P2X7 Receptors and TMEM16 Channels Form a Hub with Implications 1 for Macropore Formation and Current Facilitation 2

30 P2X7 receptors (P2X7) are cationic channels involved in many diseases. They exhibit 31 unique behaviors, such as “macropore” formation, which corresponds to enhanced large molecule cell membrane permeability, and current facilitation, which is caused by prolonged activation. These two phenomena have often been confounded, but thus far no clear mechanisms have been resolved. Here we provide evidence that current facilitation 35 and macropore formation involve functional complexes comprised of P2X7 and TMEM16, 36 a family of Ca 2+ -activated ion channel/scramblases. We found that current facilitation 37 results in an increase of complex-embedded P2X7 open channel probability, a result 38 mimicked by plasma membrane cholesterol depletion. We further show that macropore 39 formation entails two distinct large molecule permeation components, one of which 40 requires protein complexes featuring TMEM16F subtype, the other likely being direct 41 permeation through the P2X7 pore itself. Such protein complexes can be considered to 42 represent a regulatory hub intimately involved in distinct P2X7 functionalities. 43


Introduction
P2X7 26 . TMEM16F belongs to the TMEM16 family, a vast group of membrane proteins 109 characterized by a high degree of functional diversity, including roles such as Ca 2+ -110 activated Clchannels, Ca 2+ -dependent phospholipid scramblases, and dual function non-111 selective ion channel/phospholipid scramblases 29 . There are ten TMEM16 proteins 112 (TMEM16A-K, excluding I) found in humans, and recent evidence suggests that 113 TMEM16F functions as a dual Ca 2+ -activated non-selective (rather than purely Cl -) ion 114 channel and Ca 2+ -dependent phospholipid scramblase. This moonlight scramblase activity 115 led to the suggestion that TMEM16F activation by ATP-evoked P2X7 Ca 2+ flux mediates 116 downstream intracellular signaling processes, resulting in outcomes such as membrane 117 blebbing and apoptosis 26 . However, despite a possible direct interaction of P2X7 and 118 TMEM16F having been evoked 26 , no molecular complex has been identified thus far, and 119 any significance relating to macropore formation and current facilitation remains to be 120 elucidated.

122
In this work, we set out to test the hypothesis that P2X7 and TMEM16 form 123 functional complexes in HEK293T cells and Xenopus oocytes. We combined mutagenesis 124 techniques, genetic approaches, biochemical experiments, single-channel and whole-cell 125 recordings, confocal microscopy, live cell imaging and molecular modeling to show that 126 these complexes have implications in both agonist-evoked current facilitation and 127 macropore formation. Our data thus offer new insights into these enigmatic processes that 128 have proved difficult to explain for 40 years.  Next, we devised a protocol in which current facilitation was first induced by a 30 149 s perfusion of BzATP before establishing an excised membrane patch in the outside-out 150 configuration (Fig. 1B). We first verified that such long agonist perfusion induced a strong 151 whole-cell current facilitation that is similar to eight repeated short agonist applications 152 ( Fig. S1F and G), as previously described 6 . We observed some variability in the degree of  For control condition (no BzATP perfusion), two unitary current amplitudes were 157 observed (Fig. 1C,top), in agreement with previous data obtained with ATP from the 158 human (hP2X7) expressed in oocytes 33 . As we selected only those patches containing three 159 or fewer channels for analysis (see Materials and Methods), we confirm that these currents 160 did not originate from stackings of double openings, but rather reflect a dual gating 161 mechanism 33 . The simultaneous presence of these two discrete conductance levels (O1 =   Student's t-test, Fig. 1C and D, middle, n = 6 patches). However, a dramatic increase of the 171 6 mean open times of both conductance levels was recorded (~10-fold for O1, t O1 = 234 ms, 172 and ~5-fold for O2, t O2 = 211 ms), along with a modest increase of the mean closed time 173 (~1.5-fold, t C = 279 ms, Fig. 1E, middle and Fig. S1E, right, 754 events). As a result, the 174 open channel probability (NPo) for both conductance states combined increased by ~2.9-175 fold, from 0.17 to 0.47. Of note is that this increase is not only in the same range as BzATP-176 evoked current facilitation observed in the whole-cell configuration (between ~3-to ~4-177 fold), but also consistent with the increased deactivation time constant after facilitation.

