Microtubule nucleation is mediated by γ-tubulin ring complexes (γ-TuRCs). In most eukaryotes, a GCP4/5/4/6 “core” complex promotes γ-tubulin small complex (γ-TuSC) association to generate cytosolic γ-TuRCs. Unlike γ-TuSCs, however, this core complex is non-essential in various species and absent from budding yeasts. In Drosophila, Spindle defective-2 (Spd-2) and Centrosomin (Cnn) redundantly recruit γ-tubulin complexes to mitotic centrosomes. Here, we show that Spd-2 recruits γ-TuRCs formed via the GCP4/5/4/6 core, but Cnn can recruit γ-TuSCs directly via its well-conserved CM1 domain, similar to its homologs in budding yeast. When centrosomes fail to recruit γ-tubulin complexes, they still nucleate microtubules via the TOG domain protein Mini-spindles (Msps), but these microtubules have different dynamic properties. Our data, therefore, help explain the dispensability of the GCP4/5/4/6 core and highlight the robustness of centrosomes as microtubule organizing centers. They also suggest that the dynamic properties of microtubules are influenced by how they are nucleated.

During cell division, centrosomes act as major microtubule organizing centers (MTOCs) to nucleate and organize microtubules that contribute to mitotic spindle formation (Conduit et al., 2015). Centrosomes comprise a pair of centrioles that recruit and are surrounded by the pericentriolar material (PCM). The PCM is a large collection of proteins and is the predominant site of microtubule nucleation and organization during mitosis. On entry into mitosis, centrosomes expand their PCM in a process called centrosome maturation (Khodjakov and Rieder, 1999; Piehl et al., 2004). This is particularly dramatic in Drosophila cells because interphase centrosomes have very little PCM and do not organize microtubules, while mitotic centrosomes have relatively large amounts of PCM and robustly organize microtubules (Rogers et al., 2008). This makes Drosophila centrosomes ideal for the study of mitotic PCM assembly and microtubule nucleation.

γ-Tubulin ring complexes (γ-TuRCs) are important PCM clients because they template the nascent assembly of microtubules (microtubule nucleation; Tovey and Conduit, 2018; Farache et al., 2018; Kollman et al., 2011). Along with actin and Mozart proteins, they comprise a single-turn helical arrangement of 14 laterally associated “spokes,” each made from a γ-tubulin complex protein (GCP or “Grip” protein in Drosophila) and a γ-tubulin molecule. The essential subunits of γ-TuRCs are 2-spoke γ-tubulin small complexes (γ-TuSCs), made from GCP2, GCP3, and two γ-tubulins. In budding yeast, γ-TuSCs are stimulated to assemble into helical structures when bound by the conserved “Centrosomin motif 1” (CM1) domain found within the yeast spindle pole body (SPB; centrosome equivalent) proteins, Spd110 and Spc72 (Stu2, a TOG domain protein, is also required in the case of Spc72; Kollman et al., 2010; Gunzelmann et al., 2018). The “CM1 motif” within Spc110’s CM1 domain binds across adjacent γ-TuSCs, which presumably promotes the oligomerization process at the SPB (Brilot et al., 2021; Lyon et al., 2016; Kollman et al., 2010; Lin et al., 2014). In most eukaryotes, however, γ-TuSCs are stimulated to assemble into γ-TuRCs within the cytosol via a 4-spoke GCP4/5/4/6 core complex that seeds ring assembly and is absent from budding yeast (Haren et al., 2020; Würtz et al., 2022). Indeed, the depletion of GCP4, GCP5, or GCP6 strongly inhibits cytosolic γ-TuRC assembly in humans, Xenopus, Drosophila, Aspergillus, and fission yeast cells (Cota et al., 2017; Vogt et al., 2006; Vérollet et al., 2006; Xiong and Oakley, 2009; Zhang et al., 2000). Intriguingly, however, these γ-TuRC-specific proteins are not essential in Drosophila, Aspergillus, or fission yeast (Xiong and Oakley, 2009; Vogt et al., 2006; Anders et al., 2006; Vérollet et al., 2006). Consistent with this, γ-TuSCs can be recruited to Drosophila S2 cell centrosomes after the depletion of GCP4/5/4/6 core complex components (Vérollet et al., 2006) and are recruited independently of the GCP4/5/4/6 core complex to the outer SPB plaque in Aspergillus (Gao et al., 2019). Nevertheless, how γ-TuSCs are recruited to centrosomes in the absence of the GCP4/5/4/6 core remains unclear.

