein the CG pathway is up-regulated. We 263 reasoned that since the Dynamin TKO cells show a higher fluid-phase endocytosis 264 (Fig. 5c, Supplementary Fig 6c), it is likely that this would increase effective 265 membrane tension. Tether forces were indeed higher in the Dynamin TKO cells 266 compared to control cells (Fig. 5d). Consistent with the role of the CG-pathway in 267 setting membrane tension, inhibiting the CG pathway in Dynamin TKO cells by 268 GBF1 inhibition (Fig. 5c, Supplementary Fig 5h) reduced the effective membrane 269 tension below control levels ( Fig. 5d). 270 To further confirm this observation, we measured tether forces on acutely increasing 271 CG endocytosis by using BrefeldinA(BFA) as reported earlier 33 . BFA treatment 272 disrupts ER to Golgi secretion but also serves to free up ARF1, making it available at 273 the cell surface to increase CG endocytosis 33 . We further confirm that this increase is 274 mediated through a GBF1-sensitive CG endocytosis (Fig 5e). We treated the cells 275 with BFA and measured tether forces using optical tweezers when the increase in 276 endocytosis was most prominent. Tether forces were higher on treating cells with 277 BFA compared to the control case ( Fig. 5f). BFA treatment inhibits secretion 39 and 278 this could also increase the effective membrane tension due to a reduction of 279 membrane delivery from the secretory pathway independent of its effect on CG 280 endocytosis. To test this, we treated HeLa cells with BFA. BFA treatment disrupted 281 the Golgi in both CHO and HeLa cells ( Supplementary Fig. 5g) consistent with its 282 inhibition of the secretory pathway. However, neither fluid-phase uptake ( Fig. 5e) nor 283 the tether forces were affected in HeLa cells (Fig. 5f). This indicated that the increase 284 in tension in CHO cells on BFA treatment is due to an increase in CG endocytosis and 285 not due to a block in secretion in these timescales. 286 Hence, modulating the CG pathway by activating or inhibiting key regulators 287 modifies effective membrane tension directly. Since CG pathway is negatively 288 regulated by membrane tension, this indicates that CG pathway operates in a negative 289 feedback loop with membrane tension. Since CG pathway is specifically modulated 290 by tension, it is conceivable that the molecular machinery regulating CG pathway 291 would be modulated by changes in tension. 292

Mechanical manipulation of the CG endocytosis machinery 293
We tested if key regulatory molecules involved in different endocytic pathways could 294 be directly modulated by changes in tension. GBF1 is involved in the CG pathway 295 and re-localizes from the cytosol to distinct punctae at the plasma membrane upon 296 activation as visualized using TIRF microscopy 33,34 . We imaged GBF1-GFP 297 recruitment to the plasma membrane in live cells using TIRF microscopy, during a 298 hypotonic shock and after recovering from it. GBF1 punctae were lost on hypotonic 299 shock (Fig 6a and 6b) indicating a direct response by GBF1 on increasing tension. On 300 the other hand, recovery from a hypotonic shock caused the rapid assembly of GBF1 301 punctae (Fig 6a and Fig 6b). In contrast, clathrin, which is localized from cytosol to 302 membrane to help in CME is not affected by these similar changes in tension 303 ( Supplementary Fig 6a). These experiments indicated that molecular machinery 304 involved in regulating the CG pathway was modulated by membrane tension unlike 305 that of the CME pathway. 306

Vinculin serves as a mechanotransducer for CG endocytosis 307
For cells to respond to changes in tension, sensing and transduction of this 308 information must occur. Since focal adhesion related molecules help transduce and 309 respond to force 5,40,41 we hypothesized these molecules could transduce a physical 310 stimuli ( To directly test if vinculin could be involved in the tension-sensitive regulation of the 320 CG pathway, we stretched vinculin null cells. Unlike WT MEFs that shows ~82% 321 drop in uptake on stretching ( Supplementary Fig 1a), vinculin null cells show only 322 ~36% drop at the same strain (Fig 7a). Increasing the extent of hypotonic shock 323 showed a concomitant decrease in fluid-phase endocytosis of WT cells (Fig 7b). By 324 contrast, vinculin null MEFs were much more refractory to the same extent of 325 hypotonic shock (Fig 7b). Furthermore, upon strain-relaxation, fluid-phase 326 endocytosis in vinculin null MEFs did not show an increase (Fig 7a), unlike that 327 observed for CHO cells (Fig 1c) or wild type MEFs (Fig 3a). This was further tested 328 in the deadhering assay where vinculin null cells again did not show an increased 329 fluid-phase uptake, unlike the WT control cells ( Supplementary Fig 6c). Thus 330 vinculin null cells do not respond to changes in tension similar to the WT cells. 