RESEARCH ARTICLE

MZT1 regulates microtubule nucleation by linking γTuRCassembly to adapter-mediated targeting and activationRosa Ramırez Cota1, Neus Teixido-Travesa1, Artur Ezquerra1, Susana Eibes1, Cristina Lacasa1, Joan Roig1,2

and Jens Luders1,*

ABSTRACTRegulation of the γ-tubulin ring complex (γTuRC) through targetingand activation restricts nucleation of microtubules to microtubule-organizing centers (MTOCs), aiding in the assembly of orderedmicrotubule arrays. However, the mechanistic basis of this importantregulation remains poorly understood. Here, we show that, in humancells, γTuRC integrity, determined by the presence of γ-tubulincomplex proteins (GCPs; also known as TUBGCPs) 2–6, is aprerequisite for interaction with the targeting factor NEDD1, impactingon essentially all γ-tubulin-dependent functions. Recognition ofγTuRC integrity is mediated by MZT1, which binds not only to theGCP3 subunit as previously shown, but cooperatively also to otherGCPs through a conserved hydrophobicmotif present in theN-terminiof GCP2, GCP3, GCP5 and GCP6. MZT1 knockdown causes severecellular defects under conditions that leave γTuRC intact, suggestingthat the essential function of MZT1 is not in γTuRC assembly. Instead,MZT1 specifically binds fully assembled γTuRC to enable interactionwith NEDD1 for targeting, and with the CM1 domain of CDK5RAP2for stimulating nucleation activity. Thus, MZT1 is a ‘priming factor’ forγTuRC that allows spatial regulation of nucleation.

KEY WORDS: Microtubule, Nucleation, γ-tubulin, Centrosome

INTRODUCTIONThe γ-tubulin ring complex (γTuRC), composed of γ-tubulin andγ-tubulin complex proteins (GCPs, also known as TUBGCPs),nucleates microtubules by providing a template for the assembly ofα-tubulin–β-tubulin heterodimers (Kollman et al., 2011; Oakleyet al., 2015; Petry and Vale, 2015; Remy et al., 2013; Teixidó-Travesa et al., 2012). The GCPs that form the γTuRC core structureall have homologous regions known as Grip motifs (Gunawardaneet al., 2000; Kollman et al., 2011) and in human cells are namedGCP2, GCP3, GCP4, GCP5 and GCP6 (Lüders and Stearns, 2007;Murphy et al., 2001; Teixidó-Travesa et al., 2012). As a nucleator,γTuRC is not only important for generating microtubules but alsofor controlling their position and orientation, which is fundamentalto the formation of ordered microtubule arrays (Lüders andStearns, 2007). Thus, the activity of γTuRC is tightly controlledby targeting and activation factors that spatially restrict nucleation tomicrotubule-organizing centers (MTOCs) such as the centrosome.Surprisingly, whereas assembly and targeting of human γTuRCseem to involve all GCPs (Bahtz et al., 2012; Choi et al., 2010;Izumi et al., 2008; Scheidecker et al., 2015), only γ-tubulin, GCP2

and GCP3, which form the smaller subcomplex γTuSC, areessential in flies and fungi, and are sufficient for microtubulenucleation and targeting to MTOCs (Anders et al., 2006; Fujitaet al., 2002; Lin et al., 2015; Venkatram et al., 2004; Vérollet et al.,2006; Xiong and Oakley, 2009). These observations suggest thatthere are important organism-specific differences in the manner bywhich nucleation template assembly is linked to targeting andactivation.

In budding yeast, which lacks GCP4, GCP5 and GCP6,formation of higher-order γ-tubulin complexes seems to berestricted to the spindle pole body (the yeast centrosomeequivalent). Here, interactions with adapter proteins containingCM1 and SPM motifs might trigger formation of γTuSC oligomersthat are similar in structure and activity to γTuRC (Kollman et al.,2010; Lin et al., 2015, 2014). The CM1 and SPM motifs of thebudding yeast adapter Spc110 have been shown to mediateinteraction with γTuSC, through direct interaction with the N-terminus of GCP3 and possibly also GCP2 (Knop and Schiebel,1997; Lin et al., 2011; Nguyen et al., 1998). Phosphorylation of theadapter in the region between the two motifs further enhances theinteraction. As a result, γTuSC oligomerizes and becomes a moreactive nucleation template (Lin et al., 2014). Even in fission yeast,which contains GCP4, GCP5 and GCP6, formation of active higher-order γ-tubulin complexes requires interaction with the CM1-containing protein Mto1 that forms an adapter complex with Mto2at interphase MTOCs (Lynch et al., 2014). Adapter-mediatedγTuSC oligomerization at MTOCs might also explain how theexperimental loss of the γTuRC-specific subunits GCP4, GCP5 andGCP6 is compensated for in flies (Vérollet et al., 2006).

In contrast to fungi, in most animal cells γTuRC assembly,targeting and activation are not tightly coupled, and pre-assembledbut relatively inactive γTuRCs can be found in the cytosol, outsideof MTOCs. However, in human cells γTuRC pre-assembly,targeting and activation are linked in the sense that all of thesesteps seem to involve the γTuRC-specific GCP4, GCP5 and GCP6(Bahtz et al., 2012; Choi et al., 2010; Izumi et al., 2008; Scheideckeret al., 2015), but the underlying mechanism has not been revealed.In fact, a systematic comparative analysis of the roles of all humanGCPs in γTuRC assembly and function has never been conducted.

Targeting of human γTuRC can be mediated by various differentproteins. Pericentrin, AKAP9, CDK5RAP2 and myomegalin havebeen implicated in γTuRC recruitment to the centrosome (Fonget al., 2008; Takahashi et al., 2002; Zimmerman et al., 2004) and tothe Golgi, which functions as a non-centrosomal MTOC (Riveroet al., 2009; Roubin et al., 2013; Wang et al., 2010). All of theseproteins contain a CM1 motif (Lin et al., 2014, 2015; Samejimaet al., 2008; Sawin et al., 2004), whereas an SPM motif is foundonly in pericentrin and AKAP9 (Lin et al., 2014, 2015). Curiously,NEDD1 (also known as GCP-WD), which in human cells is themost crucial factor for centrosomal targeting of γTuRC (Haren et al.,Received 15 July 2016; Accepted 9 November 2016

1Institute for Research in Biomedicine (IRB Barcelona), Barcelona 08028, Spain.2Molecular Biology Institute of Barcelona (IBMB-CSIC), Barcelona 08028, Spain.

*Author for correspondence (jens.luders@irbbarcelona.org)

406

© 2017. Published by The Company of Biologists Ltd | Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

2006; Lüders et al., 2006), does not seem to contain any SPM orCM1motifs. Regardless of the presence or absence of SPM or CM1motifs, both CDK5RAP2 and NEDD1 seem to specificallyassociate with γTuRC but not subcomplexes such as γTuSC (Choiet al., 2010; Lüders et al., 2006), but how this specific recognition ofγTuRC is achieved is unclear.Apart from mediating the interaction with γ-tubulin complexes,

the CM1 motif seems to have a role in activating the nucleationactivity of γ-tubulin complexes. In yeast, activation might resultfrom γTuSC oligomerization (Kollman et al., 2010; Lin et al., 2014;Lynch et al., 2014). In human cells, however, oligomerization per semight not be sufficient for full nucleation activity: even thougholigomeric complexes exist in the form of pre-assembled γTuRCs,the 50-amino-acid CM1 motif of human CDK5RAP2 alone is ableto strongly stimulate nucleation activity of these complexes in vitroand in vivo (Choi et al., 2010). Indeed, it has been proposed that inaddition to oligomerization the GCP3 subunit has to undergo aconformational change to allow formation of an optimal nucleationtemplate (Kollman et al., 2015).Adding to the complexity of the regulatory network controlling

nucleation, the small ∼8-kDa protein MZT1, which is found inanimals, plants and most fungi, and is not related to any of the otherknown adapters, is also required for targeting γTuRC to nucleationsites (Dhani et al., 2013; Hutchins et al., 2010; Janski et al., 2012;Masuda et al., 2013; Nakamura et al., 2012). However, whetherMZT1 directly attaches γ-tubulin complexes to MTOCs andwhether this role is related to the function of other well-established targeting factors such as NEDD1 or CM1-containingadapters, is unknown. It has also not been tested whether MZT1 hasany role in γTuRC or γTuSC oligomer assembly. Given that MZT1interacts with GCP3 (Dhani et al., 2013; Janski et al., 2012;Nakamura et al., 2012), the subunit that has been suggested toundergo a conformational switch during the activation of γ-tubulincomplexes (Kollman et al., 2015), and based on circumstantialevidence, MZT1 has been proposed to stimulate nucleation activity(Masuda and Toda, 2016; Nakamura et al., 2012). The relationshipwith CM1-dependent activation, if any, remained unclear.In summary, whereas in some organisms a minimal machinery

involving adapters that interact with γTuSC might be sufficient toallow assembly, targeting and activation of higher-order γ-tubulincomplexes, in other systems, including in human cells, this processis subject to more complex regulation and remains poorlyunderstood. Here, we have addressed this issue by systematicprobing of the subunit requirements, in particular the role of MZT1,in the spatial regulation of γTuRC in human cells.

RESULTSImpairment of γ-tubulin-dependent functions scales withcompromised γTuRC integrityTo analyze and compare the roles of the different GCPs in γTuRCassembly and function, we used RNA interference (RNAi)-mediated depletion. Given that some depletion conditions,including knockdown of GCP2, resulted in strong co-depletion ofother γTuRC subunits (Fig. S1A), we first established RNAiconditions that allowed specific and efficient (>95%) depletion ofonly the target protein, as judged by western blotting of cell extracts(Fig. 1A). With the exception of GCP3 knockdown, which alwaysled to a slight co-depletion of GCP6, transfection of the selectedsmall interfering RNAs (siRNAs) for 72 h specifically reduced thelevels of the targeted GCP with minimal effects on other GCPsubunits (Fig. 1A). Using these conditions, we then determined, foreach transfection, the state of γTuRC, by performing sucrose

gradient fractionation, and its functionality, by analysis of cellularphenotypes using immunofluorescence microscopy. This allowedus, in each case, to directly correlate γTuRC integrity with function.

