Functional Analysis of γ-Tubulin Complex Proteins Indicates Specific Lateral Association via Their N-terminal Domains*

Microtubules are nucleated from multiprotein complexes containing γ-tubulin and associated γ-tubulin complex proteins (GCPs). Small complexes (γTuSCs) comprise two molecules of γ-tubulin bound to the C-terminal domains of GCP2 and GCP3. γTuSCs associate laterally into helical structures, providing a structural template for microtubule nucleation. In most eukaryotes γTuSCs associate with additional GCPs (4, 5, and 6) to form the core of the so-called γ-tubulin ring complex (γTuRC). GCPs 2–6 constitute a family of homologous proteins. Previous structural analysis and modeling of GCPs suggest that all family members can potentially integrate into the helical structure. Here we provide experimental evidence for this model. Using chimeric proteins in which the N- and C-terminal domains of different GCPs are swapped, we show that the N-terminal domains define the functional identity of GCPs, whereas the C-terminal domains are exchangeable. FLIM-FRET experiments indicate that GCP4 and GCP5 associate laterally within the complex, and their interaction is mediated by their N-terminal domains as previously shown for γTuSCs. Our results suggest that all GCPs are incorporated into the helix via lateral interactions between their N-terminal domains, whereas the C-terminal domains mediate longitudinal interactions with γ-tubulin. Moreover, we show that binding to γ-tubulin is not essential for integrating into the helical complex.

In all eukaryotes, microtubules are nucleated from specialized multiprotein complexes containing ␥-tubulin and associated proteins (1)(2)(3). These complexes resemble small rings by electron microscopy and are thus called ␥-tubulin ring complexes (␥TuRCs) 4 (4 -7). Closer inspection revealed that these ␥TuRCs are helices of one turn, with the two ends overlapping. They are ubiquitous and essential for viability: the growth of new microtubules is crucial to drive mitotic spindle formation and cell division.
␥TuRCs are mainly composed of ␥-tubulin and of proteins of the GCP (␥-tubulin complex protein) family. GCPs are characterized by sequence homology in two specific regions, also referred to as the grip1 and grip2 motifs (8). Five members of this family are known: GCPs 2, 3, 4, 5, and 6. GCPs 2 and 3 associate with ␥-tubulin to form a V-shaped subcomplex, called ␥-tubulin small complex (␥TuSC). GCPs 2 and 3 constitute the arms of the V, interacting laterally via their N-terminal domains (Fig. 1, A and B). The C-terminal domains are located at the two tips of the V, each binding one molecule of ␥-tubulin (9). In the budding yeast Saccharomyces cerevisiae, ␥TuSCs are directly recruited to the spindle pole body (yeast centrosome equivalent) by the protein Spc110. Oligomers of Spc110 interact with the basis of the V-shaped ␥TuSCs and stabilize their lateral association (10 -13). Likely, seven ␥TuSCs assemble stepwise into a helix of one turn plus a small overlap (14). In this helical array, the ␥-tubulin molecules are exposed to form a platform from which ␣/␤-tubulin dimers assemble into protofilaments. The structure of these helical complexes predefines the geometry of the microtubule, with a diameter of 25 nm and with 13 ␥-tubulin molecules per turn. S. cerevisiae represents the simplest model organism for microtubule nucleation, building ␥-tubulin complexes from ␥TuSCs in the absence of GCPs 4, 5, and 6. In most other eukaryotes, however, GCPs 4, 5, and 6 are an integral part of ␥TuRCs, albeit only one or very few copies of each protein are associated with each ring (7,8,(15)(16)(17)(18)(19). In human cells, preformed, inactive ␥TuRCs are in a soluble pool in the cytoplasm from which they are recruited to nucleation centers followed by their activation. Depletion of GCPs 4, 5 or 6 prevents the formation of stable, full-sized ␥TuRCs and prevents the targeting of ␥-tubulin to the centrosome and to spindle microtubules (20 -22). This causes errors in centriole duplication and spindle assembly, often inducing monopolar spindles (21). These phenotypes resemble defects from impaired recruitment or function of ␥-tubulin (23)(24)(25).