The predominant view of γ-TuRC recruitment involves the binding of large coiled-coil “tethering proteins” whose experimental depletion leads to measurable reductions in γ-tubulin levels at centrosomes. One of these proteins, NEDD1/Grip71, associates with preformed γ-TuRCs in the cytosol and subsequently docks γ-TuRCs to the centrosomes (Lüders et al., 2006; Haren et al., 2006, 2009; Zhang et al., 2009; Gomez-Ferreria et al., 2012a, 2012b). All other tethering proteins do not associate with cytosolic γ-TuRCs but instead localize to centrosomes and appear to “dock” incoming γ-TuRCs. These include the Pericentrin family of proteins, CM1 domain-containing proteins (e.g., human CDK5RAP2, Drosophila Centrosomin [Cnn], fission yeast Mto1, and budding yeast Spc110 and Spc72), and the Spd-2 family of proteins (CEP192 in humans; Zimmerman et al., 2004; Gomez-Ferreria et al., 2007; Haren et al., 2009; Fong et al., 2008; Sawin et al., 2004; Zhang and Megraw, 2007; Conduit et al., 2014; Dobbelaere et al., 2008; Lee and Rhee, 2011). It is difficult to determine the individual role of these proteins in γ-TuRC recruitment as they act redundantly and depend on each other for their proper localization within the PCM. We previously showed that γ-tubulin partially accumulated at mitotic centrosomes in the absence of either Cnn or Spd-2, but failed to accumulate when both proteins were removed, with centrosomes also failing to accumulate other PCM proteins and nucleate microtubules (Conduit et al., 2014). This data showed that Cnn and Spd-2 can independently recruit γ-tubulin-containing complexes (hereafter γ-tubulin complexes), but it remained unclear how.

Cnn contains the highly conserved CM1 domain (Sawin et al., 2004), which binds directly to γ-tubulin complexes in yeast and humans, respectively, (Brilot et al., 2021; Wieczorek et al., 2019; Choi et al., 2010) and has been implicated in the recruitment of γ-tubulin complexes to centrosomes in different systems (Zhang and Megraw, 2007; Lyon et al., 2016; Samejima et al., 2008; Choi et al., 2010; Muroyama et al., 2016; Fong et al., 2008). However, whether the CM1 domain is essential for Cnn to recruit γ-tubulin complexes remains unclear as the effect of removing the CM1 domain has only been tested in the presence of Spd-2 (Zhang and Megraw, 2007). In contrast to Cnn, Spd-2 does not contain a CM1 domain and so how it recruits γ-tubulin complexes remains to be established.

In this study, we investigated how γ-tubulin complexes are recruited to mitotic centrosomes by Cnn and Spd-2. We used classical Drosophila genetics to combine specific mutant alleles or RNAi constructs and examined γ-tubulin accumulation at centrosomes in larval brain cells. We show that Cnn allows the centrosomal accumulation of γ-tubulin in the absence of the GPC4/5/4/6 core and Grip71 and that this is dependent on its CM1 domain. Mutations in the CM1 domain also abolish Cnn’s ability to associate with γ-tubulin in immunoprecipitation assays. This suggests that Cnn’s CM1 domain can bind and recruit γ-TuSCs to centrosomes, similar to the CM1 domains in budding yeast Spc110 and Spc72. In contrast, we find that Spd-2 does not support the centrosomal accumulation of γ-tubulin in the absence of the GPC4/5/4/6 core and Grip71, suggesting that Spd-2 can only recruit γ-TuRCs that have preformed in the cytosol. By selectively abolishing γ-tubulin complex recruitment, we show that mitotic centrosomes can nucleate microtubules independently of γ-tubulin complexes and that this depends on the TOG domain protein Mini-spindles (Msps), consistent with the conserved ability of TOG domain proteins to promote microtubule nucleation in vitro. Moreover, the microtubules nucleated in the absence of γ-TuRCs are more cold-stable than those nucleated in the presence of γ-TuRCs, suggesting that the dynamic properties of microtubules depend in part on how the microtubules were nucleated.

We first explored how Cnn recruits γ-tubulin complexes to mitotic centrosomes by comparing the levels of centrosomal γ-tubulin at interphase and mitotic centrosomes in larval brain cells from flies depleted of Spd-2 and different γ-tubulin complex proteins. Typically, interphase centrosomes have only ∼5–20% of the γ-tubulin levels found at mitotic centrosomes, and this residual γ-tubulin is closely associated with the centrioles and is non-functional with respect to microtubule nucleation (Conduit et al., 2014). An increase in γ-tubulin signal between interphase and mitotic centrosomes indicates that γ-tubulin complexes have been recruited to the expanding mitotic PCM, at least to some degree. Similar to our previous results (Conduit et al., 2014), we found that γ-tubulin was recruited to mitotic centrosomes in spd-2 null mutant brains with an average level of ∼77% compared with wild-type brains (Fig. 1, A and B). We know that this recruitment of γ-tubulin is entirely dependent on Cnn because centrosomes in cnn,spd2 double mutants fail entirely to recruit γ-tubulin during mitosis (Conduit et al., 2014). In contrast, combining spd-2 null mutant alleles with null or severe depletion mutant alleles, or RNAi alleles, for Grip71 and the GCP4/5/4/6 core components did not prevent γ-tubulin accumulation at mitotic centrosomes. In fact, the centrosomes in spd-2,grip71,grip75GCP4,grip128GCP5-RNAi,grip163GCP6 mutant cells had ∼66% of the γ-tubulin levels found at wild-type centrosomes, only slightly lower than the ∼77% in spd-2 mutants alone (Fig. 1, A and B). Thus, the recruitment of γ-tubulin to mitotic centrosomes that occurs in the absence of Spd-2, i.e., that depends upon Cnn, does not appear to require Grip71 or the GCP4/5/4/6 core.