Sucrose gradient fractionation of the extracts shown in Fig. 1Aand probing with antibodies against γTuRC subunits revealed twodistinct peaks corresponding to γTuRC and to complexes of smallermolecular mass for control extract (Fig. 1B,C; Fig. S2A–D). Giventhat we had observed in previous work that the amplitude of theγTuRC peak was variable between experiments but consistentbetween samples that were processed in parallel in individualexperiments (Lüders et al., 2006; Teixidó-Travesa et al., 2010), wealways analyzed knockdown samples side-by-side with controlextract. In GCP3-depleted extract, γ-tubulin in both peaks wasshifted to fractions of very small molecular mass (<7S), consistentwith GCP3 being required for the assembly of both γTuRC andsmaller complexes (Fig. 1B,C). Depletion of any of the threeγTuRC-specific subunits (GCP4, GCP5 or GCP6) also reduced theamount of γTuRC present in the extract (Fig. 1B,C; Fig. S2A–D).The strongest γTuRC disruption was observed with extractsdepleted of GCP6, whereas depletion of GCP4 or GCP5 led toonly partial γTuRC disruption, with a substantial amount of GCP6still co-fractionating in a high-molecular-mass complex that seemedonly slightly smaller than γTuRC (Fig. S2A–D). Knockdowncultures derived from the same transfections were also used foranalysis by immunofluorescence microscopy. Interestingly,depletion of GCP6 caused a similar increase in the mitotic indexto that seen upon depletion of GCP3 (∼45% and ∼40%,respectively). Depletion of GCP4 or GCP5 also increased themitotic index but to a lesser extent (∼15% and 10%, respectively)(Fig. 1D). GCP6-depleted cells displayed mitotic defects that wereat least as severe as in cells depleted of GCP3. Approximately 80%of GCP6-depleted mitotic cells had monopolar-like spindleconfigurations with poorly organized microtubules, and few cellsat metaphase or in later mitotic stages were observed. Depletion ofGCP4 or GCP5 produced similar, but milder, mitotic spindledefects (Fig. 1E).

We also analyzed the samples for possible centriole duplicationdefects by counting centrin foci in mitotic cells. Again GCP6depletion produced defects that were comparable to those observedin GCP3-depleted cells. In both cases, depleted cells (∼60%) hadnot duplicated their centrioles and contained a single centriole ateach of their spindle poles (configuration ‘1+1’) compared todoublets in control cells (configuration ‘2+2’) (Fig. 1F). Depletionof GCP4 or GCP5 also impaired centriole duplication, but thedefects were milder than in GCP6-depleted cells. In 10–20% ofthe cells one of the centrioles had failed duplication (configuration‘2+1’) (Fig. 1F). Centriole duplication was recently shown torequire phosphorylation of GCP6 by Plk4 (Bahtz et al., 2012). Ourexperiments additionally suggest that centriole duplication requiresan intact γTuRC. In summary, our data shows that impairment ofγ-tubulin dependent functions is correlated with the degree ofγTuRC disruption in human cells.

Centrosomal targeting of γ-tubulin is directly correlated withγTuRC integrityTo better understand how γTuRC disruption impairs spatialregulation of nucleation, we determined γ-tubulin recruitment tocentrosomes in control and siRNA-treated mitotic cells. Depletionof GCPs reduced centrosomal γ-tubulin staining in all cases,whereas pericentrin was not affected, indicating specific loss ofγ-tubulin (Fig. 2A). Simultaneous disruption of γTuSC andγTuRC (upon depletion of GCP3) or specific disruption of only

407

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

GCP6

GCP3

-T

GCP5

GCP4

GCP2

siRNA:

250

100 100

100 75

50 37

A

top bottom -Tubulin

GCP6

GCP3

GCP2

GCP5 GCP4

Con

trol R

NA

GC

P3

siR

NA

-Tubulin

GCP6

GCP3

GCP2

GCP5 GCP4

-Tubulin

GCP3

GCP6

GCP2

GCP4

GCP5 GC

P6

siR

NA

TuRC

B

0

10

20

30

40

50

1 2 3 4 5 6 7 8 9 10 11 12 13

-Tubulin

Ban

d in

tens

ity [%

]

0

10

20

30

40

50

1 2 3 4 5 6 7 8 9 10 11 12 13

-Tubulin Control RNA

GCP6 siRNA

fractions

fractions

Ban

d in

tens

ity [%

]

Control RNA

GCP3 siRNA

C*

*

*

(kDa)

D

Mito

tic c

ells

[%]

0

10

20

30

40

50

siRNA:

****

****

****

****

Control

GCP3GCP4

GCP5GCP6

**** ****

F

0

20

40

60

80

100

2+2 2+1 1+1 1+0 >4

Mito

tic c

ells

[%]

Centriole configuration

Control RNAGCP3 siRNAGCP4 siRNAGCP5 siRNAGCP6 siRNA

****

*******

********

****

********

**nsnsns ns**

nsns ns

E

0

20

40

60

80

100 Control RNA

GCP3 siRNA

GCP4 siRNA

GCP5 siRNA

GCP6 siRNA

Mito

tic c

ells

[%]

**nsns

**

ns

ns

nsns

****

***

*

****

***

nsns

*** *****

ns

**

nsns

**

Prophase

Prometaphase

Monopolar Spindle

Metaphase

Anaphase

Telophase

***

ns

7S 19S

ubulin

GAPDH

Fig. 1. The degree of γTuRC disruption correlates with the severity of cellular defects. (A) Western blot of inputs for the extracts subjected to sucrosegradient centrifugation that are shown in B and Fig. S2A,C, demonstrating specific depletion of individual GCPs. HeLa cells were transfected with siRNA andγTuRCproteins were detectedwith specific antibodies as indicated. GAPDHwas detected as loading control. (B) Extracts from control-, GCP3- or GCP6-depletedHeLa cells shown in Awere analyzed by sucrose gradient centrifugation. Fractions were probed with antibodies against the indicated proteins. Asterisks mark anunspecific band recognized by the anti-GCP6 antibody. The vertical box indicates the γTuRC peak fractions. Aldolase (158K, 7S) and thyroglobulin (669K, 19S)were used as sizemarkers. (C) Fractionation profiles for γ-tubulin for the sucrose gradients in B. The γ-tubulin band intensities in western blots of gradient fractionswere quantified and plotted for control and depleted extracts as indicated. Values represent percentages of the total signal for all fractions combined. (D) Thepercentage of mitotic cells was determined for HeLa cells depleted as in A. After staining cells with antibodies against pericentrin, γ-tubulin or α-tubulin, and DAPI,mitotic cells were counted by immunofluorescence microscopy; n=3 experiments (n=5 experiments for control cells). (E) Mitotic cells in D were scored accordingto the mitotic stage and the presence of a monopolar spindle configuration as indicated; n=3 experiments (n=5 experiments for control cells), 100 mitotic cells percondition. (F) The centriole configuration in mitotic HeLa cells transfected as in A was scored after staining with anti-centrin antibodies. For each of the twocentrosomes, the number of centrin foci was counted; n=3 experiments (n=5 experiments for control cells), 100 mitotic cells per condition. All quantitative resultsare mean±s.e.m. ns, not significant; **P<0.01; ***P<0.001; ****P<0.0001 (unpaired t-test).

408

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

γTuRC (upon depletion of GCP6) reduced the amount ofcentrosomal γ-tubulin by ∼55% compared to control cells,whereas depletion of GCP4 or GCP5 caused milder effects(∼30% and ∼15% reduction, respectively) (Fig. 2A–C). Thereduction in centrosomal γ-tubulin correlated with a proportionalreduction in centrosomal nucleation activity in a microtubuledepolymerization–repolymerization assay (Fig. 2B,C). We notethat the measured reduction in centrosomal γ-tubulin is likely anunderestimation, as many severely affected mitotic cells, inparticular after depletion of GCP3 or GCP6, detached from thedish or displayed fragmented centrosomes and thus were not usedfor quantification. We conclude that, similar to the mitotic and

centriole duplication defects, impaired centrosomal targeting ofγ-tubulin scales with the degree of γTuRC disruption.

γTuRC integrity is crucial for binding to the targeting subunitNEDD1The phenotypes that we observed in cells with disrupted γTuRChave striking similarity with the phenotypes previously reported forknockdown of the targeting factor NEDD1 (Haren et al., 2006;Lüders et al., 2006). To investigate this, we determined the proteinlevels of NEDD1 in depleted extracts. Surprisingly, with theexception of GCP5 depletion, all siRNA treatments caused a strongco-depletion of NEDD1 (82%, 69% and 51% reduction in cells

A B

Contro

l

GCP2GCP3

GCP4GCP5

GCP6

GST

GST-C-NEDD1

GCP6GCP5GCP4

γ-Tubulin

GCP3GCP2

C siRNA:

********

****

****

******

********

DCon

trol

GCP2GCP3

GCP4GCP5

GCP6

250

100

100

150 75

50 (kDa)

Input

Pull-down

Fig. 2. Disruption of γTuRC interferes with γ-tubulin centrosomal targeting by impairing interaction with NEDD1. (A) HeLa cells depleted as in Fig. 1Awere imaged by immunofluorescence microscopy after staining with anti-γ-tubulin and anti-pericentrin antibodies, and DAPI to label DNA. Scale bar: 10 µm.(B) Microtubule depolymerization–repolymerization assay with cells shown in A. After 1 min of regrowth, cells were fixed and stained with anti-γ-tubulin andanti-α-tubulin antibodies. (C) The intensities for centrosomal γ-tubulin and microtubule asters surrounding the centrosome (α-tubulin) in cells as in B werequantified and plotted. Values were plotted as percentages of the values obtained for control cells (set to 100%); n>95 centrosomes per condition combined fromtwo independent experiments. Results are mean±s.e.m. **P<0.01, ****P<0.0001 (unpaired t-test compared with control). (D) HeLa extracts depleted forindividual GCPs were incubated with recombinant GST or GST–C-NEDD1 immobilized on glutathione–Sepharose. Beads were washed and the retained γTuRCsubunits were detected by western blotting. Pulled down GST and GST–C-NEDD1 were detected on the blotted membranes by Ponceau staining.

