A nanodomain anchored-scaffolding complex is required for PI4Kα function and localization in plants

Phosphoinositides are low-abundant lipids that participate in the acquisition of membrane identity through their spatiotemporal enrichment in specific compartments. PI4P accumulates at the plant plasma membrane driving its high electrostatic potential, and thereby facilitating interactions with polybasic regions of proteins. PI4Kα1 has been suggested to produce PI4P at the plasma membrane, but how it is recruited to this compartment is unknown. Here, we pin-point the mechanism that tethers PI4Kα1 to the plasma membrane via a nanodomain-anchored scaffolding complex. We identified that PI4Kα1 is part of a complex composed of proteins from the NO-POLLEN-GERMINATION, EFR3-OF-PLANTS, and HYCCIN-CONTAINING families. Comprehensive knock-out and knock-down strategies revealed that subunits of the PI4Kα1 complex are essential for pollen, embryonic and post-embryonic development. We further found that the PI4Kα1 complex is immobilized in plasma membrane nanodomains. Using synthetic mis-targeting strategies, we demonstrate that a combination of lipid anchoring and scaffolding localizes PI4Kα1 to the plasma membrane, which is essential for its function. Together, this work opens new perspectives on the mechanisms and function of plasma membrane nanopatterning by lipid kinases.


41
Eukaryotic cells are composed of several membrane-surrounded compartments. Each 42 compartment has a unique physicochemical environment delimited by a membrane with a 43 specific biochemical and biophysical identity (Bigay and Antonny, 2012). The membrane 44 identity includes the nature of the lipids, the curvature, the electrostaticity and the density of 45 lipids at the membrane. The identity of each membrane allows the proper localization of 46 membrane-associated proteins. 47 48 Phosphoinositides are rare anionic lipids present in membranes. for yeast survival and at least partially independent (Roy and Levine, 2004). In animal, three 62 PI4K isoforms, PI4KIIIβ/PI4KIIα/PI4KIIβ, are responsible for synthetizing PI4P at the 63 Golgi/TGN and in endosomes (Balla et al., 2002;Wang et al., 2003;Wei et al., 2002). Similar 64 to the stt4 in yeast, PI4KIIIα loss-of-function mutant is lethal in mammals (Nakatsu et al.,65 2012). In PI4KIIIα conditional mutants, the pool of PI4P disappears from the plasma 66 membrane while the TGN structures seem to remain untouched suggesting that the two pools 67 could be independent (Nakatsu et al., 2012). 68 69 In plants, PI4P massively accumulates at the plasma membrane and is less abundant at the 70 TGN (Simon et al., 2014(Simon et al., , 2016Vermeer et al., 2009). This PI4P accumulation at the cell 71 surface drives the plasma membrane electrostatic field, which in turn recruits a host of 72 signalling proteins to this compartment (Barbosa et al., 2016;Platre et al., 2018;Simon et al., 73 2016). Moreover, the plant TGN is the site of vesicular secretion but also involved in 74 endocytic sorting and recycling, which might imply regulatory mechanisms of lipid 75 exchanges or maintenance of membrane identity between plasma membrane and TGN (Noack 76 and Jaillais, 2017). 77 78 The Arabidopsis genome codes four PI4-kinases: PI4Kα1, PI4Kα2, PI4Kβ1 and PI4Kβ2 79 (Szumlanski and Nielsen, 2010 2014). In addition, pi4kβ1pi4kβ2 presents fewer and misshaped secretory vesicles at the TGN 85 (Kang et al., 2011). PI4Kβ1 and PI4Kβ2 have first been described to be localized to the 86 Trans-Golgi Network/Early Endosomes (TGN/EE) in root hairs (Preuss et al., 2006). This 87 localization, as well as its accumulation at the cell plate, has later been validated by electron 88 tomography and confocal microscopy in root meristematic cells (Kang et al., 2011;Lin et al., 89 is not homogenously present on the plasma membrane but enriched in nanodomains. 111 Although all the subunits of the complex are peripheral proteins and lack a transmembrane 112 domain, they show very little lateral mobility at the plasma membrane. These results suggest 113 that PI4Kα1 is not localized homogeneously at the plasma membrane but rather accumulates 114 in distinct hotspots at the inner leaflet of the plasma membrane. Consequently, the targeting 115 of this lipid kinase by a multiprotein scaffold might allow its precise spatiotemporal 116 recruitment in order to maintain the proper electrostatic landscape of plant cell membranes. 117 118

