Phospho-regulated auto-inhibition of Cnn controls microtubule nucleation during cell division

γ-tubulin-ring-complexes (γ-TuRCs) nucleate microtubules. They are recruited to centrosomes in dividing cells via binding to N-terminal CM1 domains within γ-TuRC-tethering proteins, including Drosophila Cnn. Binding promotes microtubule nucleation and is restricted to centrosomes, but the mechanism regulating binding remains unknown. Here we identify an extreme N-terminal “CM1 auto-inhibition” (CAI) domain within the centrosomal isoform of Cnn (Cnn-C) that inhibits γ-TuRC binding. Cnn-C is phosphorylated at centrosomes and we find that phospho-mimicking sites within the CAI domain helps relieve auto-inhibition. In contrast, the testes-specific mitochondrial Cnn-T isoform lacks the CAI domain and can bind strongly to cytosolic γ-TuRCs. Ubiquitously expressing a version of Cnn-C lacking the CAI domain leads to major cell division defects, which appears to be due to ectopic cytosolic microtubule nucleation. We propose that the CAI domain folds back to sterically inhibit the CM1 domain, and that this auto-inhibition is relieved by phosphorylation that occurs specifically at centrosomes.


Introduction
TuRCs contain 14 γ-tubulin molecules held in a single-turn helical shape by laterally 48 associating γ-tubulin complex proteins (GCPs) (Kollman et al., 2011). The γ-tubulin molecules 49 bind directly to a/b-tubulin dimers to promote new microtubule assembly (Thawani et al., 50 2020). γ-TuRCs have a low activity within the cytosol but are thought to be "activated" after 51 recruitment to MTOCs (

The N-terminal regions of Cnn isoforms differ in their ability to bind γ-TuRCs 136
We previously published evidence that different isoforms of Cnn bind γ-TuRCs within the 137 cytosolic extracts of Drosophila embryos with different affinities (Tovey et al., 2018). We found 138 that bacterially purified MBP-tagged N-terminal fragments of Cnn-T (MBP-Cnn-T-N) could 139 immunoprecipitate cytosolic γ-tubulin with a much higher affinity than the equivalent fragments 140 of Cnn-C (MBP-Cnn-C-N). Both isoforms share a short sequence just proximal to the CM1 141 domain (residues 78-97 in Cnn-C), but differ in their extreme N-terminal region, which is 77 142 residues long in Cnn-C but only 19 residues long in Cnn-T (dark blue and red, respectively, in 143 1B; Figure S1). We realised that we could use these scaffolds to assess the ability of 157 and any modified forms of Cnn-C ( Figure 1B), to bind and recruit γ-TuRCs by examining the 158 recruitment of fluorescently-tagged γ-TuRC proteins to the scaffolds. 159 160 We first compared scaffolds formed from GFP-Cnn-C-PReM m to scaffolds where the extreme 161 N-terminal region of Cnn-C-PReM m was either exchanged with the extreme N-terminal region 162 of Cnn-T (GFP-Cnn-T-PReM m ) or was removed (GFP-Cnn-C ∆1-77 -PReM m ). For simplicity we 163 refer to these as Cnn-C, Cnn-T, and Cnn-C ∆1-77 scaffolds, respectively, regardless of the 164 fluorescent tag used. In order to assess recruitment of γ-tubulin complexes to the scaffolds, 165 we endogenously tagged the maternal form of γ-tubulin with mCherry (γ-tubulin-37C-166 mCherry) and injected unfertilised eggs laid by these mothers (we refer to γ-tubulin complexes 167 rather than γ-TuRCs, as it is possible that Cnn can bind and recruit γ-TuSCs as well as γ-168 TuRCs). Initial visual observations suggested that γ-tubulin-37C-mCherry associated more 169 readily with Cnn-T and Cnn-C ∆1-77 scaffolds than with Cnn-C scaffolds ( Figure 1C-E). We 170 measured the GFP (Cnn) and mCherry (γ-tubulin-37C) fluorescence signals at multiple 171 scaffolds of different sizes and from different eggs and generated linear lines of best-fit (Figure 172 1F; see Methods for details). The slopes of these lines (S) are an estimation of the relative 173 binding affinity between the different Cnn constructs and γ-tubulin complexes. S values for the 174 various Cnn scaffolds differed significantly