331 Fluid-phase uptake in vinculin null MEFs is much higher than wild type cells (Fig  332  7c). To test if the endocytic effects of vinculin null cells are specifically due to 333 vinculin, we expressed full length vinculin in vinculin null cells. This caused a 334 decrease in fluid-phase endocytosis ( Supplementary Fig 6d). Further, inhibiting GBF1 335 in vinculin-null cells with LG186 decreased fluid-phase uptake to the same levels as 336 cells expressing vinculin, confirming that a GBF1 sensitive CG pathway is functional 337 here ( Supplementary Fig 6d). This indicates that GBF1 operates downstream of 338 vinculin and vinculin negatively regulates CG pathway. 339 Since vinculin-null cells have a higher basal endocytosis rate it is possible that they 340 are unable to increase their endocytic capacity further in response to a decrease in 341 membrane tension. However, vinculin null cells respond to BFA treatment to increase 342 their endocytic rate in a manner that is also sensitive to GBF1 inhibition, similar to 343 wild type cells ( Supplementary Fig 7a). 344 We next tested if GBF1 shows a tension-dependent membrane localization of GBF1 345 in vinculin null cells. The level of punctae remained constant and failed to respond to 346 hypotonic shock (Fig 7d) unlike that observed in WT MEF (Fig 6a). The density of 347 GBF1 punctae at the plasma membrane was also slightly higher in vinculin null cells 348 compared to WT cells ( Supplementary Fig 6b). This is consistent with higher fluid-349 phase endocytosis in vinculin null cells compared to control MEF cell line (Fig 7c). 350 Further, we tested how the steady state membrane tension in vinculin null cells is 351 compared to WT cells. Tether forces measured using optical tweezers showed a 352 higher value for vinculin null cells compared to wild type cells (Fig 8a). The high 353 tether force in cells lacking vinculin was drastically reduced on inhibiting the CG 354 pathway (Fig 8a), consistent with the role of the CG pathway in regulating the 355 effective membrane tension. These experiments show that vinculin acts as a negative 356 regulator of CG pathway and is necessary for the transduction of physical stimuli for 357 the biochemical control of the CG pathway. 358

DISCUSSION 359
Membrane tension has been long proposed to be tightly coupled to vesicular 360 trafficking through endo-exocytic pathways. However, the specific trafficking 361 mechanisms have remained elusive. Here, we show that membrane tension and CG 362 endocytosis operate in a negative feedback loop that helps restore any change from a 363 set point (model: Supplementary Fig. 8b). When membrane tension decreases, it 364 transiently triggers the CG pathway, bringing about a fast endocytic response to reset 365 the cell's resting membrane tension. On the other hand, increasing membrane tension 366 has the opposite effect; the CG endocytosis is inhibited in a proportional manner. The 367 membrane flux through the CG pathway also has an effect on the effective membrane 368 tension. Acutely lowering the CG pathway decreases membrane tension while 369 upregulating the pathway increases membrane tension. Thus, changes in membrane 370 tension lead to an inverse effect on CG endocytosis, while changes in CG endocytosis 371 lead to a direct effect on plasma membrane tension. This type of a response known as 372 a negative feedback loop is used in many different biological contexts to maintain 373 homeostasis 44 . 374 The CG endocytic pathway specifically responds to acute changes in membrane 375 tension despite multiple pathways operating simultaneously at the plasma membrane. 376 Caveolae passively buffer increases in tension 7 , while the clathrin-mediated pathway 377 concentrates specific ligands and mediates robust endocytosis despite the increase in 378 tension 14 . De-adhered cells exhibit an increased caveolin-mediated internalization that 379 persists over hours, and is crucial for the removal of specific membrane constituents 380 and anchorage-dependent growth and anoikis 30 . On the other hand, unlike the 381 caveolar pathway, the CG pathway showed a higher transient upregulation of 382 endocytosis only during deadhering which does not persist in suspension. 383 The fast response of the CG pathway on strain relaxation is lost within 90 seconds. 384 The loss of this transient response could be even faster as at present 90 seconds is the 385 dead time in our experiments. Fast clathrin-independent mechanisms have been 386 reported in different contexts. Ultrafast endocytosis occurs at synapses following a 387 synaptic vesicle fusion to retrieve the excess membrane. This fast clathrin-388 independent but dynamin-dependent process is temperature sensitive as well 45,46 . 389 Endophilin dependent FEME pathway is another clathrin-independent but dynamin-390 dependent pathway. HeLa cells seem to predominantly have AP2, GRAF1 and 391 dynamin-dependent machinery 25 . Consistent with this, inhibiting AP2 or GRAF1 in 392 HeLa cells inhibits fluid-phase endocytosis. These cells respond in a slower rate to the 393 changes in osmotic shock and shows blebbing on inhibiting this pathway due to lack 394 of endocytosis 47 . Here, we find that the GBF1/ARF1/CDC42 dependent CG pathway 395 shows a fast transient response to changes in tension and is more sensitive to lowering 396 of physiological temperature compared to the CME pathway. In the absence of such a 397 fast pathway, other slower endocytic mechanisms operate to internalize the excess 398 membrane, lack of which might lead to blebbing. The CG pathway helps to swiftly 399 respond and reset any changes from the steady state, thereby also helping to set the 400 resting membrane tension of a cell. This indicates that different endocytic pathways 401 have distinct functions and the CG pathway may be responsible for membrane tension 402 homeostasis. 