PI4Kα1 is a soluble protein with a potential lipid-binding domain 121
To determine how PI4Kα1 is recruited to the plasma membrane, we first analysed its 122 protein sequence in silico. Using TMHMM Server v. 2.0, no transmembrane helices could 123 be predicted in PI4Kα1 protein sequence suggesting that PI4Kα1 is a soluble cytosolic 124 protein. We then looked for lipid binding domains. Indeed, a Plecthrin Homology (PH) 125 domain has been found in its C-terminal, upstream from the catalytic domain (Stevenson 126 et al., 1998; Stevenson-Paulik et al., 2003;Xue et al., 1999). This PH domain was first 127 thought to localize PI4Kα1 at the plasma membrane through interaction with anionic 128 phospholipids but its role is now discussed (De Jong, F. and Munnik, T., 2021). Fat blot 129 experiments showed affinity of the putative PH domain for PI4P and in a lesser extent 130 for PI(4,5)P2 (Stevenson et al., 1998;Stevenson-Paulik et al., 2003). However, no 131 experiment in planta validates this result and using the Simple Modular Architecture 132 Research Tool (SMART) software, we were not able to retrieve the PH domain. Because 133 of the lack of predicted domain (except for the kinase domain), we decided to consider 134 other targeting mechanisms involving possible protein partners. Indeed, protein 135 targeting to a membrane can be multifactorial and may require coincidence binding of 136 lipids and protein partners. 137 138 To investigate this last hypothesis, we screened for PI4Kα1-protein partners. We performed a 139 yeast-two-hybrid screen with the large N-terminal part of PI4Kα1 (1-1468 aa) ( Figure 1A). 140 We recovered 267 in frame clones, which corresponded to 48 different proteins. Among them, 141 the screen revealed interactions between PI4Kα1 and the three members of a protein family 142 called NO POLLEN GERMINATION (NPG): NPG1 (At2g43040), NPG-Related 1 (NPGR1 143 -At1g27460) and NPGR2 (At4g28600) (Golovkin and Reddy, 2003). In the screen, we 144 retrieved 39 clones (7 independent clones) for NPG1, 32 clones (6 independent clones) for 145 NPGR1 and 2 clones (1 independent clone) for NPGR2. The clones from the NPG family 146 corresponded to about 30% of the total clones obtained from the screen, suggesting that they 147

PI4Kα1 interacts with NO POLLEN GERMINATION proteins
were over-represented. 148 149 NPG proteins contain tetratricopeptide repeats (TPR) motifs that are protein-protein 150 interaction motifs. In the yeast-two-hybrid screen, the selected interaction domain identified 151 for NPG1, NPGR1 and NPGR2 correspond to the C-terminal part of the proteins (aa 444-704  152  for NPG1; 501-694 for NPGR1 and 414-739 for NPGR2). This is also the part of the 153 sequence that contains the highest density of predicted TPR motifs suggesting that the  154  interaction between PI4Kα1 and NPGs could be mediated by the C-terminal TPR motifs.  155  156 Because all three members of the NPG family interacted with PI4Kα1 in yeast-two hybrids, 157 and given their high degree of identity and similar architecture, we decided to focus on one 158 member of the family to confirm the interaction in planta. We guided this choice based on 159 the RNAseq expression data compiled on eFP browser (https://bar.utoronto.ca/efp/cgi-160 bin/efpWeb.cgi). We chose NPGR2, as it is the family member with the highest and more 161 widespread expression. Indeed, NPG1 was predicted to be specifically expressed in the 162 pollen, while NPGR1 expression matched that of NPGR2 but was predicted to be much 163 weaker. 164 165 To confirm the interaction between NPGR2 and PI4Kα1, we produced stable transgenic lines 166 expressing UBQ10::NPGR2-mCITRINE. We raised antibodies against the native PI4Kα1 167 (residues 1 to 344 of PI4Kα1). In western blot the antibody recognized PI4Kα1 around the 168 expected size (225kDa) and the tagged version of PI4Kα1 with mCITRINE and 169 2xmCHERRY slightly higher (Supplemental Figure 1A; Table S1). We 170 immunoprecipitated NPGR2-mCITRINE or the plasma membrane protein Lti6b-CITRINE as 171 control using anti-GFP antibodies and probed whether they could co-immunoprecipitate 172 PI4Kα1, using our native antibody. We efficiently immunoprecipitated NPGR2-mCITRINE 173 or Lti6b-CITRINE, but PI4Kα1 was only co-immunoprecipitated with NPGR2-mCITRINE 174 ( Figure 1B). Together, these experiments suggest that PI4Kα1 can interact in yeast with the 175 C-terminus of all the three members of the NPG family and is at least found in complex with 176 NPGR2 in planta. 177 178  CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. we used the lines expressing mCITRINE-PI4Kα1 and NPGR2-mCITRINE to perform 201 immunoprecipitation (IP) followed by mass spectrometry analyses. Two lines expressing 202 membrane-associated (myristoylation-2xmCITRINE) and nuclear-excluded (mCITRINE-203 NES-mCITRINE) proteins were also used as generic controls for plasma membrane and 204 cytosolic proteins, respectively. In the NPGR2 IP, we found PI4Kα1, further confirming that 205 these two proteins are present in the same complex in plants. Only one common protein was 206 found in both NPGR2 and PI4Kα1 IPs but excluded from the two controls ( Figure 1C). we raised a UBQ10::HYC2-mCITRINE expressing lines and successfully isolated antibodies 215 raised against NPGR2 (residues 1 to 273 of NPGR2). The expected size of NPGR2 is 82kDa. 216 The antibody recognized a band at ca. 80kDa that is not present in npgr2-1 or npgr2-3 knock 217 out mutants or npgr1npgr2-1 double mutant (Supplemental Figure 1B; Table S1).

