Mechanochemical feedback control of dynamin independent endocytosis modulates membrane tension in adherent cells

Plasma membrane tension regulates many key cellular processes. It is modulated by, and can modulate, membrane trafficking. However, the cellular pathway(s) involved in this interplay is poorly understood. Here we find that, among a number of endocytic processes operating simultaneously at the cell surface, a dynamin independent pathway, the CLIC/GEEC (CG) pathway, is rapidly and specifically upregulated upon a sudden reduction of tension. Moreover, inhibition (activation) of the CG pathway results in lower (higher) membrane tension. However, alteration in membrane tension does not directly modulate CG endocytosis. This requires vinculin, a mechano-transducer recruited to focal adhesion in adherent cells. Vinculin acts by controlling the levels of a key regulator of the CG pathway, GBF1, at the plasma membrane. Thus, the CG pathway directly regulates membrane tension and is in turn controlled via a mechano-chemical feedback inhibition, potentially leading to homeostatic regulation of membrane tension in adherent cells.

In this manuscript, the authors address a very important and still debated question about the origin of tension homeostasis in cells. It has already been established that cells use endo-and exocytosis to regulate their membrane area and respond to membrane tension changes, but which particular pathway is used and how the tension set-point is controlled is not so clear. Thottacherry et al use different methods to change cell membrane tension, practically by inducing a membrane tension decrease either by stretching cells and relax them abruptly or by de-adhering them and observing their relaxation to basal state. They convincingly demonstrate that the CLIC/GEEC (GC) pathway is directly involved in the rapid elimination of the membrane reservoirs induced during the process, at least in cells where this pathway is operative. During cell stretching, they also observe a decrease of the uptake. The dynamin-dependent (CME or CIE) or caveolin-dependent pathways seem not to be involved in this mode of tension regulation. Their experiments show that the tension set-point can be modulated by changing the level of key regulators of the GC pathway. Moreover, they have also interestingly identified vinculin as a member of the mechanosensing machinery that responds in a negative feedback loop for the tension regulation. I find that this paper opens new perspectives on the role of non-conventional endocytic pathways on the mechanical regulation of cell plasma membrane and should eventually be published in Nature Communications, providing that some points are clarified or complemented.
1-Vinculin is normally considered as part of the focal adhesion (FA) complex. But, talin is the mechanosensitive element : when a force opens it, its activation leads to vinculin binding, thus vinculin acts downstream of talin. Would it be possible to test whether talin is also involved in tension sensing? At this stage, the connection between vinculin, the CG pathway and the GEF GBF1 appears quite phenomenological. Is there a direct coupling between adhesion and tension regulation? What would be the molecular mechanism? I guess it will require much more work to establish this, but it could be at least discussed. Moreover, in the experiments, the methods to change membrane tension have at the same time a mechanical action on FA and eventually on vinculin, even when cells are detached since both adhesion and tension are released. Is there a way to make a hypertonic shock or aspirate a cell in suspension to check if in non-adhering conditions, vinculin is still involved in the mechanosensing machinery? 2. The authors show that CG regulates tension either by resorbing large membrane reservoirs when tension amplitude changes are large, but also in the case of sub-macroscopic deformations. It is not clear for me how the same molecular endocytic machinery practically works in these very different situations, and internalize a large range of membrane patch sizes. Could the authors comment on this point? Moreover, since CG is also used for controlling the level of some specific membrane proteins in the plasma membrane, how much tension control does interfere with protein traffic?
3. Figure 5d  The manuscript by Thotacherry et al. entitled, 'Mechanochemical feedback and control of endocytosis and membrane tension', the authors present an experimental study on how membrane tension and CG pathway for endocytosis forms a mechanochemical feedback. This is an important question in the field of trafficking and the authors have conducted a considerable number of experiments to investigate how membrane tension affects the CG pathway. There has been a considerable interest in recent times about the role of membrane tension and endocytosis. Here, the authors study a dynamin and clathrin-independent pathway, CG endocytosis and conclude that tension and CG act in a feedback loop. The main novelty of this study is the two-way analysis --changing tension through multiple modes affects the extent of CG pathway and changing the extent of the CG pathway affects membrane tension. This is demonstrated quite convincingly through a multitude of experiments. I have the following comments for the authors to consider to strengthen their study. Major comments: 1) Membrane tension --in cells, membrane tension is closely related to the cortical actin organization. The authors should clarify what they mean by membrane tension and how one is to interpret this. They allude to this somewhat in line 258, but this can be discussed better and perhaps earlier.
2) Multiple modes of inducing tension: how do they affect the cytoskeleton? The authors used multiple modes of tension induction --stretch relaxation, spreading/deadhering, osmotic stress -to induce membrane tension. How do these different modes affect the underlying cortical cytoskeleton? How do we know that the only impact of these tension inducing modes in on the plasma membrane? 3) Role of tension in CME --from my reading of the manuscript, it appears that the authors suggest that tension doesn't play an important role in regulating CME. See page 2, line 58 for example. But there are many references that indicate that CME is impacted by tension, only a few of which are given below. Can the authors expand on their discussion or at least clarify the message they are seeking to provide with respect to CME? 4) There are two separate parts to this study --the feedback between CG and tension and the role of actin cytoskeleton proteins. The first part is a convincing study. However, the study on the role of vinculin appears to be only a preliminary investigation. It is not clear to me why vinculin alone was chosen for this study and how other focal adhesion proteins might play a role. I suggest that the authors conduct a separate, in-depth investigation of focal adhesion proteins rather than just adding one set of experiments here. 5) Use of different cell types --Why did the authors use HeLa and CHO cells in certain experiments ( Figure 5)? I was trying to follow the rationale for this and understand the arguments made in lines 280-285 but couldn't. The authors should clarify this. 6) A mechanistic, theoretical model that explains how membrane mechanics, tension, and the CG pathway interact could strengthen the paper. By this, I don't mean an in-depth computational study, but rather a physical model that can help develop intuition for how membrane properties, tension, and GBF1 interact that will put all the experimental observations in context.
Minor comments: 1. The authors should proofread the text carefully. There are some spelling mistakes and awkward sentences in the text throughout. 2. This is a matter of personal preference but the title has two ands and doesn't necessarily reflect the particular endocytic pathway being studied. Perhaps the authors can revise the title to be more specific? 3. Abbreviations appear in a number of places without being fully expanded upon.
In general, I find the paper quite sensible and interesting. The notion of a negative feedback between tension and endocytosis that could help maintain an homeostatic tension has been around for some time, but it is nice to see experimental evidence for it, and to have this process linked with a particular pathway, with hints of a molecular mechanism. I think this is a very useful paper. My comments are two-fold. On the one hand, I think the authors are at places over-interpreting their data, and would like to see this amended. I also find that a more precise characterisation of the dynamical nature of the mechanochemical feedback would be very valuable. These comments are developed below. 1) L.177: "Together, these experiments indicated that the clathrin, dynamin or caveolin dependent endocytic mechanisms do not exhibit a rapid respond to a reduction in membrane tension." What is shown here is that there is a rapid response to membrane tension reduction that is clathrin, dynamin, and caveolin independent, but not that these pathways do not exhibit such response. This sentence should be changed to reflect the actual findings.
2) L.201 and below. The author claim that the tension-dependent endocytosis mechanism creates a feedback loop hat et membrane tension to an homeostatic value. Is tension regulation also possible for HeLa cells, which lack CG endocytosis? Could the author show the evolution of tether force on HeLa cell upon stretch-release, or hypo-osmotic shock. It is likely that these cells also show a transient variation of tension, which would suggest, at best, the existence of redundant tension regulation mechanisms. However, the dynamics of tension relaxation could be very different, perhaps not as fast as the one shown here to depend upon CG endocytosis.
3) L.255 and Supl. Fig. 5: Calling the response to a magnetic tweezer "membrane stiffness" seems completely misleading to me. If only the membrane were involved to that response, one would expect a stiffness in the range of tens of pN/µm at the most (close to the value of membrane tension), while the one measured here is 1000 larger, and is certainly not solely due to the membrane mechanical response. While an effect can be seen upon LG186 treatment (Fig.S5 d-f), I find it very doubtful that this can be directly related to variation of membrane tension, as stated in the text. Consequently, the sentence (L:257):" That this effect was due to a reduction in membrane tension and not any effects on the cytoskeleton, was corroborated…" is, to my view, inconsistent with numerical estimates and not acceptable. Short of a clear understanding of what the so-called "membrane stiffness" measured by magnetic tweezers actually means, and how it is related to membrane mechanics, this sentence should be either heavily watered down or removed, together with the magnetic trap experiments.
4) The tether force is used to quantify membrane tension at steady-state under different conditions. I imagine that tether force measurement is done on adhered cells, although I don't think this is explicitly mentioned (it should be). As a number of different factors may affect the steady-state tension, tether force measurement should also be performed in the mechanically perturbed state (stretched, or hypo-osmotic shock), under different condition to test the mechanical feedback hypothesis. It is presumably possible to measure the temporal evolution of the tether force during the transient response of cell to perturbation, especially in the case of stretch-release cycles. Could these experiments be performed to see if there is a direct correlation between the evolution of the tether force and reservoir resorption (as in Fig.4).
The dynamics of recruitment of the CG machinery ( Fig.6) could also be monitored dynamically before, during, and after the osmotic shock. This could also be combined with optical tweezers measurement of tether force to obtain a direct dynamical correlation between the two processes. Such measurement would greatly support the claim of mechanical feedback and its involvement in setting the homeostatic tension. Fig.4. Why is there no quantification for T=37 deg.?

