A stable sub-complex between GCP4, GCP5 and GCP6 promotes the assembly of γ-tubulin ring complexes

γ-Tubulin is a protein involved in the nucleation of microtubules. It assembles with so-called ‘gamma-tubulin complex proteins’ (GCPs) into multiprotein complexes of two different sizes. A ‘γ-tubulin small complex’ (γTuSC) comprises two molecules of γ-tubulin that are bound by GCPs 2 and 3. A much larger ‘γ-tubulin ring complex’ (γTuRC) is formed by multiple γTuSCs that associate with additional GCPs 4, 5 and 6, and several smaller accessory proteins into a helical structure of 2 MDa (Kollman et al., 2011; Farache et al., 2018). A few eukaryotes, such as Saccharomyces cerevisiae or Candida albicans contain only GCPs 2 and 3. In these organisms, multiple γTuSCs are assembled that form a helix with the help of additional proteins, such as Spc110 or Mzt1 (Kollman et al., 2010; Erlemann et al., 2012; Lyon et al., 2016; Lin et al., 2014, 2016). Most eukaryotes, however, express the full complement of GCPs 2, 3, 4, 5 and 6, and form γTuRCs. It is believed that all these GCPs have similar structures. They are characterized by sequence homology of two conserved regions, the grip1 and grip2 motifs, corresponding to the N-terminal and C-terminal halves of GCP4, the smallest GCP (Gunawardane et al., 2000; Guillet et al., 2011). The crystallographic structure of GCP4 shows that these domains correspond to bundles of α-helices. The other GCPs contain additional specific sequences, mainly at the extreme N-terminus or in the region that links the grip1 and grip2 motifs, as in GCPs 5 and 6 (Guillet et al., 2011; Farache et al., 2016).

Depletion of GCP2 or GCP3 leads to severe spindle abnormalities, and depleted cells are not viable. Depletion of GCP4, 5 or 6 can be tolerated in fission yeast or in somatic cells of Drosophila but not in vertebrates, where removal of either of these GCPs prevents the formation of the γTuRC and provokes spindle defects (Anders et al., 2006; Vérollet et al., 2006; Farache et al., 2016; Cota et al., 2017). Rescue experiments with chimeric proteins containing N-terminal domains fused to C-terminal domains of a different GCP showed that the chimeras rescued the defects as long as they carried the N-terminal domain of the depleted GCP (Farache et al., 2016). Thus, the GCPs are not functionally redundant, despite their structural similarities, and the function of individual GCPs are specified by their N-terminal domains. GCP2 and GCP3 interact laterally through their N-terminal domains in γTuSC helices, whereas γ-tubulin molecules are bound by the C-terminal domains (Kollman et al., 2010). FRET experiments also demonstrated a direct lateral interaction between GCP4 and GCP5 through their N-terminal domains (Farache et al., 2016), suggesting that the N-terminal domains specify lateral binding partners and, thereby, position the GCPs within the γTuRC helix.

In this context, the specific functions of GCPs 4, 5 and 6 in γTuRC assembly need to be investigated. Because these proteins are present only in one or two copies per complex (Murphy et al., 2001; Choi et al., 2010), and because the rescue experiments with chimeras suggest that GCPs 4, 5 and 6 occupy non-random positions, their localization within the complex is of particular interest. During the course of this work, three studies have described the structure of native γTuRCs by cryo-electron microscopy, and have found a lateral association of four γTuSCs, bound to a lateral array of GCP4/GCP5/GCP4/GCP6, to which an additional γTuSC was associated (Consolati et al., 2019 preprint; Liu et al., 2020; Wieczorek et al., 2020). Altogether, this has raised the question whether GCPs 4, 5 and 6 form assembly intermediates equivalent to γTuSCs. In this study, we demonstrate biochemically that GCPs 5 and 6, and two copies of GCP4 together form a stable, KCl-resistant core within the γTuRC, which can be purified and drives the assembly of free γTuSCs into a γTuRC that is competent to nucleate microtubules.

To determine how GCPs 4, 5 and 6 assemble within the γTuRC, and to examine whether γTuSC-like intermediates are formed by these proteins, we destabilized the γTuRC by treating HeLa cytoplasmic extracts with increasing concentrations of KCl. GCPs 4, 5 or 6 were immunoprecipitated and all interacting GCPs were identified by western blotting (Fig. 1A, Fig. S1A). Whereas the full set of GCPs was immunoprecipitated at 100 mM KCl, we observed an increasing loss of γTuSCs from the immunoprecipitate at higher concentrations of KCl. At 500 mM KCl, GCPs 4, 5 and 6 remained the major constituents of the immunoprecipitate, irrespective of the antibody used for precipitation. This indicated that the binding affinities between GCPs 4, 5 and 6 are stronger than their affinities to γTuSCs. Consistently, γTuRCs have previously been found to dissociate and to release γTuSCs in response to treatment with KCl at high concentrations (Moritz et al., 1998; Oegema et al., 1999).