218
Moreover, the antibody recognized NPGR2-mCITRINE around 110kDa, but did not 219 recognize NPGR1-mCITRINE indicating that the antibody specifically recognized NPGR2 220 (Supplemental Figure 1B). We found that NPGR2 co-immunoprecipitated with HYC2-221 mCITRINE but not Lti6b-mCITRINE ( Figure 1D). Similarly, PI4Kα1 also co-222 immunoprecipitated with HYC2-mCITRINE but not with Lti6b-mCITRINE ( Figure 1E). 223 Next, we used yeast-two hybrid to check whether the two HYCCIN family members may 224 directly interact with PI4Kα1/NPGs ( Figure 1F). We found that the two isoforms that are 225 pollen specific, HYC1 and NPG1 interacted in yeast. anti-GFP antibody. We found that PI4Kα1 co-immunoprecipitated with EFOP2-mCITRINE 259 while it did not with Lti6b-mCITRINE ( Figure 1E), suggesting that EFOP2 may belong to 260 the PI4Kα1/NPGR2/HYC2 complex. 261 The summary of these interactions suggests that PI4Kα1 is part of an heterotetrameric 262 complex in which NPG proteins may act as a scaffold that bridges EFOP, HYC and PI4Kα1 263 proteins together ( Figure 1G). 264 265 pi4kα1 mutants are pollen lethal 266 Next, we took a genetic approach to confirm whether NPG, HYC and EFOP family members 267 indeed may function together with PI4Kα1 in planta. To this end, we isolated single mutants 268 for all the genes encoding for a subunit of the PI4Kα1 complex (Table S2). We started our 269 analysis with PI4Kα1 because it is the catalytic subunit and it is present as a single-copy gene 270 in the Arabidopsis genome for which we isolated two T-DNA insertion alleles. The first allele 271 (pi4kα1-1; GK_502D11) had an insertion in the first exon, while the second insertion (pi4kα1-272 2; FLAG_275H12) was located in the 20 th intron ( Figure 1A). we performed reciprocal crosses, using pi4kα1+/and wild-type as either male or female. For 281 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint both alleles, we recovered 0% resistant plants when pi4kα1+/was used as the male 282 indicating no transmission of the mutation via the pollen and thus complete male sterility 283 (Table 2). When pi4kα1+/was used as female, we obtained 39% and 9% of resistant plants 284 for each allele (Table 2). This result shows that the pi4kα1 mutation did not fully impair the 285 transmission through the female gametophyte but led to a partial distortion of the segregation. 286 287 288

291
Next, we observed pi4kα1 pollen grains using scanning electron microscopy (SEM) to test 292 whether they showed morphological defects (Figure 2A). For both alleles, half of the pollen 293 grains were shrivelled and likely not able to germinate explaining the pollen lethality ( Figure  294 2A and B). However, using Alexander staining, we observed that the pi4kα1-1 pollens were 295 still alive (Supplemental Figure 2B). DAPI staining also revealed the presence of the 296 vegetative nucleus and the two sperm cell nuclei indicating that meiosis likely occurred 297 normally (Supplemental Figure 2C). Further analysis by transmission electron microscopy 298 showed that the pi4kα1-1 pollen grains displayed an abnormally thick cell wall ( Figure 2C). 299 300 The reintroduction of a copy of PI4Kα1 under the control of its own promoter in pi4kα1-1 301 background fully complemented the pi4kα1-1 lethality as shown by the possibility to obtain 302 homozygous mutant plants (three independent complemented lines, see Supplemental 303 Figure 2A). In addition, self-fertilized pi4kα1-1-/-; PI4Kα1prom::PI4Kα1 plants showed a 304 low number of shrivelled pollen grains, comparable to control plants, indicating that a wild-305 type copy of PI4Kα1 is required for transmission through the male gametophyte and normal 306 pollen morphology (Figure 2A and B). Together, these results show that PI4Kα1 is an 307 essential gene for pollen development. 308 309 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 310 Table 2: Segregation analyses of reciprocal crosses between wild-type and the indicated mutants. n 311 represent the number of seedling analysed.

313
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 314 315

326
Disturbing subunits of the PI4Kα1 complex mimics pi4kα1 pollen phenotypes 327 Next, we isolated single mutants for all the genes encoding for NPG, HYC and EFOP 328 subunits in order to ask whether they would recapitulate pi4kα1 loss-of-function phenotype 329 (Table S2). 330 331 The npg1 mutant was previously published as not able to germinate giving its name NO 332 POLLEN GERMINATION to the family (Golovkin and Reddy, 2003). We reproduced this 333 result by characterizing two new T-DNA mutant alleles of NPG1. The self-progeny of npg1-334 1+/-and npg1-2+/-had segregation rate of 50,3% and 32% resistant seedlings, respectively, 335 indicating gamete lethality (Table 1). Reciprocal crosses confirmed their male sterility 336 phenotype, with 0% of transmission of the mutation through the pollen, while the female 337 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint gametophyte might be affected only for the second allele with a weak distortion of the 338 segregation rate (Table 2). However, the observation of npg1-1 and npg1-2 pollen grains by 339 SEM did not show any morphological defect, unlike pi4kα1 pollens ( Figure 3A-B; 340 Supplemental Figure 3A-B). The reintroduction of NPG1 fused with mCITRINE under the 341 control of its own promoter complemented the male sterility in npg1-2 background leading to 342 npg1-2 homozygous plants (Supplemental Figure 3C). Similarly, the expression of NPGR2 343 fused to the mCITRINE under the control of the NPG1 promoter also complemented npg1-2 344 male sterility. These experiments indicate that NPGR2 can substitute for NPG1 function in 345 pollen and that both NPG1-mCITRINE and NPGR2-mCITRINE fusion are fully functional 346 (Supplemental Figure 3C). 347 348 349 350