5)
Reviewer #5 (Remarks to the Author): In this paper, the authors reported that the CLIC/GEEC (CG) pathway of endocytic process is regulated by membrane tension in a vinculin dependent manner. The results are interesting and the evidences are compelling. I would like to recommend publication of this work after the authors clarify a few points and make changes accordingly.
One of the key experiments is the use of optical tweezers to measure the force in membrane tethers. The author didn't provide information on how the bead is attached to the membrane. Did they use ConA-coated bead similar to what used in their magnetic tweezers experiment? Further details of how this membrane tether was formed are needed.
In the optical tweezers experiments, the changes in the tether force were used to establish a link between the CG pathway and the membrane tension. However, the tether force depends on other conditions such as the length of the tether, the locations on the cells where the measurement is done, and membrane adhesion to cytoskeleton. The authors need to clarify how these conditions are controlled such that the level of tether forces they measured can be correlated with the endocytic process through the CG pathway.
The authors also investigated the correlation between the local stiffness of the cell and the endocytic process through the CG pathway. The local stiffness was measured by applying 0.5 nN force pulses using a magnetic tweezers device to ConA-coated magnetic beads that were attached to the cell membrane and detecting the resulting bead displacement. This force is one order of magnitude greater than the tether force measured using the optical tweezers. I wonder why such a large force didn't produce a protruding member tether. Does it imply a strong membrane attachment to cytoskeleton that prevents formation of the membrane tether? If this is the case, then this membrane stiffness measurement does not reflect membrane tension, and cannot be compared with the tether force measured using optical tweezers. The implications of the results from the magnetic tweezers measurement on the endocytic process through the CG pathway need to be re-discussed.
The authors stated "That this effect was due to a reduction in membrane tension and not any effects on the cytoskeleton, was corroborated by the lack of a change in the measured stiffness of fibronectin-coated beads attached to cells via integrin-fibronectin…" Based on this sentence, it seems that the authors suggest that the local stiffness measured by the magnetic tweezers device is not due to membrane adhesion to the cytoskeleton. If this is what the authors implied, then they need to quantitatively establish a link between the local membrane stiffness they measured and the membrane tension, and need to explain why the protruding membrane tethers did not form in the magnetic tweezers experiments.
The authors identified vinculin as a key player in the observed mutual dependence between the CG pathway of endocytic process and membrane tension, but they didn't provide a clue of how vinculin might be involved in this regulation. Although it would be too much to ask the authors to completely elucidate the mechanism of the involvement of vinculin in this regulation, it will be very helpful if they provide information whether it requires vinculin activation through binding to talin at focal adhesion. Repeating the measurement in talin-deleted cells can provide this piece of information. Reviewer #1 (Remarks to the Author): 6 7 This study investigates how endocytosis is coupled to membrane tension 8 variations induced by cell stretching or cell deadhering. The data presented 9 here convincingly indicate that among the different endocytic pathways 10 tested, the CLIC/GEEC (CG) pathway plays a prominent role in rapidly  11 retrieving the excess of plasma membrane that follows stress relaxation.

12
They further identify vinculin as a possible regulator of this process. We thank this reviewer for his/her critical review of the manuscript and for 28 appreciating a role for the CG pathway in tension response. There are three parts 29 to the significance of our finding. First, we extensively test all the major endocytic 30 pathways and identify CG endocytosis as the major pathway that acutely responds 31 to changes in tension. Second, we measure tether forces to explore the role of 32 endocytic pathways on membrane tension. Finally, we find that this pathway is 33 under a mechano-chemical control, by identifying a key mechanotransducer, 34 vinculin, in this process.

36
In the revised version we further strengthen our understanding of the molecular and couple of months before we put up our manuscript on 'bioRxiv' and 46 submission for review at Nature communications. We believe that Holst et al, is 47 looking at a qualitatively different response. Holst et al, suggests that a GRAF1 48 mediated endocytic mechanism which occurs in HeLa and certain colon cancer 49 cells suppress blebbing. This is relevant since blebbing-mediated migration may 50 be important for metastasis. They conclude that a clathrin-independent pathway 51 may be involved in this response, since GRAF1 knock down affects the hypotonic-52 isotonic (Hypo-Iso) stimulated uptake whereas AP2-knock down is less effective. 53 54 In this regard, we note that Holst et al differs from our study in two main aspects. 55 Firstly, this study uses HeLa cells which have been shown to have a predominantly 56 clathrin-mediated endocytic pathway 2 (further tested in our study and elaborated 57 below), and secondly the conditions for the tension-decrease stimulated endocytic 58 process is very different from the conditions we have used in our study. We 59 address these issues below to clarify that we are indeed looking at a very different 60 process from that explored by Holst et al 2017. 2017). This is consistent with an earlier study that conclusively show CME is the 65 predominant endocytic pathway in this cell line 2 . We further confirm this 66 observation using AP2 shRNA and find that indeed fluid-phase uptake is 67 predominantly clathrin (AP2)-mediated in these cells ( Supplementary Fig 5a).

68
Consistent with a shutting down of the clathrin-mediated pathway in HeLa cells; 69 the AP2 shRNA causes a decrease in transferrin (Tf) uptake, and an increase in 70 surface transferrin receptor (TfR). We have tested Tf uptake, TfR surface levels 71 and fluid-phase uptake in the same cells to avoid variability and this is quantified 72 following three color imaging. However, in cells that have a functioning CG 73 pathway, the same AP2 shRNA mediated knockdown while inhibiting Tf uptake 74 and enhancing TfR surface levels, results in an increase in fluid-phase uptake 75 ( Supplementary Fig 5b). This increase in fluid uptake upon AP2 knock down is 76 similar to inhibiting dynamin-mediated CME 3 (Figure 5c). In fact, this uptake takes 77 place almost entirely by the CG pathway, since it is almost completely inhibited on 78 inhibiting the CG pathway ( Figure 5c). The CG pathway takes in the major fraction 79 of extracellular fluid along with other cargoes as reported earlier while Tf that 80 traffics through CME does not colocalize with fluid in these cells. This pathway is 81 regulated by CDC42, ARF1, GBF1 and IRSP53 and is independent of dynamin, AP2 82 or Clathrin 4-7 . Further, HeLa cells do not have a robust CG pathway 8 . We find that 83 HeLa cells do not respond to inhibition of GBF1 ( Supplementary Fig 5c). GBF1, a 84 GEF for ARF1, also not seem to be present in an active form at the cell surface in 85 HeLa cells unlike CHO with bonafide CG pathway ( Supplementary Fig 5d/ Supplementary Fig 7a and 7b).