409

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

depleted of GCP6, GCP3 and GCP4, respectively) (Fig. S3A),suggesting that an intact γTuRC is required for NEDD1 stability andsupporting the notion that NEDD1 is a core subunit of γTuRC(Lüders et al., 2006; Teixidó-Travesa et al., 2010). To confirm thatthe defects that we have described above for depletion of GCPs werea direct consequence of γTuRC disruption and not a secondaryeffect caused by NEDD1 co-depletion, we performed rescueexperiments by expressing RNAi-resistant recombinant NEDD1–EGFP in siRNA-treated cells. In cells depleted of NEDD1exogenously expressed NEDD1–EGFP efficiently rescued loss ofcentrosomal γ-tubulin and mitotic arrest (Fig. S3B–E). However,expression of NEDD1–EGFP did not restore centrosomal γ-tubulinand mitotic progression in GCP3- or GCP6-depleted cells(Fig. S3C–E). This result demonstrates that the disruption ofγTuRC and not the loss of NEDD1 is the primary cause of themitotic defects in cells depleted of GCP3 or GCP6.Our finding that centrosomal targeting of human γ-tubulin

depends on γTuRC integrity, and that depletion of GCP6 producesphenotypes indistinguishable from those observed after depletion ofGCP3, contrasts with results in other organisms. In Drosophila andfungi, only γTuSC subunits were found to be essential (Anderset al., 2006; Fujita et al., 2002; Venkatram et al., 2004; Vérolletet al., 2006; Xiong and Oakley, 2009). A key to this discrepancymight be the fact that γ-tubulin targeting depends strongly onNEDD1 in human cells, whereas this protein seems to be lessimportant in flies and is not present in fungi. Given that NEDD1wasalso destabilized under conditions that disrupted γTuRC but leftγTuSC intact (Fig. S3A), we speculated that only intact γTuRC, butnot γTuSC, interacts with NEDD1. To test this, we studied theinteraction between NEDD1 and γTuRC subunits in control extractand extracts depleted of individual GCPs. To make the assayinsensitive to the destabilization of endogenous NEDD1, wesupplemented the extract with an excess of a recombinant GSTfusion of the C-terminal half of NEDD1 (GST–C-NEDD1), whichis sufficient for γTuRC binding (Haren et al., 2006; Lüders et al.,2006; Manning et al., 2010). After affinity purification onglutathione beads, GST-C–NEDD1-associated GCPs and γ-tubulin were detected by western blotting with specific antibodies.Interestingly, depletion of any of the GCPs reduced the amounts ofγ-tubulin and of all remaining GCPs that were bound to GST–C-NEDD1 (Fig. 2D). The strongest impairment was observed afterknockdown of GCP2, GCP3 or GCP6. Importantly, in depletedextracts, none of the remaining GCPs retained the strong bindingobserved in control extract (Fig. 2D). As a result, despite thedifferences in the total amount of γ-tubulin and GCPs that werebound to GST–C-NEDD1, the relative amounts of the retainedsubunits were always similar, strongly suggesting that NEDD1interacts preferentially with intact γTuRCs composed of thecomplete set of subunits. Individual subunits or subcomplexes,such as γTuSC, cannot interact efficiently with NEDD1, explainingwhy only γTuRC can target γ-tubulin to centrosomes in humancells.

γTuRC interaction with NEDD1 is promoted by MZT1How does NEDD1 selectively associate with γTuRC? Previouswork suggested that NEDD1 directly binds to γ-tubulin (Manninget al., 2010), but subcomplexes of γTuRC including γTuSC, despitethe presence of γ-tubulin, do not associate with NEDD1 (Fig. 2D)(Lüders et al., 2006). Thus interaction with γ-tubulin cannot explainthe preferential binding of NEDD1 to γTuRC. Interestingly, similarto NEDD1 the recently discovered γTuRC subunit MZT1 alsopreferentially binds to γTuRC and is required for its recruitment to

centrosomes (Hutchins et al., 2010). However, it remained unclearwhetherMZT1mediates targeting directly or indirectly, by affectingγTuRC integrity. To address this we generated an antibody againsthuman MZT1 that recognized a protein of the expected size bywestern blotting. The specificity of the band was confirmed bysiRNA-mediated depletion of MZT1 (Fig. 3F,G). We thenconfirmed that MZT1-depleted cells arrested in mitosis withmonopolar spindles and displayed a specific reduction incentrosomal γ-tubulin as described previously (Hutchins et al.,2010) (Fig. S4A–F). Interestingly, loss of centrosomal γ-tubulinwas also observed in interphase cells, indicating that MZT1 is not amitosis-specific factor but is required for γTuRC targetingthroughout the cell cycle (Fig. 3A,B). In agreement with this, wefound that MZT1 depletion also impaired centriole duplication(Fig. 3E). Compared to γ-tubulin, centrosomal localization ofNEDD1 was not sensitive or less sensitive to MZT1 depletion(Fig. 3C,D; Fig. S4E,F), suggesting that MZT1 does not controlcentrosomal targeting of NEDD1 but rather links NEDD1 toγTuRC. To test this more directly, we immunoprecipitated NEDD1from control and MZT1-knockdown cells and probed for co-precipitation of γ-tubulin and GCP6. Indeed, in the absence ofMZT1, interaction of NEDD1 with these γTuRC subunits wasstrongly diminished (Fig. 3F). Importantly, when we analyzedMZT1-depleted extracts on sucrose gradients, γ-tubulin and GCPswere still detected in fractions corresponding to the size of γTuRC,despite the reduced levels of MZT1 in these fractions (Fig. 3G,H).Thus, under MZT1 depletion conditions that cause severe cellulardefects, γTuRC is still present, but contains a reduced amount ofMZT1 and is defective in binding to the targeting factor NEDD1.

MZT1 directly binds to the N-termini of multiple GCPsGiven that MZT1 was required for NEDD1 to interact with γTuRC,we speculated that MZT1 might also be responsible for thepreferential recognition of γTuRC over subcomplexes. However, itwas previously shown that MZT1 itself directly binds to the N-terminus of GCP3, a subunit that is not specific to γTuRC (Dhaniet al., 2013; Janski et al., 2012; Nakamura et al., 2012). We decidedto re-investigate the interactions of MZT1 with γTuRC subunits. Wegenerated N-terminal fragments of each of the GCPs, comprisingthe first Grip domain and any N-terminal extensions (NTEs)(Fig. 4A). We then performed co-immunoprecipitation experimentsin human cells overexpressing FLAG-tagged N-terminal GCPfragments together with EGFP–MZT1. Surprisingly, in this assaywe observed interaction of MZT1 not only with the N-terminalfragment of GCP3, as previously described, but also with thecorresponding fragments of GCP5 and GCP6 (Fig. 4B). N-terminalfragments of GCP2 and GCP4 failed to coprecipitate with MZT1. Inthe case of GCP3, the MZT1-binding site was mapped to the NTEthat precedes the conserved first Grip domain (Dhani et al., 2013;Janski et al., 2012). Indeed, deletion of the corresponding regions inthe N-terminal fragments of GCP3, GCP5 and GCP6 diminishedthe interaction with EGFP–MZT1 in the coprecipitation assay ineach case (Fig. 4B). To test whether these newly discoveredinteractions were direct, we employed the yeast two-hybrid assay.This revealed that MZT1 strongly interacted with the N-terminalfragments of GCP3, GCP5 and GCP6, and also with GCP2,although somewhat less strongly (Fig. 4C), whereas no interactionwas observed with the N-terminus of GCP4. Serving as controls,γ-tubulin and GCP8, two subunits of the γTuRC that are unrelated toGCPs 2–6 (Teixidó-Travesa et al., 2010), also failed to interact withMZT1 (Fig. 4C). Despite the overall very low sequenceconservation in the NTEs of GCP2, GCP3, GCP5 and GCP6,

410

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

careful analysis revealed a conserved sequence motif composedmostly of hydrophobic residues (Fig. 4D). To test whether thismotif may be involved in MZT1 binding, we mutated three of theconserved residues to alanine (‘3A’ mutants). Indeed, N-terminalGCP fragments carrying these mutations did not bind MZT1 in theyeast two-hybrid assay (Fig. 4E). Together the results suggest thatMZT1 directly binds to the N-termini of multiple GCPs and that aconserved motif present in the NTEs is required for this interaction.We next asked how mutation of this motif would affect MZT1

binding in the context of full-length GCPs. GCPs with mutatedMZT1 binding motif coprecipitated with other GCPs and γ-tubulin

similar to the wild-type versions. However, in contrast to thewild-type proteins, the 3A mutants did not efficiently coprecipitateendogenous MZT1 and NEDD1 (Fig. 5A). Consistent with thisfinding, immunofluorescence analysis showed that GCP 3Amutantswere defective in targeting to centrosomes (Fig. 5B,C). We alsodetermined the functional consequences for γTuRC containing amutant GCP in the absence of the corresponding endogenousprotein. We expressed FLAG-tagged wild-type GCP3 or the 3Amutant in cells depleted of endogenous GCP3 (Fig. 5D). Whereasloss of centrosomal γ-tubulin and the increased mitotic index afterGCP3 depletion were rescued in cells expressing recombinant

A B

-Tubulin Merge/DNAC

E

0

20

40

60

80

100

120

NEDD1 -Tubulin

Fluo

resc

ence

inte

nsity

[%] Control siRNA

MZT1 siRNA

DNEDD1

F

Input IgG IPControl siRNA

GCP3GCP4GCP5GCP6

GCP2

MZT1

-Tubulin

MZT

1 si

RN

AC

ontro

l siR

NA

GCP3

GCP4GCP5GCP6

GCP2

MZT1

-Tubulin

TuRCtop bottom

NEDD1

GAPDHGCP6-Tubulin

MZT1

NEDD1 IP

Con

trol s

iRN

AM

ZT1

siR

NA

****

Con

trol s

iRN

AM

ZT1

siR

NA

-Tubulin Merge/DNAPericentrin

0

20

40

60

80

100

120

Pericentrin -Tubulin

Control siRNAMZT1 siRNA

Fluo

resc

ence

inte

nsity

[%]

****

MZT1 siRNA

Control siRNA

MZT1 siRNA

Control siRNA

MZT1 siRNA

H

******

***

0

10

20

30

40

50

60

70

80

90

2+2 2+1 2+0 1+1 1+0 1+3 >4

Mito

tic c

ells

[%]