360
Because, NPGR2 can substitute for NPG1 in pollen, we speculated that a certain degree of 361 functional redundancy between NPG1, NPGR1 and NPGR2 or compensatory effects during 362 pollen development could lead to the weaker phenotype of npg1 pollen compared to pi4kα1 363 pollen and thus explain why npg1 pollen did not present morphological defect by SEM. To 364 test this hypothesis we generated higher order mutant combination within the NPG family 365 (Table S2). The npg1-2+/-npgr1-/-mutant combination presented about 10% of pi4kα1-like 366 shrivelled pollen grains while npgr1npgr2-1 and npgr1npgr2-2 double homozygous mutants 367 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint displayed about 25% of deformed (but not shrivelled) pollen grains ( Figure 3A respectively. These data indicate that combinations of npg mutants partially mimic pi4kα1 371 pollen phenotype. 372 373 Next, we addressed the loss-of-function phenotypes of HYCCIN family members. The self-374 progeny of a hyc1+/-single mutant presented a segregation rate around 50% indicating a 375 gametophytic lethality (Table 1). As HYC1 expression is restricted to pollen, we were 376 expecting that the segregation bias was caused by defects of the male gametophyte. As 377 anticipated, reciprocal crosses showed male sterility while the T-DNA transmission through 378 the female gametophyte was not affected ( Table 2). Observation of hyc1+/-pollen grains by 379 SEM revealed that half of the pollen grains were shrivelled ( Figure 3A and B). In addition, 380 transmission electron microscopy also showed a thickening of the cell wall of the hyc1 mutant 381 pollen grains, which was similar to the phenotype observed for pi4kα1-1 ( Figure 3C). 382 Finally, the male sterility, as well as the pollen morphological defects, were complemented by 383 the reintroduction of HYC1prom::HYC1-mCITRINE (Supplemental Figure 3A-B and D). 384 All together, these data show that hyc1 knockout phenotype fully mimic pi4kα1 knockout 385 regarding pollen development. 386 387 None of the efop single mutant presented any pollen morphological defects or distortion of 388 segregation likely because of redundancy between the four members of this family 389 (Supplemental Figure 3B). We thus generated all the possible combinations of double 390 mutants (Supplemental Figure 3B; Table 3). We were able to obtain efop2efop3 double 391 homozygous mutants, suggesting no strong synthetic lethality. However, they presented from 392 19% to 25% of shrivelled pollen grains, resembling those of the pi4kα1 and hyc1 mutants, and 393 from 43% to 65% of deformed pollens (resembling those of npgr1npgr2 double mutants). 394 Thus, depending on the alleles, these double mutant combinations presented from 70% to 395 90% of abnormal pollens (Supplemental Figure 3A and B). In addition, it was not possible 396 to generate efop3efop4 double homozygous mutant no matter the alleles used. Indeed, 397 reciprocal crosses indicated 0% of transmission of efop3 mutant allele when efop3+/-efop4-/-398 plants were used as male, revealing that efop3efop4 pollens were lethal (Table 2). SEM 399 showed that about 45% of the pollen grains present abnormal morphology ( Figure 3A  Thus, these proteins likely act as a single protein complex in plants. 409 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint   . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint

Disturbing the PI4Kα1 complex results in various sporophytic phenotypes 423
The pollen lethality of pi4kα1 knockout did not allow us to further study the role of PI4Kα1 in 424 plant development. However, combinations of npg mutants presented growth defect 425 phenotypes. Indeed, npgr1npgr2-2 double mutant and npg1+/-npgr1-/-npgr2-1-/-present a 426 mild growth phenotype while npg1+/-npgr1-/-npgr2-2-/-is dwarf ( Figure 4A-B; 427 Supplemental Figure 4A-B). Reintroduction of NPGR1prom::NPGR1-mCITRINE was able 428 to rescue the growth phenotype of npgr1npgr2-2 double mutant (Supplemental Figure 4C). 429 This suggests that the PI4Kα1 complex is essential not only for pollen development but also 430 for other developmental and growth processes. 431 432 While HYC1 is specifically expressed in pollen and is male sterile, HYC2 is predicted to be 433 expressed in the sporophyte. We characterized two T-DNA alleles corresponding to two 434 putative hyc2 loss-of-function mutants. The segregation rate of hyc2-2 heterozygous plants 435 was of 60% (Table 1). Moreover, it was not possible to retrieve homozygous plants in the 436 self-progeny of both hyc2-2 and hyc2-3. Reciprocal crosses indicated a transmission of the 437 allele through the male and the female gametophytes even if a weak distortion could be 438 observed in both cases (Table 2). Siliques from hyc2-2 and hyc2-3 heterozygous plants 439 presented around 25% to 30% of aborted seeds (Figure 4C-D; Supplemental Figure 4D). 440 Observations of the embryo after clearing showed that in those siliques, some embryo stopped 441 their development at the globular stage before degenerating while the rest of the embryos 442 pursued their development normally ( Figure 4E). This phenotype was lost and homozygous 443 mutant plants were obtained when HYC2-mCITRINE or HYC2-2xmCHERRY were 444 reintroduced under the control of the HYC2 promoter ( Figure 4D; Supplemental Figure 4D-445 E). Thus, the loss of HYC2 leads to embryo lethality at the globular stage, suggesting that 446 HYC1 is essential for the male gametophyte, while HYC2 is essential for sporophytic 447 development. These results are consistent with the idea that the four-subunit PI4Kα1 complex 448 is essential in plants.