While we cannot rule out that residual amounts of GCP4/5/4/6 core components in spd-2,grip71,grip75GCP4,grip128GCP5-RNAi,grip163GCP6 mutant cells may support a certain level of γ-TuSC oligomerization in the cytosol, we favor the conclusion that Cnn can recruit γ-TuSCs directly to centrosomes in the absence of the GCP4/5/4/6 core for several reasons: first, the alleles used for grip71 and grip75GCP4 are null mutants and the allele for grip163GCP6 is a severe depletion allele (see Materials and methods), and even individual mutations in, or RNAi-directed depletion of, Grip75GCP4, Grip128GCP5 or Grip163GCP6 are sufficient to strongly reduce the presence cytosolic γ-TuRCs (Vogt et al., 2006; Vérollet et al., 2006). Second, spd-2,grip71,grip75GCP4,grip128GCP5-RNAi,grip163GCP6 mutant cells are depleted for all structural γ-TuRC components except for γ-TuSCs and Actin (note that Mozart1 [Mzt1] is not expressed in larval brain cells [Tovey et al., 2018] and that Mzt2 has not been identified in flies). In human and Xenopus γ-TuRCs, Actin supports γ-TuRC assembly via interactions with a GCP6-N-term-Mzt1 module (Liu et al., 2019; Wieczorek et al., 2019, 2020; Zimmermann et al., 2020; Consolati et al., 2020), and so Actin alone is unlikely to facilitate assembly of γ-TuSCs into higher order structures. Third, our data agree with the observation that near complete depletion of Grip71, Grip75GCP4, Grip128 GCP5, and Grip163GCP6 from S2 cells does not prevent γ-tubulin recruitment to centrosomes (Vérollet et al., 2006). Fourth, given the strength of mutant alleles used, one would have expected a much larger decrease in centrosomal γ-tubulin levels in spd-2,grip71,grip75GCP4,grip128GCP5-RNAi,grip163GCP6 mutant cells if Cnn were not able to recruit γ-TuSCs directly to centrosomes. Thus, Cnn appears to recruit γ-TuSCs to centrosomes without a requirement for them to first assemble into higher-order complexes.

The recruitment of γ-TuSCs to centrosomes by Cnn appears to reflect the natural situation in budding yeast, where homologs of the GCP4/5/4/6 core, Grip71 and Mzt1, are absent and where γ-TuSCs are recruited to the SPB by direct binding of Spc110 and Spc72’s CM1 domain. We, therefore, reasoned that Cnn’s recruitment of γ-TuSCs may also be mediated by its CM1 domain. A previous study, however, had shown that replacing the endogenous cnn gene with an ectopically expressed UAS-GFP-Cnn construct lacking the CM1 domain led to a reduction, but not elimination, of γ-tubulin at centrosomes in syncytial embryos (Zhang and Megraw, 2007). Along with the potential effects caused by Cnn over-expression, we now know that Spd-2 can recruit γ-tubulin complexes independently of Cnn (Conduit et al., 2014), making it hard to evaluate the true effect of deleting the CM1 domain without also removing Spd-2. We, therefore, deleted the CM1 domain (amino acids 98–167, inclusive) from the endogenous cnn gene (Fig. S1, A–C; see Materials and methods) and combined this mutant allele with the spd-2 null mutant allele. Note that the Cnn∆CM1 protein was produced at similar levels to wild-type Cnn (Fig. S1 D). We found that γ-tubulin no longer accumulated at mitotic centrosomes in these cnn∆CM1,spd-2 mutant cells (Fig. 1, C and D), showing definitively that Cnn’s CM1 domain is essential for Cnn to recruit γ-tubulin complexes to mitotic centrosomes.

We also tested whether the CM1 domain was required for Cnn to associate with γ-tubulin complexes by comparing the ability of bacterially purified MBP-tagged Cnn fragments to coimmunoprecipitate γ-tubulin from wild-type cytosolic embryo extracts. We recently showed that Cnn’s centrosomal isoform (Cnn-C) is autoinhibited from binding cytosolic γ-tubulin complexes by an extreme N-terminal “CM1 auto-inhibition” (CAI) domain, but that this autoinhibition can be relieved by introducing T27E and S186D phospho-mimetic mutations (Tovey et al., 2021). These mutations were therefore included in the fragments to enable Cnn binding in “control” conditions (Cnn-C-NT27E,S186D). To identify mutations predicted to perturb CM1 binding, we used cross-species protein sequence alignments and identified F115, R101, and E102 as equivalent to residues important for γ-tubulin complex binding in humans (F75; Choi et al., 2010) and budding yeast (K120 and E121; Lin et al., 2014; Gunzelmann et al., 2018; Fig. 1 E). We mutated these residues in the Cnn-C-NT27E,S186D fragments (R101Q, E102A, and F115A) to mimic the mutations previously used in yeast and human experiments. Introducing all three mutations, or only F115A, abolished the ability of the Cnn fragments to coimmunoprecipitate γ-tubulin, while introducing E102A and F115A reduced but did not abolish coimmunoprecipitation (Fig. 1, F and G). Thus, F115 within Cnn’s CM1 domain is essential for Cnn binding to γ-TuRCs, as is true of the equivalent F75 residue in human CDK5RAP2.

