Tubulin isotype substitution revealed that isotype combination modulates microtubule dynamics in C. elegans embryos

Microtubules (MTs) are dynamic polymers composed of αβ-tubulin heterodimers that play diverse roles in processes such as cell division, cell motility and intracellular transport. MTs are dynamic structures that shift rapidly between growth (polymerization) and shrinkage (depolymerization) at both their plus and minus ends (Mitchison and Kirschner, 1984). This property − called dynamic instability − is important for diverse MT functions, and is temporally and spatially regulated.

In many organisms, both α- and β-tubulin are encoded at multiple genomic loci. 40 years ago, the ‘multi-tubulin hypothesis’ was proposed, which postulated that different tubulin isotypes are required for specialized MT functions (Fulton and Simpson, 1976). Consistent with this hypothesis, distinct roles of tubulin isotypes have been implicated. In humans, mutations of different tubulin isotypes cause distinct neurological disorders (for reviews, see Chakraborti et al., 2016; Romaniello et al., 2015). In Drosophila, a testis-specific isotype cannot be replaced by a somatic isotype (Fackenthal et al., 1993; Hoyle and Raff, 1990). In C. elegans, the expression of specific α- and β-tubulin isotypes in neurons generates MTs with a specific structure (Chalfie and Thomson, 1982; Fukushige et al., 1999; Savage et al., 1989). In vitro studies have demonstrated the differences of polymerization properties between tubulin isotypes (Banerjee et al., 1992; Lu and Luduena, 1994; Panda et al., 1994). Two α-tubulins in S. cerevisiae confer distinct dynamics on MTs in vitro, although they are interchangeable for viability in vivo (Bode et al., 2003; Schatz et al., 1986). Despite these implications, how multiple tubulin isotypes within a cell contribute to specific MT dynamics remains unclear, partly due to technical difficulties in the visualization and manipulation of endogenous tubulin isotypes within live organisms.

Of the nine α-tubulins and six β-tubulins in C. elegans (Gogonea et al., 1999), two α-tubulins (tba-1 and tba-2; hereafter referred to as α1 and α2) and two β-tubulins (tbb-1 and tbb-2; hereafter referred to as β1 and β2) are the principal tubulin isotypes during early embryogenesis; they differ in 12 and 14 amino acids, respectively (Baugh et al., 2003) (Fig. 1A). The amino acid differences of α1 and α2 are spread out over the protein, but are concentrated in the N-terminal region (Fig. 1A; ‘38-43’ region). By contrast, β1 and β2 are most different in their C-terminal tails (Fig. 1A; ‘tail region’). Mutant phenotype analyses in early embryos indicate that, whereas α1 and α2 are regarded to be functionally equivalent, β1 and β2 have distinct roles in early embryos (Ellis et al., 2004; Lu and Mains, 2005; Phillips et al., 2004; Wright and Hunter, 2003). Whereas α1-, α2- or β1-depleted embryos are fully viable and their stereotypic oscillation of mitotic spindles during anaphase is almost normal, β2-depleted embryos show exaggerated oscillation of mitotic spindles, resulting in partial embryonic lethality (Ellis et al., 2004; Lu and Mains, 2005; Lu et al., 2004; Phillips et al., 2004; Wright and Hunter, 2003). In addition, the MT-severing heteromeric complex katanin (comprising MEI-1 and MEI-2 in C. elegans) preferentially interacts with α2 and β2, rather than with α1 and β1 (Lu and Mains, 2005; Lu et al., 2004). Furthermore, each α- or β-tubulin isotype differently affects embryonic viability (Ellis et al., 2004; Lu and Mains, 2005; Wright and Hunter, 2003). Although these reports imply that these co-expressed isotypes have some distinct properties, whether and how each isotype contributes to MT dynamics is still unclear.

Here, by using genome editing, RNA interference (RNAi) resulting in the depletion of each isotype and quantitative live imaging, we demonstrate that tubulin isotypes cooperatively contribute to MT dynamics in live C. elegans embryos. The four isotypes are expressed at different levels and show distinct MT incorporation rates in early embryos. Combination of genome-editing-mediated isotype substitution and depletion of each isotype in response to RNAi reveal that each tubulin isotype confers distinct MT dynamics in vivo, and that any alteration in the combination (ratio and concentration) of the four tubulin isotypes modulates MT dynamics and mitotic spindle oscillation.