Centriole configuration

Control siRNA

MZT1 siRNA

*

*

7S 19S

Control siRNA

MZT1 siRNAInput

GAPDH

GCP6

-Tubulin

MZT1

GCP5GCP3

GCP4

GCP2

250

100 100 100 75

50

37

(kDa)

10

75 10

50250

(kDa)

37

G

Fig. 3. Interaction of γTuRC with the targetingfactor NEDD1 requires MZT1. (A) Interphasecontrol and MZT1-depleted cells were stained withanti-γ-tubulin and anti-pericentrin antibodies asindicated. DNA was stained with DAPI. Scale bar:10 µm. (B) The intensities for centrosomalγ-tubulin and pericentrin staining in A werequantified and plotted as percentages of the signalin control cells (set to 100%); n>100 centrosomesper condition combined from three independentexperiments. (C) Interphase control and MZT1-depleted cells were stained with anti-NEDD1 andanti-γ-tubulin antibodies as indicated. DNA wasstained with DAPI. Scale bar: 10 µm. (D) Theintensities for centrosomal NEDD1 and γ-tubulinstaining in C were quantified and plotted aspercentages of the signal in control cells (set to100%); n>100 centrosomes per conditioncombined from three independent experiments.(E) The centriole configurations in control andMZT1-depleted mitotic HeLa cells were quantifiedas in Fig. 1F; n=3 experiments, 100 mitotic cellsper condition. (F) Endogenous NEDD1 wasimmunoprecipitated from control and MZT1-depleted extracts. Precipitation with unspecific IgGserved as a control. The precipitates were probedwith antibodies against the indicated proteins bywestern blotting. Probing for GAPDH served as aloading control. (G) Control and MZT-depletedHeLa cell extracts were analyzed by westernblotting with the indicated antibodies. (H) Thecontrol and MZT1-depleted cell extracts shownin G were analyzed by sucrose gradientcentrifugation as in Fig. 1B. Asterisks mark anunspecific band recognized by the anti-GCP6antibody. The vertical box indicates the γTuRCpeak fractions. Aldolase (158K, 7S) andthyroglobulin (669K, 19S) were used as sizemarkers. All quantitative results are mean±s.e.m.***P<0.001; ****P<0.0001 (unpaired t-testcompared with control).

411

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

wild-type GCP3, these defects were still observed in cells expressingthe GCP3 3A mutant (Fig. 5D–F). Taken together, these resultssuggest that MZT1 specifically recognizes γTuRC through binding

to multiple GCPs, and that mutation of the MZT1-binding motif inonly one of the GCPs was sufficient to reduce MZT1 and NEDD1binding, and interfere with centrosomal targeting.

B pGAD-MZT1

pGBK--Tubulin

pGBK-N-GCP2

pGBK-N-GCP3

pGBK-N-GCP4

negativecontrol

positivecontrol

pGBK-N-GCP5

pGBK-N-GCP6

pGBK-GCP8

pGBK-MZT1

negativecontrol

positivecontrol

C

WT WT WT

mut mut mut

MZT1 bdg. motif mutant: A A A

NTE Grip domain 1 Grip domain 2GCP2

GCP5

GCP3

GCP6

GCP4

A

D

EpGAD-MZT1 +pGBK-N-GCP3 pGBK-N-GCP5 pGBK-N-GCP6

“N-GCP”“N-GCPNTE”

Input FLAG IP

GFP-MZT1

GFP-MZT1 +

FLAG

PFG-

GALF

2PC

G-N-

GALF

2PC

G-N-

G

ETN

ALF 3P

CG-

N-G

ALF 3P

CG-

N-G

ET

NALF

4PC

G-N-

GALF

5PC

G-N-

GALF

5PC

G-N-

G

ETN

ALF 6P

CG-

N-G

ALF 6P

CG-

N-G

ET

NALF

PFG-

GALF

2PC

G-N-

GALF

2PC

G-N-

G

ETN

ALF 3P

CG-

N-G

ALF 3P

CG-

N-G

ET

NALF

4PC

G-N-

GALF

5PC

G-N-

GALF

5PC

G-N-

G

ETN

ALF 6P

CG-

N-G

ALF 6P

CG-

N-G

ET

NALF

H.s. GCP3 1-249A.t. GCP3 1-188D.m. GCP3 1-229A.n. GCP3 1-265S.p. GCP3 1-192H.s. GCP2 1-209H.s. GCP5 1-269H.s. GCP6 1-316

H.s. GCP3 92-101A.t. GCP3 104-113D.m. GCP3 110-119A.n. GCP3 139-148S.p. GCP3 95-104H.s. GCP2 80-89H.s. GCP5 109-118H.s. GCP6 109-118

(kDa)

37

37

50

75

GFP

Fig. 4. MZT1 binds to a conservedmotif present in the N-termini of GCP2, GCP3, GCP5 and GCP6. (A) Schematic representation of the domain structure ofthe human GCPs. The positions of the N- and C-terminal Grip domains are indicated by black boxes. The N-terminal extensions (NTEs) that precede theN-terminal Grip domains in GCP2, GCP3, GCP5 and GCP6 are colored in beige. The domains comprised by the constructs used in the interaction assays areindicated by black lines. (B) Co-immunoprecipitation (IP) assay to reveal interactions between MZT1 and the indicated GCP fragments. GFP–MZT1 andFLAG-tagged GCP fragments were coexpressed in HEK293 cells. After FLAG immunoprecipitation samples were analyzed by western blotting with anti-FLAGand anti-GFP antibodies. (C) Yeast two-hybrid interactions between MZT1 and N-terminal fragments of the GCPs. Positive interaction is revealed by growthand blue color. (D) Alignment of the amino acid sequences of the NTEs of humanGCP2, GCP3, GCP5, andGCP6 with the NTEs of GCP3 orthologs from variousspecies (A.t., Arabidopsis thaliana; D.m., Drosophila melanogaster; A.n., Aspergillus nidulans; S.p., Schizosaccharomyces pombe). The numbers indicatethe amino acid positions. The magnified region shows a conserved hydrophobic motif. Alanine replacement mutations used in MZT1-binding mutants areindicated. The alignment was made with the software Geneious using the MUSCLE algorithm (Edgar, 2004). (E) Yeast two-hybrid assay revealing the lossof interaction between MZT1 and N-terminal GCP fragments (wild type, WT) upon mutation of the MZT1-binding motif (mut). Growth and blue color indicatespositive interaction. +, positive control; –, negative control.

412

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

MZT1 promotes not only γTuRC targeting but also activationOnly a relatively small percentage of total γTuRC localizes to thecentrosome and the large majority is present in the soluble fractionof the cytoplasm (Bauer et al., 2016; Moudjou et al., 1996). Giventhat microtubule nucleation occurs predominantly at thecentrosome, the cytosolic γTuRC is presumably relativelyinactive. Therefore, apart from targeting, spatial regulation ofγTuRC should also involve activation of its nucleation activity.Indeed, expression of a 50-amino-acid fragment of CDK5RAP2containing the conserved CM1 motif (CDK5RAP2 51–100)strongly stimulates the nucleation activity of cytosolic γTuRC(Choi et al., 2010). Using the same assay, we asked whether theactivation of cytosolic γTuRC by CDK5RAP2 51–100 requiredMZT1. Cells transfected with plasmid encoding EGFP or EGFP–CDK5RAP2 51–100 and with siRNA to knockdown γTuRCsubunits were subjected to cold to completely depolymerizemicrotubules and then incubated at 37°C for 10 s to allowmicrotubule nucleation to occur. After fixation, cytoplasmicmicrotubules were analyzed by performing immunofluorescencemicroscopy and then quantified. As expected, the basal nucleationactivity of γTuRC in control cells expressing EGFP was low, butcould be stimulated by expression of EGFP–CDK5RAP2 51-100,as described previously (Fig. 6A,B) (Choi et al., 2010). Strikingly,this stimulatory effect was eliminated by simultaneous knockdownofMZT1. Complete γTuRC disruption upon depletion of GCP2 hada similar effect (Fig. 6A,B). Interestingly, the basal activity inEGFP-expressing control cells was also reduced by depletion ofMZT1, Again, this was similar to cells in which γTuRC wasdisrupted by GCP2 knockdown (Fig. 6A,B). As an additionalcontrol, we also tested whether MZT1 knockdown affected thelevels of endogenous CDK5RAP2, but this was not the case(Fig. 6D). Taken together, these results suggested that activation ofγTuRC nucleation activity required MZT1.Similar to NEDD1, CDK5RAP2 also binds specifically to

γTuRC rather than to smaller subcomplexes, and the CM1 domainof CDK5RAP2 is sufficient for this interaction (Choi et al., 2010).We speculated that MZT1 might not only promote interaction ofγTuRC with the C-terminus of NEDD1, but also with the CM1domain of CDK5RAP2. To test this, we immunoprecipitatedtransiently expressed EGFP–CDK5RAP2 51-100 from control cellsand cells depleted of MZT1. Supporting our hypothesis, EGFP–CDK5RAP2 51-100 efficiently coprecipitated γTuRC subunitsfrom control extracts, whereas in MZT1-depleted extract thisinteraction was diminished (Fig. 6C). This result strongly suggeststhat the inability of EGFP–CDK5RAP2 51-100 to stimulate γTuRCnucleation activity in the absence of MZT1 (Fig. 6A,B) was due toimpaired γTuRC binding.

DISCUSSIONHere, we have elucidated the spatial regulation of microtubulenucleation in human cells. Side-by-side analysis of GCP4-, GCP5-and GCP6-knockdown cells revealed differences in the requirementof these GCPs for γTuRC integrity. In all cases, the degree ofγTuRC disruption was correlated with impairment of all tested γ-tubulin-dependent functions, including centrosomal nucleation,centriole duplication, assembly of the mitotic spindle and mitoticprogression. Cells were particularly sensitive to knockdown ofGCP6 with phenotypes indistinguishable from defects after GCP3knockdown, suggesting that in human cells γTuSC alone is not ableto support γ-tubulin-dependent processes. Mechanistically, thisdependence on γTuRC integrity is also rooted in the fact that onlyintact γTuRC can interact with the targeting factor NEDD1. The

importance of GCP6 might be related to its central role in γTuRCassembly, as suggested by previous work in fungi in which GCP6was shown to be required for the incorporation of GCP4 and GCP5into complexes with γTuSC (Anders et al., 2006) and targeting ofAspergillus GCP4 and GCP5 to the spindle pole body was found torequire GCP6, but GCP6 spindle pole body localization was foundto be independent of GCP4 and GCP5 (Xiong and Oakley, 2009).However, whereas in control extracts GCP6 co-fractionates almostexclusively with γTuRC, GCP4 and GCP5 are also present incomplexes smaller than γTuRC. If GCP4 or GCP5 knockdownpreferentially affect this non-γTuRC-associated pool of the proteins,this might leave some γTuRC intact and reduce the severity of theknockdown phenotypes compared to cells depleted of GCP6. Thus,to unequivocally determine the relative importance of GCP4, GCP5and GCP6 for γTuRC integrity and activity, reconstitution ofγTuRC from recombinant proteins might be required.