We conclude that the CM1 domain is essential for Cnn binding to γ-TuRCs. Moreover, taken together, our data strongly indicate that, similar to Spc110 and Spc72 in budding yeast, Cnn can bind and recruit γ-TuSCs to centrosomes directly from the cytosol without the need for them to preform into γ-TuRCs in the cytosol.

To explore how Spd-2 recruits γ-tubulin complexes even though (unlike Cnn) it lacks a CM1 domain, we compared the levels of centrosomal γ-tubulin at interphase and mitotic centrosomes in larval brain cells from flies lacking Cnn and different γ-tubulin complex proteins. In cnn mutants alone, Spd-2 levels at mitotic centrosomes are reduced to ∼60% (Conduit et al., 2014), and this Cnn-independent pool of Spd-2 recruits γ-tubulin to on average ∼22–23% of wild-type levels (Fig. 2, A and B; Conduit et al., 2014). We were therefore testing which γ-TuRC components, when removed in addition to Cnn, reduced the mitotic centrosomal level of γ-tubulin further, such that there was no significant accumulation of γ-tubulin above interphase levels. Of note, the Cnn-dependent pool of Spd-2 also recruits some γ-tubulin complexes because γ-tubulin levels were slightly higher at centrosomes in Cnn∆CM1 mutant cells compared with cnn null mutant cells (Fig. S1, E and F). The absence of a large increase may be because deleting the CM1 domain appears to affect the ability of Cnn to form a robust scaffold and therefore maintain Spd-2 in the PCM, as we noticed that the γ-tubulin signals were often offset from the centriole signal (Fig. S1 E), as is the case in cnn mutant cells (Fig. 2 A; Fig. S1 E; Lucas and Raff, 2007).

We predicted that Spd-2 recruits γ-tubulin complexes via Grip71 because the human homolog of Spd-2, CEP192, associates with the human homolog of Grip71, NEDD1 (Gomez-Ferreria et al., 2012a). We found, however, that γ-tubulin could still accumulate at mitotic centrosomes relatively well in cnn,grip71 mutant cells, being only slightly lower than the γ-tubulin levels in cnn mutant cells (Fig. 2, A and B). Thus, Spd-2 relies only partly on Grip71 to recruit γ-tubulin complexes. There was a stronger reduction, however, when we removed Cnn and members of the GCP4/5/4/6 core, Grip75GCP4, and Grip163GCP6, either individually or in combination (Fig. 2, A and B). Given that the GCP4/5/4/6 core is required for the assembly of γ-TuRCs within the cytosol, this result suggests that Spd-2 (unlike Cnn) predominantly recruits preformed γ-TuRCs rather than γ-TuSCs. We found, however, that γ-tubulin accumulation at mitotic centrosomes was only abolished after the additional removal of Grip71 i.e., in cnn,grip71,grip163GCP6 mutant cells (Fig. 2, A and B), a phenotype that was not due to a failure of Spd-2 to accumulate at mitotic centrosomes (Fig. 2, C and D). Thus, Spd-2 appears to recruit a very small amount of γ-TuSCs (which may, or may not, be present as larger assemblies due to an association with Grip128-γ-tubulin) via Grip71 (i.e., the recruitment that occurs in cnn,grip75GCP4,grip163 GCP6 cells), but its recruitment of γ-tubulin complexes relies predominantly on the GCP4/5/4/6 core.

Intriguingly, γ-tubulin could still accumulate at mitotic centrosomes to some degree in cells lacking Cnn, Grip71, and Grip75GCP4, showing that removal of Grip75GCP4 does not perfectly phenocopy the removal of Grip163GCP6, and therefore suggesting that Grip163GCP6 may still be able to promote at least partial γ-TuRC assembly in the absence of Grip75GCP4. This is consistent with observations in human cells, where GCP6 depletion has a greater effect on cytosolic γ-TuRC assembly than GCP4 depletion (Cota et al., 2017). Alternatively, Spd-2 may interact with Grip163GCP6 and so be able to recruit its associated γ-tubulin independent of Grip75GCP4.

In summary, Spd-2’s recruitment of γ-TuRCs relies strongly on the presence of the GCP4/5/4/6 core, and therefore on γ-TuRC assembly within the cytosol, but the additional removal of Grip71 is required to entirely prevent accumulation of γ-tubulin at mitotic centrosomes. In contrast, Cnn’s conserved CM1 domain can mediate the recruitment of γ-TuSCs directly from the cytosol. The requirement of Spd-2 for the GCP4/5/4/6 core aligns with the absence of Spd-2 and GCP4/5/4/6 core component homologs in lower eukaryotes. In addition, the ability of Cnn to recruit γ-TuSCs may explain why the GCP4/5/4/6 core is not essential in several species studied so far, particularly if all CM1 domain proteins are able to stimulate γ-TuSC assembly into ring-like structures, as is the case for yeast Spc110 and Spc72.