overall (p<0.001, F2,354=1645) and also in pairwise 175 comparisons: the S value for Cnn-T scaffolds (0.55) was ~26-fold higher than the S value for 176 Cnn-C scaffolds (0.02) (p<0.001, t350=-57) ( Figure 1F), indicating that the N-terminal region of 177 Cnn-T binds γ-tubulin complexes with ~26-fold higher affinity than the N-terminal region of 178 Cnn-C, and confirming the results from our previous in vitro binding experiments (Tovey et al., 179 2018). Importantly, the S value for Cnn-C ∆1-77 scaffolds (0.17) was ~8-fold higher than the S 180 value for Cnn-C scaffolds (p<0.001, t357=36) (Figure 1F), showing that removing the extreme 181 N-terminal region of Cnn-C increases binding affinity to γ-tubulin complexes by ~8 fold. This 182 therefore confirms that the extreme N-terminal region of Cnn-C is inhibitory for binding to γ-183 tubulin complexes. Consistent with this, we also found that MBP-tagged N-terminal fragments 184 of Cnn-C ∆1-77 (MBP-Cnn-C-N ∆1-77 ) could immunoprecipitate γ-tubulin from embryo extracts 185 more efficiently than N-terminal fragments of Cnn-C ( Figure 1G Figure 1H). 189 Thus, while it is known that the CM1 domain is required for binding γ-tubulin complexes, we 190 find that the extreme N-terminal region of Cnn-C inhibits binding to γ-tubulin complexes; we 191 therefore name this region the "CM1 auto-inhibition" (CAI) domain. 192 193 Intriguingly, the in vivo assay also showed that the S value for Cnn-C ∆1-77 scaffolds was ~3.  fold lower than that of Cnn-T scaffolds (p<0.001, t357=-21) ( Figure 1F), suggesting that the 195 extreme N-terminal region of Cnn-T (red in Figure 1A) promotes binding to γ-tubulin 196 complexes. 197

Cnn-T scaffolds organise microtubules more robustly than Cnn-C scaffolds 218
If Cnn-T scaffolds can recruit functional γ-tubulin complexes, the scaffolds should be able to 219 nucleate and organise microtubules. We had previously shown that Cnn-C scaffolds could 220 organise microtubule asters, but only when these scaffolds were large (Conduit et al., 2014a). 221 To test whether microtubules are organised more robustly by Cnn-T scaffolds (which recruit 222 more γ-tubulin complexes than Cnn-C scaffolds) we formed scaffolds within eggs expressing 223 the microtubule binding protein Jupiter-mCherry ( Figure 3A,B). We used a blind analysis to 224 quantify the propensity of these Cnn scaffolds to organise microtubules by categorising eggs 225 into those where the scaffolds organised "strong", "medium", "weak", or no microtubule asters. 226 We also included a "tubulin overlay" category, where the Jupiter-mCherry signal did not extend 227 beyond the GFP scaffold signal. For simplicity, we refer to eggs containing Cnn-C and Cnn-T 228 scaffolds as Cnn-C and Cnn-T eggs, respectively. Consistent with the increased recruitment 229 of γ-tubulin complexes to Cnn-T scaffolds, we found that a higher proportion of Cnn-T eggs 230 (43.8%) than Cnn-C eggs (11.5%) contained scaffolds that organised strong or medium 231 microtubule asters ( Figure 3C). Moreover, while 69.2% of Cnn-C eggs contained scaffolds 232 that did not organise any visible microtubule asters, none of the Cnn-T eggs fell into this 233 category ( Figure 3C). Based on these differences, we conclude that the γ-tubulin complexes 234 recruited to Cnn scaffolds are able to nucleate microtubules, at least to some extent. The 235 ability of some Cnn-C scaffolds to organise microtubules is not unexpected, as these scaffolds 236 can still recruit low levels of γ-tubulin complexes ( Figure 1C,F). 