102
In the Holst protocol which represents a drastic change in osmolarity, we find that 103 many cells are distorted and dextran-filled structures are located outside phase 104 contrast areas demarcating cells (Arrows and Insets, Supplementary Fig 7a and  105 7b). This suggests that in many instances the membrane may have been 106 completely separated from the cortex, and a massive endocytic engulfment may 107 have initiated as a result. Thus, we conclude that the mechanism being assessed 108 by Holst is very different from the responses we have addressed. It is also likely 109 that such a mechanism may also preferentially utilize GRAF1, but does not appears 110 to utilize the CG machinery.

112
Therefore, we suggest that it is likely that cells may utilize different mechanisms 113 depending on the extent of the hypotonic shock, and the resultant change in 114 membrane tension. Importantly, under the conditions of modulating membrane 115 tension as described in our studies, the CG pathway is the dominant mechanism 116 for the restoration of membrane tension and area. Blebbing appears to be a 117 cellular response to the inhibition of all endocytic mechanisms 9 and Holst et al 118 2017 seem to have uncovered the importance of this in the migration of cancer 119 cells. In the manuscript we discuss the significance of the Holst et al. study, and 120 propose an alternative explanation for their results (see page no 6, line no 217 and 121 discussion-page 11, line 473).

123
We have extensively characterized the role of different endocytic pathways using 124 multiple means of modulating tension and more importantly we measure tether 125 forces to quantify membrane tension. Our study proposes to understand the role 126 of these endocytic processes in responding to changes in membrane tension. Our 127 work also brings out a novel link between the focal adhesion protein, vinculin, in 128 regulating the CG pathway by a mechano-chemical feedback mechanism. This is a 129 hitherto unappreciated connection between mechanical inputs and chemistry. We 130 are confident that the readers of Nature Communications would find the 131 mechanochemical control of an endocytic pathway and its importance in 132 maintaining membrane homeostasis, not only noteworthy, but of broad 133 significance. 134 135 -The authors refer to VLD as the "vacuole like dilations" that are formed 136 upon recovery from hypo-osmotic shock. It is unclear how the CG pathway is 137 related to VLDs. Indeed VLDs are not pictured on the final model presented 138 in Fig. 8 We have used multiple means to change membrane tension, namely 'stretch-149 relax', deadhering, and osmotic shocks to study morphological changes in the cell 150 membrane and endocytic responses ensuing. 'Reservoirs' and 'VLDs' represent 151 similar types of membrane responses that differ in shape (see our previous 152 manuscript; , formed by stretch-relax and hypotonic-153 isotonic shift, respectively 10 . Subsequent to both these responses we observe a 154 very similar CG endocytic response. The final model in the original manuscript 155 tries to assimilate these multiple responses to change in tension and hence VLDs 156 are not specifically depicted in our figure. (Note that in the revised manuscript, 157 the final model is modified to reflect the theoretical model that integrates our 158 findings of endocytic response and membrane tension.) To be clear, VLDs are a 159 fast 'passive' membrane invagination in response to changes in osmolarity that is 160 followed by 'active' endocytic response to take in the excess membrane, 161 accumulated to accommodate the hypotonic treatment. Since VLDs are passive 162 membrane invaginations at the cell surface, cargo endocytosed by CG cannot be 163 delivered to VLDs. Please refer to the reviewer's figure that explains this 164 (Reviewer Fig 1). Our experiments indicate that formation of this passive 165 structure is not necessary for the ensuing endocytic response. By growing cells on 166 a hydrogel, the water that is pumped out of the cell during the restoration of 167 osmolarity, is resorbed by the hydrogel, preventing the generation of VLDs as we 168 observed previosly 10 ( Supplementary Fig 9b). However, the excess membrane 169 continues to be retrieved via upregulation of CG pathway even in the absence of 170 VLD formation ( Supplementary Fig 9c). We thank the reviewer for pointing this 171 out and hope that the rewritten version helps the readers as well.

173
As previously mentioned, the role of GRAF1 has mostly been studied in HeLa cells 174 that lack the CG pathway. In Holst et al, GRAF1 levels (either on inducing or 175 knocking down the same) appear to correspondingly modulate AP2 levels that 176 could explain its effect on fluid uptake since fluid uptake is AP2 dependent in HeLa 177 cells (Holst et al., Fig 1E) 1 . GRAF1 associates with CME derived endosomes 2 and 178 CG-derived endosomes and affect the maturation (and probably recycling) of 179 endosomes 11 . GRAF1 also appears to be recruited to the plasma membrane and 180 VLDs during the extreme osmotic shock protocol in Holst et al. that is independent 181 of CG pathway. The exact role of GRAF1 in CG and CME pathway would require an 182 in-depth study of its own which is not the focus of this manuscript. Here we 183 propose a role for CG endocytic pathway in mediating an acute response to 184 moderate and possibly more physiological membrane tension changes. It is 185 important to note that we propose that a constitutive endocytic pathway that is 186 dynamin, clathrin independent, but dependent on CDC42, ARF1 (and its GEF, 187 GBF1), as well as IRSp53, is upregulated to respond to acute but moderate 188 alterations in membrane tension, mediated by different mechanisms. Whilst in the 189 Holst protocol, cells that do not exhibit this form of CG endocytic mechanism 190 (HeLa cells and IRSp53 null cells) also exhibit a similar endocytic response to the 191 extreme change in osmolarity. 192

194
Reviewer Figure 1: The cells on shifting to isotonic from hypotonic medium  Iso) shows formation of VLD 10 . Further, there is an increase in endocytosis following 196 which the membrane morphology is restored. We would like to clarify that the fluid-phase uptake response that we measure 205 occurs contemporaneously with strain relaxation. Our assay for dextran uptake is 206 for 90 seconds to integrate the endocytic uptake over this time-period, even in the 207 control cells at steady state. Importantly, we find that the increase in endocytosis 208 is a fast transient event that is lost in 90 seconds (Fig 1c; Stretch-Relax-Wait).

209
Further, this also coincides with the time it takes to remove the membrane 210 reservoirs (Compare Fig 1c with Fig 4b). Attempts at imaging these two fast 211 processes have been difficult at present, and a microfluidic chamber to carry out 212 rapid pH buffer change using SEC-GFP-GPI 7,12 while changing tension would be 213 necessary for this purpose. While these experiments are currently planned, it 214 requires the development of new instrumentation in our laboratories and is 215 currently beyond the scope of this manuscript. 216 217 -Membrane tension measurements based on membrane tether pulling is an 218 important piece of information in this study. The measurements however 219 are performed at steady state after up or down regulation of the CG pathway. 220 To better link these measurements to the process studied here, tether forces 221 should be measured during the stretch-relax procedure.

223
We measure steady state tension of a cell to understand the importance of 224 endocytic process since we see that CG pathway can respond to changes in tension.

225
Cells at steady state would be maintained at a specific membrane tension due to 226 the simultaneous functioning of a number of endo-exocytic processes, adhesion. 227 Thus it is necessary for us to look at the importance of modulating the endocytic 228 process on the steady state tension, as measured by monitoring tether pulling 229 force. 230 However, we agree with the reviewer's suggestion of monitoring tension of a cell 231 during the stretch-relax and other procedures, and we are developing the 232 instrumentation to do the same. Currently, the PDMS-based stretch-relax device 233 is incompatible with the optical tweezer set up, due to the inherently scattering 234 nature of the PDMS on which the cells have to be grown.

236
The case of osmotic shock is simpler but still poses problems as a sudden change 237 of medium causes large enough flow disturbance to detach the bead from the 238 tether. We need to improve the techniques to enable the measurements the 239 referee has suggested and feel that these experiments are also beyond the scope 240 of this current manuscript. 241 242 -Although the data presented here seem to exclude caveolae, it is still 243 intriguing that they are not involved in this regulation. the lack of any role for caveolae this fast active retrieval of plasma membrane (Fig  265  3a), since endocytic uptake via caveolae is dynamin-dependent 13 .