In the course of examining cnn,grip71,grip163 mutants, we observed that their mitotic centrosomes, which fail to accumulate γ-tubulin but still accumulate Spd-2, were still associated with microtubules during prophase and localized to spindle poles during mitosis (Fig. 3 A). This is in contrast to centrosomes in cnn,spd-2 mutant cells, which lack PCM entirely, fail to nucleate or organize microtubules, and do not associate with spindle poles (Conduit et al., 2014). Thus, mitotic centrosomes can organize microtubules independently of γ-TuRCs so long as the PCM can at least partially assemble. To test whether these microtubules are actually nucleated at centrosomes (rather than being nucleated elsewhere and then attaching to the centrosomes), we performed a cooling–warming microtubule repolymerization assay. We depolymerized microtubules by cooling larval brains on ice for ∼40 min and then either chemically fixed samples on ice or allowed them to warm up for 30 s before rapid chemical fixation. ∼40 min of cooling efficiently depolymerized microtubules at most centrosomes in wild-type and cnn,grip71,grip163 mutant centrosomes (Fig. 3 B). After 30 s warming, all wild-type and cnn,grip71,grip163 centrosomes had an associated α-tubulin signal, either as asters or as part of a reformed mitotic spindle (Fig. 3 C), strongly suggesting that an accumulation of γ-tubulin at mitotic centrosomes is not necessary for these centrosomes to nucleate microtubules.

To better understand microtubule dynamics at wild-type and cnn,grip71,grip163 centrosomes, we established a system to image cells live while cooling and warming the sample. We generated stocks containing fluorescent markers of microtubules (Jupiter-mCherry) and centrosomes (GFP-PACT) with and without the cnn, grip71, and grip163 mutations and used a microscope-fitted heating–cooling device (CherryTemp) to modulate the temperature of larval brain samples during recording. We imaged samples for ∼30 s before cooling them to 5°C for 3 min to depolymerize microtubules and then rapidly warming them to 20°C to observe microtubule regrowth. When cells were cooled to 5°C, the centrosomal Jupiter-mCherry signal decreased toward cytosolic background levels at both control and cnn,grip71,grip163 centrosomes (Fig. 3, D–F; and Videos 1 and 2). In a subset of cells, this centrosomal signal reached cytosolic levels (i.e., disappeared) after 3 min of cooling (Fig. S2, A and B). On warming to 20°C, there was an immediate increase in the centrosomal Jupiter-mCherry signal at all control and cnn,grip71,grip163 centrosomes (Fig. 3, D–F; and Fig. S2, A and B), confirming that microtubules can be nucleated at mitotic centrosomes that have not accumulated γ-tubulin. The dynamics of the microtubules differed, however (Fig. 3 F). On cooling to 5°C, microtubules depolymerized faster at control centrosomes (Fig. 3 F)—fitting “one-phase exponential decay” models to the depolymerization phases produced half-lives of 21.79 and 44.88 s and decay rate constants of 0.0361 and 0.0154 for control and cnn,grip71,grip163 centrosomes, respectively. On warming to 20°C, microtubules also polymerized faster at control centrosomes (Fig. 3 F)—fitting “exponential plateau” models produced growth rate constants of 0.0759 and 0.0536, respectively, which when normalized to the YM values (the maximum plateau values) showed an ∼3.4-fold difference in growth rate (Fig. 3 H). Differences in microtubule dynamics were also apparent when imaging EB1-GFP comets, which mark growing microtubule plus ends. EB1-GFP comets emerging from control centrosomes disappeared immediately after cooling to 5°C and then reappeared immediately after warming to 20°C (Fig. 3 I and Video 3), but comets emerging from mutant centrosomes took longer to disappear, and fewer reappeared, during cooling and warming cycles (Fig. 3 J and Video 4). Moreover, it was easier to observe EB1-GFP comets emerging from chromatin regions in these cnn,grip71;grip163 mutant cells (Fig. 3 J and Video 4), presumably because the centrosomes were no longer such dominant sites of microtubule nucleation. Thus, microtubules depolymerize faster and are then nucleated and/or polymerized faster at control centrosomes compared to at cnn,grip71,grip163 centrosomes.