237 238 Filming the scaffolds through time revealed that scaffolds could merge and could also be quite 239 mobile, especially those that had microtubules emanating from just one side (Video 1). In 240 these instances, the microtubules appeared to push the scaffolds through the cytosol. Most 241 intriguingly, we could observe events where spindle-like structures formed between adjacent 242 Cnn-T scaffolds ( Figure 3D). This was unexpected, but suggested that the microtubules 243 organised by the scaffolds are dynamic and can be regulated by motor proteins. We managed 244 to film the formation of these transient spindle-like structures (Video 2; Video 3), and in one 245 example multiple spindle-like structures formed simultaneously in close proximity and were 246 organised by a nearby group of coalescing scaffolds into a flower-like arrangement; the 247 spindles then quickly disappeared (Video 3). On some occasions, we could observe giant 248 Cnn-T scaffolds that organised large bundles of microtubules (Video 4). One of these giant 249 scaffolds rotated, dragging its associated microtubule bundles, indicating that the microtubules 250 were robustly anchored to the scaffolds, presumably via linkage by γ-tubulin complexes. In 251 summary, we conclude that Cnn-T scaffolds can recruit γ-tubulin complexes that are capable 252 of nucleating and anchoring microtubules. 253 254

Phosphorylation of a conserved helix within the CAI domain, including a conserved 255
Polo kinase site, helps relieve auto-inhibition 256 The data above shows that the extreme N-terminal CAI domain of Cnn-C inhibits binding to 257 γ-tubulin complexes. Given the similar length of the CAI and CM1 domains, we speculated 258 that the CAI domain may fold back over and inhibit the CM1 domain ( Figure 4A). Given that 259 Cnn-C recruits γ-TuRCs to centrosomes (Zhang and Megraw, 2007;Conduit et al., 2014b) 260 and that Cnn-C is phosphorylated specifically at centrosomes and not within the cytosol 261 (Conduit et al., 2014a), we reasoned that CAI domain inhibition could be relieved by 262 phosphorylation, with negatively charged phosphate groups driving the separation of the CAI 263 and CM1 domains ( Figure 4A). We therefore used sequence alignments and secondary 264 structure predictions of amino acids 1 to ~255 of Cnn-C from various Drosophila species to 265 search bioinformatically for potential phosphorylation sites ( Figure S3). We identified three 266 putative phosphorylation "patches" based on a high concentration of often conserved serine 267 and threonine residues ( Figure 4A a previous study that mutated T 27 to alanine within Cnn-C constructs reported microtubule 297 defects at centrosomes in Drosophila embryos (Eisman et al., 2015). We therefore generated 298 three different phospho-mimetic fragments, where either the proximal three residues (S 21 , S 22 , 299 T 27 ) were mimicked (MBP-Cnn-C-N P1a ), the distal three residues (T 31 , T 33 , S 34 ) were mimicked 300 (MBP-Cnn-C-N P1b ), or only the predicted Polo site was mimicked (MBP-Cnn-C-N T27 ). We 301 found that MBP-Cnn-C-N P1a and MBP-Cnn-C-N T27 bound more γ-tubulin than MBP-Cnn-C-N, 302 but that this was not true of MBP-Cnn-C-N P1b ( Figure 4C). Quantification of γ-tubulin band 303 intensities from 5 experimental repeats indicated that MBP-Cnn-C-N T27 immunoprecipitated 304 ~34% of γ-tubulin compared to MBP-Cnn-T-N, which is more than for MBP-Cnn-C-N (~18%) 305 although less than when all sites within P1 are mimicked (~44%). This data suggests that 306 phosphorylation of multiple residues within P1, including the predicted Polo site at T 27 , helps 307 to relieve CAI domain inhibition to allow binding to γ-tubulin complexes. This is consistent with 308 the microtubule defects previously observed at centrosomes in T 27 A mutant Drosophila 309 embryos (Eisman et al., 2015).  Figure 5A). We decided to use the N-terminal region of Cnn-T, rather than simply 320 removing the CAI domain or introducing phospho-mimetic mutations, as our in vivo and in vitro 321 assays had shown that this would allow the strongest and most consistent binding. As a control 322 we generated a line ubiquitously expressing untagged wild-type Cnn-C (pUbq-Cnn-C), whose 323 binding to cytosolic γ-tubulin complexes should be restricted by the CAI domain. 324 325 Consistent with the prediction that pUbq-Cnn-C T would bind and activate cytosolic γ-tubulin 326 complexes, we found it difficult to generate a viable pUbq-Cnn-C T line. We quickly discovered 327 that the pUbq-Cnn-C T stock that we had generated was difficult to maintain and combine with 328 other alleles, presumably due to toxic effects caused by pUbq-Cnn-C T expression. For 329 example, we were unable to generate a stock where pUbq-Cnn-C T was expressed in a cnn 330 mutant background. Thus, all of the following experiments were performed with the pUbq 331 constructs expressed in the presence of endogenous Cnn. 332