To investigate whether GCPs 4, 5 and 6 are associated in a single sub-complex within the γTuRC, we fractionated cytoplasmic extracts on gradients of 5–40% sucrose containing KCl (100 mM or 500 mM; Fig. 1B, Fig. S1B). We noticed that the cell line used in these experiments (HeLa Flp-In T-REx) contains high levels of GCP4 protein, part of which sedimented independently of γTuRCs in low-density-fractions (Fig. S1B; Farache et al., 2016). To test for mutual binding of GCPs 4, 5 and 6, we performed immunoprecipitation from each individual fraction in the gradient by using antibodies against GCP4 (Fig. 1C) or GCP5 (Fig. 1E, left panel). At 100 mM KCl, γTuSCs and γTuRCs sedimented mainly in fractions 3 and 7, at a mean density of 13% and 28% sucrose, respectively. GCP2 to GCP6 co-immunoprecipitated efficiently from fraction 7 (Fig. 1C, first lane). By contrast, at 500 mM KCl, we observed the disappearance of γTuRCs, and immunoprecipitation revealed that GCPs 4, 5 and 6, and γ-tubulin associated with each other in fractions of intermediate size, excluding GCPs 2 and 3 (Fig. 1C,E; fractions 4 and 5 at 15–23% sucrose). Interestingly, when the concentration of KCl was decreased to 200 mM or 100 mM KCl before gradient sedimentation, the peaks of GCP2 to GCP6 shifted back to higher fractions, and GCPs 4, 5 and 6 re-associated with GCPs 2 and 3 in fractions 6 and 7 at 23–30% sucrose (Fig. 1D,E, Fig. S1C). This suggests that the sub-complex of GCPs 4, 5 and 6 can re-assemble with γTuSCs into γTuRCs.

We then purified the sub-complex comprising GCP4, GCP5 and GCP6 (hereafter referred to as GCP4/5/6) to obtain an estimation of its size and stoichiometry. We constructed a HEK293 cell line expressing GCP6 with a C-terminal GST-hexa-histidine (6his) tag, by modifying the TUBGCP6 gene using CRISPR-Cas9. Consecutive steps of affinity-purification over glutathione sepharose and over Ni-NTA agarose were carried out in the presence of 500 mM KCl. The eluate was analyzed by SDS–PAGE and silver staining, and scanned signals were normalized to the molar lysine content of each protein (Dion and Pomenti, 1983; Fig. 1F). We realized that small amounts of GCP2 or GCP3 are still present in the purified sub-complex, which were probably underestimated in western blotting experiments (Fig. 1F). Quantifications from four independent experiments indicated a molar ratio of two copies of GCP4, one of GCP5 and one of GCP6, with three copies of γ-tubulin and one equivalent of either GCP2 or GCP3. Although it was impossible to distinguish GCP2 and GCP3 by one-dimensional electrophoresis, western blotting proved that both proteins were present within a single band (Fig. 1F,G). Taking this stoichiometry into account, the estimated size of the complex should be ∼750 kDa. Size-exclusion chromatography confirmed elution at a size close to thyroglobulin (∼700 kDa) (Fig. 1G). These results suggest that dissociation by salt (i.e. KCl) produced a heterogeneous mixture, containing a core complex made of GCPs 4, 5 and 6, stochastically associated with either GCP2 or GCP3, and γ-tubulin.

To determine whether GCPs 4, 5 and 6 associate independently of γTuRCs, we used RNA interference (RNAi) to deplete GCP2 (Fig. 2A). Consistent with previously published findings on the co-regulation of GCPs (Vérollet et al., 2006), we noticed that siRNAs against individual GCPs also affected the protein levels of others (GCP2 siRNA also decreased levels of GCP3, and levels of GCP6 were slightly affected by several siRNAs). Most importantly, loss of GCP2 caused the disappearance of γTuRCs in sucrose gradients (Fig. 2B, Fig. S2) and GCPs 4, 5 and 6 immunoprecipitated together with γ-tubulin in intermediate fractions. Here, the sub-complex peaked at a sucrose density of ∼21% (fraction 5), whereas disassembly of γTuRCs at 500 mM KCl yielded the strongest peak at 17% sucrose in fraction 4 (compare Fig. 1C with Fig. 2C). This difference might be due to the loss of interactors at high concentrations of KCl.