Why do animal cells contain pre-assembled cytosolic γTuRCs?Limiting the assembly of such higher-order γ-tubulin complexes toMTOCs might satisfy the needs of relatively simple microtubulenetworks consisting of only a few microtubules, such as those inyeast cells. In more complex animal cells that contain hundreds ofmicrotubules, pre-assembled γTuRCs might be more suitable forresponding quickly to sudden changes in the demand for nucleators,for example, when cells initiate mitotic spindle assembly. Spatialcontrol over nucleation requires γTuRC pre-assembly to beseparated from targeting and activation, to ensure thatincompletely assembled γTuRC or its subcomplexes are notrecruited to MTOCs, and fully assembled γTuRCs become activeonly upon interaction with MTOCs. We found that a fundamentalelement in this regulation is the small subunit MZT1, which, on theone hand allows specific recognition of intact γTuRC through amulti-subunit binding mode, and, on the other hand, enables γTuRCto interact with NEDD1 and CDK5RAP2, to allow targeting andactivation, respectively. Owing to its central role in γTuRCregulation, one can speculate that MZT1 activity (controlled at thelevel of available MZT1 protein or by post-translationalmodification) will determine the amount of γTuRC that can betargeted and activated to nucleate microtubules. Other γTuRCregulators could also be subject to this type of regulation. Consistentwith this notion, the regulatory factors NEDD1 and CDK5RAP2have a significantly higher turnover rate compared to the γTuRCstructural core subunits γ-tubulin, GCP2, GCP3, GCP4, GCP5 andGCP6 (Jakobsen et al., 2011). Moreover, NEDD1 is phosphorylatedat multiple sites to specifically control γTuRC function in distinctmitotic nucleation pathways (Gomez-Ferreria et al., 2012; Harenet al., 2009; Johmura et al., 2011; Lüders et al., 2006; Pinyol et al.,2013; Scrofani et al., 2015; Sdelci et al., 2012; Teixidó-Travesaet al., 2010, 2012). Future work will show whether human MZT1 issubject to similar regulation.

Based on genetic evidence, it has been recently shown that GCP6and MZT1 in fission yeast act synergistically in mitotic spindleassembly, but the underlying mechanism remained unclear(Masuda and Toda, 2016). Our findings now allow a mechanisticinterpretation of this synergistic effect: GCP6 is not only crucial forγTuRC integrity but also interacts directly with MZT1. In fact, ourdata indicates that MZT1 binds to all GCPs but GCP4, and not onlyto GCP3, as previously suggested. MZT1 binding is mediated by aconserved motif present in the NTEs of GCP2, GCP3, GCP5 andGCP6. The individual binary interactions appear to be relativelyweak and in some cases were barely detectable in yeast two-hybridor co-immunoprecipitation assays, which might explain why theywere not revealed in previous studies. In this context, it is interesting

413

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

Input FLAG IP

NEDD1

GAPDH

FLAG

γ-Tubulin

GCP2GCP4

MZT1

FLAG

γ-Tubulin

GCP5

MZT1

Input FLAG IP

GAPDH

GCP5

A

B

C

FLA

GP

eric

entri

nM

erge

/DN

A

GCP3 GCP3 3A GCP5 GCP5 3A GCP6 GCP6 3A

0

40

80

120

Fluo

resc

ence

inte

nsity

[%]

Pericentrin FLAG

GCP3 3A

GCP3

GCP5 3A

GCP5

GCP6 3A

GCP6

********

****

FLAG-GFP

FLAG-GCP-3

FLAG-GCP3 3A

FLAG-GCP5

FLAG-GCP5 3A

FLAG-GFP

FLAG-GCP-3

FLAG-GCP3 3A

FLAG-GCP5

FLAG-GCP5 3A

FLAG-GCP-6

FLAG-GCP6 3A

FLAG-GCP-6

FLAG-GCP6 3A

(kDa) (kDa)100 150

100 75

50

37

150

75 10

37

250

50

10010

37

E

D

F

0

50

100

150

Cen

troso

mal

γ-tu

bulin

inte

nsity

[%]

********

0

5

10

15

20

25

Mito

tic In

dex

[%]

******

Ctrl siRNA GCP3 siRNA+EGFP +EGFP +GCP3 +GCP3 3A

Ctrl siRNA GCP3 siRNA+EGFP +EGFP +GCP3 +GCP3 3A

GCP3 siRNA + GCP3 GCP3 siRNA + GCP3 3Ainterphase mitosis mitosisinterphase

FLA

Gγ-

Tubu

linN

ED

D1

Fig. 5. See next page for legend.

414

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

to note that one study showed weak interaction of the ArabidopsisMZT1 homologs GIP1a and GIP1b with GCP2 (Nakamura et al.,2012). However, the authors did not interpret this as a positiveinteraction, possibly due to a much stronger interaction with GCP3that was observed in the same assay.Although binary interactions betweenMZT1 and GCPs appear to

be weak, binding of MZT1 to γTuRC is robust: in sucrose gradientsthe majority of total MZT1 is consistently found to be associatedwith γTuRC and only a small amount is present in lower molecularmass fractions. Our data is consistent with a cooperative bindingmode that allows strong and specific interaction of MZT1 withγTuRC by binding to multiple GCPs. This is in agreement with ourobservation that mutation of the MZT1-binding motif in only asingle GCP impairs the interaction of γTuRC with MZT1 and, as aresult, NEDD1-dependent centrosomal targeting. Considering itssmall size (∼8 kDa), binding of MZT1 to multiple GCPs likelyinvolves multiple MZT1 molecules, possibly in an oligomeric form.Indeed, purified recombinant MZT1 has been previously shown toform oligomers in solution (Dhani et al., 2013).Based on our results, we propose that the function of MZT1 does

not require interaction with specific GCPs but rather with thecommon conserved motif in the NTEs, either in the form of γTuRCor, in organisms that contain MZT1 but lack GCP4, GCP5, andGCP6 (Lin et al., 2015), in the form of γTuSC oligomers (Fig. 7).The essential role ofMZT1 does not seem to be γTuRC assembly,

given that MZT1 depletion conditions that produced severe cellulardefects, still allowed detection of γTuRC by sucrose gradientfractionation. InsteadMZT1 seems to ‘prime’ γTuRC for interactionwith other regulatory factors. An open question is how MZT1promotes these interactions. We envision two scenarios that are notmutually exclusive. MZT1might directly interact with targeting andactivation factors to promote their binding to γTuRC or may actindirectly, by inducing a conformational change in the γTuRCstructure that increases affinity or exposes a binding site for theseproteins (Fig. 7).In summary, our work demonstrates that human MZT1 is central

to the regulation of microtubule nucleation by linking specificrecognition of fully assembled γTuRC to attachment and activation

at MTOCs. Although we cannot rule out that MZT1 also directlyregulates γTuRC, our data suggests that MZT1 functions as a‘priming’ factor that prepares γTuRC for interaction with otherproteins to control targeting and activation.

MATERIALS AND METHODSMolecular biologyFull-length MZT1 cDNA (Human MGC Verified FL clone, ID: 40122127,accession: BC125183) was amplified by PCR and inserted into the plasmidpEGFP-C1 (Clontech) containing a modified multiple clone site with FseIand AscI restriction sites. To express GCP2, GCP3, GCP4, GCP5 and GCP6full-length proteins and fragments tagged with FLAG, the respectivecDNAs (Murphy et al., 2001, 1998) were PCR amplified and cloned into apCS2+-based vector encoding an N-terminal 3xFLAG tag and carrying amodified cloning site with FseI and AscI restriction sites. For the expressionof His- or GST-tagged proteins in Escherichia coli, cDNAs encodingMZT1, GCP2 1–506, GCP3 1–552, GCP4 1–347, GCP5 1–713 and GCP61–710 were inserted into pET28a with a modified cloning site using FseIand AscI or into pGEX-4X-1 (GE Healthcare, Piscataway, NJ) using EcoRIand XhoI restriction sites. Previously described plasmids were used forexpression of GST–C-NEDD1 (amino acids 344–667) (Lüders et al., 2006),EGFP–CDK5RAP2 51-100 (Choi et al., 2010) and NEDD1–EGFP (Harenet al., 2009). For expression in yeast, MZT1 and the N-terminal fragments ofGCP2–GCP6 where inserted into plasmids pGBKT7 and pGADT7(Clontech) using modified cloning sites with FseI and AscI restrictionsites. GCP3 I93A/L94A/L97A, GCP5 I110A/L111A/L114A, and GCP6V110A/L111A/L114A mutants, and RNAi-resistant mutants of NEDD1and of GCPs generated by introduction of multiple silent mutations in thetarget sequences where made by following the QuikChange Site-DirectedMutagenesis protocol (Agilent Technologies). Sequence analysis andalignments were performed with Geneious software (Biomatters).

For RNAi-mediated depletion of GCP2, GCP3, GCP4, GCP5, GCP6 andMZT1, we used RNA oligonucleotides with the following targetingsequences: GCP2 #1, 5′-GAGCUAUGCCUGUACCUAA-3′ (#s21284,Ambion/Thermo Fisher); GCP2 #2: 5′-GGCUUGACUUCAAUGGUUU-3′ (#s21286, Ambion/Thermo Fisher); GCP3, 5′-GGACUUGCUAAAACCAGAA-3′ (Silencer Select Pre-designed siRNA, #s20395, Ambion/Thermo Fisher); GCP4, 5′-GCAAUCAAGUGGCGCCUAA-3′ (Choiet al., 2010); GCP5, 5′-GGAACAUCAUGUGGUCCAUCA-3′ (Izumiet al., 2008); GCP6, 5′-AAACGAGACUACUUCCUUA-3′ (SilencerSelect Pre-designed siRNA, #s2249992, Ambion/Thermo Fisher); andMZT1, 5′-GCUUUAUCAUCGGUUAUUA-3′ (Ambion ID s54042).RNAi-mediated depletion of NEDD1 was performed as previouslydescribed (Lüders et al., 2006).