To understand the contribution of four tubulin isotypes (α1, α2, β1 and β2) to MT dynamics in C. elegans embryo, we first quantitatively analyzed their expression levels and their incorporation into MTs in live C. elegans embryos. We used strains in which the GFP-coding sequence is directly integrated into the 5′-end of each tubulin isotype coding region through the CRISPR/Cas9 system. Expression levels of each isotype in whole animals were confirmed by immunoblotting with anti-GFP antibodies (Fig. S1A). Live imaging of GFP-tagged tubulins in early embryos (either heterozygotes or homozygotes) confirmed that their incorporation into spindle MTs is likely to be equivalent to that of endogenous non-tagged tubulins (Fig. S1B,C). Notably, when the majority of untagged β-tubulins was depleted by RNAi, GFP-tagged tubulins did not assemble MTs (Fig. S1D), suggesting that a certain amount of untagged β-tubulins is necessary for GFP-tagged β-tubulins to participate in MT assembly. Therefore, to measure expression levels and MT incorporation rates of tubulin isotypes, these GFP-insertion alleles were used only in the presence of untagged tubulins.

The amount of each tubulin isotype in live one-cell embryos was quantified by the total GFP fluorescence intensity in a whole embryo (Fig. 1B,C). The total amount of α-tubulins (α1 and α2) was approximately four-fold higher than that of β-tubulins (β1 and β2) (Fig. 1C). Within the same family of tubulins, the level of β2 [median of relative fluorescence intensity; 1983 arbitrary units (a.u.)] was 1.7-fold higher than β1 (1150 a.u.), and that of α1 (7038 a.u.) was 1.2-fold higher than that of α2 (5971 a.u.). The proportion of tubulins incorporated into MTs was estimated by calculating the ratio of centrosomal GFP signal (incorporated into spindle MTs) to that of cytosolic GFP (unincorporated free tubulins) during metaphase (Fig. 1D). Consistent with the higher level of α-tubulins compared with that of β-tubulins, the proportion of tubulin proteins incorporated into MTs was higher in β than α (Fig. 1D). These results indicate that the amount of β-tubulins is likely to be the limiting factor for MT formation in early embryos.

Within the same family of tubulins, β2 and α1 (3.35 and 1.93 a.u., respectively) were more readily incorporated into MTs than β1 and α2 (2.85 and 1.54 a.u., respectively) (Fig. 1D), implying that each isotype has distinct properties for polymerization/depolymerization and/or stability of MTs.

To analyze how the combination (ratio and concentration) of β-isotypes affects MT behaviors in vivo, we used a series of strains that express β1 and/or β2 at different levels (Fig. 2A). The wild-type strain expresses both β1 and β2, whose genotype is hereafter presented as [β1;β2]. The β1 or β2 null mutant strains are [Δβ1;β2] (tbb-1(gk207)) and [β1;Δβ2] (tbb-2(gk129)) (Wright and Hunter, 2003), and express only β2 or β1, respectively, during early embryogenesis (Fig. 2A). The total β-tubulin levels estimated by immunoblotting of [Δβ1;β2] and [β1;Δβ2] adult worms were 70% and 19%, respectively, of wild-type [β1;β2] (Fig. 2B), which correlated with the expression levels of each isotype described above (Fig. 1C).

To compare in vivo properties of β1 and β2 at the same concentration, we generated the isotype substitution strains [β2;β2] (tbb-1(tj59)) and [β1;β1] (tbb-2(tj41)), in which the coding sequence of β1 or β2 was genome edited to encode β2 or β1, respectively (Fig. 2A). In order to minimize alteration of expression levels from the modified loci, only the relevant 14 codons were substituted, and original exon-intron structures were maintained. As expected, the total amount of β-tubulins in [β1;β2], [β2;β2] and [β1;β1] strains was almost equal − as confirmed by immunoblotting (100%, 106% and 113%, respectively) (Fig. 2B).

Because [β1;Δβ2] embryos had been reported to show partial lethality (Ellis et al., 2004; Lu and Mains, 2005; Lu et al., 2004; Wright and Hunter, 2003), we first examined the effect of isotype substitution on embryonic lethality (Fig. 2C). Although, as reported, [β1;Δβ2] embryos were partially lethal (47%), [β1;β2], [Δβ1;β2], [β2;β2] and [β1;β1] embryos showed rates of low lethality (1.7%, 2.6%, 1.2% and 1.2%, respectively).