Since Wieczorek et al. (2020) and Liu et al. (2020) suggested the existence of γTuSC-like structures, containing complexes GCP4/5 or GCP4/6, we tested how the depletion of individual GCPs 4, 5 or 6 affects the composition of the GCP4/5/6 sub-complex, and whether the proposed γTuSC-like structures do, indeed, exist. We depleted each of these three GCPs individually by using RNAi, which, in each case, resulted in the loss of γTuRCs and the accumulation of γTuSCs, as seen in sucrose gradients (Fig. 2A,B, Fig. S2). Next, we immunoprecipitated the proteins from the gradient fractions, using antibodies against GCP4, GCP5 or GCP6 (Fig. 2D). GCPs 4, 5 and 6 were systematically co-precipitated in the absence of GCP2. GCP4 and GCP6 co-precipitated together even in the absence of GCP5. However, GCP5 bound to GCP4 or GCP6 only in the presence of all three proteins. This suggests a hierarchy of assembly, with a small and stable GCP4/6 complex enabling the association with GCP5.

We tested whether elevated amounts of GCPs 4, 5 and 6, or combinations thereof, promote the sequestration of free γTuSCs and their incorporation into γTuRCs. The Flp-In T-REx cell line allows inducible expression of transgenes in response to treatment with doxycycline (Tighe et al., 2008). This cell line was used to induce high levels of GCP5 and/or GCP6, on the high endogenous background of GCP4 (Fig. S1B). GCP6-inducible clones were generated by stable transfection, whereas overexpression of GCP5 was obtained in response to transient transfection of wild-type or GCP6-inducible cells. This created cells producing an excess of either GCP4 alone, or GCPs 4 and 5, GCPs 4 and 6 or GCPs 4, 5 and 6 (Fig. 3A). Excess of GCPs 4 and 5 did not alter the ratio between γTuSCs and γTuRCs (Fig. 3B,C). By contrast, excess of GCPs 4 and 6 reduced the peak of γTuSCs (fraction 3), and shifted the protein signals of GCPs 2, 3 and γ-tubulin towards high-density fractions (Fig. 3B–D; fractions 4–7, 15–30% sucrose).

Immunoprecipitation of the gradient fractions revealed that this shift was due to the binding of γTuSCs to assemblies of GCPs 4 and 6 (Fig. 3G). This is consistent with the idea of GCPs 4 and 6 forming an intermediate that can bind γTuSC, whereas GCP5 cannot efficiently bind to GCP4 in the absence of GCP6 (Fig. 3F). When an excess of all three GCPs (i.e. GCPs 4, 5 and 6) was created, immunoprecipitation of GCP5 yielded co-precipitation of GCPs 4 and 6 (Fig. 3H), supporting the idea of a hierarchy of assembly, as proposed above, with GCP4/6 as the ultimate core. Interestingly, high amounts of GCPs 4, 5 and 6 resulted in the disappearance of the peak of γTuSCs, and led to an increase of γTuRCs (Fig. 3B–E). These results indicate that the GCP4/5/6 sub-complex binds γTuSCs, thereby leading to their assembly into γTuRCs.

To evaluate the capacity of the GCP4/5/6 sub-complex to promote the assembly of γTuRCs, we loaded the sub-complex onto beads and complemented these beads with γTuSCs, using microtubule nucleation assays and western blotting to assess the presence and activity of reconstituted γTuRCs (Fig. 4A). Dynabeads coupled to anti-GCP5 antibody were used to immunoprecipitate either γTuRC or the GCP4/5/6 sub-complex (using 100 mM or 600 mM KCl, respectively) from a HEK293 cytoplasmic extract (hereafter referred to as γTuRC extract). We also used cells depleted of GCP2 to prepare extracts at 600mM KCl, to ensure that the sub-complex was efficiently stripped of γTuSCs. The obtained beads bound to GCP4/5/6 were washed and incubated at 100 mM KCl with a second extract, prepared from cells depleted of GCPs 4, 5 and 6 (hereafter referred to as γTuSC extract, Fig. 4B). Reconstitution of γTuRCs was monitored by western blot (Fig. 4C), and by measuring the nucleation activity of the beads following incubation with pure tubulin (Fig. 4D–F). Compared with γTuRC beads, GCP4/5/6-beads were lacking GCPs 2 and 3, and were unable to nucleate microtubules. Addition of the γTuSC extract resulted in re-incorporation of γTuSCs, and in recovery of the nucleation capacity of the beads. Used as a control, the γTuSC extract did not show any binding to beads without GCP4/5/6, and these beads failed to nucleate any microtubules. Quantification of the number of microtubules nucleated per bead showed that reconstituted GCP4/5/6-beads with γTuSCs reached up to 78% of the nucleation capacity of the positive control, i.e. beads bound to native γTuRC (Fig. 4F). The difference in nucleation capacity compared to the positive control might reflect a decrease in the number of bound complexes (in particular after GCP2-depletion). Altogether, our results showed that the GCP4/5/6 sub-complex can trigger the assembly of nucleation-competent γTuRCs.