The role of GCPs 4, 5, and 6 in ␥TuRC assembly is not well understood. Original observations by electron microscopy suggested that they could form a scaffolding cap at the base of the ␥TuSC helix (6,7,26). More recent structural data have challenged this model. Resolution of the atomic structure of GCP4 has revealed a remarkable conservation with the structures of GCP2 and GCP3 of S. cerevisiae. Cryo-EM reconstructions of ␥TuSC helices and in silico modeling has shown that the atomic structure of GCP4 can be fitted into the helical wall in the place of GCPs 2 or 3 (11,27,28). In addition, biochemical experiments showed that GCP4 interacts with ␥-tubulin via its C-terminal domain, as previously shown for GCP2 and GCP3 (9,10,27). Drosophila GCPs 5 and 6 were also found to bind ␥-tubulin (29). These results led to the prediction that GCPs 4, 5, and 6 integrate into ␥TuRCs in an equivalent manner to GCPs 2 and 3. In this model, all GCPs would interact laterally via their N-terminal domains. Longitudinal interactions would bind ␥-tubulin molecules to the C-terminal domains of each GCP (1). To test this idea experimentally, we have compared the function of the N-and C-terminal domains of GCPs 4, 5, and 6 to the equivalent domains of GCPs 2 and 3, and we have performed FLIM-FRET analysis to determine the mode of interaction between GCPs 4 and 5. We have found that the N-terminal domains define the identity of the GCP, driving lateral interactions similar to the interactions found in ␥TuSCs. This is the first experimental approach supporting that GCPs 4, 5, and 6 are part of the helical wall of the ␥TuRC.

Results
The Particular Role of Each GCP Is Specified by Its N-terminal Domain-The atomic structure of GCP4 shows that the protein contains two independent domains corresponding to the N-terminal and C-terminal halves (Fig. 1A). All GCP proteins possess structural similarities in their N-and C-terminal domains, highlighted in dark blue and red in Fig. 1B (27). Nevertheless, depletion of any individual GCP from cells reduces the amount of ␥TuRCs and leads to mitotic defects, suggesting that the various GCPs are not functionally redundant (19 -21). To investigate potential functional differences between the protein domains of individual GCPs, we developed an assay involving RNA silencing of specific GCPs and rescuing function by the expression of siRNA-resistant chimeric proteins. Our experimental setup involved stable transfection of GCP constructs into HeLa cells, taking advantage of the Flip-in T-Rex system that allows site-specific integration and inducible expression of the transgenes. To validate the system, we first verified that expression of WT GCP4 was able to rescue the depletion of the endogenous protein. We transfected an siRNA targeting GCP4 with induction of a siRNA-resistant construct at levels similar to the endogenous protein (Fig. 1C). Under these conditions, we measured the amount of soluble ␥TuRCs and their recruitment to spindles in mitosis using ␥-tubulin as a marker (Fig. 1, D-F). In sucrose gradients, the percentage of ␥-tubulin sedimenting as ␥TuRCs was quantified relative to total ␥-tubulin levels (because the siRNA treatments delayed cell growth, total amounts of proteins varied depending on the condition) (Fig. 1, D and E). Depletion of GCP4 drastically reduced the amount of ␥TuRCs, whereas concomitant induction of the siRNA-resistant GCP4 restored the full-sized complexes. Depletion also reduced the recruitment of ␥-tubulin to mitotic centrosomes and to spindle microtubules, which was rescued by siRNA-resistant GCP4 (Fig. 1F).
In a second step cells expressing chimeric proteins were analyzed. In these proteins, the N-terminal or the C-terminal domain of GCP4 was replaced by the corresponding domain of another GCP, i.e. 2, 3, 5, or 6 ( Fig. 2). The atomic structure of GCP4 shows that the two domains are connected at a central "hinge," located between helices ␣11 and ␣12 (Fig. 1, A and B). The location of the hinge in different GCPs was determined by multiple sequence alignments (Fig. 1B) and was chosen as a site of recombination for the construction of chimeras. Sucrose gradient analysis showed that all chimeras carrying the N-terminal domain of GCP4 were able to rescue ␥TuRC formation after depletion of the endogenous GCP4 irrespective of the origin of the C-terminal domain to which they were fused ( Fig.  2A). In contrast, chimeras with the C-terminal domain of GCP4 fused to the N-terminal domain of any other GCP failed to rescue ␥TuRC formation after depletion of GCP4 (Fig. 2B). In general, the chimeras were able to rescue only if they carried the N-terminal domain identical to the depleted endogenous GCP (Fig. 2, A and B). Quantification of ␥-tubulin levels showed that rescue-competent chimeras restored nearly fully the original amounts of ␥TuRCs (Fig. 2C). They also rescued the recruitment of ␥-tubulin in mitosis (Fig. 3A). The results are summarized in Fig. 3B. Occasionally, chimeras were expressed at very low levels, leading to ambiguous answers (marked with ? in Fig.  3B; expression levels are presented in Fig. 4). In particular, insufficient expression might explain why the chimera carrying the N-terminal domain of GCP3 does not rescue the depletion of GCP3. Alternatively, GCP3 might behave differently from the other GCPs and require its own C-terminal domain to function. Successful rescue also diminished the amount of aberrant, monopolar spindles (Fig. 3C), suggesting that not only the structure but also the function of ␥TuRCs was restored. The results show that the N-terminal domains define the functional identity of the different GCPs, whereas the C-terminal domains are exchangeable.