178
This suggests that in HEK293T cells, agonist-evoked current facilitation predominantly 179 originates from a large increase of rP2X7 open channel probability, but not of single-180 channel conductance, thereby disfavoring the pore dilation theory.  open probability, we next asked whether this increase is related to macropore formation. Because CaCC inhibitors have been previously shown to inhibit P2X7-dependent dye 211 uptake 22,26 , we tested whether they are able to inhibit P2X7 currents which have been 212 facilitated by four repeated applications of BzATP. We focused on tannic acid (TA), 213 flufenamic acid (FFA) and 9-anthracene-carboxylic acid (9-AC), for which functional 214 modulation of TMEM16 activity, including TMEM16A 34,35 , TMEM16B 36 and 215 TMEM16F 26 , has been recently shown in recombinant systems. We found that application 216 of 20 µM TA alone for 60 s (referred to as "application alone") to facilitated receptors 217 abolished subsequent BzATP-evoked whole-cell currents in rP2X7-expressing HEK293T 218 cells, while 1 mM 9-AC reduced subsequent currents by ~35% ( Fig. 2A and B). For FFA 219 (100 µM), no inhibition was found in these conditions, however, a single co-application of 220 FFA and BzATP at the end of FFA perfusion reduced subsequent BzATP-evoked currents 221 to ~35% ( Fig. 2A and B). Contrary to TA and 9-AC, FFA inhibitions were fully reversible 222 upon washout. Importantly, no inward current was induced by application of TA, 9-AC, or 223 FFA alone (Fig. S2A). Thus, these data suggest that facilitated P2X7 currents are fully 224 inhibited by TA, and partially inhibited by 9-AC and FFA.

225
To exclude a possible direct modulation of P2X7 activity by selected CaCC 226 inhibitors, we investigated the effects of these inhibitors in Axolotl oocytes (Ambystoma 227 mexicanum), which are reported to be electrophysiologically void of CaCCs 37,38 . Using 228 two-electrode voltage-clamp (TEVC) and the "application alone" protocol, no P2X7 229 current inhibition was detected by 9-AC or FFA ( Fig. 3A and B), while 50% inhibition was 230 observed for both inhibitors in Xenopus oocytes ( Fig. 3C and D), which do express 231 endogenous CaCCs 37 . Note that for FFA, inhibition took place using the "application 232 alone" protocol directly, suggesting that washout dissociates bound FFA more slowly in 233 Xenopus oocytes than in HEK293T cells. Current facilitation was observed in Xenopus 234 oocytes, albeit not systematically, but not in Axolotl oocytes, suggesting that the absence 235 of current facilitation is not correlated to an absence of CaCC inhibition. These data, 236 therefore, strongly suggest that 9-AC and FFA do not act as inhibitors of rP2X7 itself, 237 suggesting that their partial inhibition of facilitated rP2X7 currents in HEK293T cells and 238 Xenopus oocytes occurs through a functional interplay between endogenous human (h) 239 hTMEM16 and rP2X7 that we named thereafter P2X7/TMEM16.  CaCC inhibitors. We found that TA and 9-AC not only strongly inhibited BzATP-evoked 251 8 currents (measured at the fourth application), but also prevented current facilitation (Fig. 252 2C and D). FFA was not different from the control, and accordingly no inhibition was 253 detected (Fig. 2D). Given the fact that 9-AC has been shown to not act on P2X7 itself, the 254 above data strongly suggest that the functional interplay between P2X7 and TMEM16 255 contributes to current facilitation.  Single-channel inhibitions were reversible upon washout, even for TA, and levels of 267 inhibition were similar to those obtained by co-applying the selective P2X7 inhibitor 268 AZ10606120 with BzATP. Overall, 9-AC inhibition data indicate that the interplay 269 between rP2X7 and hTMEM16 remains stable in excised patches, and therefore does not  demonstrate that TA is a powerful YO-PRO-1 quencher, and therefore should not be used 291 as a pharmacological inhibitor in dye uptake assays.

292
Next, we found that 9-AC significantly reduced BzATP-stimulated YO-PRO-1 293 uptake, by ~40% while no dye uptake inhibition was observed with FFA ( Fig. 4A-C). 294 These results suggest that the functional interaction between P2X7 and TMEM16 295 contributes partially to dye uptake. Because TMEM16F was previously shown to mediate addressing dye uptake, we first tested whether rP2X7 expression is influenced by 306 endogenous hTMEM16F, by comparing rP2X7 subunit expression in both HEK293T and 307 16F-null cell lines. We used a rP2X7 construct that was tagged at its C-terminus with a 308 myc sequence, and found a significant increase of both total and cell surface expression of 309 rP2X7 subunit in 16F-null cell lines compared to HEK293T cells ( Fig. 5B and C). These 310 data reveal that the presence of endogenous hTMEM16F channels reduces rP2X7 subunit 311 expression when this latter is ectopically expressed in HEK293T cells.