Since GCP6 appears to play a central role in the assembly of the γTuRC, and since this protein is characterized by extensive non-grip sequences in its N-terminal region and in the domain connecting the grip1 and grip2 motifs (Figs 5A and 6A), we constructed deletion mutants, to evaluate the importance of these sequences. The deletion mutants were expressed in HeLa cell lines in an inducible manner, following depletion of endogenous GCP6 by using RNAi. We measured the potential of the mutants to rescue the assembly and function of γTuRCs by quantifying the amount of γTuRCs in sucrose gradients, and the formation of bipolar spindles in mitotic cells. In agreement with published data, depletion of GCP6 induces the loss of γTuRCs, the delocalization of γ-tubulin from centrosomes and from the mitotic spindle, and inhibits the separation of the spindle poles (Bahtz et al., 2012; Farache et al., 2016; Cota et al., 2017). Deletion mutants were designed on the basis of structure prediction. The wide central insertion (residues 675–1501) can be divided into three main amino acid regions. The first region (675–816) is predicted to be a continuum of the helix α11 seen in GCP4, with a potential coiled-coil structure (residues 730–760). The middle region (816–1400) is unstructured, and contains a repeated sequence described to be phosphorylated by Plk4 (residues 1027–1269, Bahtz et al., 2012). The third region (1400–1501) is composed of multiple small helices (Fig. 5A). Combined deletions of these domains were evaluated for rescue of spindle bipolarity (Fig. 5B–D). Depletion of GCP6 in wild-type cells resulted in >70% of mitotic cells with a monopolar spindle. The deletion mutants rescued the bipolarity of spindles, except when the third region (1400–1501) was absent. GCP6Δ675-1400 fully rescued spindle bipolarity, whereas GCP6Δ675-1501 showed no rescue at all (Fig. 5D). Moreover, GCP6Δ675-1400 restored the recruitment of γ-tubulin to the centrosomes and to the mitotic spindle (Fig. 5C), and rescue was also observed in the absence of induction, due to a leak of the promoter (Fig. 5C–E). Sucrose gradient profiles confirmed that γTuRCs were fully recovered when only region 675–1400 was deleted (Fig. 5F,G). These results show that most of the central insertion of GCP6 is not required for γTuRC assembly, except for the third region containing the last 100 amino acids. We immunoprecipitated the GCP4/5/6 complex at 500 mM KCl from cells expressing GCP6Δ675-1400 in low amounts (without addition of doxycycline). This deletion mutant of GCP6 still associated with GCP4 and GCP5, thus, maintaining stable interactions at high concentrations of KCl (Fig. 5G).

To investigate the role of the N-terminal extension of GCP6, it was gradually shortened by cutting between its multiple predicted helices (Fig. 6A). Deletion of the entire extension (GCP6 352-1819) abolished the capacity of GCP6 to rescue spindle bipolarity and prevented the assembly of γTuRCs. All other mutants rescued to various extents with rescue efficiency gradually decreasing with the length of the deletion (Fig. 6B–D; Fig. S3). GCP6 280-1819 generated low amounts of γTuRCs, and complexes of intermediate size accumulated on sucrose gradients. These complexes may represent partially assembled γTuRCs, or unstable γTuRCs that started to disassemble during sedimentation. To evaluate whether GCP6 mutants 280-1819 or 352-1819 affected the interactions with other GCPs of the γTuRC, we overexpressed them and performed immunoprecipitations from gradient fractions as described for Fig. 3. GCP6 280-1819 co-immunoprecipitated with GCP4 and with γTuSCs, to a degree similar to that of wild-type protein, showing that the protein interactions were maintained. By contrast, the majority of GCP6 352-1819 failed to interact with the other GCPs (Fig. 6E,F). We then immunoprecipitated GCP6 280-1819 from cytoplasmic extracts at increasing concentrations of KCl. Although GCPs 2, 3, 4 and 5 efficiently co-precipitated at 100 mM KCl, their interaction was lost at 500 mM KCl. At 200 mM KCl, GCP2 and GCP3 still interacted with wild-type GCP6; however, their interaction with the 280-1819 mutant was strongly reduced (Fig. 6G). Taken together, these results suggest that the GCP6 amino acid region 280–352 is necessary for interactions with GCP4 and γTuSCs, and that the first 280 residues of GCP6 stabilize its interactions with the γTuRC.

In analogy to our deletion experiments on GCP6, we also attempted to delete the N-terminal extension of GCP5. However, due to the unavailability of antibodies recognizing GCP5 outside its N-terminal domain, this experiment was not controllable. We successfully deleted the internal insertion of GCP5 (residues 627–704), and noticed that the resulting mutant was still able to rescue spindle bipolarity and γTuRC assembly in GCP5-depleted cells (data not shown).

We demonstrate that GCPs 4, 5 and 6 form a stable sub-complex that permits the association with γTuSCs into a functional γTuRC. Within the sub-complex, the stoichiometric ratio of these GCPs matches the values found in recently reported structures of native γTuRCs (Consolati et al., 2019 preprint; Liu et al., 2020; Wieczorek et al., 2020). We find that GCPs 2 and 3 are also present in our preparations of the sub-complex, albeit at a molar ratio equivalent to half of γTuSC. We believe that this reflects stochastic binding of remnant γTuSCs after incomplete salt-stripping at 500 mM KCl. More-stringent conditions with even higher concentrations of KCl failed to increase the purity of the sub-complexes but, rather, led to their disassembly (Fig. 1A).