267
We have tried to discuss the roles of different endocytic pathways in the 268 discussion to highlight this (see page 11, line 460). By forming membrane 269 invaginations caveolae help counter increase in tension by passively releasing 270 extra membrane and preventing rupture in a passive manner 14 . Regarding the 271 increase in caveolar endocytosis on deadhering 15 , it is to be noted that the 272 caveolar uptake appears to be triggered once the cells are placed in suspension 273 and this continues for several minutes and up to couple of hours, taking in 274 membrane of specific composition, containing GM1 and cholesterol. The kinetics 275 are therefore very different to the CG pathway response shown here, as discussed 276 in the revised manuscript. The caveolar pathway may act as a timer for the time 277 the cell is in suspension helping to control anoikis 15 , and may function in parallel 278 to this fast transient endocytic response. On the contrary, the CG pathway is 279 acutely upregulated only on reducing tension by stretch-relaxation process or 280 during deadhering but swiftly returns to steady state once the excess membrane 281 is removed ( Fig. 1C and 1D). However, we find that if cells are left in suspension 282 for 10 minutes, the fluid uptake reduces in contrast to the increase in endocytosis 283 expected from cargo that is internalized via the caveolar pathway (Supplementary 284 Fig. 10d). Thus, various endocytic pathways with different time scales appear to 285 operate, potentially for a variety of reasons. Communications, providing that some points are clarified or complemented.

313
We thank the reviewer for her/his support of our work. We are pleased that our 314 data was able to convince this reviewer of the importance of the CG pathway and 315 its relationship to a mechanosensing machinery. We have done further 316 experiments to complement the present study and clarified the points below. 317 318 1-Vinculin is normally considered as part of the focal adhesion (FA) 319 complex. But, talin is the mechanosensitive element : when a force opens it, 320 its activation leads to vinculin binding, thus vinculin acts downstream of 321 talin. Would it be possible to test whether talin is also involved in tension 322 sensing? At this stage, the connection between vinculin, the CG pathway and 323 the GEF GBF1 appears quite phenomenological. Is there a direct coupling 324 between adhesion and tension regulation? What would be the molecular 325 mechanism? I guess it will require much more work to establish this, but it 326 could be at least discussed. Moreover, in the experiments, the methods to 327 change membrane tension have at the same time a mechanical action on FA 328 and eventually on vinculin, even when cells are detached since both 329 adhesion and tension are released. Is there a way to make a hypertonic 330 shock or aspirate a cell in suspension to check if in non-adhering conditions, 331 vinculin is still involved in the mechanosensing machinery? 332 333 We thank the reviewer for the set of constructive comments. We have now tested 334 the role of talin, and focal adhesions in general, for regulating this process.

336
Vinculin along with Talin is part of the mechanosensitive module of the focal 337 adhesion 16 and is known to be important for mechanotransduction 17 . An obvious 338 experiment to test a role for Talin requires depleting both talin 1 and 2. 339 Unfortunately, this causes cells to round up in culture 18 making it difficult to do 340 any endocytic experiments that involves multiple washes. Therefore, we took an 341 alternative approach and have tested different vinculin mutants (Vin-CA: 342 Constitutively active and does not require Talin to activate it, VinA50I: Talin  343 binding mutant, Vin-A50I-CA: Constitutively active and cannot bind talin) in the 344 Vinculin -/-back ground 19 . We find that Vin-CA mutant inhibit CG pathway and 345 does not show a response to decrease in tension unlike WT ( Fig. 7c/7e). Vin-CA 346 that cannot bind to Talin (Vin-A50I-CA) shows a similar phenotype (Fig. 7e), 347 however a wild type Vin which is unable to bind Talin and therefore be activated 348 by Talin is not able to restore the sensitivity to the change in membrane tension 349 (Fig. 7d). This indicates that the talin is required to activate vinculin in response 350 to changes in tension while active vinculin is required to negatively regulate the 351 CG endocytic pathway (Also see Page 9, Line 358).

353
We also looked at cells in suspension to further dissect the endocytic response to 354 changes in tension. Cells in suspension lack any focal adhesions and interestingly, 355 showed no increase in endocytic uptake on hypo-iso shift unlike the attached cells 356 ( Supplementary Fig. 10d). These experiments indicate the focal adhesion 357 signaling in general and talin in particular are important for activating vinculin to 358 show this transient endocytic response. We have now expanded the discussion to 359 reflect the possible ways vinculin could regulate the CG pathway through GBF1. This reviewer has raised an interesting point. The early endosomes in the CG 371 pathway seem to be pleomorphic 20 and makes vesicles of different sizes. This is 372 likely to internalize differential amount of material depending on the conditions. 373 We have now done high resolution imaging to understand the size and number of 374 endosomes during an osmotic release. We find that the number of endosomes as 375 well as the size increases ( Supplementary Fig. 3b/3c/3d, Movie 1 and 2).

377
Since the CG pathway has machinery to substantially upregulate its endocytic 378 capacity, it is consistent with the increase in uptake on changing tension. CG 379 pathway is only transiently upregulated when the tension is reduced and comes 380 back to the steady state (Fig. 1C, 1D). The increase in size of endosome points to a 381 mechanism that could regulate the timing of scission of endosome to control size 382 of an endosome in relation to the tension. However, further study is required to 383 understand the mechanism. There is a substantial amount of recycling from the 384 CG pathway and it could help in sorting the contents of the endosome 20 . It is likely 385 that this sorting event would help restore the change in membrane composition if 386 any. However, the lipid composition regulation through this high capacity 387 pathway is an open question and we are pursuing this. 388 389 390 3. Figure 5d ( show that if the CG pathway is absent, such effects on tether forces are absent as 448 well (Fig. 5f). We thank this reviewer for her/his constructive comment. We have explained this 472 in reply to Reviewer number 2 in the first major comment. 473 474

5) Use of different cell types --Why did the authors use HeLa and CHO cells 475
in certain experiments ( Figure 5) We have proof read the text and have asked an outside lab colleague to help 500 improve the readability of the manuscript. 501 502 2. This is a matter of personal preference but the title has two ands and 503 doesn't necessarily reflect the particular endocytic pathway being studied. 504 Perhaps the authors can revise the title to be more specific? 505 506 The title is now changed to increase the specificity and remove repeating words 507 as follows "Mechanochemical feedback control of dynamin independent 508 endocytosis modulates membrane tension in adherent cells". 509 510

Abbreviations appear in a number of places without being fully expanded 511
upon.

513
We have now made sure that the abbreviations are listed and fully expanded on 514 being used at the first time. Hope this reviewer comment helps make it easier for 515 the readers. This paper reports evidence that a reduction of membrane tension triggers 545 a fast and transient increase of endocytosis and fluid uptake. Tension 546 variation is obtained by different means: strech-release cycles, de-547 adhesion, or osmotic shock. Evidence is shown that the increased uptake 548 upon membrane tension reduction is neither due to clathrin mediated 549 endocytosis, nor is it dynamin dependent. The authors link the increased 550 uptake to the so-called CLIC/GEEC endocytosis pathway, and identify a 551 molecular mechanism for the mechano-sensitivity of through the 552 involvement of vinculin. In general, I find the paper quite sensible and 553 interesting. The notion of a negative feedback between tension and 554 endocytosis that could help maintain an homeostatic tension has been 555 around for some time, but it is nice to see experimental evidence for it, and 556 to have this process linked with a particular pathway, with hints of a 557 molecular mechanism. I think this is a very useful paper.

559
We thank the reviewer for his/her encouraging remarks on the manuscript. 560 561 My comments are two-fold. On the one hand, I think the authors are at places 562 over-interpreting their data, and would like to see this amended. I also find 563 that a more precise characterisation of the dynamical nature of the 564 mechanochemical feedback would be very valuable. These comments are 565 developed below. 566 567 1) L.177: "Together, these experiments indicated that the clathrin, dynamin 568 or caveolin dependent endocytic mechanisms do not exhibit a rapid respond 569 to a reduction in membrane tension."What is shown here is that there is a 570 rapid response to membrane tension reduction that is clathrin, dynamin, 571 and caveolin independent, but not that these pathways do not exhibit such 572 response. This sentence should be changed to reflect the actual findings.