GCP4 and GCP5 Are Associated Laterally via Their N-terminal Domains-Whereas lateral interaction between the N-terminal domains of GCP2 and GCP3 has been shown previously (9,10,30), experimental evidence is missing on how GCPs 4, 5, and 6 bind to the remainder of the ␥TuRC. Modeling of the GCP4 structure into reconstructions of ␥TuSCs from cryoelectron microscopy has strongly suggested that GCP4 is capable of interacting laterally with other GCPs, equivalent to GCPs 2 and 3. However, this has never been verified by any biochemical or biophysical experiments. To determine whether GCP4 is involved in lateral interactions, we performed FLIM-FRET analysis. Co-immunoprecipitation experiments in Schizosaccharomyces pombe have suggested that GCP4 and GCP5 can bind to each other independently of ␥TuSCs (17). We co-expressed GCP4 and GCP5 tagged with GFP and mCherry and looked for energy transfer between the fluorescent proteins incorporated into ␥TuRCs at centrosomes. If the two proteins interact, the transfer of energy from the donor (GFP) to the acceptor (mCherry) decreases the fluorescence lifetime of the donor. The relative difference of lifetime is a measure of FRET efficiency and depends on the distance between the donor and the acceptor. Initial attempts of expressing fluorescent GCPs in cell lines yielded centrosomal fluorescence that was too weak ␥-Tubulin Complex Proteins and ␥-TuRC Assembly OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44 for measuring fluorescence lifetimes. To circumvent this problem, we used a previously established cell line stably overexpressing ninein-like protein (Nlp) (31). Increased amounts of Nlp form large centrosomal assemblies in interphase that enrich ␥TuRCs and thereby boost microtubule nucleation. We verified by immunofluorescence that these assemblies recruit all different endogenous GCPs (Fig. 5A). Moreover, depletion of GCP4 or GCP5 blocked the recruitment of the other GCPs to these centrosomal assemblies, indicating that GCPs are incorporated only in the form of ␥TuRCs ( Fig. 5B and data not shown). The sequences at the chimeric junction are shown relative to GCP4 helices ␣11 and ␣12. C, Western blot analysis of GCPs expression in HeLa cells treated with or without GCP4 siRNA with or without induction of the rescue construct resistant to the siRNA (GCP4 siR). D, comparison of extracts from control cells and cells treated with GCP4 siRNA with or without induction of GCP4 siR after fractionation in gradients of 5-40% sucrose. ␥TuRC sedimentation is visualized using a ␥-tubulin antibody; fraction 3 contains the majority of ␥TuSCs, fraction 7 contains the majority of ␥TuRCs. WB, Western blot. E, quantification of ␥-tubulin intensities in the fractions of the gels shown in D, expressed as a percentage of total ␥-tubulin (sum of the 10 fractions). The values in fraction 7 are shown on the right, relative to control cells. F, comparison of mitosis in control cells or siRNA-treated cells with or without induction, stained for ␥-tubulin (top panels and red), microtubules (green), and DNA (blue). Scale bar, 10 m. Top right, quantification of ␥-tubulin intensities at centrosomes and along spindle microtubules (Ͼ25 cells scored per condition; error bars, mean Ϯ S.D.). Bottom right, percentage of cells with defective recruitment of ␥-tubulin (mean of five experiments, 500 total cells scored per condition; error bars, S.E., 95% confidence interval). ns, not significant; ****, p Ͻ 0.0001.
Our approach was designed to monitor FRET only between adjacent proteins within ␥TuRCs. Using the fitting models of GCP4 structure into the ␥TuSC helix, we predicted the distances between the extremities of the GCPs (11,28). Because of the V-shape of ␥TuSCs, the N termini of adjacent GCP4 molecules were found in close contact (Ͻ2.5 nm), but the C termini were separated by an average of 6 nm. The distances between non-adjacent C termini exceeded 11 nm. As energy transfer is only possible if the distance between the fluorophores is below 10 nm, fluorescent tags positioned C-terminally should restrict FRET to direct neighbors. Modeling also predicted that the C termini extend outside the helix, away from the lateral surfaces of the GCPs, suggesting that the addition of the fluorescent proteins should not interfere with protein-protein interactions ␥-Tubulin Complex Proteins and ␥-TuRC Assembly OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44 and should not inhibit the assembly of the complex. Indeed, the tagged GCPs were recruited to the centrosomal Nlp assemblies, and their presence did not alter microtubule nucleation, indicating that they were functional (Fig. 5C).