Structural and biochemical work has shown that γ-tubulin complexes in budding yeast are formed by lateral assembly of γTuSCs (Kollman et al., 2010). Lateral association and oligomerization into a helically shaped template are supported by targeting factors, such as Spc110, that promote the recruitment of γTuSCs to the spindle pole body and, thus, couple localization to the assembly into a nucleation-competent complex (Kollman et al., 2010; Lyon et al., 2016; Lin et al., 2014, 2016). Other targeting factors have been identified in different species of yeast, and include Spc72, Mozart1 and Mto1/2 (Lin et al., 2016; Masuda and Toda, 2016; Lynch et al., 2014; Leong et al., 2019). In most eukaryotes, fully assembled, helically shaped complexes are already present as soluble entities in the cytoplasm in the form of γTuRCs. Nevertheless, these soluble γTuRCs remain inactive unless recruited to specific sites of microtubule nucleation, thus, allowing tight spatial and temporal control of microtubule formation (Farache et al., 2018). GCPs 4, 5 and 6 are essential for the assembly and/or stabilization of γTuRCs, as depletion of either component causes a reduction of γTuRCs both at the centrosome and in the cytoplasm (Izumi et al., 2008; Bahtz et al., 2012; Scheidecker et al., 2015; Farache et al., 2016; Cota et al., 2017). Besides GCPs 4, 5 and 6, additional factors might also be needed, such as Mozart1, actin or other proteins that correspond to unassigned densities in cryo-electron microscopy-obtained structures of native γTuRCs (Lin et al., 2016; Cota et al., 2017; Consolati et al., 2019 preprint; Liu et al., 2020; Wieczorek et al., 2020). It is now clear that GCPs 4, 5 and 6 integrate into the helical wall of the γTuRC, and that they are laterally bound to γTuSCs, but their specific role within the γTuRC still remains to be determined (Consolati et al., 2019; Liu et al., 2020; Wieczorek et al., 2020). Our results show that they form a nucleus promoting the stable assembly of the 2 MDa complex. During the formation of γTuRCs, complexes comprising GCPs 4, 5 and 6 may act as building blocks that recruit γTuSCs by lateral association and thereby initiate γTuRC-assembly. This hypothesis is directly supported by our experiments with salt-stripped (KCl-treated) sub-complexes of GCPs 4, 5 and 6 that can drive the formation of functional γTuRCs, after incubation with γTuSC-containing cytoplasm (Fig. 4).

The assembly of γTuRCs mediated by GCPs 4, 5 and 6 may be important for the regulation of its microtubule-nucleation activity: cryo-electron microscopy structures show an asymmetric architecture, incompatible with the geometry of microtubules (Consolati et al., 2019 preprint; Liu et al., 2020; Wieczorek et al., 2020). The presence of GCPs 4, 5 and 6 might, therefore, prevent soluble γTuRCs from acquiring an active conformation, unless bound to additional activating factors, such as Cdk5rap2/Cep215, at designated microtubule-organizing centers. Consistently, the GCP5 subunit was identified in two different conformations at position 10 within the γTuRC, and conformational changes propagated towards positions 11–14 of the γTuRC-helix, suggesting that these regulate the overall structure and the activation of the γTuRC (Wieczorek et al., 2020). In addition, specific non-grip domains in GCP6 might also be involved in the binding to regulatory factors. The large central insertion in GCP6, including nine tandem repeats of 27 amino acids, has been proposed to be regulated by Plk4-dependent phosphorylation (Bahtz et al., 2012). Contrary to this view, our deletion experiments show that the majority of the central insertion – i.e. residues 675–1400 that comprise the tandem repeats – can be removed from GCP6 without impacting the assembly or the activity of the γTuRC because a Δ675-1400 deletion mutant permits the assembly of regular mitotic spindles. Moreover, neither the recruitment of γTuRCs to the centrosome nor to the mitotic spindle were affected by this deletion. Eventually, this GCP6 region might be involved in recruitment to specific non-centrosomal microtubule-organizing centers (Oriolo et al., 2007).