333
We first tested the fertility rates of males and females bred at 25˚C, comparing them to the 334 pUbq-Cnn-C control stock. We quantified the hatching rate of embryos that were generated 335 when pUbq-Cnn-C or pUbq-Cnn-C T males or females were crossed to w 1118 "wild-type" flies. 336 Young (0-1 week old) pUbq-Cnn-C T females produced embryos of which only 55% hatched, 337 compared to an average hatching rate of ~85% for embryos from young pUbq-Cnn-C females 338 ( Figure 5B; p<0.001). Hatching rates for older (1-2 week old) females of both genotypes were 339 similar to the younger females (52.2% for pUbq-Cnn-C T , 74.8% for pUbq-Cnn-C) and were 340 also significantly different from each other ( Figure 5B; p<0.001). An even larger reduction in 341 hatching rate between pUbq-Cnn-C and pUbq-Cnn-C T flies was observed when crossing 342 males. When young (0-1 week old) pUbq-Cnn-C T males were crossed to wild-type females 343 only 40.3% of embryos hatched, compared to 84.7% when crossing pUbq-Cnn-C males 344 (Figure 5B; p<0.001). When older (1-2 week old) pUbq-Cnn-C T males were crossed to wild-345 type females only 9.9% of embryos hatched, compared to 90.5% for pUbq-Cnn-C males 346 ( Figure 5B; p<0.001). Thus, the ability of both males and females to generate progeny is 347 reduced when ubiquitously expressing Cnn-C T compared to Cnn-C, and males are more 348 strongly affected. 349 350 Given this difference between females and males, we tested whether the expression of the 351 two different pUbq constructs varies within female and male germlines. We used two different 352 Cnn antibodies to probe western blots for Cnn protein: one raised against the N-terminal half 353 of Cnn-C (aa1-660), the other against the C-terminal half (aa661-1147). The N-terminal 354 antibody recognises fewer unspecific bands on western blots, but its ability to bind Cnn-C T 355 could be affected by the differences in the extreme N-terminal region of Cnn-T. Western 356 blotting extracts from wild-type, pUbq-Cnn-C, and pUbq-Cnn-C T embryos or testes with either 357 the N-terminal or C-terminal Cnn-C antibody showed that each extract contained a Cnn-C 358 band at ~150kDa ( Figure 5C). This band represents the endogenous form of Cnn-C in both 359 the wild-type and pUbq-Cnn-C T extracts (black arrowheads, Figure 5C), but is a combination 360 of endogenous Cnn-C and pUbq-driven Cnn-C in the pUbq-Cnn-C extract (blue arrowhead, 361 Figure 5C), explaining why this band is much stronger in the pUbq-Cnn-C extract lanes. The 362 lower band in the pUbq-Cnn-C T extract lanes is of an appropriate size for the slightly smaller 363 Cnn-C T protein (red arrowheads, Figure 5C). The C-terminal antibody generated a weaker, 364 non-specific band of similar size to the Cnn-C T band, which could be observed in the wild-type 365 and pUbq-Cnn-C lanes (asterisks). The unspecific nature of this band was highlighted by its 366 absence on membranes probed with the N-terminal Cnn antibody. Nevertheless, to confirm 367 that the lower bands in the pUbq-Cnn-C T lanes do indeed represent Cnn-C T , we generated an 368 isoform-specific antibody raised against a short peptide within the isoform-specific extreme N-369 terminal exon of Cnn-T (anti-Cnn-T N ). This antibody recognised a single band of the expected 370 size only in the pUbq-Cnn-C T extract lanes ( Figure 5C), thus confirming the identity of the Cnn-371 C T band. 372 373 Importantly, there was a clear difference in the relative levels of