403 Similar to the endocytic response, exocytic processes in a cell could modulate and 404 respond to changes in tension. Exocytic processes in a cell help in addition of 405 membrane to the cell surface and reduction of membrane tension 9 . Unlike 406 endocytosis, exocytosis seems to be positively regulated by membrane tension. High 407 membrane tension increases the exocytic rate and could regulate the mechanism of 408 vesicle fusion 48,49 . Increase in membrane tension during cell spreading activates 409 exocytosis to increase spread area through a GPI-anchored protein rich endocytic 410 recycling compartment 13,15 . This increase in area is independent of secretory pathway 411 or other exocytic mechanisms. CG pathway takes in a major fraction of GPI-anchored 412 proteins 27 and recycles a huge fraction of its endocytic volume 32 . On increasing 413 tension, we find that CG endocytosis is downregulated preventing further increase in 414 tension but it could be recycling through CG pathway that helps add membrane to 415 restore the steady state tension (Cartoon: Fig 8b). Thus, regulation of membrane 416 cycling through the endo-exocytic leg of CG pathway could be important in 417 membrane homeostasis and further research would help bring about a complete view 418 of this mechanism. 419 We find that the active response by CG pathway follows the passive response via 420 membrane invaginations (i.e. reservoirs and VLD) and helps in efficient resorption of 421 these passive local membrane structures. There could be other active cellular 422 mechanisms driving the flattening of these invaginations as well. However, these 423 membrane invaginations are not necessary for the creation of the CG endosomes. 424 Thus, following a reduction in tension these are two parallel responses by the cell, one 425 passive and the other active eventually leading to excess membrane internalization 426 through CG endosomes. 427 Similar to the passive response, physical parameters could directly regulate the active 428 endocytic machinery by influencing the extent of membrane deformation needed to 429 make an endocytic vesicle. A higher membrane tension makes it more difficult to 430 deform the membrane, thus producing fewer endosomes, and vice versa, alleviating 431 the need for a specific mechanotransduction machinery. However, our results from 432 studying the vinculin-null cells suggest otherwise. Vinculin, a key focal adhesion 433 protein, transduces many mechanical inputs at the site of the focal adhesion into 434 information for the cell to process 41,43,50 . In this context, it appears that vinculin plays 435 a central role in transducing the increase (or decrease) in membrane tension to the CG 436 pathway to help inhibit (or activate) its endocytic mechanism (Cartoon: Fig 8b). This 437 appears to be effected by its control of a key regulator of CG endocytosis, GBF1, the 438 GEF for ARF1. In WT cells, GBF1 forms tension-sensitive punctae at the cell surface 439 wherein increasing tension abolishes these punctae and decreasing tension increases 440 it. In the vinculin-null cells, this tension-dependent regulation is lost and CG 441 endocytosis appears to be uncoupled from tension regulation. Thus, vinculin is 442 important for a tension-sensitive negative regulation of a key effector of CG pathway, 443 translating mechanical information into a biochemical read out to influence the 444 endocytic rate. This negative feedback loop between effective membrane tension and 445 CG pathway thus is mediated through vinculin and maintains the cells at a lower 446 effective membrane tension. Different functional modules operate in a focal adhesion 447 for mechanotransduction 43 . However, the precise mechanism behind the ability for 448 vinculin to regulate the availability of GBF1 at the cell surface is not yet understood 449 and is a subject of further investigation. 450 Modulations in membrane tension are used in multiple cellular processes 4,51-54 and 451 CG pathway could have a role in these. In migrating fibroblasts, CG endosomes are 452 localized to the leading edge and transient ablation of these endosomes inhibits 453 efficient migration 26 . Increase in the membrane tension at the leading edge keeps the 454 neutrophil cells polarized and helps in its migration 54 . In a separate study in 455 neutrophils, GBF1 localizes to the leading edge by binding to products of 456 phosphatidylinositol 3-kinase (PI3K), recruits ARF1, and this localization is needed 457 for unified cell polarity 55 . We have found that PI3K products help recruit GBF1 to the 458 plasma membrane and this is necessary for CG endocytosis 56 . Thus, one could 459 speculate that a polarized CG pathway and its regulators operating in migrating cells 460 could be modulating membrane tension by regulating membrane trafficking. 