Besides any regulatory role, the GCP4/5/6 sub-complex might contribute to the stabilization of γTuRCs post assembly by preventing the loss of γTuSCs from the helical complex. In particular, the N-terminal extension of GCP5 might be part of a lumenal bridge within the γTuRC and might thereby fulfil a stabilizing role (Wieczorek et al., 2020). Moreover, the N-terminal extension of GCP6 (amino acids 1–352) might correspond, at least in part, to a stabilizing ‘plug’ seen in the cryo-electron microscopy structure of the γTuRC (Wieczorek et al., 2020). In accordance, we observed that progressive shortening of the N-terminal extension of GCP6 destabilizes the γTuRC and renders it more sensitive to treatment with KCl. For example, the deletion of the first 279 amino acids leads to the loss of γTuSCs at 200 mM KCl, and to the loss of GCPs 4 and 5 at 400 mM KCl, whereas wild-type γTuRCs remain stable under these conditions (Fig. 6G). Additional stabilization of the γTuRC might be provided by contacts between the central insertion in GCP6 (amino acids 675–1501) and GCP2 in position 13, since GCP6 deletion mutants that lack this insertion (Δ675-1501) prevent the formation of stable γTuRCs (Fig. 5F; Wieczorek et al., 2020).

Overall, the sub-complex of GCPs 4, 5 and 6 might promote lateral associations with γTuSCs through the grip1 domains of its peripheral constituents GCP4 and GCP6, and might stabilize interactions within a larger γTuRC through the N-terminal extensions of GCPs 5 and 6.

The question has been raised whether γ-tubulin and one copy of GCP4 assemble with either GCP5 or GCP6 into intermediate heterotetramers, such as ‘γ-TuG4/5’ and ‘γ-TuG4/6’, equivalent to γTuSCs (Liu et al., 2020; Wieczorek et al., 2020). Our data do not provide direct evidence for this mechanism, since extraction using KCl or GCP2 depletion yields ‘monolithic’ sub-complexes, in which GCPs 4, 5 and 6 are stably bound to each other. In fact, extraction by using KCl demonstrates that the affinities between the GCPs of this sub-complex are higher than the affinities between γTuSCs. Nevertheless, we have seen that γ-TuG4/6 intermediates form in the absence of GCP5, and that these are sufficient to establish lateral contacts with because excess protein levels of GCPs 4 and 6 are able to immunoprecipitate together with GCPs 2 and 3, without GCP5 (lanes 4, 5 in Fig. 3G). By contrast, GCP5 needs the presence of both GCPs 4 and 6 to form any higher-order complex, suggesting that, in human cells, a γ-TuG4/5 does not exist (Fig. 2). The situation might be different in other species, such as fission yeast, where the GCPs 4 and 5 orthologs Gfh1p and Mod21p, respectively, can be co-immunoprecipitated in the absence of the GCP6 ortholog Alp16 (Anders et al., 2006). In both experimental systems, though, the GCP6 ortholog needs GCP4 to associate with GCP5, suggesting that the spatial arrangement of GCPs 4, 5 and 6 within γ-tubulin complexes is conserved across species. On the basis of our experiments, using depletion or overexpression of individual GCPs (Figs 2 and 3), we propose a hierarchy of assembly, in which GCPs 4 and 6 together enable the recruitment of GCP5 into a stable sub-complex that drives lateral association with γTuSCs into a full-sized γTuRC (Fig. 7). Studies by Cota et al. (2017) and Farache et al. (2016) have shown that depletion of either GCP4 or GCP5 leads to milder defects than depletion of GCP6, confirming a central role of GCP6 in γTuRC-assembly. This raises the possibility that partial or unstable γ-tubulin complexes can still form through lateral interactions between GCP6 and γTuSCs, or that γTuSCs assemble into larger complexes in situ, e.g. at centrosomes or spindle pole bodies, partially stabilized by GCP6 and with restricted nucleation capacity (Vérollet et al., 2006; Xiong and Oakley, 2009; Masuda and Toda, 2016).

To gain further insights into the assembly of γTuRCs, it would be of interest to perform controlled reconstitution experiments of the multiprotein complex using purified recombinant components.

HeLa Flp-In T-REx (a gift from Stephen Taylor, University of Manchester, UK; Tighe et al., 2008), and HEK293 FT (Invitrogen, Carlsbad, CA) cells were grown in Dulbecco's modified Eagle's medium (DMEM), supplemented with 10% fetal bovine serum at 37°C, under 5% CO₂. Cells were authenticated and tested for contamination. For CRISPR/Cas9 targeting of GCP6 in HEK293 FT cells, a target guide RNA overlapping the stop codon (5′-AACTACTACCAGGACGCCTG-3′, computed from http://crispr.mit.edu/) was inserted into pSpCas9(BB)-2A-Puro (Addgene plasmid #48139). The homologous recombination donor was a DNA fragment consisting of a 1445 bp left homology arm (GCP6 exon 20–25), a 2244 bp insertion sequence (i.e. GST-6his coding sequence in-frame with exon 25 of GCP6, followed by a puromycin-resistance cassette), and a 1577 bp right-homology arm (GCP6 exon 25 and 3′UTR sequence). Cells were transfected using Lipofectamine 2000 (Invitrogen, Carlsbad, CA). Limiting dilution cloning was performed 3 days post transfection, in conditioned medium supplemented with 2 μg/ml puromycin. Targeted clones were identified by PCR, and verified by sequence analysis and by western blotting using an anti-GST antibody (Roche Applied Science, Basel, Switzerland).