pUbq-Cnn-C T between 374 embryo and testes extracts. In the embryo extracts, the pUbq-Cnn-C T band was much weaker 375 than the endogenous Cnn-C band, which is unusual for pUbq-driven Cnn constructs (P. 376 Conduit unpublished observations); presumably the intensity of the pUbq-Cnn-C T band 377 observed when using the C-terminal antibody would have been even lower were it not for the 378 overlapping unspecific band recognised by this antibody. Indeed, the relative band intensity 379 of pUbq-Cnn-C T is much lower when using the N-terminal antibody, although this could also 380 reflect a difference in the ability of this antibody to recognise the modified N-terminal region of 381 pUbq-Cnn-C T . In contrast to embryo extracts, the pUbq-Cnn-C T band was of a similar intensity 382 to, if not higher than, the endogenous Cnn-C band in the testes extracts. We therefore 383 conclude that, relative to endogenous Cnn-C, pUbq-Cnn-C T is weakly expressed within the 384 maternal germline but is expressed to levels similar to endogenous Cnn within the testes. 385 While other factors could be involved, such as cell-specific effects of CM1 domain binding on 386 γ-tubulin complexes, differences in expression levels of pUbq-Cnn-C T between cells could 387 explain the difference in the ability of male and female flies to generate progeny. Consistent 388 with this, it was easier to maintain the pUbq-Cnn-C T stock at lower temperatures (where 389 expression is likely reduced). 390 391 pUbq-Cnn-C T binds to cytosolic γ-tubulin complexes with a higher affinity than pUbq-392

Cnn-C 393
We performed IP experiments to confirm that pUbq-Cnn-C T binds γ-TuRCs more efficiently 394 than pUbq-Cnn-C, which would presumably account for the toxic effects. We tried using beads 395 coated with the anti-Cnn-T N antibody but found that it did not pull down any protein (data not 396 shown), presumably as this antibody was raised against a peptide antigen and recognises 397 only denatured pUbq-Cnn-C T on western blots. We therefore used beads coupled to Cnn-C 398 C-terminal antibodies, which should recognise both pUbq-Cnn-C and pUbq-Cnn-C T equally 399 well. In order to compare the amount of γ-tubulin bound by each type of Cnn molecule, we 400 adjusted gel loading to reflect the differences in the expression levels of pUbq-Cnn-C and 401 pUbq-Cnn-C T within embryos. We found that more γ-tubulin was immunoprecipitated with 402 pUbq-Cnn-C T than with pUbq-Cnn-C, as expected ( Figure 5D, right panels). 403 404

Mis-regulation of binding to γ-tubulin complexes results in ectopic microtubule 405 nucleation and defects during cell division 406
The failure to generate normal numbers of progeny suggested that ectopic binding of Cnn to 407 γ-tubulin complexes leads to cellular defects during germline or embryo development. We 408 therefore carried out immunostainings to examine directly any potential defects. We first fixed 409 and stained oocytes for markers of polarity, where specific microtubule arrangements are 410 required to establish and maintain polarity (Bastock and Johnston, 2008). In wild-type oocytes, 411 the nucleus is positioned in the dorsal corner from stage 8 to 10 and Staufen protein localises 412 in the centre of the oocyte at stage 8 and then at the posterior in stage 9 and 10. This was 413 true of all pUbq-Cnn-C (n=35, stage 8; n=35, stage 9; n=30, stage 10) and all pUbq-Cnn-C T 414 (n=40, stage 8; n=50, stage 9; n=40, stage 10) oocytes ( Figure S4A,B). Gurken protein is 415 normally positioned close to the nucleus in the dorsal corner of the oocyte and its mis-416 positioning or its absence results in abnormal dorsal appendages that protrude from the 417 surface of the egg. Once again, we found that Gurken protein was localised normally in all 418 pUbq-Cnn-C (n=30) and all pUbq-Cnn-C T (n=35) stage 9 oocytes ( Figure S4C), and the dorsal 419 appendages were normal on all pUbq-Cnn-C (n=724) and all pUbq-Cnn-C T (n=488) eggs. We 420 therefore conclude that there are no severe microtubule organisation defects in pUbq-Cnn-C T 421 oocytes and that polarity is normally established. 