This is despite the fact that both [β1;Δβ2] and [β1;β1] embryos express β1 only, with levels of β-tubulin compared with wild type 19% and 113%, respectively. These results, therefore, indicate that the lethality of [β1;Δβ2] is likely to be caused by the paucity of the total β-tubulins in these embryos, rather than the absence of a β2-specific function. Next, by using the above strains, we analyzed the effect different combinations (ratio and concentration) of β-isotypes have on MT behavior.

Using the five strains ([β1;β2], [Δβ1;β2], [β1;Δβ2], [β2;β2] and [β1;β1]) described above, we quantitatively analyzed the stereotypic mitotic spindle oscillation during anaphase in one-cell embryos as a readout of MT behaviors (Fig. 3). First, by using strains expressing a single β-isotype, the effect of β-tubulin concentration on spindle oscillation was examined (Fig. 3). As mentioned above, β-tubulin levels of [β1;Δβ2] and [β1;β1] (expressing only β1) were 19% and 113%, respectively. However, spindle oscillation in [β1;Δβ2] was much more exaggerated − with a median of the maximum posterior centrosome oscillation of 10.7 µm – than in [β1;β1] (median: 6.37 µm) (Fig. 3B). Likewise, [Δβ1;β2] and [β2;β2] that express only β2, were compared (β level: 70% and 106%, respectively, of the wild type, see above). Despite the lower level of β-tubulin in [Δβ1;β2] than [β2;β2], no significant difference in spindle oscillation was detected (6.14 µm and 4.97 µm, respectively). These results indicate that, when the total concentration of β-tubulin was severely decreased in [β1;Δβ2], its lower concentration resulted in more active spindle oscillation. However, in [Δβ1;β2], the total concentration of β-tubulin might be still sufficient and spindle oscillation might remain normal.

Next, by using strains that express similar amount of total β-tubulins ([β1;β2], [β1;β1] and [β2;β2]), we examined whether β1 and β2 distinctly affect mitotic spindle oscillation. Compared to the spindle oscillation of [β2;β2] embryos (4.97 µm), [β1;β1] embryos showed exaggerated oscillation (6.37 µm) (Fig. 3B). Thus, at an equivalent concentration of total β-tubulins, mitotic spindles composed of β1-MTs oscillate more actively than those composed of β2-MTs. When β1 and β2 are mixed, the spindles oscillate at a level between the two.

These results indicate that β1-composed and β2-composed MTs have distinct properties that affect mitotic spindle oscillations, and that the ratio and concentration of the isotypes contribute to create specific MT behaviors.

To test the hypothesis that β1 and β2 have distinct properties that trigger specific MT behaviors, we analyzed MT dynamics in vivo by using GFP-tagged EBP-2, a homologue of the human EB1 protein (also known as MAPRE1), that binds to plus-ends of growing MTs (Srayko et al., 2005). From the time-lapse images of EBP-2::GFP comets on mitotic astral MTs in one-cell embryos, the speed and lifetime of comets were measured, which correspond to MT growth rate and duration of growth (Fig. 4A).

First, the effect of β-tubulin concentration on the MTs composed of a single β-isotype was examined in strains in which β1 or β2 was depleted by using RNAi (hereafter referred to as [^RNAiβ1;β2] or [β1;^RNAiβ2], respectively) (Fig. S2A). RNAi depletion of β1 or β2 resulted in levels of embryonic lethality that were similar to those of corresponding deletion mutants (Fig. S2B). The efficiency and specificity of β1 or β2 depletion by using RNAi was confirmed by measuring the decrease of the GFP signal when GFP was tagged to β1 or β2. The GFP::β1 signals were greatly decreased following RNAi depletion of β1 but not following that of β2, whereas those of GFP::β2 were decreased following depletion of β2 (Fig. S2C-F), indicating the high specificity and efficiency of isotype-specific RNAi.