One caveat with the experiments above is that centrosome assembly is strongly perturbed in cells lacking the centrosome scaffold protein Cnn (Lucas and Raff, 2007; Conduit et al., 2014), potentially impacting the γ-TuRC-independent ability of centrosomes in cnn,grip71;grip163 mutant cells to nucleate and organize microtubules. We, therefore, generated cnn∆CM1,grip71,grip163 mutants with or without GFP-PACT and Jupiter-mCherry, allowing us to examine microtubule dynamics at mitotic centrosomes that did not accumulate γ-tubulin (Fig. 4, A and B) but that still had Cnn to help assemble the PCM (although we note that PCM assembly appears perturbed to some degree in Cnn∆CM1 mutant cells—see Fig. S1, E and F). Prior to cooling, centrosomes in cnn∆CM1,grip71,grip163 mutant cells had a Jupiter-mCherry signal that was on average slightly higher than in control cells, suggesting robust microtubule organization (Fig. 4, C–E). Similar to cnn,grip71,grip163 mutants, microtubules depolymerized slower at cnn∆CM1,grip71,grip163 centrosomes compared with controls (Fig. 4, C–E). Fitting models to the data revealed half-lives of 16.7 and 37.7, and decay rate constants of 0.0416 and 0.0184 for control and cnn∆CM1,grip71,grip163 mutants, respectively (Fig. 4 G). However, the signal plateaued at a relatively high value despite cooling for 5 min as opposed to 3 (Fig. 4 E), suggesting that a larger proportion of microtubules are cold-stable at cnn∆CM1,grip71,grip163 mutant centrosomes compared with cnn,grip71,grip163 mutant and control centrosomes. On warming to 20°C, microtubules polymerized at cnn∆CM1,grip71,grip163 centrosomes but again at a slower rate than at control centrosomes (Fig. 4, C–E): growth rate constants normalized to the YM values showed an ∼3.4-fold difference in growth rate (Fig. 4 H), very similar to the growth rate for cnn,grip71,grip163 mutant centrosomes. The absence of increased microtubule nucleation from cnn∆CM1,grip71,grip163 centrosomes may in part reflect the fact that Cnn∆CM1 does not appear to support PCM assembly as well as wild-type Cnn (Fig. S1 E). Moreover, we note that the absence of Grip71 may impact the ability of Augmin to amplify the microtubules being nucleated from centrosomes, thereby reducing nucleation efficiency compared with controls.

We conclude that centrosomes can nucleate microtubules independently of γ-TuRCs, but these microtubules are nucleated slower or grow slower and are more cold-stable than microtubules nucleated from wild-type centrosomes. This suggests that different modes of microtubule nucleation generate microtubules with different properties.

We next addressed which proteins promote γ-TuRC-independent microtubule nucleation at mitotic centrosomes. We did not observe any clear enrichment of α-tubulin at centrosomes after microtubule depolymerization (Fig. 3 B and Fig. S3), ruling out the possibility that a high local concentration of α/β-tubulin accounts for or contributes to γ-TuRC-independent microtubule nucleation. Proteins of the chTOG/XMAP215 and TPX2 protein families have been reported to promote γ-TuRC-independent microtubule nucleation. These proteins promote microtubule nucleation in a range of species both in vitro and in vivo, including in the absence of γ-TuRCs (see Discussion and references therein). The Drosophila homolog of chTOG is Minispindles (Msps), which binds microtubules, localizes to centrosomes and spindle microtubules, and is required for proper spindle formation, mitotic progression, and chromosome segregation (Cullen et al., 1999). Msps has also been reported to stabilize the minus ends of microtubules when bound and recruited to centrosomes by TACC (Barros et al., 2005; Lee et al., 2001). Msps is also part of a group of proteins that organize microtubules independently of γ-tubulin at the nuclear envelope of fat body cells (Zheng et al., 2020). Moreover, the TOG1 and TOG2 domains of Msps promote microtubule nucleation in vitro (Slep and Vale, 2007). The putative Drosophila TPX2 homolog is Mei-38 and, while its depletion results in only mild spindle defects, Mei-38 binds microtubules, localizes to centrosomes and spindle microtubules, and is required for microtubule regrowth from kinetochores (Popova et al., 2022; Goshima, 2011). CAMSAP/Patronin/Nezha protein family members have also been implicated in γ-TuRC-independent microtubule nucleation and organization at non-centrosomal sites (Akhmanova and Kapitein, 2022), and CAMSAP2 condensates can stimulate microtubule nucleation in vitro (Imasaki et al., 2022).

To test the role of these proteins in γ-TuRC-independent nucleation from centrosomes, we combined mutant or RNAi alleles with the cnn, grip71, and grip163 mutant alleles and analyzed microtubule organization at centrosomes during prophase, when microtubule asters are most robust (Conduit et al., 2014). We were unable to obtain third instar larvae when combining the cnn, grip71, and grip163 mutant alleles with patronin mutant or RNAi alleles, presumably due to severe microtubule defects that prevented development, and thus could not test the role of Patronin. We could, however, obtain larvae when combining the cnn, grip71, and grip163 mutant alleles with mutant alleles for msps or tacc or an RNAi allele for Mei38. A clear association of microtubules with centrosomes was observed in 100% of wild-type prophase cells and in 96.8% of cnn,grip71,grip163 mutant prophase cells (Fig. 5, A, B, and F), consistent with our observations above that cnn,grip71,grip163 centrosomes can nucleate and organize microtubules. In contrast, a clear association of microtubules with centrosomes was observed in only 55.3% of cnn,grip71,grip163,msps mutant prophase cells, and in 70.7% and 81.4% of cnn,grip71,grip163,tacc and cnn,grip71,grip163,mei-38-RNAi mutant cells, respectively (Fig. 5, C–F). Moreover, the cnn,grip71,grip163,msps centrosomes tended to be positioned further from the spindle poles than the cnn,grip71,grip163 centrosomes (Fig. 5, G and H), which is indicative of a reduced capacity to organize microtubules. Positioning of centrosomes in cnn,grip71,grip163,tacc and cnn,grip71,grip163,mei-38-RNAi mutant cells was less affected, presumably due to the less severe defects in microtubule organization at cnn,grip71,grip163,tacc and cnn,grip71,grip163,Mei38-RNAi centrosomes.