422 423 We next stained early embryos laid by either wild-type, pUbq-Cnn-C, or pUbq-Cnn-C T females 424 for DNA, microtubules, and γ-tubulin. Prior to cellularisation, embryos go through 13 rounds 425 of rapid and near-synchronous mitotic divisions within a syncytium. Centrosomes are 426 constantly in a "mature" state and centrosomal microtubules help coordinate rearrangements 427 of the actin network that in turn ensures the correct spacing of nuclei and microtubule-based 428 spindles. In theory, ectopic nucleation of cytosolic microtubules could interfere with this 429 process, as well as spindle formation. We therefore performed a blind analysis of pUbq-Cnn-430 C or pUbq-Cnn-C T embryos categorising each embryo into those with severe, moderate, mild, 431 or no defects, depending on both the broadness of defects across the embryo and how severe 432 individual defects appeared. We found that a higher percentage of both pUbq-Cnn-C and 433 pUbq-Cnn-C T embryos fell into the severe and moderate categories compared to wild-type 434 embryos, but that a higher proportion of pUbq-Cnn-C T embryos than pUbq-Cnn-C embryos 435 fell into the severe category ( Figure 5E). Broadly, the categorisation reflects the observed 436 hatching rates in Figure 5B. Cytosolic non-centrosomal microtubules appeared to be present 437 within ~10.8% of pUbq-Cnn-C T embryos, but this was very similar in pUbq-Cnn-C embryos 438 (~9.6%). Nevertheless, this suggests that ectopic binding of γ-tubulin complexes by Cnn 439 cytokinesis also lead to variations in nebenkern size and number. We therefore quantified and 470 compared the variability in nuclear size within cysts (as a marker of karyokinesis defects) and 471 the nucleus:nebenkern ratio (as a marker of major karyokinesis defects and/or cytokinesis 472 defects). We found that the variability in nuclear size was larger within pUbq-Cnn-C T testes 473 compared to pUbq-Cnn-C testes ( Figure 5H). Moreover, while the nucleus:nebenkern ratio 474 was always close to 1 in pUbq-Cnn-C testes, it varied between 0.5 and 1.75 in pUbq-Cnn-C T 475 testes ( Figure 5H). Thus, major karyokinesis and cytokinesis defects exist within pUbq-Cnn-476  spermatocytes at different cell cycle stages ( Figure S5). We conclude that expressing pUbq-491 Cnn-C T , which can bind γ-tubulin complexes away from centrosomes, leads to major defects 492 during male meiosis due to ectopic nucleation of microtubules within the cytosol. 493 494 Discussion 495 We propose a molecular model for the spatiotemporal regulation of γ-tubulin complex binding 496 by Cnn-C during cell division in Drosophila. In this model, Cnn-C is prevented from binding γ-497 tubulin complexes within the cytosol by its extreme N-terminal region that we name the CM1-498 autoinhibition (CAI) domain. We propose that the CAI domain folds back to sterically inhibit 499 the CM1 domain, and that this inhibition is relieved only once Cnn-C is recruited to 500 centrosomes and phosphorylated at sites within the CAI domain, including the predicted Polo 501 site T 27 . This mechanism explains why Cnn-C does not normally bind γ-Tubulin complexes 502 within the cytosol, which can lead to ectopic microtubule nucleation and major defects during 503 cell division. 504 505 In contrast to Cnn-C, we show that the testes-specific isoform of Cnn, Cnn-T, binds cytosolic 506 γ-tubulin complexes with high affinity. We now show that this difference is largely due to the 507 presence of the CAI domain at the extreme N-terminus of Cnn-C, which is absent from Cnn-508 T due to differential exon splicing. Removal of the CAI domain from Cnn-C increases binding 509 to cytosolic γ-TuRCs. Moreover, phospho-mimicking sites in a conserved a-helix within the 510 CAI domain, including the putative Polo kinase site at T 27 , also allows increased binding to γ-511 tubulin complexes, presumably by