574
We have now changed the wordings accordingly in the main text of the manuscript 575 (see page 5, line 171). 576 577 2) L.201 and below. The author claim that the tension-dependent 578 endocytosis mechanism creates a feedback loop hat et membrane tension to 579 an homeostatic value. Is tension regulation also possible for HeLa cells, 580 which lack CG endocytosis? Could the author show the evolution of tether 581 force on HeLa cell upon stretch-release, or hypo-osmotic shock. It is likely 582 that these cells also show a transient variation of tension, which would 583 suggest, at best, the existence of redundant tension regulation mechanisms. 584 However, the dynamics of tension relaxation could be very different, 585 perhaps not as fast as the one shown here to depend upon CG endocytosis. 586 587 We agree with the reviewer on this. There could be other mechanisms at work 588 here using cortical cytoskeleton etc. It is also possible that the exocytic 589 mechanisms that we have not probed here could have a role in maintaining 590 homeostasis. We find that HeLa cells do respond to extreme perturbations that is 591 independent of CG pathway ( Supplementary Fig. 7a, 7b). Even vinculin null cells 592 respond to higher changes in tension (Fig. 7b). These observations indicate that 593 there could be multiple mechanisms responding to changes in tension. We are 594 currently trying to create a set up to measure tension live while being able to 595 modulate the strain but these experiments are beyond the scope of the current 596 manuscript. In magnetic tweezers experiments, concanavalin A-coated beads are attached to 616 the cell membrane via glycoproteins, and were then pulled. Since the membrane 617 is linked to the cytoskeleton, we fully agree with the reviewer that the 618 measurement reflects both membrane and cytoskeletal properties. We note that 619 the experiments are not directly comparable to optical tweezers measurements 620 for two reasons. First, the beads are bigger (4 um) and attached to cells via a large 621 part of their surface, increasing mechanical resistance. Second and most 622 important, magnetic tweezers exert force largely in the plane of the membrane, 623 whereas optical tweezers experiment pulls membrane tethers perpendicular to 624 the membrane plane. Cell membrane bilayers may be extended in the 625 perpendicular direction, they are likely to be extremely stiff in-plane, due to the 626 effects of the membrane and the cytoskeleton interaction. As the reviewer notes, 627 we performed experiments by directly linking beads to the cytoskeleton through 628 fibronectin and integrins, which abolishes the measured differences. In our view, 629 the difference we observe are related to alterations in membrane tension, and the 630 discrepancy with tether values is due to the very different experimental setups. 631 However, we agree with the reviewer that this is not definitive proof, and 632 therefore we conducted optical tweezers experiments. 633 In the interest of preventing any confusion, we have taken the advice of the 634 reviewer and have removed the magnetic tweezer data. We also provide a better direct correlation between the evolution of the tether force and reservoir 649 resorption (as in Fig.4). The dynamics of recruitment of the CG machinery 650 (Fig.6) could also be monitored dynamically before, during, and after the 651 osmotic shock. This could also be combined with optical tweezers 652 measurement of tether force to obtain a direct dynamical correlation 653 between the two processes. Such measurement would greatly support the 654 claim of mechanical feedback and its involvement in setting the homeostatic 655 tension.

657
Yes, the tether forces are measured on adherent cells to understand the effect of 658 modulating endocytic pathways on steady state tension. We have now modified 659 the text to convey this message more clearly (Page 7, Line 272). The cells are 660 grown on a PDMS substrate for the stretch-relax experiments. Unfortunately, 661 PDMS scatters the laser light preventing accurate measurement of tension, in the 662 configuration that we make the measurements. Therefore, it is currently not 663 possible to measure tension while cells are grown on PDMS, and we work in an 664 inverted microscope configuration. We understand that measuring tether forces 665 while making changes to tension in a live cell is important and we are developing 666 the right tools to measure tension changes while making other perturbations. We 667 hope to discuss these results and more in future manuscripts. 668 669

5) Fig.4. Why is there no quantification for T=37 deg.? 670 671
The values were normalized to the T= 37C. To avoid confusion, we have now 672 replotted the figure to show the quantification and normalization clearly. We have now added a more detailed explanation of the experiment (see page 7, 694 line 272 and Methods -Page 23, Line 948).

696
We have used ConA coated beads in magnetic tweezer and also uncoated 697 polystyrene beads that binds to the plasma membrane due to non-specific 698 interactions in optical tweezer. However, similar beads are used between control 699 and test conditions wherever it has been used and these have been mentioned in 700 the Methods.

702
In the optical tweezers experiments, the changes in the tether force were 703 used to establish a link between the CG pathway and the membrane tension. 704 However, the tether force depends on other conditions such as the length of 705 the tether, the locations on the cells where the measurement is done, and 706 membrane adhesion to cytoskeleton. The authors need to clarify how these 707 conditions are controlled such that the level of tether forces they measured 708 can be correlated with the endocytic process through the CG pathway.

710
In cells the steady-state tether force, as measured after the initial force during 711 formation has relaxed, is independent of the tether length. This is presumably 712 because of the tension regulation mechanisms operating in the cells. So one is not 713 working at constant lipid number. This has been studied previously 21,22 . We do not 714 know if the tension is going to be different at different locations. However, we do 715 not use polarized cells in our study that could have different tension at different 716 location (apical vs basal, leading edge vs trailing edge etc.). We consistently pull 717 tethers from lamellipodia and average our data over several tethers. This should 718 average out such dependencies. Spatial dependency is of interest in its own right. 719 It has been shown previously that the steady state tether force depends on the 720 adhesion to the actin cortex. The force is slightly reduced when the cortex is 721 removed. This has been attributed to membrane-cortex adhesion. In our 722 experiments we are only making relative comparisons of tether force between 723 cells with active CG pathway and those with this pathway inhibited. As opposed to 724 absolute measurement of membrane in-plane tension, the cortex effects are 725 expected to be negligible in such membrane tether measurements. We have now 726 clarified all three points in the main text.

728
The authors also investigated the correlation between the local stiffness of 729 the cell and the endocytic process through the CG pathway. The local 730 stiffness was measured by applying 0.5 nN force pulses using a magnetic 731 tweezers device to ConA-coated magnetic beads that were attached to the 732 cell membrane and detecting the resulting bead displacement. This force is 733 one order of magnitude greater than the tether force measured using the 734 optical tweezers. I wonder why such a large force didn't produce a 735 protruding member tether. Does it imply a strong membrane attachment to 736 cytoskeleton that prevents formation of the membrane tether? If this is the 737 case, then this membrane stiffness measurement does not reflect membrane 738 tension, and cannot be compared with the tether force measured using 739 optical tweezers. The implications of the results from the magnetic tweezers 740 measurement on the endocytic process through the CG pathway need to be 741 re-discussed 742 The authors stated "That this effect was due to a reduction in membrane 743 tension and not any effects on the cytoskeleton, was corroborated by the 744 lack of a change in the measured stiffness of fibronectin-coated beads 745 attached to cells via integrin-fibronectin…" Based on this sentence, it seems 746 that the authors suggest that the local stiffness measured by the magnetic 747 tweezers device is not due to membrane adhesion to the cytoskeleton. If  748 this is what the authors implied, then they need to quantitatively establish 749 a link between the local membrane stiffness they measured and the 750 membrane tension, and need to explain why the protruding membrane 751 tethers did not form in the magnetic tweezers experiments.

753
In the interest of preventing any confusion, we have taken the advice of this and 754 reviewer 4 (response 3) and have removed the magnetic tweezer data.

756
The authors identified vinculin as a key player in the observed mutual 757 dependence between the CG pathway of endocytic process and membrane 758 tension, but they didn't provide a clue of how vinculin might be involved in 759 this regulation. Although it would be too much to ask the authors to 760 completely elucidate the mechanism of the involvement of vinculin in this 761 regulation, it will be very helpful if they provide information whether it 762 requires vinculin activation through binding to talin at focal adhesion. 763 Repeating the measurement in talin-deleted cells can provide this piece of 764 information.