Antibodies and reagentsAnti-MZT1, anti-GCP2, anti-GCP3, anti-GCP5 and anti-GCP6 rabbitpolyclonal antibodies were generated against human His–MZT1 or His-tagged N-terminal fragments of the human GCPs (GCP2 1–506, GCP3 1–552, GCP5 1–713 and GCP6 1–710) expressed in ArticExpress cells(Agilent Technologies), solubilized in 8 M urea and affinity-purified underdenaturing conditions using Ni-Sepharose beads (GE Healthcare). Theproteins were used for immunization of rabbits (Antibody ProductionService, Facultat de Farmàcia, Universitat de Barcelona, Spain). Anti-MZT1specific antibodies were purified by acid elution after binding to GST–MZT1 immobilized on Affi-Gel 15 resin (Biorad), the anti-GCP2-6antibodies were affinity-purified using the antigens subjected to PAGEand blotted on membranes.

Other antibodies used in this study were: mouse anti-γ-tubulin (GTU-88;Sigma-Aldrich; western blotting, 1:10,000), mouse anti-γ-tubulin (TU-30,Exbio; immunofluorescence, 1:500), rabbit anti-γ-tubulin (Sigma-Aldrich:immunofluorescence, 1:250), mouse anti-α-tubulin (DM1A, Sigma-Aldrich;immunofluorescence, 1:2000), mouse anti-NEDD1 (7D10, Abnova; westernblotting, 1:2000), rabbit anti-NEDD1 (immunofluorescence, 1:500; westernblotting, 1:2500; Lüders et al., 2006), rabbit anti-GCP3 (15719-1-AP,Proteintech; western blotting, 1:1000), rabbit anti-GCP4 (GTX115949,Genetex; western blotting, 1:500), rabbit anti-GCP5 (BS3303, Bioworld;

Fig. 5. GCPs carrying mutations in the MZT1-binding motif are defectivein binding to NEDD1 and targeting to centrosomes. (A) FLAG-tagged wild-type or MZT1-bindingmutant (‘3A’) of full-length GCP3, GCP5 andGCP6wereexpressed in HEK293 cells and immunoprecipitated (IP). The associatedproteins were identified by western blotting with specific antibodies asindicated. GAPDH was detected as a loading control and to confirm thespecificity of the immunoprecipitation. (B) FLAG-tagged constructs as in Awere expressed in U2OS cells. After staining with anti-FLAG and anti-pericentrin antibodies, and DAPI to label DNA, cells were analyzed byimmunofluorescence microscopy. Scale bar: 10 µm. (C) Centrosomal FLAGand pericentrin signal intensities of cells stained as in B were quantified andplotted as percentages relative to control cells expressing wild-type proteins(set to 100%); n>55 centrosomes combined from three independentexperiments. (D) Immunofluorescence images of HeLa cells depleted of GCP3and expressing RNAi-resistant FLAG-tagged wild-type or 3A mutant GCP3.Cells were stained with anti-FLAG, anti-NEDD1 and anti-γ-tubulin antibodies.Scale bar: 10 µm. (E) Centrosomal γ-tubulin staining was quantified ininterphase HeLa cells transfected with control or GCP3 siRNA, and withplasmid expressing FLA–EGFP, and FLAG–GCP3 or FLAG–GCP3 3Amutant. Mean intensities were plotted relative to control cells (set to 100%);n=60 centrosomes per condition combined from three independentexperiments. (F) The mitotic index was scored in cells transfected with controlor GCP3 siRNA, and with plasmid expressing FLAG–EGFP, and FLAG–GCP3or FLAG-GCP3 3A mutant; n=3 experiments, >200 cells per condition in eachexperiment. All quantitative results aremean±s.e.m. ***P<0.001; ****P<0.0001(unpaired t-test compared with wild type or as indicated).

415

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

western blotting, 1:500), rabbit anti-GCP6 (AB95172, Abcam; westernblotting, 1:2000), rabbit anti-pericentrin (1:500; Lüders et al., 2006), mouseanti-GFP (3E6, Thermo Fisher; western blotting, 1:2000;immunofluorescence, 1:500), rabbit anti-GFP (TP401, Torrey PinesBiolabs; western blotting, 1:2000; immunofluorescence, 1:500), mouseanti-Centrin 3 (H00001070-M01, Abnova; immunofluorescence, 1:500),mouse anti-FLAG (F3165, Sigma-Aldrich; immunofluorescence, 1:1000),rabbit anti-FLAG (F7425, SigmaAldrich; immunofluorescence, 1:200),mouse anti-GAPDH (sc-47724, Santa Cruz Biotechnology; westernblotting, 1:10,000). Dilutions of affinity-purified, custom-made antibodiesvaried from preparation to preparation, but were typically used 1:500 forimmunofluorescence and 1:2000 for western blotting.

Alexa-Fluor-488, Alexa-Fluor-568-, and Alexa-Fluor-350-conjugatedsecondary antibodies used for immunofluorescence microscopy were fromThermo Fisher (1:200). For co-labeling with two mouse primary antibodies,isotype-specific secondary antibodies were used. Horseradish-peroxidase-coupled secondary antibodies for western blotting were from JacksonImmunoResearch Laboratories.

Cell culture, transfection and microtubule regrowthHek293, U2OS and HeLa (all ATCC) cells were grown in Dulbecco’smodified Eagle’s medium (DMEM; Invitrogen) supplemented with 10%fetal calf serum, at 37°C with 5% CO2, and were routinely confirmed to befree of contamination. Cells were transfected with plasmid usingLipofectamine 2000 or, in the case of Hek293 cells, with calciumphosphate, and siRNA using Lipofectamine RNAiMAX (Thermo Fisher).For rescue experiments, cells were first transfected with siRNA and 48 hlater with plasmid. At 24 h after the second transfection cells were fixed andanalyzed.

For mitotic microtubule regrowth experiments, dishes containingcoverslips with transfected HeLa cells were incubated with 100 ng/mlnocodazole for ∼16 h. Nocodazole was then washed out and cells wereincubated for 30 min in an ice-water bath to depolymerize microtubules. Forregrowth experiments in interphase, transfected U2OS cells without drugtreatment were incubated 30 min in an ice-water bath. To allow microtubuleregrowth, coverslips with cells were moved to medium at 37°C, followed bymethanol fixation.

A

B

0

100

200

300

400

500

600

700

Control GCP2 MZT1 Control GCP2 MZT1

Cyt

opla

smic

Flu

ores

cenc

e In

tens

ity [%

]

siRNA:

GFP GFP-CDK5RAP2 (51-100)

siRNA: Control GCP2 MZT1

GFP

α-T

ubul

in

GFP

Control GCP2 MZT1

GFP-CDK5RAP2 (51-100)

******

**** ****

C

GCP6

GCP2

GFP

MZT1

Input IP

Control siRNA

GFPMZT1 siRNA

GFP-C5R2 (51-100)

+ - + -- + - ++ + - -- - + +

+ - + -- + - ++ + - -- - + +

D

250

100

25

(kDa)

10

ControlMZT1siRNA:

CDK5RAP2

MZT1GAPDH

250

10

37

(kDa)

Fig. 6. MZT1 is required for the activation of γTuRC nucleation activity by mediating interaction with the CDK5RAP2 51–100 fragment.(A) Immunofluorescence images of U2OS cells were transfected with siRNA and with GFP or GFP–CDK5RAP2 51-100, as indicated. Microtubules weredepolymerized by being subjected to cold and allowed to regrow for 10 s before fixation. Cells were stained with anti-GFP and anti-α-tubulin antibodies. Themagnified regions show nucleated cytoplasmic microtubules for each condition. (B) The amount of nucleated microtubules in the entire non-centrosomalcytoplasmic area of cells as in Awas quantified and plotted as percentages of the amount of microtubules measured in cells transfected with control siRNA andGFP (set to 100%); n>76 cells per condition combined from three independent experiments. Results aremean±s.e.m. ***P<0.001; ****P<0.0001 (unpaired t-test).(C) HEK293 cells transfected as in A were subjected to immunoprecipitation (IP) with anti-GFP antibodies. Samples were analyzed by western blotting withantibodies against the indicated proteins. (D) Western blot of MZT1-depleted extract probed with the indicated antibodies.

416

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

Yeast two-hybrid assayTo test the interactions between GCPs and MZT1 the Matchmaker two-hybrid system was used according to the manufacturer’s protocol(Clontech). Briefly, plasmids containing a GAL4 DNA-binding domain(pGBKT7 backbone) and plasmids containing a GAL4 activation domain(pGADT7 backbone) carrying the desired cDNAs were transformed intoyeast strains AH109 and Y187, respectively. Yeast mating was performedin 2× YPDA medium for 16–20 h, and diploids were selected by leucineand tryptophan prototrophy. Protein interactions were identified selectingfor histidine and adenine prototrophy. Mel1 activity of yeast strainswith positive protein interactions was assayed by adding the substrateX-α-Gal (5-bromo- 4-chloro- 3-indolyl-α-D-galactopyranoside, Clontech)directly to the plates. Yeast strains carrying pGADT7-T (SV40large T-antigen 84-708) and pGBKT7-53 (encoding murine p53 72–390),and pGADT7-T (encoding SV40 large T-antigen 84-708) and pGBKT7-Lam (encoding human lamin C 66–230) served as positive and negativecontrols, respectively.

Immunoprecipitation and GST pulldownTransfected cells were washed in PBS and lysed in buffer (50 mM HEPES,pH 7.5, 150 mM NaCl, 1 mM MgCl2, 1 mM EGTA, 0.5% NP-40 andprotease inhibitors) for 10 min on ice. Cleared lysates were obtained bycentrifugation for 15 min at 16,000 g at 4°C. For immunoprecipitation,lysates were incubated with antibodies for 1 h at 4°C. Sepharose–protein-Gbeads (GE Healthcare) were added and the mixture was incubated for anadditional hour at 4°C. For GST pulldown, purified recombinant GST orGST–C-NEDD1 (Lüders et al., 2006) pre-bound to glutathione–Sepharosebeads (GE Healthcare) were added to cleared lysates and incubated for 2 h.The beads were pelleted and washed three times with lysis buffer. Sampleswere prepared for SDS-PAGE by boiling in sample buffer.