Given that Msps appeared to be most important for γ-TuRC-independent nucleation of microtubules from centrosomes, we tested its role directly by performing a cooling/warming microtubule nucleation assay (similar to the fixed cell assay performed in Fig. 3, B and C) and compared the recovery of microtubules at cnn,grip71,grip163 and at cnn,grip71,grip163,msps centrosomes 30 s after warming. We categorized cells as those with or without centrosomes (some cells lack centrosomes due to mis-segregation during mitosis) and those that had or had not yet formed spindles; the proportion of these categories was similar in both mutant types (Fig. S4). There were, however, differences between the mutant types within each category. Of the cells that contained centrosomes but had not yet formed a spindle, centrosomes organized microtubules in ∼94.3% of cnn,grip71,grip163 mutant cells, the majority of which were scored as having strong or medium asters, but centrosomes organized microtubules in only ∼37.7% of cnn,grip71,grip163,msps mutant cells, the majority of which were scored as having weak asters (Fig. 5, I and J). This difference appeared to affect spindle formation because, of the cells that had centrosomes and that had formed a spindle structure, spindles were scored as being of “high” or “medium” quality (based on their morphology and density) in ∼67.1% of cnn,grip71,grip163 mutant cells but in only ∼28% of cnn,grip71,grip163,msps mutant cells (Fig. S5, A and B). This was specific to centrosomes because there was a similarly high proportion of cells containing low-quality spindles in both mutant types when cells lacked centrosomes (Fig. S5 C). For comparison, spindles were scored as being of “high” or “medium” quality in ∼95.3% of wild-type cells (Fig. S5 A). Note also that the absence of Grip71 abolishes the Augmin-mediated nucleation pathway necessary for efficient spindle assembly (Reschen et al., 2012; Chen et al., 2017b; Dobbelaere et al., 2008; Vérollet et al., 2006), but as both mutant types lacked Grip71 this cannot explain the differences observed between the mutants.

In summary, our data show that centrosomes lacking γ-tubulin complexes can still nucleate microtubules, despite having reduced PCM, and that the TOG domain protein Msps plays an important role in this γ-TuRC-independent microtubule nucleation pathway.

How centrosomes nucleate and organize microtubules is a long-standing question. Centrosomes contain hundreds of proteins, many of which associate with microtubules, meaning that understanding how centrosomes nucleate and organize microtubules is not trivial. Prior to our current work, we had identified Cnn and Spd-2 as the two key PCM components in flies—remove one and the other could support partial PCM assembly and microtubule organization; remove both and PCM assembly and microtubule organization fail (Conduit et al., 2014). We found that γ-tubulin could still accumulate at mitotic centrosomes after the removal of either Cnn or Spd-2, showing that both proteins could mediate the recruitment of γ-tubulin complexes, but it remained unclear how. The work presented here shows that Cnn and Spd-2 recruit different types of γ-tubulin complex, with Cnn able to recruit γ-TuSCs and Spd-2 recruiting predominantly preformed γ-TuRCs. Moreover, by preventing γ-tubulin recruitment but not PCM assembly, we have shown that centrosomes still nucleate microtubules in the absence of γ-TuRCs and that this γ-TuRC-independent mode of microtubule nucleation is stimulated by the TOG domain protein Msps (Fig. 6).

By using classical genetics, we have found that Cnn can recruit γ-tubulin complexes independently of Grip71 and the GCP4/5/4/6 core, meaning that it must be able to recruit γ-TuSCs. This is consistent with previous observations showing that γ-TuSCs could still be recruited to mitotic Drosophila centrosomes in S2 cells lacking the GCP4/5/4/6 core components (Vogt et al., 2006), although it was unknown at that time that this recruitment was dependent on Cnn. Our data here also show that this occurs in vivo. The ability of Cnn to recruit γ-TuSCs is similar to its budding yeast homologs’ ability, where Grip71 and the GCP4/5/4/6 core are naturally absent. Consistent with this, we show that Cnn’s binding and recruitment of γ-tubulin complexes relies entirely on its highly conserved CM1 domain, which binds across the inter γ-TuSC interface in budding yeast complexes (Brilot et al., 2021). The binding of the CM1 domain in budding yeast stimulates the oligomerization of γ-TuSCs into γ-TuRCs (Kollman et al., 2010; Lyon et al., 2016; Brilot et al., 2021; Lin et al., 2014; Gunzelmann et al., 2018), but whether this is true of Cnn’s CM1 domain, or CM1 domains in other eukaryotes, remains to be determined. Consistent with this possibility, however, Cnn’s CM1 domain is more similar to Spc110’s rather than Spc72’s CM1 domain, which unlike Spc72 does not require the TOG domain protein Stu2 for efficient oligomerization of γ-TuSCs. Moreover, Grip71 and the GCP4/5/4/6 core components are neither essential in flies (Reschen et al., 2012; Vogt et al., 2006) nor in Aspergillus or S. pombe (Xiong and Oakley, 2009; Anders et al., 2006), suggesting that there must be ways to assemble ring-like templates in these organisms in the absence of the GCP4/5/4/6 core. We speculate that this “other way” is via CM1-mediated oligomerization of γ-TuSCs (Fig. 6). Nevertheless, Cnn and other CM1 domain proteins can also bind γ-TuRCs formed via the GCP4/5/4/6 core (Muroyama et al., 2016; Choi et al., 2010; Tovey et al., 2021; Wieczorek et al., 2019, 2020) and so it remains unclear whether Cnn recruits γ-TuSCs only in the absence of preformed γ-TuRCs.