766
This is an important point raised by this reviewer and reviewer 2, and has been 767 clarified above (see point 1; reviewer 2). With respect to the Holst paper, I partially disagree with the interpretation made by the authors. While this is correct that GRAF1 depletion increased cell blebbing, it was also shown that a decrease in surface tension was buffered by clathrin-independent endocytosis, a process regulated by GRAF1. Whether this is a peculiarity of HeLa cells is not so clear as the role of clathrinindependent endocytosis in buffering membrane tension decrease was also established in colon cancer cells in the same study. We obviously disagree on this point and this is why the simplest answer to this was to test GRAF1.
While it may be indeed difficult to adapt the PDMS-based stretch-relax device to nanotube pulling, several groups have used hypo-iso osmotic cycles to measure the dynamics of membrane tension variations. I disagree therefore that these experiments are beyond the scope of this study as the coupling between endocytosis and membrane tension variations is intrinsically dynamic and is key to the processes investigated here.
Reviewer #2 (Remarks to the Author): I think that the authors did a great job answering my comments as well as those of my colleagues.
The new data on the role of Talin and of the activation of Vinculin are convincing. I also like very much the mechanical/chemical feedback model. It helps understanding how the tension set point could be regulated. There is still work to do to refine the model and understand how to relate time scales to the actual experiments, but I think that at this stage the paper can be published Reviewer #3 (Remarks to the Author): In the revised version of this manuscript, the authors have addressed all of my previous comments satisfactorily. It would have been easier to re-review if the authors would have marked up the changes to the text in a different color.
Reviewer #4 (Remarks to the Author): The authors have taken into account some of my comments regarding over-interpreting data, and have argued that the dynamical assessment of tether force that I proposed is not feasible at the moment. I am willing to accept this response. However, upon re-reading the manuscript and the response to reviewer 2 (question 1), it appears to me that there could be a major problem with the interpretation of the data, which I unfortunately overlooked upon my first reading, and which casts serious doubts on the design of the feedback loop shown in Fig.8.
The new results linking vinculin activation to Talin, and showing that cells in suspension do not show an increase in endocytosis upon hypo-iso shift are to my mind quite significant. So far, the discussion of this paper relies on a mechano-chemical feedback loop between membrane tension and endocytosis rate. I tend to think that the absence of such feedback in suspended cell suggest that the feedback does directly involve membrane tension, but rather the tension on Focal Adhesion (FA) elements in adherent cells. While the two types of tension could be related, and can both be decreased during stretch-release, de-adhesion, or hypo-iso shift experiments, it is not clear at all that there is a direct correlation between FA tension and membrane tension. This suggests that the feedback model (Fig.8b) is missing a key element, which is the tension on the adhesion sites, and which is related in an unclear and potentially complex way to membrane tension in adherent cells. As far as I am concerned, this represents a major difference from the direct feedback loop proposed in this paper. Indeed, one could imagine treatment that would affect the tension on adhesion sites (such as inhibition of myosin contractility) could also lead to upregulation of the CG endocytotic pathway without relying on obvious variation of the membrane tension itself. A crucial control in that regard, would be to measure the changes in membrane tension of cells in suspension during a hypo-iso shift. I would expect that membrane tension does indeed decrease in such condition, without triggering the feedback on endocytosis described in this paper, since we already know that such cells do not show the modulation of GC endocytosis. That would show that variation of membrane tension may not be what is measured in the feedback loop, thus invalidating the direct robust feedback proposed in Fig.8. In fact, as far as I know, it is rather unclear to see at the molecular level how membrane tension, which is a force (per unit length) actin in the plane of the membrane, could modulate the force felt by Talin/Vinculin complexes, which presumably has a sizeable component perpendicular to the membrane.
These considerations makes me hesitant to recommend publication without discussing the effect of cell adhesion and of FA tension. The fact that the title was amended to specifically restrict the study to adherent cells is good. The paper surely shows a correlation between reduction of membrane tension and increase of CG endocysis, and this finding is worth publication. However, the causality aspect which is essential to talk about feedback loop has, in my opinion, not been clearly demonstrated. I am conscient of the fact that a discussion has been added on the "real" meaning of tension, which is now called an "apparent tension" that also involves membranecytoskeleton adhesion (l.277). I would like to stress that my reservations are of a different nature, and cannot be satisfied by simply stating that Talin/Vinculin tension also contributes to the apparent, or effective tension discussed in this article.
As I realise that changing the focus of the paper to FA tension might be too much to ask, I thus propose that the paper be amended so that all implication of causality between a decrease of membrane tension and an increase of endocytosis are removed, and replaced by a discussion of the existence of a clear correlation between the two quantities. For instance (the list is not exhaustive) l.74: "we find that a clathrin, caveolin and dynamin-independent endocytic mechanism, the CLIC/GEEC (CG) pathway, rapidly responds to changes in membrane tension, acting to restore it to a specific set point." should thus be modified to reflect the actual finding, for instance: we find that a clathrin, caveolin and dynamin-independent endocytic mechanism, the CLIC/GEEC (CG) pathway, correlates with changes in membrane tension, and might be involved in setting a specific membrane tension set point." or l. 171: "171 Together, these experiments indicated that there is a rapid endocytic response to reduction in membrane tension that is clathrin-, dynamin-, and caveolin-independent." should thus be modified, for instance: there is a rapid endocytic response that correlates with a reduction in membrane tension, but relies on the activation of proteins involved in cell adhesion.

Point by point response to reviewer's comments:
(Note that all the revisions and text mentioned here have also been highlighted in the main text)

1) Reviewer #1 (Remarks to the Author):
With respect to the Holst paper, I partially disagree with the interpretation made by the authors. While this is correct that GRAF1 depletion increased cell blebbing, it was also shown that a decrease in surface tension was buffered by clathrin-independent endocytosis, a process regulated by GRAF1. Whether this is a peculiarity of HeLa cells is not so clear as the role of clathrin-independent endocytosis in buffering membrane tension decrease was also established in colon cancer cells in the same study. We obviously disagree on this point and this is why the simplest answer to this was to test GRAF1.
We thank the reviewer for her/his comments on our revised manuscript.
However, we wish to indicate (and reiterate) important points that this reviewer has failed to appreciate.
1) The protocols used for the change in osmolarity in Holst et al., and our manuscript are drastically different. While the protocol utilized by Holst creates a drastic change in osmolarity (either just water addition or >6x dilution of isotonic medium) for 10 min prior to reverting to isotonicity (Hypo-6x-Iso), the protocol adopted for the bulk of our work, is a 1 minute hypotonic (~ 50 % Isotonic) treatment followed by isotonicity (Hypo-iso). Upregulation of the CG endocytic activity under Hypo-Iso conditions is similar to what we see in cell detachment, and cell stretch-relax conditions ( Supplementary Fig. 1c, Supplementary Fig. 6a, Fig. 1b, and Fig. 1c). On the other hand, cell lines that do not exhibit the CG pathway (HeLa Cells or the IRSp53 KOs), remain responsive only to the Hypo-6x-iso condition (Supplementary Fig. 7a and 7b). This indicates that the response being monitored in Holst et al, is functioning via a different mechanism, which may be GRAF1 and potentially AP2 dependent.
Osmotic changes could have multiple effects 1 , which is why we have taken pains to establish three separate methods to test the role of tension on endocytic processes. Holst et al either uses just water or 6x-Hypo medium for 10 minutes to alter membrane properties. This causes dramatic morphological changes which persist even after a recovery to isotonic conditions (Supplementary Fig. 7a and 7b) and is clearly not the same process that we are studying at more moderate changes in osmolarity (Supplementary Fig. 1c).
2) This reviewer also fails to distinguish between clathrin-independent endocytosis as a generic catch all phrase for all endocytic pathways that function in the absence of clathrin, and a specific clathrin-independent endocytic pathway, namely the CLIC/GEEC (CG) pathway that functions in the absence of both clathrin and dynamin.
A confusion that exists in categorizing various endocytic pathways is also reflected by this reviewer's response to our statements about the role of GRAF1. From available literature, GRAF1 has been implicated in endocytic pathways that function in Hela cells [2][3][4][5] . These cells lack a ARF1/CDC42 regulated CG pathway as shown by us previously 6 , (and this study) and also more extensively in Bitsikas et al 7 . Thus, perturbing GRAF1 is likely to have more widespread effects, and could affect more than one mechanism for endocytosis. The confusion arises when this reviewer equates the CG pathway with any GRAF1-sensitive clathrin-independent (or dependent) pathway.
The CG pathway whose role we are addressing in the control of membrane tension, requires the function of GBF1, ARF1, CDC42 and now the recently shown, IRSp53, but importantly not dynamin [8][9][10][11][12] . While it may also utilize GRAF1, for the reasons cited above, the perturbation of GRAF1 and monitoring its effect on fluid phase endocytosis will not provide any further clarification of the endocytic mechanism influencing the Hypo-Iso treatment condition utilized by us.
3) Thus, we would like to reiterate (as we have in our previous response to this reviewer), since GRAF1 participates in both clathrin-dependent 7 and clathrin-independent pathways 2 , ascertaining whether GRAF1 plays a role in the CG pathway that responds to specific alterations in membrane tension, is not the goal of this study. We have other much clearer means to identify the endocytic mechanism that shows us that the CG pathway is a central player in this response. a) we have tested the roles of 3 key players (GBF1, CDC42 and IRSP53) to be necessary for the membrane tension modulated operation of the CG pathway in this manuscript. b) by using a combination of specific null mutants of dynamin and caveolin, and experiments addressing the endocytic uptake of ligands trafficking through CME and CG pathway, we conclude that the CG pathway specifically responds to tension changes we have utilized. c) We also uncover the mechano-chemical transducer of this response, providing evidence that Vinculin, a mechanotransducer, mediates this strain-sensitive response.
Therefore, we believe testing yet another molecule such as GRAF1 will provide no additional insight into the understanding of the mechanism. More importantly, this is not the focus of the paper or part of the concerns raised by any of the other 4 reviewers. Further, reviewer 1 expressed agreement with our conclusion that CG pathway is important for this response in the previous round. Thus, we are quite perplexed about this repeated demand made on us to prove the role of GRAF1 in our pathway.