relieving the auto-inhibition to some degree. We previously 512 We note that phospho-mimicking either all sites within P1, or T27 alone, did not result in 531 increased recruitment of γ-tubulin to Cnn scaffolds in the majority of eggs in our in vivo 532 recruitment assay (data not shown). It is therefore possible that in the presence of full length 533 Cnn-C that is oligomerising into scaffolds, other regulatory events become even more 534 important. Nevertheless, our data still support a model in which phosphorylation of residues 535 within P1, including T 27 , help to relieve CAI-domain mediated auto-inhibition. although a later study questioned the effect of these phosphorylation sites (Lyon et al., 2016). 557 The binding between human CDK5RAP2 and γ-TuRCs also appears to be regulated by 558 phosphorylation. It was shown that depletion of LLRK1 kinase, which functions downstream 559 of Plk1, reduced the ability of N-terminal fragments of CDK5RAP2 (aa51-200) to co-560 immunoprecipitate γ-tubulin from HEK293 cells, and also reduced the ability of these 561 fragments to promote ectopic microtubule nucleation within the cytosol (which is dependent 562 on γ-TuRC binding) (Hanafusa et al., 2015). In contrast to Drosophila Cnn-C, however, the 563 phosphorylation site identified as being important for γ-TuRC binding is located downstream 564 of the CM1 domain (S140). Nevertheless, it remains possible that this downstream region, in 565 place of, or in combination with, the region upstream of the CM1 domain, could fold back over 566 the CM1 domain and function in an equivalent manner to the CAI domain within Drosophila 567 Cnn. Clearly, CDK5RAP2 must somehow be prevented from binding cytosolic γ-TuRCs, and 568 this will have to be addressed in future. 569 570 While we have focussed on how the binding between Cnn-C and γ-tubulin complexes is 571 regulated during cell division, our data also highlights differences in how binding is regulated 572 between cell types and MTOCs. We have shown both here and previously (Tovey et al., 2018) 573 that the testes specific Cnn-T isoform, which lacks the CAI domain, binds efficiently to γ-tubulin 574 complexes in the apparent absence of any upstream regulatory events. Cnn-T is expressed 575 primarily within developing sperm cells and isoform-specific C-terminal exons mediate its 576 recruitment to mitochondria, where it binds and recruits γ-tubulin complexes (Chen et al., 577 2017). The mitochondrial surface is very different from mature centrosomes, which 578 concentrate a selection of kinases, including Polo. It therefore seems appropriate that Cnn-T 579 isoforms splice out the exons that comprise the CAI domain to ensure that phosphorylation is 580 not required for γ-tubulin complex binding. Presumably, binding and potential activation of γ-581 tubulin complexes within the shrinking cytosol of developing sperm cells is not detrimental to 582 sperm development (and may even be important for amplifying cytoplasmic microtubules), 583 unlike in dividing cells where our data shows that spindle formation and cytokinesis are clearly 584 perturbed. 585 586 Intriguingly, our data suggest that Cnn-T binds γ-tubulin complexes with a higher affinity than 587 a truncated version of Cnn-C lacking the inhibitory CAI domain. This shows that the Cnn-T 588 specific extreme N-terminal exon (red in Figure 1A) promotes binding. It also suggests that 589 binding within sperm cells might be stronger than binding at centrosomes within dividing cells. 