450
The PI4Kα1 complex is associated with the plasma membrane 451 To confirm the localization of PI4Kα1 at the plasma membrane, we first raised stable 452 transgenic lines expressing PI4Kα1 tagged with mCITRINE at either its Nor C-terminal 453 ends and the red fluorescent protein 2xmCHERRY at the C-terminal end under the control of 454 either its own promoter or the UBQ10 promoter. Consistent with the hypothesis that PI4Kα1 455 acts at the plasma membrane, the three constructs mCITRINE-PI4Kα1, PI4Kα1-mCITRINE, 456 and PI4Kα1-2xmCHERRY localized at the plasma membrane and in the cytosol ( Figure 5A). 457 However, the introduction PI4Kα1::PI4Kα1-mCITRINE, PI4Kα1::mCITRINE-PI4Kα1 or 458 PI4Kα1::PI4Kα1-2xmCHERRY constructs in the pi4kα1-1+/mutant background failed to 459 complement the pollen lethality and we never recovered pi4kα1-1-/plants (data not shown). 460 We used the same PI4Kα1 promoter used for the rescue experiment with the untagged 461 PI4Kα1 PI4Kα1-6xHA or Flag-PI4Kα1) also failed to complement pi4kα1-1 (data not shown). 464 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 465 466

481
To confirm the localization obtained with mCITRINE fusion, we used the antibodies against 482 the native PI4Kα1 and performed whole mount immunolocalization in roots. Similar to the 483 mCITRINE-PI4Kα1 and PI4Kα1-mCITRINE fusions, we observed again a signal at the 484 plasma membrane ( Figure 5B). We also noticed signal in the cytosol and in nuclei ( Figure  485 5B). To further confirm the preferential association of PI4Kα1 with the plasma membrane, we 486 used cellular fractionation and PEG/Dextran phase partition of whole seedling and compared 487 the signal obtained on a purified plasma membrane or whole microsomal fractions. We 488 confirmed the presence of proteins in the two fractions using an antibody against V-type 489 ATPase. The purity of the plasma membrane fraction was evaluated with antibodies against 490 the PIP1,2 aquaporin, a known plasma membrane resident protein ( Figure 5C). When loading 491 the same amount of protein in each fraction, this experiment revealed the presence of a band 492 around 225kDa, corresponding to PI4Kα1 in the plasma membrane fraction and only a very 493 faint signal in the total microsomal fraction, showing that PI4Kα1 is enriched in the plasma 494 membrane fraction ( Figure 5C). Together, fluorescent fusion, immunolocalization and 495 cellular fractionation show that PI4Kα1 is associated with the plasma membrane. 496 497 We next addressed the subcellular localization of NPG, HYC and EFOP proteins at the 498 plasma membrane. NPG1 was previously found to be an extracellular protein in pollen grains 499 (Shin et al., 2014). In our hand, NPG1-mCITRINE, NPGR1-mCITRINE and NPGR2-500 mCITRINE all localized at the periphery of the cell in root meristem ( Figure 5A). To 501 distinguish between the plasma membrane and cell wall, we co-expressed NPGR2-502 mCITRINE with Lti6b-mCHERRY. We observed that the two signals perfectly colocalized 503 indicating that NPGR2 is present at the plasma membrane ( Figure 5D). Furthermore, we 504 performed plasmolysis of the epidermal root cell by addition of sorbitol. In this context, the 505 plasma membrane detaches from the cell wall. We observed that the signal of NPGR2-506 mCITRINE remains at the plasma membrane and is not present in the cell wall ( Figure 5E). 507 Moreover, in western blot using an anti-NPGR2 antibody, NPGR2 was found enriched in the 508 plasma membrane fraction compared to the microsomal fractions ( Figure 5C). 509 510 Similarly, HYC1-mCITRINE, HYC2-mCITRINE, EFOP1-mCITRINE, EFOP2-mCITRINE 511 and EFOP3-mCITRINE expressed under the control of the UBQ10 promoter were found at 512 the plasma membrane ( Figure 5A and D). In addition to the plasma membrane localization, 513 we noticed that EFOP1 and EFOP2 were associated with intracellular compartments, which 514 were more prominently labelled for EFOP1 than EFOP2. Also, NPG1 and HYC1 were highly 515 soluble in the cytosol. As HYC1 and NPG1 are normally mainly expressed in pollens, they 516 might need each other to localize at the plasma membrane. This could explain why when we 517 overexpressed them alone in root epidermal cells, we saw most of the proteins in the cytosol. 518 519 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint Upon plasmolysis, PI4Kα1-mCITRINE, mCITRINE-PI4Kα1, HYC2-mCITRINE and 520 EFOP2-mCITRINE signals are found inside the cell and absent from the cell wall. EFOP2 521 remained associated with the plasma membrane while PI4Kα1 and HYC2 seem to delocalize 522 to internal compartments ( Figure 5E). In any case, all the protein classes in the putative 523 PI4Kα1 complex were associated to some extent with the plasma membrane. 524 525 The PI4Kα1 complex is present in plasma membrane nanodomains 526 Using confocal microscopy, we noticed that for several of the translational reporters of the 527 PI4Kα1 complex, the signal at the plasma membrane was not continuous, raising the question 528 of a possible subcompartimentalization of the proteins. This is notably the case for PI4Kα1-529 mCITRINE, NPG1-mCITRINE, NPGR2-mCITRINE, HYC2-mCITRINE, EFOP2-530 mCITRINE, and EFOP3-mCITRINE ( Figure 5A and D and FIGURE 6A). Using plant co-531 expressing Lti6b-2xmCHERRY and NPGR2-mCITRINE, HYC2-mCITRINE or EFOP2-532 mCITRINE, we observed that NPGR2-mCITRINE, HYC2-mCITRINE and EFOP2-533 mCITRINE signals along the plasma membrane is less homogeneous than Lti6b, and 534 accumulated in patches of high intensity ( Figure 5D). We calculated the linear regression of 535 each signal along the membrane and observed that the R square of NPGR2-mCITRINE, 536 HYC2-mCITRINE and EFOP2-mCITRINE are always smaller than the one of Lti6b-537 2cmCHERRY indicating a higher dispersion of the intensity. In addition, plants co-expressing 538 NPGR2-2xmCHERRY and EFOP2-mCITRINE show similar intensity pattern with signals 539 partially localized along the plasma membrane ( Figure 5D). Similarly, when we observed the 540 plasma membrane in tangential sections, NPGR2-2xmCHERRY and EFOP2-mCITRINE 541 subdomains were partially colocalized ( Figure 6A). As control, EFOP2-mCITRINE 542 containing plasma membrane domains did not colocalize with the mostly uniformed 543 localization of Lti6b-2xmCHERRY ( Figure 5D and Figure 6A). In order to get a better axial 544 resolution, we used Total Internal Reflection Fluorescence Microscopy (TIRF) microscopy 545 and confirmed that NPGR2-mCITRINE, HYC2-mCITRINE and EFOP2-mCITRINE were 546 present in nanodomains of the plasma membrane ( Figure 6B). 547 548 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint

561
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 562 PI4Kα1 complex-containing nanodomains are static at the plasma membrane 563 Next, we investigated the lateral dynamics of the PI4Kα1 complex at the plasma membrane. 564 To do so, we used fluorescence recovery after photobleaching (FRAP). After bleaching, the 565 signal of NPGR2-mCITRINE, HYC2-mCITRINE and EFOP2-mCITRINE did not recover 566 after 2 min of acquisition ( Figure 6C-E; Supplemental Figure 5). In comparison, PI4P 567 sensors (P4M and PH FAPP1 ) recover in less than a minute after bleaching. Accordingly, the 568 mobile fraction calculated of NPGR2-mCITRINE and EFOP2-mCITRINE was low (around 569 20%) while the mobile fraction of the sensor reached 100% ( Figure 6F). This indicates that if 570 PI4P sensors are rapidly diffusing at the membrane, the PI4Kα1 complex is relatively static. 571 Furthermore, the identical dynamics of NPGR2-mCITRINE, HYC2-mCITRINE and EFOP2-572 mCITRINE further reinforce the notion that these subunits are part of a single protein 573 complex in vivo. 574 575 EFOPs localize at the plasma membrane via a lipid anchor 576 We next decided to investigate the mechanism by which the PI4Kα1 complex is targeted at 577 the plasma membrane. The four subunits of the PI4Kα complex are soluble proteins, without 578 known lipid binding domains. The efop3-1efop4-4 and efop3-2efop4-4 mutants showed the 579 same pollen lethality phenotype as the efop3-1efop4-2 and efop3-2efop4-2 mutant ( Figure  580 3A-B, Supplemental Figure 3A-B). While efop4-2 led to a very small-truncated protein (42 581 aa), the efop4-4 allele led to near full-length protein with only a small in frame deletion of 19 582 residues close to EFOP4 N-terminus. This suggested that this N-terminal region is crucial for 583 EFOP4 function (Figure 7, A). 584 585 The residues corresponding to this deletion are well conserved among the four EFOPs and 586 include both a cysteine-rich motif, which could be S-acylated, and a polybasic region, which 587 could contact anionic lipids at the plasma membrane ( Figure 7A). We thus tested the 588 potential role of those two elements in the regulation of EFOP localization and potentially the 589 recruitment of the PI4Kα1 complex at the plasma membrane. 590 591 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 592 593