HeLa Flp-In T-REx overexpressing cell lines were obtained as described (Farache et al., 2016). GCP5 and GCP6 (resistant to siRNA) were expressed from pCDNA5/FRT/TO (Invitrogen). Internal deletions of GCP6 were constructed using the Gibson kit (NEB, Ipswich, MA), whereas N-terminal deletions were generated by PCR (using restriction sites BamH1 and Ale1 in the sequences of the vector and the GCP6 DNA, respectively). Briefly, GCP6 constructs were co-transfected with pOG44 (Invitrogen) expressing the Flp recombinase by using CaCl₂. Resistant clones at 200 μg/ml hygromycin B were picked and expanded to obtain clonal cell lines. The phenotypes of at least two independent clones were compared for each construct. Transient transfection of the GCP5 construct was performed using Lipofectamine 2000. Transgene expression was induced using 1 μg/ml doxycycline. For RNA interference (RNAi), 10 nM siRNA were transfected into HeLa Flp-In T-REx cells, using Lipofectamine RNAi max (Invitrogen; Farache et al., 2016). After 24 h, medium was replaced with medium containing or not containing doxycycline. Cells were harvested or fixed 72 h post transfection.

Cells homozygous for GCP6-GST-6his were harvested by trypsinization, rinsed with PBS and stored at −80°C. Pellets were resuspended in 50 mM HEPES-KOH pH 7.2, containing 1 mM MgCl₂ and 1 mM EGTA (HB) plus 100 mM KCl (HB100), and supplemented with 0.1 mM GTP, 1 mM DTT, 1 mM PMSF and complete protease inhibitor cocktail (Roche, Basel, CH). Cells were disrupted by sonication and centrifuged for 30 min at 30,000 g. Protein precipitation from the supernatant was performed by adding 30% of a saturated (NH₄)₂SO₄ solution, followed by solubilization in HB plus 500 mM KCl (HB500), supplemented with 1 mM DTT, 1 mM PMSF, protease inhibitors, 15 mM imidazole and 0.05% IGEPAL CA-630 detergent. The solution was added to glutathione Sepharose 4B (GE Healthcare, Chicago, IL) and incubated 4 h at 4°C under agitation. Beads were washed twice with HB500 containing supplements, and twice with the same buffer without PMSF and protease inhibitors. Elution was performed with the same buffer containing 40 mM reduced glutathione, pH 7.2. Buffer was exchanged by passing the eluate through a desalting PD-10 column (GE Healthcare, Chicago, IL), equilibrated with 50 mM HEPES-KOH pH 7.2, containing 1 mM MgCl₂, 500 mM KCl, 0.5 mM DTT and 0.05% IGEPAL CA-630. The solution was then incubated overnight with Ni-NTA agarose (Qiagen, Hilden, Germany) at 4°C, washed twice with the same equilibration buffer, and twice with equilibration buffer supplemented with 15 mM imidazole. Elution was performed using 200 mM imidazole.

Eluted proteins were analyzed on western blots or polyacrylamide gels and stained using the PlusOne silver staining kit (GE Healthcare, Chicago, IL). Size-exclusion chromatography was performed on Superdex™ 200 Increase 5/150 GL columns (GE Healthcare, Chicago, IL), using thyroglobulin and ferritin as markers for calibration.

Sucrose gradients were performed as described by Farache et al. (2016) in HB100 or by using varying concentrations of KCl.

Cytoplasmic lysates were produced from trypsinized HeLa Flp-In T-Rex cells, lysed in HB100 supplemented with 1 mM GTP, 1 mM DTT, 1 mM PMSF, protease inhibitors, 1% IGEPAL CA-630 and 10% glycerol. After 5 min on ice, cells were centrifuged for 10 min at 16,000 g. Aliquots of the supernatant containing 500 μg of protein were diluted in 100 μl HB100 containing supplements. These cytoplasmic lysates where then supplemented with KCl to increase the concentration as needed, before incubation with 1–2 μg of anti-GCP4, anti-GCP5 or anti-GCP6 antibodies for 2 h at 4°C. A50 μl aliquot of protein A-dynabeads (Invitrogen) was added for 1 h at 4°C, and proteins were eluted in gel-loading buffer, after two washes with HB100 containing 10% glycerol.

Immunoprecipitation from the gradient fractions was performed as above, i.e. by addition of antibodies to the fractions and incubation for 2 h at 4°C, followed by incubation with protein A-dynabeads for 1 h. The beads were then washed twice in HB100 containing 10% glycerol and samples were eluted in gel-loading buffer.