461 Endocytic pathways are proposed to be at the core of a eukaryotic cellular plan 462 integrating multiple inputs over spatio-temporal scales 57 . They are also necessary for 463 tissue patterning: the CG pathway is utilized for Wingless signaling for patterning of 464 the Drosophila wing disc during larval development 56 . Here we find that a high 465 capacity CG pathway that turns over a huge fraction of the plasma membrane 26,32 and 466 sensitive to membrane composition 32 is modulated by temperature and 467 mechanochemical inputs. A vinculin-mediated negative feedback loop between 468 membrane tension and the CG pathway helps maintain the cell at a lower tension set 469 point (Cartoon : Fig 8b). This could also help in increasing the potential for 470 modulating membrane tension to regulate other cellular processes. Thus, the CG 471 pathway responds and coordinates a variety of cellular inputs including membrane 472 tension and is likely to function in multiple physiological contexts. 473

ACKNOWLEDGEMENTS 474
We Research (  vacuum below the circular PDMS sheet, which stretches it in a calibrated manner. 667 Releasing the vacuum relaxes the strain on PDMS thus relaxing the cell. Cells plated 668 on PDMS can be imaged in an upright or inverted microscope as required. (c) 669 Endocytic response on strain relaxation. CHO cells were pulsed for 90 sec with TMR-670 Dex at steady state (steady state), immediately on relaxing the stretch (stretch-relax), 671 or after a waiting time of 90 seconds after relaxing the stretch (stretch-relax-wait). 672 After the pulse, cells were quickly washed with ice cold buffer and fixed, followed by 673 imaging on a wide field microscope. Images show representative cells used to receptor endocytosis in the de-adhered condition normalized to those measured in the 708 Spread condition (grey dashed line). In each experiment, the data represent the mean 709 intensity per cell (± S.D) from two different experiments with duplicates containing at 710 least 100 cells per experiment. *: P < 0.001, ns: not significant. Scale bar, 10 µm. 711 Reservoir resorption on inhibiting CG pathway and decreasing temperature. The 742 reservoir fluorescence intensity after stretch relax of CHO cells transfected with a 743 fluorescent membrane marker (pEYFP-mem) was quantified as a function of time at 744 37 o C in the absence (37°C control) or presence of LG186 (37°C inhibitor), or at room 745 temperature (26°C control). Each point represents mean ± S.E.M from more than 100 746 reservoirs from at least 10 cells. Scale bar, 10 µm. 747 microscopy. They exhibit GBF1-punctae at the plasma membrane, which is 782 modulated by alterations in osmolarity obtained by changing the media from isotonic 783 (Iso) to 40% hypotonic (Hypo) and back to isotonic (Iso). (b) Quantification of the 784 number of punctae per cell during hypotonic shock and subsequent shift to isotonic 785 medium. The GBF1 spots upon hypotonic shock and subsequent shift to isotonic 786 medium is normalized to number of spots in the respective cell determined before 787 hypotonic shift and plotted as a box plot. Each data point is a measurement from a 788 single cell and box plot shows data of 12 cells from two independent experiments. 789 Scale bar, 10 µm. 790 (Vin +/+ with LG186), 25 (Vin -/-) and 29 (Vin -/-with LG186)). Vinculin null cells 820 shows a higher basal membrane tension compared to WT MEF; inhibiting the CG 821 pathway drastically reduced membrane tension in both cell lines. (b) CG pathway and 822 membrane tension operates in a vinculin dependent negative feedback loop to 823 maintain homeostasis. Reduction of tension from its steady state leads to a passive 824 response by the formation of reservoirs or VLDs. The decrease in the effective 825 tension inactivates a vinculin-dependent machinery, resulting in an increase in active 826 GBF1, which increases the CG pathway and rapid internalization of the excess 827 membrane. This is a fast transient response that appears to restore the steady state. On 828 the other hand, increasing the membrane tension from steady state activates vinculin 829 dependent machinery, inhibiting the CG pathway, via the reduction of GBF1 830 recruitment. The increase in effective tension could also activate exocytic machinery 831 which adds membrane resulting in restoration of the steady state. Thus, a vinculin 832 dependent mechanochemical-regulation of the CG pathway through a negative 833 feedback loop helps in maintaining plasma membrane tension homeostasis. Shifting from hypotonic to isotonic media results in an increase in fluid-phase uptake. 848 Cells were pulsed with TMR-Dex for 1 minute either in steady state (Iso) or after 849 hypotonic shock (hypo-iso) for one minute. Images ( uptake and Tf uptake normalized to surface receptor level shows that the effect of 902 LG186 was only on the fluid-phase uptake but not on TfR endocytosis. phase uptake that the WT cells, this uptake is sensitive to the inhibition of GBF1, 992 confirming that it is the CG endocytosis that has a higher activity in Dynamin TKO 993 cells (quantified in Fig. 5c). Scale bar, 10 µm.