596
Bold cysteines are predicted as S-acetylated on the SwissPalm database with high (blue) and medium (orange)

608
First, we evaluated the role of the polybasic patch in the N-terminus of EFOP proteins. 609 Indeed, this region could be involved in targeting the entire PI4Kα1 complex to the plasma 610 membrane through electrostatic interactions with anionic lipids, notably PI4P. In EFOP2, this 611 region goes from the aa 27 to the aa 39 and contains 7 positively charged residues ( Figure  612 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 7A). We mutated all lysines/arginines into neutral glutamines and generated a 613 UBQ10prom::EFOP2-7Q-mCITRINE construct. We observed that EFOP2-7Q-mCITRINE 614 was soluble when transiently expressed in Nicotiana benthamiana leaf cells while wild-type 615 EFOP2-mCITRINE was localized to the plasma membrane indicating that polybasic patch in 616 EFOP2 could be essential for plasma membrane targeting ( Figure 7B). We next introduced 617 the UBQ10prom::EFOP2-7Q-mCITRINE construct in Arabidopsis epidermal root cells. 618 However we did not retrieve any lines with a detectable fluorescent signal. It is likely that 619 EFOP2-7Q-mCITRINE is unstable either because of miss folding or because EFOP2 needs to 620 be associated with membrane to remain stable when expressed in Arabidopsis. Finally, we 621 directly investigated the role of PI4P in the recruitment of the PI4Kα1 complex, by using 622 PAO, a PI4K inhibitor. In this condition, the PI4P sensor is detached from the plasma 623 membrane and relocalized in the cytosol ( Figure 7C) We then investigated the role of the Cys-rich motif, which was deleted in the efop4-4 allele. 631 Such motif could be a site of S-Acylation; a lipid posttranslational modification that can 632 anchor protein to the plasma membrane (Zaballa and Goot, 2018). Indeed, according to the 633 SwissPalm prediction software, this motif is predicted as S-acetylated with a high (in blue) or 634 medium level of confidence (in orange) ( Figure 7A). Confirming this hypothesis, all four 635 Arabidopsis EFOP proteins were found to be S-Acylated in a recent proteomic study (Kumar 636 et al., 2020). Notably, all Cys-residues (underlined in Figure 7, A) within the Cys-rich region 637 of EFOP3 and EFOP4 were found to be S-acylated with high confidence in planta (Kumar et  638 al., 2020). To experimentally test the importance of such lipid modification in EFOP 639 localization, we mutated the two conserved cysteines (C20 and C23) into serine and generated 640 a UBQ10prom::EFOP2-CC-mCITRINE construct. Similar to EFOP2-7Q-mCITRINE, we 641 observed that EFOP2-CC-mCITRINE was soluble when transiently expressed in Nicotiana 642 benthamiana leaf cells ( Figure 7B). Next, we transformed the UBQ10prom::EFOP2-CC-643 mCITRINE construct into Arabidopsis and found that EFOP2-CC-mCITRINE was not 644 localized at the plasma membrane of root meristematic cells and instead accumulated in 645 intracellular structures ( Figure 7B). All together, this data suggest that EFOP2 is likely 646 targeted to the plant plasma membrane using lipid acylation anchoring. 647 648 The EFOP/NPG/HYC complex targets PI4Kα1 to the plasma membrane 649 We then asked if EFOP proteins were sufficient to determine the localization of PI4Kα1 in the 650 cell. Taking advantage of the EFOP2-CC construct localized in intracellular structures, we 651 introgressed NPGR2-2xmCHERRY or PI4Kα1-2xmCHERRY in plants expressing  CC-mCITRINE and analyzed if EFOP2-CC was able to recruit NPGR2 or PI4Kα1 in those 653 intracellular structures. We observed a weak signal of NPGR2 labelling intracellular 654 compartments that partially colocalize with EFOP2-CC-containing structures. Similarly, 655 PI4Kα1 was not only at the plasma membrane and soluble in the cytosol but also associated 656 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint with the EFOP2-CC-containing structures ( Figure 7D). This showed that EFOP is able to 657 recruit NPGR2 and PI4Kα1 in different compartments of the cell. 658 659 Because NPG proteins likely bridges PI4Kα1 to the membrane-binding EFOP subunits, we 660 reasoned that they should contribute to the targeting of PI4Kα1 at the plasma membrane. To 661 test this hypothesis, we generated a fusion between PI4Kα1-mCITRINE and the 662 transmembrane protein Lti6b in order to artificially target PI4Kα1 at the plasma membrane 663 and thus bypass the role of the NPG/HYC/EFOP complex. We transformed the npg1-2 664 mutant with the PI4Kα1prom::PI4Kα1-mCITRINE-Lti6b construct, and found a line able to 665 complement the npg1-2 mutant ( Figure 7E). This indicates that npg1 pollen lethality is likely 666 due to the absence of PI4Kα1 at the plasma membrane during pollen development. In these 667 plants, PI4Kα1-mCITRINE-Lti6b is located in clusters at the plasma membrane while Lti6b is 668 known as being rather homogenously localized at the plasma membrane ( Figure 5D and 7G). 669 The The plant PI4Kα1 complex is essential for cell survival 679 In this study, we showed that the loss-of-function of the PI4Kα1 leads to lethality of the male 680 gametophyte. Similarly, knockouts of HYC1 and EFOP proteins mimic this phenotype 681 supporting the idea that these proteins act as a complex. Surprisingly, npg1 single mutant is 682 pollen lethal but do not present the same morphological defects that pi4kα1 or hyc1 mutants. 683 The showed that NPGR2 complements the npg1 phenotype when expressed under the NPG1 696 promoter, which suggest that the difference between NPG1 and NPGR2 function is mostly at 697 the transcriptional level. 698 699 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint In addition to gametophytic phenotypes, we also observed that the loss of HYC2 induces 700 embryo lethality while npg triple sesquimutant display severe growth phenotypes. This is 701 concordant with the idea that PI4Kα1 has a critical role for the cell function, not only during 702 gametophytic but also sporophytic development. PI4P is crucial for the plasma membrane phenotypes. This contrasts with yeast, in which the loss of either pik1 or sst4 is lethal 716 (Audhya et al., 2000). It is thus possible that in plants, the lack of PI4Kβs at the TGN is 717 compensated by PI4Kα1 activity, perhaps through the endocytosis of plasma membrane PI4P. 718 In fact, a large portion of the pi4kβpi4k1β2 double mutant phenotype can be attributed to its 719 function in cytokinesis (Lin et al., 2019). Thus, PI4Kα1 activity is not able to compensate for 720 PI4Kβs function during cell division. 721 722