312
Next, we measured BzATP-induced dye uptake in the rP2X7-expressing 16F-null  Nevertheless, dye uptake was not abolished in 16F-null cells, raising the possibility 320 that the remaining dye uptake involves other TMEM16 members and/or intrinsic passage 321 through the P2X7 pore. To further probe these possibilities, we asked whether CaCC 322 inhibitors were able to inhibit this remaining dye uptake. Co-application of FFA did not 323 inhibit BzATP-induced YO-PRO-1 uptake in rP2X7R-expressing 16F-null cells (Fig. 5G), 324 as previously observed in HEK293T cells. Interestingly, 9-AC no longer inhibited dye 325 uptake either (Fig. 5G), suggesting that the inhibition of cell permeabilization observed 326 with 9-AC in HEK293T cells was related to hTMEM16F (Fig. 4C). These results suggest 327 that 9-AC is a selective inhibitor of hTMEM16F channels involved in dye uptake, and 328 support the hypothesis that remaining dye uptake is likely related to the direct passage 329 through the P2X7 pore.

330
To further validate TMEM16F contribution to dye uptake, we asked whether re-331 expression of TMEM16F along with P2X7 in 16F-null cell lines is able to rescue dye 332 uptake. By assaying varying cDNA ratios between rP2X7 and rat TMEM16F 333 (rTMEM16F), we found that a ratio of 1:0.05 rP2X7 to rTMEM16F (2 µg rP2X7 and 0.1 334 µg rTMEM16F) fully rescued dye uptake (Fig. 5F). These results therefore not only 335 confirm the contribution of TMEM16F to dye uptake, regardless of the species (human or 336 rat), but also reveal an optimal non-stoichiometric cDNA ratio for restoring dye uptake.

337
Although the underlying mechanism is elusive, these data are coherent with our findings 338 showing reduced rP2X7 expression in the presence of TMEM16F, which may explain why 339 a cDNA ratio largely skewed in favor of P2X7 is needed to rescue dye uptake. To confirm a close physical proximity of P2X7 and TMEM16F, we carried out co-343 immunoprecipitation in 16F-null cells overexpressing myc tagged rP2X7 and rTMEM16F 344 that was also tagged at its C-terminus with a myc-DDK sequence. Because we found that 345 dye uptake was fully rescued at a cDNA ratio of 1:0.05 in 16F-null cells, and given the fact 346 that P2X7 expression is modulated by the presence of TMEM16F, we asked whether 347 varying this ratio from 1:0.05 to 1:1 may also influence protein expression. We found that 348 co-expressing rP2X7 and rTMEM16F at a 1:1 ratio dramatically reduced total rP2X7 349 expression in 16F-null cells, but not at a 1:0.05 ratio compared to control (Fig. 6A). Thus, 350 in agreement with our previous dye uptake data, these results further confirm the negative 351 effect of TMEM16F on rP2X7 expression, and suggest that the ratio 1:0.05 is more suitable 352 for co-immunoprecipitation. In these conditions, we found that anti-FLAG antibody, which 353 recognizes DDK tag of myc-DDK tagged rTMEM16F, specifically and efficiently co-354 immunoprecipitated rP2X7 subunits, despite rTMEM16F expression being 20 times less 355 than that of rP2X7 (Fig. 6B). These data therefore indicate a close physical proximity 356 between rTMEM16F and rP2X7.

357
Next, we asked whether the two proteins co-localize in 16F-null cells by confocal 358 microscopy. We designed a fluorescent rP2X7 that was tagged at its C-terminus with 359 mScarlet, a bright monomeric red fluorescent protein, and transiently co-expressed rP2X7-360 mScarlet with rTMEM16F at a ratio of 1:0.05 in 16F-nulls cells. By subsequently staining 361 rTMEM16F with a primary anti-h16F antibody coupled to a secondary Alexa 488 362 conjugated antibody, we showed that both proteins fully co-localized in 16F-null cells 363 (Pearson's coefficient = 0.897 ± 0.014; n = 8 cells from 5 transfections; Fig. 6C). Although 364 most of the co-localization appears to be intracellular, putative plasma membrane co-365 localization was also detected (Fig. 6C).