Western blottingCells were washed in PBS and lysed in buffer (50 mM HEPES pH 7.5,150 mM NaCl, 1 mM MgCl2, 1 mM EGTA, 0.5% NP-40 and proteaseinhibitors) on ice. Cleared extracts were prepared by centrifugation and, afterdetermination of protein concentration by Bradford assay (Biorad),subjected to SDS-PAGE. Proteins were transferred to PVDF ornitrocellulose membranes by tank blotting with transfer buffer containing20% methanol and 0.1% SDS for 90 min and probed with antibodies. Fordetection of MZT1 transfer buffer without SDS and PVDF membrane with0.2 µm pore size was used, and transfer time was reduced to 60 min.

Sucrose gradient centrifugation300 µl of cell extract prepared as above was loaded on a 4.2 ml 10–40%sucrose gradient prepared in 50 mM HEPES, pH 7.5, 150 mM NaCl, 1 mMMgCl2 and 1 mM EGTA and centrifuged in a SW-55Ti rotor (Beckman) for4 h at ∼366,000 g at 4°C. Fractions were collected from the top of thegradient by pipetting and analyzed by western blotting. As molecular massstandards, aldolase (158K, 7S) and thyroglobulin (669K, 19S) (GEHealthcare) were analyzed in gradients prepared and run under identicalconditions.

Fluorescence microscopyHeLa and U2OS cells grown on coverslips were fixed in methanol at −20°Cfor at least 5 min and processed for immunofluorescence. Fixed cells wereblocked in PBS-BT (1× PBS, 0.1% Triton X-100 and 3% BSA) andincubated with antibodies in the same buffer. Images were acquired with anOrca AG camera (Hamamatsu) on a Leica DMI6000Bmicroscope equippedwith 1.4 NA 63× and 100× oil immersion objectives. AF6000 software(Leica) was used for image acquisition. Image processing and quantificationof fluorescence intensities was performed with ImageJ software. Intensitieswere measured in images acquired with constant exposure settings andwere background-corrected. To quantify non-centrosomal, cytoplasmicmicrotubules in regrowth experiments, the cell boundaries and the areaaround the centrosome containing astral microtubules were outlined inImageJ. The intensity value obtained for astral microtubules around thecentrosome was subtracted from the intensity value obtained for totalmicrotubule staining within the cell boundaries.

Statistical analysisStatistical analysis was performed using Prism 6 software. Two-tailed,unpaired t-tests were performed to compare experimental groups. Theresults are reported in the figures and figure legends.

AcknowledgementsWe are grateful to visiting student Eva Domenjo for her help withimmunoprecipitation experiments.

Competing interestsThe authors declare no competing or financial interests.

Author contributionsConceptualization: R.R.C., N.T.-T., J.R. and J.L.; Methodology, formal analysis,investigation: R.R.C., N.T.-T., A.E., S.E. and C.L.; Resources: R.R.C., N.T.-T., C.L.;Writing – original draft preparation: J.L.; Writing – review and editing: R.R.C., N.T.-T.,

active

inactive

γTuSC

active

adapters

γTuRC

MZT1

‘priming’

targeting& activation

inactive

γTuSC-like?

Fig. 7. Model for the role of MZT1 in linking γTuRC assembly to interactionwith adapters that mediate targeting and activation. Cytosolic γTuRC andsubcomplexes, such as γTuSC and hypothetical γTuSC-like complexes(Farache et al., 2016), are inactive as nucleators. Interaction of γTuRCsubcomplexes with MZT1 is weak (dashed arrows). Only fully assembledγTuRC interacts efficiently with MZT1 owing to a cooperative binding modeinvolving multiple MZT1 molecules and oligomeric GCPs. The positions ofGCP4, GCP5 and GCP6 in the γTuRC are unknown and speculatively shownto be at the seam. Binding of MZT to the N-termini of the GCPs is a ‘priming’step that allows γTuRC to interact with different regulatory adapter proteins.MZT1 mediates the interaction with adapters either directly or indirectly, forexample, by inducing a conformational change that reveals or increases theaffinity of a binding site for the adapter (bright spot; position hypothetical). Theadapters might also interact as oligomers, but for simplicity only monomericadapters are shown. The adapters serve as targeting and activation factors thatattach γTuRC at MTOCs, such as the centrosome, and activate its nucleationactivity. Targeting and activation might be mediated by the same adapter or bydistinct adapters as recently suggested (Muroyama et al., 2016).

417

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

A.E., J.R. and J.L.; Visualization: R.R.C., N.T.-T., A.E. and J.L.; Supervision, projectadministration, funding acquisition: J.R. and J.L.

FundingThis study was supported by grants from the Ministerio de Economıa yCompetitividad (MINECO) (BFU2009-08522, BFU2012-33960 and BFU2015-69275-P to J.L. and BFU2014-58422-P to J.R.), and IRB Barcelona intramuralfunds. J.L. acknowledges additional support from the Ramon y Cajal Programme(RYC-2010-07108 MINECO, Spain). R.R.C. was supported by a PhD fellowshipfrom the Consejo Nacional de Ciencia y Tecnologıa (Mexican National Council forScience and Technology; CONACYT) (CVU 269229).

Supplementary informationSupplementary information available online athttp://jcs.biologists.org/lookup/doi/10.1242/jcs.195321.supplemental

ReferencesAnders, A., Lourenço, P. C. C. and Sawin, K. E. (2006). Noncore components ofthe fission yeast gamma-tubulin complex. Mol. Biol. Cell 17, 5075-5093.

Bahtz, R., Seidler, J., Arnold, M., Haselmann-Weiss, U., Antony, C., Lehmann,W. D. and Hoffmann, I. (2012). GCP6 is a substrate of Plk4 and required forcentriole duplication. J. Cell Sci. 125, 486-496.

Bauer, M., Cubizolles, F., Schmidt, A. and Nigg, E. A. (2016). Quantitativeanalysis of human centrosome architecture by targeted proteomics andfluorescence imaging. EMBO J. 35, 2152-2166.

Choi, Y.-K., Liu, P., Sze, S. K., Dai, C. and Qi, R. Z. (2010). CDK5RAP2 stimulatesmicrotubule nucleation by the gamma-tubulin ring complex. J. Cell Biol. 191,1089-1095.

Dhani, D. K., Goult, B. T., George, G. M., Rogerson, D. T., Bitton, D. A., Miller,C. J., Schwabe, J. W. R. and Tanaka, K. (2013). Mzt1/Tam4, a fission yeastMOZART1 homologue, is an essential component of the -tubulin complex anddirectly interacts with GCP3Alp6. Mol. Biol. Cell 24, 3337-3349.

Edgar, R. C. (2004). MUSCLE: multiple sequence alignment with high accuracy andhigh throughput. Nucleic Acids Res. 32, 1792-1797.

Farache, D., Jauneau, A., Chemin, C., Chartrain, M., Remy, M.-H., Merdes, A.and Haren, L. (2016). Functional analysis of gamma-tubulin complex proteinsindicates specific lateral association via their N-terminal domains. J. Biol. Chem.291, 23112-23125.

Fong, K.-W., Choi, Y.-K., Rattner, J. B. and Qi, R. Z. (2008). CDK5RAP2 is apericentriolar protein that functions in centrosomal attachment of the gamma-tubulin ring complex. Mol. Biol. Cell 19, 115-125.

Fujita, A., Vardy, L., Garcia, M. A. and Toda, T. (2002). A fourth component of thefission yeast gamma-tubulin complex, Alp16, is required for cytoplasmicmicrotubule integrity and becomes indispensable when gamma-tubulin functionis compromised. Mol. Biol. Cell 13, 2360-2373.

Gomez-Ferreria, M. A., Bashkurov, M., Helbig, A. O., Larsen, B., Pawson, T.,Gingras, A.-C. and Pelletier, L. (2012). Novel NEDD1 phosphorylation sitesregulate γ-tubulin binding and mitotic spindle assembly. J. Cell Sci. 125,3745-3751.

Gunawardane, R. N., Martin, O. C., Cao, K., Zhang, L., Dej, K., Iwamatsu, A. andZheng, Y. (2000). Characterization and reconstitution of Drosophila gamma-tubulin ring complex subunits. J. Cell Biol. 151, 1513-1524.

Haren, L., Remy, M.-H., Bazin, I., Callebaut, I., Wright, M. and Merdes, A. (2006).NEDD1-dependent recruitment of the gamma-tubulin ring complex to thecentrosome is necessary for centriole duplication and spindle assembly. J. CellBiol. 172, 505-515.

Haren, L., Stearns, T. and Luders, J. (2009). Plk1-dependent recruitment ofgamma-tubulin complexes to mitotic centrosomes involves multiple PCMcomponents. PLoS ONE 4, e5976.

Hutchins, J. R. A., Toyoda, Y., Hegemann, B., Poser, I., Heriche, J.-K., Sykora,M. M., Augsburg, M., Hudecz, O., Buschhorn, B. A., Bulkescher, J. et al.(2010). Systematic analysis of human protein complexes identifies chromosomesegregation proteins. Science 328, 593-599.

Izumi, N., Fumoto, K., Izumi, S. and Kikuchi, A. (2008). GSK-3beta regulatesproper mitotic spindle formation in cooperation with a component of the gamma-tubulin ring complex, GCP5. J. Biol. Chem. 283, 12981-12991.

Jakobsen, L., Vanselow, K., Skogs, M., Toyoda, Y., Lundberg, E., Poser, I.,Falkenby, L. G., Bennetzen, M., Westendorf, J., Nigg, E. A. et al. (2011). Novelasymmetrically localizing components of human centrosomes identified bycomplementary proteomics methods. EMBO J. 30, 1520-1535.

Janski, N., Masoud, K., Batzenschlager, M., Herzog, E., Evrard, J.-L., Houlne,G., Bourge, M., Chaboute, M.-E. and Schmit, A.-C. (2012). The GCP3-interacting proteins GIP1 and GIP2 are required for γ-tubulin complex proteinlocalization, spindle integrity, and chromosomal stability. Plant Cell 24,1171-1187.