Function of the NPG-HYC-EFOP complex 723
Our study also showed that PI4Kα1's plasma membrane localization is likely mediated by 724 interactions with NPG and EFOP proteins rather than by a PH domain. At first, this PH . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. showed that these nanodomains are immobile in plants, despite the fluidity of the plasma 779 membrane (Jaillais and Ott, 2020). It is possible that unknown interactions with 780 transmembrane protein, cytoskeleton or lipid rafts stabilize the PI4Kα1 complex. 781 782 Do these nanodomains correspond to a functional unit? Among many possibilities, they could 783 correspond to ER-plasma membrane contact sites. Indeed, in yeast, Stt4 reside at these 784 contacts (Omnus et al., 2018). Another hypothesis is that they could be a zone of attachment 785 between the plasma membrane and the actin cytoskeleton. PI4Kα1 has been purified from F-786 actin-rich fraction from carrots and to associate with polymerized F-actin in vitro (Stevenson 787 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint Lennox, #214010, 20 g/L, 15% agar, Difco™ Bacto Agar) containing the Kanamycin at 849 50mg.L -1 . Plasmids were purified using the Nucleospin Plasmid kit (Macherey-Nagel®) and 850 inserts sequenced. Expression vectors containing the promoter-gene-fluorescent tag cassette 851 or 3'UTR were obtained by using LR clonase-based three-fragment recombination system 852 (Invitrogen®), the pB7m34GW/pH7m34GW/pK7m34GW/pS7m43GW/ 853 pLOK180_pR7m34g (gift from L. Kalmbach) destination vectors, and the corresponding 854 entry vectors (Table S3-S4). Only the EFOP2-7Q was introduce in the destination vector 855 pK7FWG2 using the LR clonase-based one-fragment recombination system (Invitrogen®) 856 (Karimi et al., 2002). Thermocompetent DH5α cells were transformed and selected on LB 857 plate (Difco™ LB Broth, Lennox, #214010, 20 g/L, 15% agar, Difco™ Bacto Agar) 858 containing 100mg.L -1 of Spectinomycin. Plasmids were purified using the Nucleospin 859 Plasmid kit (Macherey-Nagel®) and inserts sequenced. 860 861

Site directed mutagenesis 862
Plasmids were amplified by PCR using the high fidelity Phusion Hot Start II (Thermo 863 Fisher Scientific®) taq polymerase and the indicated primers carrying mutations (Table  864 S6). PCR products were digested using and cutsmart buffer (Biolab®

Plant transformation and selection 885
Agrobacterium were spun and resuspended in 5% sucrose and 0.02% Silwet L-77 886 detergent. Arabidopsis were transformed by floral dipping. 887 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. Yeast-two-hybrid 927 The initial screen was performed by hybrigenics services (https://www.hybrigenics-928 services.com/contents/our-services/interaction-discovery/ultimate-y2h-2), using the 929 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020.  (Table S7). 952

T-DNA and Crispr Mutant Genotyping 954
Leaf tissue from wild-type plants, T-DNA insertion lines and Crispr lines were collected, 955 frozen in liquid nitrogen and ground to powder. The powder was resuspended in the 956 extraction buffer (200mM of Tris pH 7.5, 250mM of NaCl, 25mM of EDTA, 0.5% of SDS).

957
DNA was precipitated with isopropanol and the DNA pellet was washed with 75% 958 ethanol before resuspension in water.

959
Plants were genotyped by PCR using the GoTaq® polymerase (Promega®) and the indicated 960 primers (Table S8). PCR products were migrated on 1% agarose gel or the percentage 961 indicated (Table S8). When sequencing was required, the bands were purified using the 962 NucleoSpin Gel and PCR Clean kit (Macherey-Nagel®) and sequenced. 963 964 965

Pollen observation by SEM 966
Pollen grains from mutant or WT flowers were placed on tape and observed using the 967 mini SEM Hirox® 3000 at -10°C, 10kV. 968 969 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint

Pollen inclusion and observation by TEM 970
For transmission electron microscopy, anthers were placed in a fixative solution of 3.7% 971 paraformaldehyde and 2.5% glutaraldehyde, Na2HPO4 0. . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint

1402
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1403 1404 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1408 1409 1410 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1413 1414

1415
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1416 1417 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1420 1421 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1424 1425 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1429 1430 . CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint 1433 1434

1435
. CC-BY 4.0 International license perpetuity. It is made available under a preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in The copyright holder for this this version posted December 9, 2020. ; https://doi.org/10.1101/2020.12.08.415711 doi: bioRxiv preprint