366
Taken together, these data are consistent with the idea that P2X7 and TMEM16F 367 form protein complexes. Having revealed the existence of P2X7/TMEM16F protein complexes, we next 371 asked whether P2X7 functions, other than those involved in dye uptake, are also affected 372 in 16F-null cells. We found that current facilitation remained similar to that recorded in 373 HEK293T cells, albeit exhibiting seemingly slower kinetics, as well as current density (Fig. 374 7A, left and middle). A similar variability in the degree of current facilitation was also 375 noticed ( Fig. 7A, right), as well as a significant increase in the deactivation time constants 376 of facilitated currents ( Fig. S4A and B). Equally, no change in the profile of whole-cell 377 inhibition by CaCC inhibitors was observed compared to that of HEK293T cells (Fig. 7B 378 and C and Fig. S4C and D). These data indicate that the absence of hTMEM16F has no 379 apparent impact on whole-cell current properties. comprised of the P2X7 closed channel state ( Fig. 8A and Fig. S5). These modeling data 407 are therefore in agreement with experimental data indicating that P2X7 and TMEM16F are 408 able to form a physical association within the membrane.

411
In this work, we shed new light on the molecular mechanisms underlying 412 macropore formation and agonist-evoked current facilitation, two hallmarks of P2X7 that 413 have for several decades remained enigmatic. We reveal that both phenomena implicate 414 functional complexes formed between P2X7 and members of the TMEM16 family that we 415 named P2X7/TMEM16. We also provide evidence that, contrary to current beliefs 12,13 , 416 current facilitation does not stem from pore dilation.

418
The first important finding of this study is that when ectopically expressed in 419 HEK293T cells, P2X7 forms protein complexes with TMEM16F channels (and possibly 420 with other members of the TMEM16 family). We also reveal a functional coupling between 421 P2X7 and CaCCs in Xenopus oocytes, but no conclusive evidence can be provided on the 422 existence of such complexes in oocytes. A previous work has provided evidence that such 423 protein complexes are not formed in Xenopus oocytes, although a similar functional 424 coupling has been described between P2X7 and TMEM16A, but not with TMEM16F 28 .

425
The reason of this discrepancy is unclear, but may be related to a very low surface 426 expression of TMEM16F and/or to looser protein complexes in oocytes. However, in

427
HEK293T cells it appears that these protein complexes formed with TMEM16F have  (Fig. 8A), but the underlying molecular mechanism is currently unknown.

439
The existence of such protein complexes also provides a rational explanation of 440 how P2X7 current inhibition can be mediated by the non-selective 9-AC inhibitor. The lack 441 of action of 9-AC on BzATP-induced P2X7 currents in CaCC-void Axolotl oocytes 442 provides strong evidence that this inhibitor does not directly bind to P2X7, but instead 443 binds to endogenous TMEM16 channels and inhibits P2X7 activity via an allosteric 444 interaction due to the close proximity of the two channels. Our data further reveal its 445 selective inhibition of TMEM16F channels involved in dye uptake (Fig. 8B). Yet, the fact 446 that 9-AC no longer inhibits dye uptake in 16F-null cells, while still inhibiting BzATP-447 evoked currents, suggests that another TMEM16 subtype is coupled to P2X7 ion channel 448 function, including current facilitation, but not to dye uptake. Importantly, because 9-AC 449 does not abolish BzATP-evoked currents, even after a 1-min long application (Fig. 2C), 450 the possibility that a fraction of "free" P2X7 channels (i.e. those not embedded in a 451 complex) resides at the cell membrane cannot be excluded (Fig. 8B).

453 13
Conversely, the mode of action of FFA and TA remains unclear. Although FFA 454 inhibition was observed in Xenopus oocytes with no direct action on P2X7, its inhibitory 455 effect was quite variable in HEK293T cells, and no definitive conclusion can be drawn.

456
For TA, however, P2X7 currents were systematically abolished following application of 457 the inhibitor, even in Axolotl oocytes, suggesting that it may either directly inhibit P2X7 458 channel or act non-specifically through a currently unknown mechanism. Regardless of the 459 mechanism at play, this result is in apparent disagreement with previous work having found 460 that co-application of TA and ATP to human microglia did not inhibit endogenous P2X7 461 channel activity 27 . The reasons behind this discrepancy are unknown, but may result from 462 factors differing between recombinant and native systems. Finally, our data demonstrate 463 that TA is not suitable for dye uptake experiments because of its fluorescence quenching 464 properties, a finding that is also supported by a very recent study 41 .