Dynabeads were washed and incubated with the anti-GCP5 antibody in PBS, containing 0.02% Tween-20, for 2.5 h at 4°C. After three additional washes, HEK293 FT cytoplasmic extracts (γTuRC extracts) were added to the beads and incubated for 2 h at 4°C. Extracts were prepared as described in the immunoprecipitation protocol, in the presence of 100 or 600 mM KCl. For each reaction, we used cells grown to confluence on a 100-mm dish (yielding 4 mg of protein) that were lysed in 100 μl buffer per 5 μl beads. After three washes with HB100+10% glycerol, a freshly prepared cytoplasmic extract depleted of GCPs 4, 5 and 6 (γTuSC extract) was added to the beads. Again, cells from one 100 mm dish were lysed in 100 μl HB100 containing supplements. After a 2 h incubation at 4°C, beads were washed three times with Brinkley renaturing buffer (BRB80; 80 mM PIPES pH 6.9, 1 mM EGTA, 1 mM MgCl₂) and resuspended in 12.5 μl BRB80. Nucleation was tested by incubating 2.5 μl beads for 3 min at 37°C with a solution containing 1 μl tubulin at 10 mg/ml, 1 μl carboxytetramethylrhodamine (TAMRA)-labeled tubulin (a gift from Nathalie Morin, Centre de Recherche en Biologie cellulaire de Montpellier, Montpellier, France) at 2 mg/ml, and 0.5 μl 10 mM GTP (modified from Choi and Qi, 2014). The reaction was stopped by adding 45 μl of 1% glutaraldehyde in BRB80, for 5 min at 37°C. Beads were diluted in BRB80 and layered onto a cushion of 10% glycerol in BRB80, centrifuged onto coverslips and mounted in Vectashield solution (Vector Laboratories, Peterborough, UK). Proteins were eluted from the remaining beads with gel-loading buffer and analyzed by western blotting.

Proteins were detected using an Odyssey imaging system (Li-cor Biosciences, Lincoln, NE) according to the manufacturer's protocol, with IRDye 800CW- and 680CW-conjugated secondary antibodies (Invitrogen). Protein levels were quantified using Odyssey 2.1 software. The Odyssey fluorescence system provides a linear relationship between signal intensity and antigen loading. Band intensities were measured after background subtraction. For the quantification of γ-tubulin in fractions of sucrose gradients, the band intensities of individual fractions were normalized to the total amount of γ-tubulin in the experiment, corresponding to the sum of the intensities of all fractions of the sucrose gradient.

Cells grown on coverslips were fixed in methanol at −20°C, and processed for immunofluorescence, following standard protocols. Fluorescence microscopy was performed on a wide-field microscope (Axiovert 200M; Carl Zeiss, Oberkochen, Germany) equipped with a Z motor, using a 63× (Plan Apo, 1.4 NA) objective. Images were acquired with an MRm camera and Axiovision software. Images of microtubule-nucleating beads were obtained with a LSM 710 confocal microscope (Carl Zeiss), using a 63× (Apo, 1.4 NA) objective, and an excitation wavelength of 561 nm. Images shown in Figs 4 and 5 correspond to single sections.

Image processing was performed using Adobe Photoshop. Identical settings of exposure and contrast were used for corresponding experiments and controls.

Rabbit anti-GCP6 (for immunoprecipitation, Abcam, Cambridge, UK); mouse anti-γ-tubulin TU-30 (for western blotting 1:5000, Exbio, Vestec, CZ); rabbit anti-GCP4 (for western blotting 1:1000; Fava et al., 1999); rabbit anti-GCP2 (for western blotting 1:700; Haren et al., 2006); rabbit anti-γ-tubulin R75 (for immunofluorescence 1:1000; Julian et al., 1993); mouse anti-GCP3 C3 (for western blotting 1:700), mouse anti-GCP5 E1 (for immunoprecipitation), rabbit anti-GCP5 H300 (for western blotting 1:500), mouse anti-GCP6 H9 (for western blotting, 1:1000, Santa Cruz Biotechnology, Santa Cruz, CA); mouse anti-α-tubulin (for immunofluorescence 1:1000, Sigma Aldrich, St Louis, MO). Quantities of antibodies used for immunoprecipitation were as reported in the Immunoprecipitation section above.

We thank Valérie Guillet (Toulouse) for help with size exclusion chromatography, Stephen Taylor (Faculty of Life Sciences, University of Manchester, Manchester, UK) for the gift of HeLa Flp-in T-Rex, Nathalie Morin (Centre de Recherche en Biologie cellulaire de Montpellier, Montpellier, France) for the gift of TAMRA-tubulin and Jens Lüders (Institute for Research in Biomedicine, Barcelona, Spain) for fruitful discussions during the course of this work. We thank all our colleagues at the Université Paul Sabatier for critical discussion and for technical help. The project was in part supported by grant 13-BSV8-0007-01 from Agence Nationale de la Recherche, and by grant SFI20121205511 from Fondation ARC pour la recherche sur le cancer.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.244368.reviewer-comments.pdf