590 Our preferred interpretation, however, is that the CAI domain, once unfolded from the CM1 591 domain, may promote γ-tubulin complex binding once its inhibitory role has been relieved, 592 possibly by direct binding or by binding to other proteins that support γ-tubulin complex 593 binding. This is consistent with observations in budding yeast, where removal of SPC110's N-594 terminal region up to the CM1 domain reduces its binding affinity to γ-TuSCs The idea that CM1-domain binding may promote γ-tubulin complex activity in a context-624 specific manner is consistent with our in vivo data showing that ectopic binding of cytosolic γ-625 TuRCs by Cnn leads to major defects during male meiosis but apparently to less severe 626 defects in embryos and no obvious polarity defects in oocytes. These differences, however, 627 could be due to the observed differences in expression levels of pUbq-Cnn-C T between cell 628 types. They may also be due to variation in the ability of different cell types to cope with 629 increased cytosolic microtubule nucleation. Orbitrap mass spectrometer (Thermo Fisher Scientific Inc, Waltham, MA, USA). Separation 784 of peptides was performed by reverse-phase chromatography at a flow rate of 300 nL/min 785 using a Thermo Scientific reverse-phase nano Easy-spray column (Thermo Scientific PepMap 786 C18, 2 μm particle size, 100A pore size, 75 μm i.d. x 50cm length). Peptides were initially 787 loaded onto a pre-column (Thermo Scientific PepMap 100 C18, 5 μm particle size, 100Å pore 788 size, 300 μm i.d. x 5mm length) from the Ultimate 3000 autosampler with 0.1% formic acid for 789 3 minutes at a flow rate of 10 μL/min. After this period, the column valve was switched to allow 790 elution of peptides from the pre-column onto the analytical column. Solvent A was water + 791 0.1% formic acid and solvent B was 80% acetonitrile, 20% water + 0.1% formic acid. The   and coupled to a RetigaR1 monochrome camera (QImaging) and a CoolLED pE-300 Ultra 889 light source using a 63X 1.3NA oil objective or a 40X 0.55NA air objective, respectively. 890 891

Fertility tests 892
Cages that were sealed with apple juice agar plates with a spot of dried yeast paste were set 893 up at 25°C containing ~50 newly-hatched test flies (e.g. pUbq-Cnn-C/ -C T ) and ~50 newly-894 hatched wild-type males or virgin females. The apple juice agar plates were exchanged with 895 fresh plates 2-4 times a day, and the removed plates were kept at 25°C for at least 25 hours 896 before the proportion of hatched eggs was calculated. 897 898

Image analysis 899
All images were processed using Fiji (ImageJ). Maximum intensity Z-plane projections were 900 used to quantify the intensity of Cnn and γ-TuRC components at Cnn scaffolds, and intensities 901 of each channel at each scaffold were corrected for cytoplasmic background intensity. 902 Background intensity subtraction was also performed for quantification of western blot band 903 intensities. Within each experiment, the intensities of the γ-tubulin IP bands were normalised 904 to the intensity of the γ-tubulin band in the MBP-Cnn-T-N IP. Images of Cnn scaffolds 905 organising microtubules within eggs and of stained pUbq-Cnn-C or pUbq-Cnn-C T embryos 906 were analysed blind to allow unbiased categorisation. 907 908

Statistical analysis 909
Most statistical analysis and all graph production were performed using GraphPad Prism 7 or 910 8. Analysis of Cnn scaffolds was carried out in the R programming language (https://www.r-911 project.org/) using the emmeans package. The fluorescent signals at Cnn scaffolds were 912 collected from multiple embryos but included measurements from multiple scaffolds from 913 individual embryos. For linear regression analysis we therefore needed to use a mixed effects 914 model to take account of the 'repeated' measurements from individual embryos. The raw data 915 was not Normally distributed and was log transformed prior to performing the linear regression 916 analysis. An iterative process was used to explore the best model, which revealed that a 917 straight-line relationship best fitted the raw data and that regression lines should pass through 918 the origin (0;0). This could have been expected given the biological prediction that an increase 919 in Cnn intensity should result in a linear increase in the intensity of a γ-TuRC component. For 920 the final model, we transformed the data by taking the log of the γ-tubulin/Cnn ratio.