Johmura, Y., Soung, N.-K., Park, J.-E., Yu, L.-R., Zhou, M., Bang, J. K., Kim,B.-Y., Veenstra, T. D., Erikson, R. L. and Lee, K. S. (2011). Regulation of

microtubule-based microtubule nucleation by mammalian polo-like kinase 1.Proc. Natl. Acad. Sci. USA 108, 11446-11451.

Knop, M. and Schiebel, E. (1997). Spc98p and Spc97p of the yeast gamma-tubulincomplex mediate binding to the spindle pole body via their interaction withSpc110p. EMBO J. 16, 6985-6995.

Kollman, J. M., Polka, J. K., Zelter, A., Davis, T. N. and Agard, D. A. (2010).Microtubule nucleating gamma-TuSC assembles structures with 13-foldmicrotubule-like symmetry. Nature 466, 879-882.

Kollman, J. M., Merdes, A., Mourey, L. and Agard, D. A. (2011). Microtubulenucleation by γ-tubulin complexes. Nat. Rev. Mol. Cell Biol. 12, 709-721.

Kollman, J. M., Greenberg, C. H., Li, S., Moritz, M., Zelter, A., Fong, K. K.,Fernandez, J.-J., Sali, A., Kilmartin, J., Davis, T. N. et al. (2015). Ring closureactivates yeast γTuRC for species-specific microtubule nucleation. Nat. Struct.Mol. Biol. 22, 132-137.

Lin, T.-C., Gombos, L., Neuner, A., Sebastian, D., Olsen, J. V., Hrle, A., Benda,C. and Schiebel, E. (2011). Phosphorylation of the yeast γ-tubulin Tub4 regulatesmicrotubule function. PLoS ONE 6, e19700.

Lin, T.-C., Neuner, A., Schlosser, Y. T., Scharf, A. N. D., Weber, L. and Schiebel,E. (2014). Cell-cycle dependent phosphorylation of yeast pericentrin regulatesγ-TuSC-mediated microtubule nucleation. eLife 3, e02208.

Lin, T.-C., Neuner, A. and Schiebel, E. (2015). Targeting of γ-tubulin complexes tomicrotubule organizing centers: conservation and divergence. Trends Cell Biol.25, 296-307.

Luders, J. and Stearns, T. (2007). Microtubule-organizing centres: a re-evaluation.Nat. Rev. Mol. Cell Biol. 8, 161-167.

Luders, J., Patel, U. K. and Stearns, T. (2006). GCP-WD is a gamma-tubulintargeting factor required for centrosomal and chromatin-mediated microtubulenucleation. Nat. Cell Biol. 8, 137-147.

Lynch, E. M., Groocock, L. M., Borek, W. E. and Sawin, K. E. (2014). Activation ofthe γ-Tubulin Complex by the Mto1/2 Complex. Curr. Biol. 24, 896-903.

Manning, J. A., Shalini, S., Risk, J. M., Day, C. L. and Kumar, S. (2010). A directinteraction with NEDD1 regulates gamma-tubulin recruitment to the centrosome.PLoS ONE 5, e9618.

Masuda, H. and Toda, T. (2016). Synergistic role of fission yeast Alp16GCP6 andMzt1MOZART1 in γ-tubulin complex recruitment to mitotic spindle pole bodiesand spindle assembly. Mol. Biol. Cell 27, 1753-1763.

Masuda, H., Mori, R., Yukawa, M. and Toda, T. (2013). Fission yeast MOZART1/Mzt1 is an essential -tubulin complex component required for complex recruitmentto the microtubule organizing center, but not its assembly. Mol. Biol. Cell 24,2894-2906.

Moudjou, M., Bordes, N., Paintrand, M. and Bornens, M. (1996). Gamma-Tubulinin mammalian cells: the centrosomal and the cytosolic forms. J Cell Sci 109,875-887.

Muroyama, A., Seldin, L. and Lechler, T. (2016). Divergent regulation offunctionally distinct γ-tubulin complexes during differentiation. J. Cell Biol. 213,679-692.

Murphy, S. M., Urbani, L. and Stearns, T. (1998). The mammalian gamma-tubulincomplex contains homologues of the yeast spindle pole body components spc97pand spc98p. J. Cell Biol. 141, 663-674.

Murphy, S. M., Preble, A. M., Patel, U. K., O’Connell, K. L., Dias, D. P., Moritz, M.,Agard, D., Stults, J. T. and Stearns, T. (2001). GCP5 and GCP6: two newmembers of the human gamma-tubulin complex. Mol. Biol. Cell 12, 3340-3352.

Nakamura, M., Yagi, N., Kato, T., Fujita, S., Kawashima, N., Ehrhardt, D. W. andHashimoto, T. (2012). Arabidopsis GCP3-interacting protein 1/MOZART 1 is anintegral component of the γ-tubulin-containing microtubule nucleating complex.Plant J 71, 216-225.

Nguyen, T., Vinh, D. B., Crawford, D. K. andDavis, T. N. (1998). A genetic analysisof interactions with Spc110p reveals distinct functions of Spc97p and Spc98p,components of the yeast gamma-tubulin complex. Mol. Biol. Cell 9, 2201-2216.

Oakley, B. R., Paolillo, V. and Zheng, Y. (2015). γ-Tubulin complexes inmicrotubule nucleation and beyond. Mol. Biol. Cell 26, 2957-2962.

Petry, S. and Vale, R. D. (2015). Microtubule nucleation at the centrosome andbeyond. Nat. Cell Biol. 17, 1089-1093.

Pinyol, R., Scrofani, J. and Vernos, I. (2013). The role of NEDD1 phosphorylationby Aurora A in chromosomal microtubule nucleation and spindle function. Curr.Biol. 23, 143-149.

Remy, M.-H., Merdes, A. and Gregory-Pauron, L. (2013). Assembly of gamma-tubulin ring complexes: implications for cell biology and disease. Prog. Mol. Biol.Transl. Sci. 117, 511-530.

Rivero, S., Cardenas, J., Bornens, M. and Rios, R. M. (2009). Microtubulenucleation at the cis-side of the Golgi apparatus requires AKAP450 and GM130.EMBO J. 28, 1016-1028.

Roubin, R., Acquaviva, C., Chevrier, V., Sedjaï, F., Zyss, D., Birnbaum, D. andRosnet, O. (2013). Myomegalin is necessary for the formation of centrosomal andGolgi-derived microtubules. Biol. Open 2, 238-250.

Samejima, I., Miller, V. J., Groocock, L. M. and Sawin, K. E. (2008). Two distinctregions of Mto1 are required for normal microtubule nucleation and efficientassociation with the gamma-tubulin complex in vivo. J. Cell Sci. 121, 3971-3980.

418

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience

Sawin, K. E., Lourenço, P. C. C. and Snaith, H. A. (2004). Microtubule nucleationat non-spindle pole body microtubule-organizing centers requires fission yeastcentrosomin-related protein mod20p. Curr. Biol. 14, 763-775.

Scheidecker, S., Etard, C., Haren, L., Stoetzel, C., Hull, S., Arno, G., Plagnol, V.,Drunat, S., Passemard, S., Toutain, A. et al. (2015). Mutations in TUBGCP4alter microtubule organization via the γ-tubulin ring complex in autosomal-recessive microcephaly with chorioretinopathy. Am. J. Hum. Genet. 96, 666-674.

Scrofani, J., Sardon, T., Meunier, S. and Vernos, I. (2015). Microtubule nucleationin mitosis by a RanGTP-dependent protein complex. Curr. Biol. 25, 131-140.

Sdelci, S., Schutz, M., Pinyol, R., Bertran, M. T., Regue, L., Caelles, C., Vernos, I.and Roig, J. (2012). Nek9 phosphorylation of NEDD1/GCP-WD contributes toPlk1 control of gamma-tubulin recruitment to the mitotic centrosome. Curr. Biol.22, 1516-1523.

Takahashi, M., Yamagiwa, A., Nishimura, T., Mukai, H. and Ono, Y. (2002).Centrosomal proteins CG-NAP and kendrin provide microtubule nucleation sitesby anchoring gamma-tubulin ring complex. Mol. Biol. Cell 13, 3235-3245.

Teixido-Travesa, N., Villen, J., Lacasa, C., Bertran, M. T., Archinti, M., Gygi,S. P., Caelles, C., Roig, J. and Luders, J. (2010). The gammaTuRC revisited: acomparative analysis of interphase and mitotic human gammaTuRC redefines the

set of core components and identifies the novel subunit GCP8.Mol. Biol. Cell 21,3963-3972.

Teixido-Travesa, N., Roig, J. and Luders, J. (2012). The where, when and how ofmicrotubule nucleation - one ring to rule them all. J. Cell Sci. 125, 4445-4456.

Venkatram, S., Tasto, J. J., Feoktistova, A., Jennings, J. L., Link, A. J. andGould, K. L. (2004). Identification and characterization of two novel proteinsaffecting fission yeast gamma-tubulin complex function. Mol. Biol. Cell 15,2287-2301.

Verollet, C., Colombie, N., Daubon, T., Bourbon, H.-M., Wright, M. andRaynaud-Messina, B. (2006). Drosophila melanogaster gamma-TuRC isdispensable for targeting gamma-tubulin to the centrosome and microtubulenucleation. J. Cell Biol. 172, 517-528.

Wang, Z., Wu, T., Shi, L., Zhang, L., Zheng, W., Qu, J. Y., Niu, R. and Qi, R. Z.(2010). Conserved motif of CDK5RAP2 mediates its localization to centrosomesand the Golgi complex. J. Biol. Chem. 285, 22658-22665.

Xiong, Y. and Oakley, B. R. (2009). In vivo analysis of the functions of gamma-tubulin-complex proteins. J. Cell Sci. 122, 4218-4227.

Zimmerman, W. C., Sillibourne, J., Rosa, J. and Doxsey, S. J. (2004). Mitosis-specific anchoring of gamma tubulin complexes by pericentrin controls spindleorganization and mitotic entry. Mol. Biol. Cell 15, 3642-3657.

419

RESEARCH ARTICLE Journal of Cell Science (2017) 130, 406-419 doi:10.1242/jcs.195321

Journal

ofCe

llScience