4)
We are frankly unsure of GRAF1's role in the CG pathway and testing its role in is a whole new project by itself and currently beyond the scope of the current manuscript. Our reasoning behind our lack of clarity about the role of GRAF1 is the following: a) GRAF1 has been shown to act in clathrin mediated endocytosis (CME) in addition to the papers which implicate it in clathrin independent endocytosis (CIE) 7 . b) A major and serious concern about all GRAF1 published papers 2-5 that suggest its role in CIE is that key experiments have been done in cells that lack bonafide CG endocytosis (eg. HeLa cells as extensively detailed in Bitsikas et al 7 , previously observed by us 6 , and further tested in the current manuscript). Importantly, Holst et al 5 also shows that the basal fluid uptake is almost completely dependent on AP2, a CME regulator, raising questions about the existence of a CIE mechanism in these cells.

5)
Equally important is the technical feasibility of the GRAF1 depletion studies. Unless we develop new reagents, this is somewhat beyond our immediate scope for the following reasons: Most GRAF1 studies have appeared from Richard Lundmark group 2-5 and there are no commercial antibodies that are known to work to test its depletion. However, we also note that most of the experiments in the Holst et al paper from Richard's laboratory are conducted in HeLa cells stably expressing GRAF1-GFP and its depletion is proxy for endogenous GRAF1 depletion. siRNA against GRAF1 is made-to-order and is not readily available. We have ordered a set a few months ago from the same source and are yet to receive the same. Finally, these experiments by Lundmark group have all been conducted in HeLa cells where there is no evidence for the functioning of the CG pathway (our manuscript and previous studies 6,7 ).
Due to the abovementioned concerns, we believe role of GRAF1 in different endocytic pathway(s) requires extensive and careful characterization in multiple cell lines and context. We plan to undertake this as a separate study to understand the role of GRAF1 in CME and CIE endocytosis in multiple cell lines. Our current MS looks at the response of a specific endocytic pathways to changes in membrane tension and vice versa, and in this context we think exploring the role of GRAF1 will not further the scientific understanding in any meaningful way. 6) We believe that we addressed Reviewer 1's concerns with battery of experiments detailed in our revised manuscript, which were perhaps overlooked (by this reviewer). We have successfully replicated some of the experiments that were performed in Holst et al paper, for instance experiments using AP2 shRNA, and osmotic shock. We then go on to then show that the conditions that we employ in our MS are quite different from what was used in the Holst et al protocols (Supplementary 6a, 7a and 7b). We also show that AP2 depletion that inhibits CME pathway inhibits fluid uptake in HeLa cells similar to Holst et al. in direct contrast to the increase in uptake observed in cell line with a robust CG pathway (see Supplementary Fig. 5a and 5b). This underlies the fact that we are talking about different physiological context in our MS and again are quite perplexed by the attempts of Reviewer 1 to club our work with Holst et al.

7)
Minor comment about the reviewer's statement -"Whether this is a peculiarity of HeLa cells is not so clear as the role of clathrin-independent endocytosis in buffering membrane tension decrease was also established in colon cancer cells in the same study". We urge reviewer 1 to show us where this data is documented in Holst et al., since we were unable to find this reference. The only reference to colon cancer cells exists w.r.t 3D invasion assay (there are no experiments in colon cancer cells testing endocytic response to changes in membrane tension) which again is not in the scope of our manuscript. Furthermore, Holst et al., fail to make any attempt at measuring membrane tension (or surface tension as this reviewer puts it).
We hope that with the above-mentioned arguments it is evident that understanding the role of GRAF1 firstly in CG and CME endocytosis, then going on to show its role in membrane tension regulation is quite a major undertaking by itself. This would require considerable time and more importantly, deflect the main focus of this manuscript.
While it may be indeed difficult to adapt the PDMS-based stretch-relax device to nanotube pulling, several groups have used hypo-iso osmotic cycles to measure the dynamics of membrane tension variations. I disagree therefore that these experiments are beyond the scope of this study as the coupling between endocytosis and membrane tension variations is intrinsically dynamic and is key to the processes investigated here.
Dynamic tether measurement along with modulating tension is an interesting and exciting experiment. This very challenging experiment requires an ability to dynamically modulate tension (either by changing osmolarity or stretching) while being able to maintain tether and measure tether forces. Although this has been shown before 13 , it unfortunately requires extensive modification from a standard optical tweezer set up that we have developed for our purposes.
Tether force measurement is extremely sensitive to various factors such as air draft (air conditioning/person entering the room) and even the fan running in the EM-CCD camera in our hands. Thus holding a thin membrane tether from a cell and measuring forces is an extremely sensitive and difficult process. On top of it, if we need to flow in any medium while holding a membrane tether, we would need to extensively modify our current microscope stage and construct a very precise microfluidic chamber. This is quite a major and extensive technical hurdle that needs to be overcome but at present is beyond the scope of this study. Nevertheless, we are actively trying to build the setup and hopefully the results from it would be part of a future manuscript.
We would like to point out that Reviewer 4 has agreed that dynamical tether force measurement is not feasible at this stage while rest of the three reviewers have not considered dynamic measurement of tether forces necessary for the conclusions made in this paper.

2) Reviewer #2 (Remarks to the Author):
I think that the authors did a great job answering my comments as well as those of my colleagues. The new data on the role of Talin and of the activation of Vinculin are convincing. I also like very much the mechanical/chemical feedback model. It helps understanding how the tension set point could be regulated. There is still work to do to refine the model and understand how to relate time scales to the actual experiments, but I think that at this stage the paper can be published We thank the reviewer for the support and for recommending the manuscript for publication.

3) Reviewer #3 (Remarks to the Author):
In the revised version of this manuscript, the authors have addressed all of my previous comments satisfactorily. It would have been easier to re-review if the authors would have marked up the changes to the text in a different color.
We apologize for the inconvenience and we thank the reviewer for the support and for recommending the manuscript for publication.