MTs in [β1;^RNAiβ2] grew significantly slower (0.66 µm/s) compared with those in [β1;β1] (0.87 µm/s) (Fig. 4B, P<0.05), and the comet lifetime was also shorter (blue bars indicate the percentage of comets showing a lifetime of ≤1.4 s, i.e. 71% and 58%, respectively, P<0.01) (Fig. 4C), suggesting that β1-composed MTs grow slower and are less stable when β1 concentration is low. These MT properties in [β1;Δβ2] resulted in shorter astral microtubules (Fig. S3), which might account for hyperactive spindle oscillation during anaphase (Fig. 3A). Alternatively, because spindle oscillation has been reported to depend on interaction of astral MTs with the cell cortex (Kotak and Gönczy, 2013; Spiró et al., 2014), loss of β2 might affect specific interaction with cortical proteins, such as dynein – which could indirectly affect spindle oscillation. In the case of β2, the comet lifetime was shorter in [^RNAiβ1;β2] compared with [β2;β2] (≤1.4 s, blue bars: 51% and 37%, respectively, P<0.01) (Fig. 4C), while growth rates of [^RNAiβ1;β2] and [β2;β2] were similar (Fig. 4B). Thus, for both isotypes, reduced concentration of β-tubulins resulted in more dynamic MTs.

Next, the three strains [β1;β2], [β2;β2] and [β1;β1], which contain similar amounts of β-tubulins, were compared (Fig. 4B,C). Although their growth rates were almost similar (median 0.92, 0.86 and 0.87 µm/s, respectively) (Fig. 4B), their comet lifetime was different (Fig. 4C). The proportion of long-lifetime comets (≥8.4 s, red bars) was significantly larger in [β2;β2] (7.1%) than in [β1;β2] (0.9%, P<0.01) and [β1;β1] (1.3%, P<0.01). However, the proportion of short-lifetime comets (≤1.4 s, blue bars) was smaller in [β2;β2] (37%) and larger in [β1;β1] (58%) than in [β1;β2] (48%, either of them P<0.05). Thus, under the same total concentrations of β-tubulins, the frequency of switching from growth to shrinkage of MTs was higher for β1-composed MTs than for β2-composed MTs. In other words, β1 formed more dynamic MTs, whereas β2 formed more stably elongating MTs. MTs composed of both β1 and β2 showed intermediate behaviors.

Taken together, these results suggest that β1 and β2 confer distinct properties to MTs, and that their distinct combination (ratio and concentration) within a cell can create diverse behaviors of MTs.

Properties of α1 and α2 have been regarded as almost equivalent in previous reports (Phillips et al., 2004; Wright and Hunter, 2003). However, previous reports suggested that each α- or β-isotype potentially has a different effect on embryonic lethality (Lu and Mains, 2005; Phillips et al., 2004; Wright and Hunter, 2003). Therefore, we examined the effect of α1 and α2 on β1-composed MTs. α1 or α2 was RNAi depleted in [β1;Δβ2] embryos (hereafter [β1;Δβ2];^RNAiα1 and [β1;Δβ2];^RNAiα2). The efficiency and the specificity of the isotype-specific RNAi depletion of α1 or α2 were confirmed by live imaging analysis (Fig. S4). In these embryos, only a single type of heterodimer, α2β1 or α1β1, would be present. Immunoblotting confirmed that, in ^RNAiα1 and ^RNAiα2 worms, the total amount of α-tubulin compared to that of wild-type was reduced to almost the same extent (52% and 46%, respectively) (Fig. 5A). Thus, the concentration of α2β1 in [β1;Δβ2];^RNAiα1 and of α1β1 in [β1;Δβ2];^RNAiα2 in one-cell embryos was expected to be very similar.

[β1;Δβ2] embryos that contain both α2β1 and α1β1 showed partial embryonic lethality (44%) (Fig. 5B). Interestingly, additional RNAi depletion of either α1 or α2 resulted in opposite effects: embryonic lethality of [β1;Δβ2];^RNAiα1 was enhanced (71%), whereas that of [β1;Δβ2];^RNAiα2 was suppressed (15%) (Fig. 5B) – as previously reported (Wright and Hunter, 2003) – indicating that α2β1 and α1β1 have distinct properties to form MTs.