We first used GCP2 and GCP3 as controls. We fused GFP to the C terminus of GCP3 and mCherry to the C termini of GCP2 and GCP3. Centrosomal assemblies containing GCP3-GFP displayed an average lifetime of 2.414 ns. When co-expressed with GCP2-mCherry, the lifetime decreased to 1.981 ns, but only to 2.293 ns with GCP3-mCherry (Table 1, Fig. 5, D-F). This corresponds to FRET efficiencies of 18% for the GCP3/GCP2 pair and 5% for the GCP3/GCP3 pair. For comparison, the use of free mCherry as an acceptor led to 3.7% that of FRET efficiency (Table 1). Only values Ͼ7% are considered significant (see "Experimental Procedures" for details). These results are consistent with the existing model of ␥TuSC helices, with GCP3 in direct contact with GCP2 but with distances between GCP3 C termini exceeding the radius for efficient FRET (Fig. 5, E and F).
Using GCP4-GFP paired with GCP5-mCherry or GCP5-GFP paired with GCP4-mCherry, we obtained FRET efficiencies between 8 and 9% (Table 1, Fig. 6, A and B). In contrast, pairing GCP4-GFP with GCP4-mCherry or GCP5-GFP with GCP5-mCherry led to Ͻ3% of FRET efficiency (Table 1, Fig. 6C). These values are consistent with a close and specific proximity of the C termini of GCP4 and GCP5. Note that FRET is more efficient with the GCP2/GCP3 pair than with the GCP4/GCP5 pair, because GCP3-GFP has the potential to transfer energy simultaneously to two molecules of GCP2-mCherry in the helix, which potentiates the energy transfer (Fig. 5E). GCP4 can only have one GCP5 neighbor because there is only one copy of GCP5 per complex (16,19). To verify that the interaction revealed by FRET is direct as expected, we performed crosslinking experiments in vitro on purified ␥TuRC. Despite the large size and low abundance of the complex, we were able to reveal by Western blot a cross-linked product containing both proteins at the size of the heterodimer (Fig. 6D). This confirms that GCP4 and GCP5 are direct neighbors within the complex.
We then used GCP4 and GCP5 fluorescently tagged at their N termini. Again, we measured significant FRET (9.5-12%), in contrast to pairs of one protein tagged at its C terminus and the FIGURE 3. Chimeric GCPs rescue ␥TuRC recruitment and function, in addition to ␥TuRC assembly. A, percentage of cells with defective recruitment of ␥-tubulin (mean of three experiments, 300 total cells scored per condition). B, summary of the function of the chimeras in ␥TuRC assembly as well as recruitment in mitosis (ϩ, rescue; Ϫ, no rescue; ?, ambiguous; gray squares, not evaluated). C, comparison of mitosis in control cells or siRNA-treated cells with induction of the chimeras, stained for ␥-tubulin (top panels and red), microtubules (green), and DNA (blue). Scale bar, 10 m. The graphs show the percentage of monopolar spindles. Results are shown for the N42C/N24C and N46C/N64C chimeras treated with siRNAs against GCP2 and GCP6, respectively (mean of three experiments, 300 total cells scored per condition). Because the yield of monopolar spindles is lower after GCP5 siRNA, results for the N45C/N54C chimeras are not shown, nor are they for the N43C/N34C chimeras that are not rescue-competent. Error bars, S.E., 95% confidence interval.

␥-Tubulin Complex Proteins and ␥-TuRC Assembly
other at the N terminus (0.5-3.5%) ( Table 1, Fig. 7, A and B). The latter arrangement should not permit FRET if GCPs 4 and 5 were associated in a ␥TuSC-like manner, in which the N terminus of one GCP is separated by an average distance of 12 nm from the C terminus of the neighboring GCP. Altogether, our combined results provide direct evidence for lateral interaction between GCP4 and GCP5, equivalent to GCPs 2 and 3 in ␥TuSCs. To test whether the GCP4-GCP5 interaction requires the specific N-terminal domain of each partner, we used the chimeric proteins with the N-and C-terminal domains of GCP4 and GCP5 swapped, tagged C-terminally with GFP. We measured significant FRET between the donor chimera carrying the N-terminal domain of GCP4 and GCP5-mCherry as an acceptor (9%) but not with GCP4-mCherry (4%) ( Table 1, Fig. 7, C and D). Likewise, we measured significant FRET using the chimera carrying the N-terminal domain of GCP5 paired with GCP4-mCherry (8%) but not with GCP5-mCherry (2%) ( Table  1, Fig. 7, E and F). These results confirm the role of the N-terminal domains in the interaction between GCP4 and GCP5.