466
The second important finding of the present work is that P2X7/TMEM16 largely 467 contributes to current facilitation. This is supported by the fact that 9-AC strongly impairs

479
In close agreement with previous data 7,8 , our results suggest that plasma membrane 480 cholesterol is a negative allosteric modulator of P2X7. It maintains the complex-embedded 481 P2X7, and possibly "free" P2X7, in a low channel activity state in response to initial, brief 482 agonist application. Longer or repeated agonist application then switches channels to 483 higher activity states, which, in turn, underlies the observed increased current amplitudes 484 and slowed deactivation kinetics. As suggested previously 6,8 , we too propose that this effect 485 is likely attributable to a progressive dissociation of cholesterol from channels (Fig. 8B). The third important finding of this study is that large molecule permeability 498 proceeds through at least two components in HEK293T cells (Fig. 8B). The first one likely 499 entails a direct passage through the P2X7 pore, as suggested previously 8,24 . Cryo-EM showing that long agonist exposure increases P2X7 NPo, we suggest that components 1 518 and 2 are boosted during long agonist exposure, which both contribute to the enhanced 519 levels of large molecule cell membrane permeability observed in dye uptake experiments 520 (Fig. 8B).

522
The precise permeation pathway underlying large molecule permeation of 523 component 2 is currently unknown, but may involve the TMEM16F pore as it functions as 524 a poorly selective ion channel 40,44 , enabling the passage of small organic cations, such as 525 NMDG + (ref. 40 ). Another possibility is that dye uptake shares the same mechanisms as 526 those implicated in TMEM16F scramblase activity 39,40,45 . Whatever the mechanism, 527 identification of the dye uptake pathway, as well as the molecular arrangement and 528 stoichiometry of P2X7/TMEM16F merits further study.

530
In conclusion, our data reveal that P2X7 can form functional complexes with 531 TMEM16 channels in HEK293T cells. Interestingly, a very recent work has also reported 532 similar modulation of P2X7 activity with another channel, TMEM163 (ref. 46 ). We propose 533 that P2X7/TMEM16 forms a regulating hub, which, upon P2X7 activation and dependent 534 on surrounding membrane cholesterol level, orchestrate a hive of activity, eliciting not only 535 channel gating (e.g. efflux of K + and Ca 2+ signaling), but also other cell-specific signaling, 536 including membrane blebbing, interleukin release and phospholipid scrambling 9 . Given the 537 critical roles of P2X7 and TMEM16 channels in many diseases, this platform may 538 potentially be involved in important pathophysiological processes that lead, for example, 539 to inflammation 47 and mechanical allodynia 48,49 , and may therefore be therapeutically 540 important. Future work is needed to unravel its broader implications in disease.  for all solutions used, and if necessary, adjusted carefully using NaOH to pH 7.32-7.33.

606
All solutions were maintained at approximately 300 mOsm.

607
Cells were voltage-clamped to -60 mV (whole-cell) or -120mV (outside-out) using the 608 EPC10 amplifier (HEKA), and data were recorded with PATCHMASTER software. Data 609 were acquired at 10 kHz and low-pass filtered at 2.9 kHz. Applications of agonist and/or 610 inhibitor were carried out by perfusion, using three capillary tubes placed directly over the 611 cell/patch of interest. These capillaries are displaced using the SF-77B Perfusion Fast Step  Agonist and drugs were applied using a computer-driven valve system (Ala Scientific).  where f(I) is the total probability density of a given amplitude value I, A i is the ith channel 744 amplitude, s i is the standard deviation of the ith channel amplitude, and a i is the fraction 745 of the data represented by the ith amplitude. Conductance was determined by dividing 746 current amplitudes by the holding potential (-120 mV).

747
For mean open time analysis, only patches featuring 3 or fewer channels were analyzed, 748 and analysis of stacked events resulting from simultaneous channel openings was less than 749 5% of the total events analyzed. The number of channels present within a given patch can 750 be determined by visual inspection of the maximum number of coinciding stacking events.

751
These stacked openings are designated as such, in order to verify the proportion that they 752 represent of the total events analyzed. For control condition, we recorded from 7 different 753 outside-out patches, in which 1,518 events were analyzed. Each patch contained between 754 1 to 15 sweeps. For facilitated condition, we recorded from 5 outside-out patches, in which 755 754 events were analyzed (ranged between 3 to 19 sweeps). For MCD-treated cells, we 756 recorded single-channel currents from 7 outside-out patches excised from cells that were 757 pre-treated with MCD, and in which 396 events were analyzed (ranged between 1 to 13 758 sweeps). weighting is calculated for recordings in patches exposed solely to BzATP, and then for 778 subsequent recordings of the same patch exposed to BzATP and CaCC inhibitor co- where t is the time, t d is time constant of deactivation, and A is the current amplitude.

790
For fluorescence experiment analysis, Fiji and Igor Pro were used. For co-localization 791 analysis Pearson's correlation coefficient was measured using Coloc2 in ImageJ.