We examined whether this embryonic lethality correlated with the MT dynamics formed by α2β1 and α1β1 using the same method as described above (Fig. 4A). MT growth rate in [β1;Δβ2] (containing α1β1 and α2β1; median 0.71 µm/s) was further lowered in [β1;Δβ2];^RNAiα1 embryos (containing α2β1 only) to 0.65 µm/s (P<0.05), whereas it was increased in [β1;Δβ2];^RNAiα2 embryos (containing α1β1 only) to 0.75 µm/s (P<0.05) (Fig. 5C). The comet lifetime in these three conditions were very similar (Fig. 5D). Thus, the higher embryonic lethality correlates with the lesser growth rate of MTs. These results demonstrate that α1 and α2 have distinct properties that confer significantly different growth rates to β1-composed MTs.

By using tubulin isotype locus manipulation through genome editing and quantitative in vivo imaging analyses, we demonstrated that each α- and β-tubulin isotype has distinct properties, and that their combination affects MT dynamics in C. elegans one-cell embryos. We propose a model in which the combination (ratio and and concentration) of tubulin isotypes that have distinct properties confer specific MT dynamics (Fig. 6). Individual αβ-heterodimers confer distinct MT dynamics and, when multiple isotypes are co-expressed within a cell, their ratio and concentration determine specific MT dynamics, which affects cellular events, such as mitotic spindle oscillation. Previous in vitro experiments, in which different β-tubulin isotypes had been affinity-purified from bovine brain, demonstrated that MT dynamics are modulated by the isotype combination (Panda et al., 1994). Our data are consistent with the in vitro analysis, and further demonstrate in vivo that a genetic alteration of isotype combination can impact on MT-mediated cellular events.

The previous observation that C. elegans β2 (TBB-2) mutants cause an increase in the spindle oscillation activity in early embryos led to the hypothesis that β2 is specialized to interact with the cell cortex (Wright and Hunter, 2003). Our analysis, using isotype substitution between β1 and β2 demonstrated an alternative possibility, i.e. that the active spindle oscillation is caused by more dynamic but slow-growing MTs assembled under conditions of low β1 concentration (19% of total β-tubulins in wild type). When β1 was expressed at a level similar to the total amount of β-tubulins in wild type, the embryonic lethality was completely rescued. Thus, despite the distinct dynamics of β1- and β2-composed MTs, both MTs can assemble functional mitotic spindles when sufficient amounts of either isotype is present.

In previous reports, α1 and α2 were regarded to be functionally redundant (Phillips et al., 2004; Wright and Hunter, 2003). However, our data suggest that they have a distinct effect on β1-composed MTs. The growth rate of MTs composed of α2β1 heterodimers is significantly slower than that of those composed of α1β1 heterodimers. The slow growth rate of α2β1 heterodimers would be detrimental to the formation of mitotic spindles – as indicted by increased embryonic lethality of [β1;Δβ2];^RNAiα1. However, this negative effect of the α2β1 heterodimer is generally masked in the presence of other heterodimer combinations. Thus, the presence of multiple tubulin isotypes within a cell might hedge the risk of some ‘poisonous’ αβ−heterodimers.

Our data demonstrate that C. elegans embryonic tubulin isotypes show distinct MT polymerization properties. Amino acid sequences of tubulin isotypes are generally highly conserved, and only 14 out of 450 residues are different between β1 and β2. Moreover, the majority (10 out of 14) reside in tubulin C-termini, which are located at the MT outer surface and are important for the interaction with microtubule-associated proteins (MAPs) and motor proteins (Janke and Bulinski, 2011; Lu et al., 2004; Ludueña, 2013; Sirajuddin et al., 2014). Thus, β1 and β2 might distinctively affect the interaction of MTs with other proteins their C-termini. In contrast, differences between α1 and α2 are more spread out throughout the proteins. Differences in the structural core region (four and ten amino acids in β- and α-tubulins, respectively) may affect MT polymerization/depolymerization through tubulin-tubulin interactions (Pamula et al., 2016). Additional studies – e.g. making chimeric isotypes by genome editing – will be needed to identify which of the amino acid differences are responsible for the distinct properties of each isotype.

During mitosis of C. elegans early embryos, MTs are assembled through γ-tubulin-dependent- and γ-tubulin-independent- (Aurora A-dependent) mechanisms (Motegi et al., 2006; Toya et al., 2011). Astral MTs that are assembled from centrosomes are mainly γ-tubulin-dependent MTs, whereas condensed chromatin-induced MTs are γ-tubulin-independent MTs that are likely to contribute to kinetochore MTs (Toya et al., 2011). GFP-tagged α1, α2, β1 and β2 seem equally incorporated into both astral and kinetochore MTs, implying that these isotypes are used non-selectively in both MT assembly pathways.