Interaction of GCP4 with ␥-Tubulin Is Not Necessary for Incorporation into the ␥TuRC-Our results show that the C-terminal domains of the GCPs are exchangeable, suggesting that they perform the same function, i.e. binding to ␥-tubulin. We wanted to address the significance of the GCP/␥-tubulin interaction in ␥TuRC assembly. For this, we mutated the ␥-tubulin binding region of GCP4 and evaluated its function in ␥TuRC assembly and recruitment along spindles. Based on the models of the GCP4 structure fitted into the reconstruction of ␥TuSC helices, we predicted the regions of GCP4 involved in the interaction with ␥-tubulin (27,28,32). Helices ␣16 and ␣20 of the C-terminal domain of GCP4 were found at the protein interface with ␥-tubulin (Fig. 8A). We mutated both helices and evaluated the binding of the mutants to ␥-tubulin in vitro by isolating stable complexes of FLAG-tagged ␥-tubulin and V5-tagged GCP4 on anti-FLAG affinity beads. Multiple mutations in helix ␣16 inhibited binding of GCP4 to ␥-tubulin, whereas mutating helix ␣20 showed no effect (Fig. 8, A and B).
Refining the mutations revealed that mutating residue Phe-522 of ␣16 to Arg reduced the binding and that adding the F626R mutation in ␣20 was fully inhibitory (Fig. 8, A and B).
We tested the ability of the double mutant F522R/F626R to complement the depletion of the endogenous GCP4. A stable cell line expressing the siRNA-resistant GCP4 F522R/F626R still assembled ␥TuRCs after siRNA treatment, as shown after sucrose gradient sedimentation, and ␥-tubulin was still recruited to the mitotic spindle (Fig. 8, C-E, upper panels). In contrast, deletion of the whole C-terminal domain of GCP4 was unable to rescue (Fig. 8, C-E, lower panels). We observed that GCP4 was present in low density fractions of the sucrose gradients, independent of ␥TuRCs (Fig. 8F). Immunoprecipitation of GCP4 from these fractions indicated a specific interaction with ␥-tubulin (Fig. 8G). When WT GCP4 was replaced by the double mutant F522R/F626R, the interaction was lost (Fig. 8H), confirming that the F522R/F626R mutations inhibited the binding of GCP4 to ␥-tubulin not only in vitro but also in cells where ␥TuRC assembly is rescued. Moreover, the lateral association with GCP5 was not altered, as controlled by FLIM-FRET (Table 1). These results indicate that the C-terminal domain of GCP4 is needed for incorporation into the ␥TuRC, but not the interaction with ␥-tubulin.

Discussion
In this work we provide the first experimental evidence supporting the prediction that all GCPs assemble into the helical wall of the ␥-TuRC. Using chimeric proteins formed by the N-terminal domain of one GCP fused to the C-terminal domain of another GCP, we show that the N-terminal domains determine the identity of the GCPs. Moreover, our FLIM-FRET experiments indicate that GCP4 and GCP5 establish ␥TuSClike lateral interactions within the complex. Previous studies in S. pombe have suggested an interaction between GCP4 and GCP5, independent of the presence of GCP2 and GCP3 (17), and our cross-linking results confirm a direct association of GCP4 and GCP5 within ␥TuRCs. Using chimeric proteins with the N-or C-terminal domains of GCP4 and 5 swapped, we show that the N-terminal domains carry the specificity of the interaction. Altogether, our results suggest that the N-terminal Western blot analysis of extracts from the stable cell lines treated with or without siRNA against GCPs 2, 3, 4, 5, or 6 with or without induction of the chimera. The levels of induction were grossly similar to the levels of the endogenous proteins, except for the chimeras N34C and N45C that were barely detectable after treatment with GCP3 and GCP5 siRNA respectively (*). In contrast, induction of the chimera N64C was around 15ϫ more than the endogenous GCP6 (**). OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44 domains of all GCPs are mediating specific lateral interactions. Thereby, they define the direct neighbors and position the GCPs within the ␥TuRC helix.

␥-Tubulin Complex Proteins and ␥-TuRC Assembly
In addition to the conserved sequences, including the grip1 motif, the N-terminal domains carry specific insertions of various sizes depending on the GCP, i.e. internal insertions or N-terminal extensions (27). These insertions may equally contribute to the function of individual GCPs as they have been implied in specific interactions with regulatory or structural proteins. For instance, GCP6 carries a large internal insertion phosphorylated by Plk4 (21) and containing a domain of interaction with keratins (33), whereas the N-terminal extension of GCP3 interacts with the recruitment protein MOZART1 (34,35).