4) Reviewer #4 (Remarks to the Author):
The authors have taken into account some of my comments regarding over-interpreting data, and have argued that the dynamical assessment of tether force that I proposed is not feasible at the moment. I am willing to accept this response.
We thank the reviewer for the critical reading of the rebuttal and agreeing that this finding is worth publication and for concurring that dynamical tether force measurement is not feasible at the moment.
However, upon re-reading the manuscript and the response to reviewer 2 (question 1), it appears to me that there could be a major problem with the interpretation of the data, which I unfortunately overlooked upon my first reading, and which casts serious doubts on the design of the feedback loop shown in Fig.8. The new results linking vinculin activation to Talin, and showing that cells in suspension do not show an increase in endocytosis upon hypo-iso shift are to my mind quite significant. So far, the discussion of this paper relies on a mechano-chemical feedback loop between membrane tension and endocytosis rate. I tend to think that the absence of such feedback in suspended cell suggest that the feedback does directly involve membrane tension, but rather the tension on Focal Adhesion (FA) elements in adherent cells.
We are glad that the reviewer has brought up this issue, since this is a key finding from our work. Since the reviewer was confused in her/his first reading of this manuscript, we have now taken pains to emphasize what we are attempting to say. And have modified the text according to the suggestions made by this reviewer. Please see regions in the text that have been highlighted by yellow highlights.
We completely agree with the interpretation of this reviewer and have incorporated this idea in the discussion (see line 447, page 11). The discovery of the molecular mechanism behind the change in tension triggered endocytic response in attached cells, is indeed a very interesting result. It speaks to the mechano-chemical nature of this interplay. Precisely the point we have been trying to convey in our manuscript.
While the two types of tension could be related, and can both be decreased during stretchrelease, de-adhesion, or hypo-iso shift experiments, it is not clear at all that there is a direct correlation between FA tension and membrane tension. This suggests that the feedback model (Fig.8b) is missing a key element, which is the tension on the adhesion sites, and which is related in an unclear and potentially complex way to membrane tension in adherent cells. As far as I am concerned, this represents a major difference from the direct feedback loop proposed in this paper. Indeed, one could imagine treatment that would affect the tension on adhesion sites (such as inhibition of myosin contractility) could also lead to up-regulation of the CG endocytotic pathway without relying on obvious variation of the membrane tension itself.
This reviewer is correct his her/his interpretations : please also see recent manuscripts where researchers have shown that membrane tension could regulate the molecular tension across integrin 14 and regulate adhesion positioning through vinculin 13 and we have added this in the discussion now (Line 447).
A crucial control in that regard, would be to measure the changes in membrane tension of cells in suspension during a hypo-iso shift. I would expect that membrane tension does indeed decrease in such condition, without triggering the feedback on endocytosis described in this paper, since we already know that such cells do not show the modulation of GC endocytosis. That would show that variation of membrane tension may not be what is measured in the feedback loop, thus invalidating the direct robust feedback proposed in Fig.8.
We agree with the reviewer here. Indeed, when we stretch-relax vinculin null cells, we continue to visualize the passive changes in the membrane morphology as protrusions (reservoirs or VLDs 15 ) that serve as a proxy for the lowering of tension (see Supplementary Fig. 10c). This result also indicates that change in membrane tension is not sufficient to generate the endocytic response. Further, the plasma membrane of cells in suspension also exhibit similar membrane projections in response to changes in osmotic shock and recovery indicating they do indeed exhibit a lowering of membrane tension in response to the conditions we have utilized (see Movie 3 and montage of images from the videos in Supplementary Fig. 10f). We indicate in our feedback loop that mechanotransduction through vinculin is necessary for the feedback control of CG pathway (See modified Figure 8b).
In fact, as far as I know, it is rather unclear to see at the molecular level how membrane tension, which is a force (per unit length) actin in the plane of the membrane, could modulate the force felt by Talin/Vinculin complexes, which presumably has a sizeable component perpendicular to the membrane.
In the Wang et al paper 14 , the authors do make some headway in addressing this vexing point. We now discuss this point in our discussion and refer the readers to this manuscript (line 447; page 11).
These considerations makes me hesitant to recommend publication without discussing the effect of cell adhesion and of FA tension. The fact that the title was amended to specifically restrict the study to adherent cells is good. The paper surely shows a correlation between reduction of membrane tension and increase of CG endocysis, and this finding is worth publication. However, the causality aspect which is essential to talk about feedback loop has, in my opinion, not been clearly demonstrated. I am conscient of the fact that a discussion has been added on the "real" meaning of tension, which is now called an "apparent tension" that also involves membrane-cytoskeleton adhesion (l.277). I would like to stress that my reservations are of a different nature, and cannot be satisfied by simply stating that Talin/Vinculin tension also contributes to the apparent, or effective tension discussed in this article.
As I realise that changing the focus of the paper to FA tension might be too much to ask, I thus propose that the paper be amended so that all implication of causality between a decrease of membrane tension and an increase of endocytosis are removed, and replaced by a discussion of the existence of a clear correlation between the two quantities. For instance (the list is not exhaustive) l.74: "we find that a clathrin, caveolin and dynamin-independent endocytic mechanism, the CLIC/GEEC (CG) pathway, rapidly responds to changes in membrane tension, acting to restore it to a specific set point." should thus be modified to reflect the actual finding, for instance: we find that a clathrin, caveolin and dynamin-independent endocytic mechanism, the CLIC/GEEC (CG) pathway, correlates with changes in membrane tension, and might be involved in setting a specific membrane tension set point." or l. 171: "171 Together, these experiments indicated that there is a rapid endocytic response to reduction in membrane tension that is clathrin-, dynamin-, and caveolin-independent." should thus be modified, for instance: there is a rapid endocytic response that correlates with a reduction in membrane tension, but relies on the activation of proteins involved in cell adhesion. I think this is a crucial point.
We have made the amendments in the paper to reflect the reviewers concerns and indicate correlation between tension modulation of endocytosis. We also include a discussion about membrane tension and tension across focal adhesion. These changes are made at the locations listed here and highlighted in the manuscript as well. We thank the reviewer for his or her support and for recommending the manuscript for publication.
I have no further comments about this manuscript.
Reviewer #4 (Remarks to the Author): In their response to my comments, the authors agreed wth my interpretation of their results, in which membrane tension is replaced by mechanical tension on adhesion sites. As I clearly explained in my report, this casts doubts on the existence of a direct mechano-chemical feedback loop between membrane tension and endocytosis, and in particular on the fact that this feedback to fix the homeostatic tension of cells. Indeed, the loop is not closed in their story, since adhesion molecules are also involved. Nevertheless, this idea is still written as such in the paper, and in particular in the abstract of the paper. This needs to be changed, so that the abstract reflects the actual findings and the interpretations of the results that we now agree upon.
To be specific, I have a problem with the sentences: "We find that vinculin, a well-known 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 mechanochemical feedback loop with membrane tension, potentially leading to homeostatic regulation of plasma membrane tension." This sentence is confusing, because "tension-dependent regulation of CG pathway" is used together with "homeostatic regulation of plasma membrane tension." while it is accepted by the author that whatever tension regulates the CG pathway IS NOT the membrane tension. The author must find a way to convey this message in the abstract. They should also remove "feedback loop" from the abstract since the existence of a loop has not been demonstrated. Finally, the abstract must also contain the sentence written by the authors in their response to my comments, but that I couldn't find in the text: "This result also indicates that change in membrane tension is not sufficient to generate the endocytic response" Without these changes, I find the abstract misleading.
Other changes that are need: line 76: change :"is altered upon changes in membrane tension" by "correlates with changes of membrane tension" line 308 It is not clear what the authors mean by this sentence: "These results show that, modulating the CG pathway by activating or inhibiting key regulators modifies the membrane tension. Membrane tension, on the other hand, negatively correlates with the operation of the CG endocytic pathway." Why "on the other hand", while the negative correlation IS A DIRECT CONSEQUENCE of the decrease of membrane tension by CG exocytosis. I would say: "These results show that, modulating the CG pathway by activating or inhibiting key regulators modifies the membrane tension. This leads to a negative correlation between membrane tension and the operation of the CG endocytic pathway."