In contrast to Cnn, Spd-2 (which does not contain a CM1 domain) requires Grip71 and the GCP4/5/4/6 core to recruit γ-tubulin complexes to mitotic centrosomes i.e., it is unable to bind and recruit γ-TuSCs directly or mediate their recruitment by another protein. Whether Spd-2 binds directly to preformed γ-TuRCs remains unclear. Grip71 associates with preformed γ-TuRCs in the cytosol and the human homolog of Grip71, NEDD1, has been reported to interact with the human homolog of Spd-2, CEP192 (Gomez-Ferreria et al., 2012a). Thus, Spd-2 might recruit γ-TuRCs via binding to Grip71, but since we show that Spd-2 can recruit γ-TuRCs in the absence of Grip71, it must also be able to recruit γ-TuRCs in a different way. Our data show that Grip163GCP6 is more important than Grip75GCP4 in this respect because removing Grip71 and Grip75GCP4 does not completely abolish γ-tubulin accumulation. Perhaps, therefore, Spd-2 recruits γ-TuRCs via an interaction with Grip163GCP6: or alternatively Grip163GCP6, but not Grip75GCP4, is essential for the assembly of preformed γ-TuRCs that are in turn necessary for Spd-2-mediated recruitment. This latter possibility would be consistent with findings in human cells, where depletion of GCP6 is more disruptive to γ-TuRC assembly (Cota et al., 2017). So far, our attempts to identify direct interactions between Spd-2 and γ-TuRC components have failed, so it is possible that an intermediary protein links Spd-2 to γ-TuRCs.

The finding that Cnn and Spd-2 recruit different types of γ-tubulin complexes to centrosomes fits well with recent observations that not all γ-TuRCs within a given species or cell type have the same protein composition. This was shown by analyzing the γ-TuRC protein Mzt1 in Drosophila, Caenorhabditis elegans, fission yeast, and Aspergillus, where Mzt1 is either not present or not necessary at certain MTOCs (Tovey et al., 2018; Gao et al., 2019; Huang et al., 2020; Sallee et al., 2018). For example, we have shown that Drosophila Mzt1 is expressed only in developing sperm cells and is required for γ-TuRC recruitment to basal bodies but not mitochondria (Tovey et al., 2018). Moreover, in mouse epithelial cells, γ-TuRCs are bound and recruited either by CDK5RAP2 (Cnn homolog) or by NEDD1 (Grip71 homolog), and this influences the nucleation and anchoring ability of the γ-TuRCs (Muroyama et al., 2016). Whether other forms of γ-TuRCs also exist and how this affects their function remains to be explored.

In addition to revealing details of centrosomal recruitment of γ-tubulin complexes, we have also shown that microtubules can be nucleated in their apparent absence. We’ve known for some time that microtubules are present within cells after depletion of γ-tubulin or other key γ-TuRC proteins (Hannak et al., 2002; Strome et al., 2001; Sampaio et al., 2001; Sunkel et al., 1995; Tsuchiya and Goshima, 2021; Sallee et al., 2018; Rogers et al., 2008; Nakaoka et al., 2015; Wang et al., 2015; Gunzelmann et al., 2018) and that certain non-centrosomal MTOCs naturally lack γ-tubulin (Nashchekin et al., 2016; Yang and Wildonger, 2020; Zheng et al., 2020; Mukherjee et al., 2020; Kitamura et al., 2010). Mounting evidence, including our work here, suggests that the ch-Tog/XMAP215/Msps/Alp14/Stu2 TOG domain family of proteins (which have microtubule polymerase activity) and the TPX2 family of proteins (which have microtubule stabilization activity) are important for microtubule nucleation. Depletion of TOG domain proteins from Xenopus egg extracts, Drosophila S2, fat body cells, fission yeast cells, and budding yeast cells, and depletion of TPX2 from Xenopus egg extracts severely impairs microtubule nucleation or organization (Popov et al., 2002; Thawani et al., 2018;