The N-terminal domains can be fused to C-terminal domains of other GCPs without affecting their function and specificity. This was shown for the N-terminal domains of GCPs 2, 4, 5, and 6 but remains uncertain for GCP3. The poor induction of GCP3 chimera did not allow us to test unambiguously whether it rescues the depletion of the endogenous GCP3. Either the expression level was insufficient for rescuing, or the C-terminal domain of GCP3 was irreplaceable. One difference between GCP3 and the other GCPs is the flexibility of the GCP3 hinge linking the N-and C-terminal domains. This flexibility is needed to close the ␥TuRC in a conformation compatible with the geometry of the microtubule, increasing its nucleation capacity (11). In the chimera composed of the N-terminal domain of GCP3 and the C-terminal domain of GCP4, the flexibility or the positioning of the C-terminal domain might be altered and impact the assembly of the complex. The observation that the C-terminal domains of all other GCPs can be switched fits with the structural data, which predict an interaction with ␥-tubulin via the conserved grip2 motif (27,32,36). They could thus be exchangeable if the GCPs were all binding similarly to ␥-tubulin. Surprisingly, we have found that a mutant of GCP4 lacking affinity to ␥-tubulin was still able to incorporate into ␥TuRCs and to rescue their assembly. Thus binding to ␥-tubulin might not be essential for all GCPs, and low copy number GCPs might participate in ␥TuRC assembly even if they do not carry their own copy of ␥-tubulin. ␥TuRCs assembled with the mutant GCP4 were recruited to mitotic spindles and did not induce monopolarity, suggesting that they were functional. However, we do not know whether ␥TuRCs with mutant GCP4 can "fill the gap" by integrating ␥-tubulin without longitudinal GCP4 interaction, simply by lateral interactions with neighboring ␥-tubulins.

␥-Tubulin Complex Proteins and ␥-TuRC Assembly
GCP4 or GCP5 with tagged forms suggested that more than one copy of GCP4 but only one of GCP5 are present per complex (16). Estimation from gels of purified ␥TuRCs also suggested an average of 2.5 copies of GCP4, one copy of GCP5 and Ͻ1 copy of GCP6 (although GCP6 is essential for ␥TuRC assembly, this result is still to be confirmed; 19). Because of their low stoichiometry and because they are all essential, we think that GCP 4, 5, and 6 are positioned at specific locations, critical for the formation or stability of the helix. Our limited analysis revealed that the positions of GCP4 and GCP5 were not random and that the two proteins were adjacent within ␥TuRCs. Experiments in fungi suggest that GCP 4, 5, and 6 might form a subcomplex within ␥TuRCs (17,37,38). We failed to detect an interaction with GCP6 by FLIM-FRET using C-terminally tagged GCP 4, 5, and 6 (see Table 1), suggesting that GCP6 is not laterally associated with GCP4 or GCP5. Alternatively, the wide insertion present in the middle of GCP6 might position its C terminus away from the neighboring protein, preventing FRET. We propose that GCP 4, 5, and 6 position together at the ends of the helix, which would allow them either to initiate or terminate the self-assembly of ␥TuSCs and/or to stabilize the structure by bridging the two ends. Although several models can be proposed, we favor a model in agreement with our FLIM-FRET results in which GCP4 and -5 interact laterally at one end of the helix and GCP6 is located at the other end. In this configura-tion, a longitudinal interaction between GCP6 and GCP4/5 might close and stabilize the complex (Fig. 9). Finding the precise positions of GCP 4, 5, and 6 will be essential to understand their contribution to the function of ␥TuRCs.

Experimental Procedures
Plasmid Constructs-Human GCPs 2, 3, 4, 5, and 6 cDNAs were cloned into pCDNA5/FRT/TO (Invitrogen) using high fidelity PCR with oligonucleotides with tailored restriction sites: GCP2 and GCP3 (39), GCP4 (15), GCP5 (image clone 5297149), and GCP6 (21). Five silent point mutations within the siRNA targeting sequence of the GCPs cDNAs were introduced by PCR to generate resistance. For FLIM-FRET analysis, the sequences of AcGFP and mCherry were added at the 5Ј or 3Ј end of the GCPs cDNAs by PCR with a linker of six residues (AcGFP/mCherry-AGGGGG-GCP or GCP-GGGGGA-AcGFP/ mCherry). Mutations within the C-terminal domain of GCP4 were introduced into the pET26B-GCP4-V5 (27) by PCR. The chimeric constructs were generated by introducing a NheI restriction site at the junction of the N-terminal and C-terminal domains of the GCPs by PCR and by swapping the domains using the sites flanking the constructs in pCDNA5/FRT/TO. For the GCP3 chimera, the domains were also swapped using the Gibson kit (NEB) without introduction of the NheI restriction site.   OCTOBER 28, 2016 • VOLUME 291 • NUMBER 44

JOURNAL OF BIOLOGICAL CHEMISTRY 23119
Cell Culture and Generation of Stable Cell Lines-HeLa Flp-in T-Rex (a gift from S. Taylor, University of Manchester, UK), U2OS T-Rex Myc Nlp (from E. Nigg, University of Basel, Switzerland), and HeLa S3 ␥-tubulin-Myc-TAP (from J. Lüders, IRB Barcelona, Spain) were grown at 37°C in a 5% CO 2 atmosphere in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (31,40). HeLa Flp-in T-Rex were cultured with Zeocin (100 g/ml) and blasticidin (5 g/ml). The pCDNA5/FRT/TO vectors containing the GCPs cDNAs were co-transfected with the Flp recombinase encoding plasmid pOG44 (Invitrogen) using CaCl 2 . Hygromycin B-resistant clones (200 g/ml) were picked and expanded to obtain clonal cell lines. Transgene expression was induced using 1 g/ml doxycycline. Two independent clones with levels of expression similar to endogenous levels (except for the N64C chimera that was strongly overexpressed) were systematically chosen and analyzed. The results were similar between clones. U2OS T-Rex Myc Nlp were cultured with Zeocin (100 g/ml) and hygromycin B (50 g/ml). Nlp overexpression was induced using 1 g/ml doxycycline. HeLa S3 ␥-tubulin-Myc-TAP were cultured with 400 g/ml Geneticin.
Sucrose Gradients-Cells were trypsinized, rinsed with PBS, and resuspended in gradient buffer (50 mM HEPES, pH 7.4, 100 mM KCl, 1 mM EGTA, 1 mM MgCl 2 , 1 mM GTP, 1 mM DTT) with 1% Nonidet P-40 and complete protease inhibitor mix (Roche Applied Science). After 10 min centrifugation at 16,000 Ϫ g, soluble extracts were fractionated on gradients of 5-40% sucrose in gradient buffer, in a SW55-Ti Rotor (Beckman coulter) at 50,000 rpm for 4.5 h. Fractions were precipitated in methanol and resuspended in gel loading buffer containing SDS. Proteins were detected by Western blot using the Odyssey imaging system (Li-cor Biosciences), with IRDye 800CW-and 680CW-conjugated secondary antibodies (Invitrogen). ␥-Tubulin levels were quantified using the Odyssey 2.1 software: the Odyssey fluorescence system provides a linear relationship between signal intensity and antigen loading. Band intensities were measured with background subtraction, and the values were normalized relative to the total ␥-tubulin

␥-Tubulin Complex Proteins and ␥-TuRC Assembly
amount, corresponding to the sum of the intensities of all the fractions. For immunoprecipitation, gradient fractions were incubated with anti-GCP4 FL antibody for 2 h at 4°C and with protein A magnetic beads (Dynal) for an additional 2 h. The beads were washed twice in gradient buffer, and samples were eluted in gel loading buffer.
Cross-linking Experiments-␥TuRCs were purified from a HeLa S3 cell line expressing ␥-tubulin-Myc-TAP as described in Teixidó-Travesa et al. (40). Purified complexes were crosslinked with 0.5 mM DSS (disuccinimidyl suberate) for 7 min at 20°C. The cross-linking was stopped by adding 50 mM Tris, pH 8, for 15 min at 4°C. Cross-linked complexes were analyzed by Western blot after denaturing SDS-PAGE.
Immunofluorescence-Cells grown on coverslips were fixed in Ϫ20°C methanol and processed for immunofluorescence following conventional protocols. For detection, secondary antibodies conjugated to Alexa 488 or 568 (Invitrogen) were used. Fluorescence microscopy was performed on a widefield microscope (Axiovert 200 M; Carl Zeiss MicroImaging, Inc.) equipped with a Z motor using a 40ϫ (Plan Neofluar, 1.3 NA) or 63ϫ (Plan Apo, 1.4 NA) objective. Images were acquired with a MRm camera and Axiovision software (Carl Zeiss MicroImaging, Inc.). Image processing was done using Adobe Photoshop. Quantification of fluorescence was done using Axiovision; intensities at centrosomes and along spindle microtubules were measured on images acquired under constant exposure, within circular areas of 2 m Ø. An adjacent area of the same dimension was quantified and subtracted as background. For microtubule regrowth, cells were transferred into pre-cooled medium on ice for 1 h and then 15 s into prewarmed medium at 37°C. Regrowth was stopped by methanol fixation at Ϫ20°C.
For all the quantifications, error bars correspond to confidence interval calculated for a probability of 95% (1.96 ϫ S.E.).
Fluorescence Lifetime-For FLIM, the light source was a mode-locked Ti:sapphire laser (Tsunami, model 3941, Spectra-Physics) pumped by a 10-watt diode laser (Millennia Pro, Spectra-Physics) and delivering ultrafast femtosecond pulses of light with a fundamental frequency of 80MHz. A pulsepicker (model 3980, Spectra-Physics) was used to reduce the repetition rate to 2 MHz to satisfy the requirements of the triggering unit (working at 2 MHz) and to avoid exposition of the specimen to unwanted excitation pulses. All the experiments were carried out at ϭ 850 nm, the optimal wavelength to excite GFP in multiphoton mode (12). All images were acquired with a 60ϫ lens (Plan Apo, 1.4 NA, IR) mounted on an inverted microscope (Eclipse TE2000E, Nikon, Japan) coupled to the FLIM system. In this setup the femtosecond-pulsed laser light is scanned into the left camera port of the microscope via a dichroic mirror ( Ͻ 750 nm). The fluorescence emission is directed back into the detection unit through the same camera port and selected by a band pass filter (520/25 nm). Briefly, the FLIM unit is composed of two galvo mirrors, scanning along the x and y axes, a relay lens, and a streak camera (Streakscope C4334, Hamamatsu Photonics, Japan) coupled to a fast and high sensitivity CCD camera (model C8800-53C, Hamamatsu). Fluorescence decays were recorded during a time interval of 20 ns. A 4 ϫ 4 binning was used to reduce the acquisition time. The MCP-PMT gain and the exposure time of the CCD were adjusted to avoid overexposure of the CCD. For each image, average fluorescence decay profiles were calculated, and the lifetimes were estimated by fitting these data with a mono-or bi-exponential function using the Levenberg-Marquart non-linear least squares estimation procedure. FRET efficiency was calculated by E ϭ 1 Ϫ ( DA )/( D ), with DA and D the mean lifetime of the donor in the presence or absence of the acceptor, respectively. Statistical comparisons were performed by Student's t test. The relation between the FRET efficiency and the distance r between the donor and the acceptor is given by equation E ϭ 1/ (r/R 0 ) 6 , with R 0 the Förster distance to obtain 50% of FRET. Based on the Förster theory, if the distance r is Ͼ1.5R 0 , no energy transfer occurs. Taking into account the GFP-mCherry couple of donor and acceptor, we obtained a value close to 7%, meaning that FRET is considered significant only above this value.
Miscellaneous-Molecular graphics were performed with the UCSF Chimera package. Chimera is developed by the Resource   FIGURE 8. The C-terminal domain of GCP4, but not the interaction with ␥-tubulin, is necessary to rescue ␥TuRC assembly. A, ribbon view of GCP4 and representation of the GCP4/␥-tubulin interface modeled by fitting the GCP4 atomic structure into the S. cerevisiae ␥TuSC helix. Mutated residues within helices ␣16 are colored in light blue and within ␣20 are in dark blue. Phe-522 and Phe-626 are colored in orange. B, pulldown (PD) assays using ␥-tubulin-FLAG as bait with GCP4 mutants (␣16 mut: A511Q, N518R, A521E, F522W, D525R, N526F, Y529R, Y530A; ␣20 mut: A608W, A609R, S612E, S619W, S623Y). Shown are Western blots (WB) of eluted fractions from beads rinsed with HEPES buffer followed by consecutive treatments with 0.5 M NaCl, 0.1% Nonidet P-40, radioimmune precipitation assay buffer, 0.1% SDS, and gel loading buffer, probed with antibody against GCP4. C, Western blot analysis of expression of the F522R/F626R GCP4 and the 1-349 truncated GCP4 siR mutants in HeLa cells with or without depletion of the endogenous protein and induction of the mutant. ␥-Tubulin is shown as a loading control. D, sucrose gradient fractionation of soluble cell extracts. ␥TuRC sedimentation is visualized using antibody against ␥-tubulin. Quantifications of ␥-tubulin levels in fraction 7 are shown on the right. E, percentage of cells with defective recruitment of ␥-tubulin (mean of 3 independent experiments, 300 total cells scored per condition). Error bars: S.E., 95% CI. F, sucrose gradient fractionation of a soluble extract showing GCP4 species in small fractions in addition to ␥TuRCs. G, immunoprecipitation (IP) of GCP4 from fractions 3-7 of the gradient. Fraction 3 contains the majority of ␥TuSCs, and fraction 7 contains ␥TuRCs. H, immunoprecipitation of GCP4 after fractionation of soluble extracts from stable cell lines expressing GCP4 siR WT or F522R/F626R and depleted of the endogenous GCP4. ␥-Tubulin is co-precipitated with GCP4 WT in fractions 3-5 but not with GCP4 F522R/F626R, although it is always co-precipitated from ␥TuRCs in fraction 7. We propose that they are located at the overlapping ends of the helix, allowing them to initiate or terminate the assembly of the complex. As we have not detected a proximity between the C termini of GCP4/5 and GCP6, we propose that GCP4/5 position at one end of the helix, whereas GCP6 positions at the other end. GCP6 might interact longitudinally with GCP4/5 and bridge the ends of the helix, to stabilize its structure.

␥-Tubulin Complex Proteins and ␥-TuRC Assembly
for Biocomputing, Visualization, and Informatics at the University of California, San Francisco (supported by NIGMS, National Institutes of Health Grant P41-GM103311). PDB structures used were 3RIP for GCP4 atomic structure and 5FLZ for the interface between GCP4 and ␥-tubulin.