In our study, we have focused on early embryonic roles of α1, α2, β1 and β2; however, these isotypes are expressed in a wide range of tissues throughout development and at different expression levels. In the nervous system, α1 and β2 are widely expressed (Lockhead et al., 2016; Lu et al., 2004 and Y.H., K.T. and A.S., unpublished data), whereas α2 expression is limited to specific neurons (Fukushige et al., 1993; Lockhead et al., 2016 and Y.H., K.T. and A.S., unpublished data). Thus, the combination of α1, α2, β1 and β2 in neuronal cells is likely to be different to that in embryos, and would make their MT dynamics and functional redundancy of heterodimers distinct compared with those in embryos. Consistently, a mutation in the α1 encoding gene (tba-1(ju89)) apparently compromises the α1β2 heterodimer and causes defects in the disruption of motor neuron synapses and axonal development (Baran et al., 2010), implying that, unlike embryonic cells, other heterodimers within the cell cannot compensate for this defect.

In addition to the isotypes analyzed in this study, C. elegans has a further seven α- and four β-tubulins, some of which are known to be differentially expressed (Fukushige et al., 1999; Hamelin et al., 1992; Hurd et al., 2010 and Y.H., K.T. and A.S., unpublished data). For example, touch receptor neurons (TRN) specifically express MEC-12 α-tubulin and MEC-7 β-tubulin, in addition to α1, α2, β1 and β2, and these TRN-specific isotypes are responsible for specific MT structures (Chalfie and Thomson, 1982; Fukushige et al., 1999; Lockhead et al., 2016; Savage et al., 1994). Further quantitative in vivo analysis of different isotope combinations and MT dynamics in TRN and other tissues will clarify how tissue-specific compositions of tubulin isotypes contribute to the generation of diverse MT functions.

C. elegans strains were cultured as described (Brenner, 1974). Strains are listed in Table S1. Pie-1 promoter (germline-specific promoter)-driven GFP::TBB-1-, GFP::TBB-2-, GFP::TBA-1-, GFP::TBA-2-, mCherry::TBG-1, mCherry::H2B- and EBP-2::GFP-expressing alleles were constructed by microparticle bombardment as described (Praitis et al., 2001). Other alleles were constructed by CRISPR/Cas9-mediated genome editing (see below).

For CRISPR/Cas9-mediated genome editing, purified Cas9 proteins and single-guide RNAs (sgRNAs) were used. The coding sequence of the Cas9 protein was PCR-amplified from pMJ806 (Addgene #39312) with an N-terminal 10× His-tag and a C-terminal NLS (SRADPKKKRKV), and inserted into a pCold-based vector (TAKARA Bio, Shiga, Japan). His-tagged Cas9 proteins were expressed in E. coli strain BL21(DE3) and purified with a His-Trap HP column (GE Healthcare, Little Chalfont, UK). To synthesize sgRNAs, template DNA fragment was prepared as described (Bassett et al., 2013). In brief, template DNA was amplified by PCR without templates by using the following primer set: a gene-specific forward primer containing a T7 polymerase-binding site and sgRNA target sequence, and a common reverse primer containing the overlap sequence of forward primer and the remainder of the sgRNA sequence. Primers are listed in Table S2. sgRNAs were transcribed by using the RiboMAX T7 Express System (Promega, Madison, WI) and purified by phenol-chloroform extraction.

The gfp::tba-1 and gfp::tba-2 strains were constructed as described (Dickinson et al., 2015). For each gene, homologous repair templates were constructed by modifying the pDD282 vector (Addgene #66823) using Gibson assembly. Primers are listed in Table S3. Although TBA-1 is predicted to have two different start sites, we inserted the GFP coding sequence into just before the start site of the shorter isoform (isoform a) because almost all cDNA clones for tba-1 (isolated by Yuji Kohara, National Institute of Genetics, Mishima, Japan) are the shorter isoform.

The gfp::tbb-1 strain was constructed by two-step genome editing. For the first editing, tbb-1 ORF was replaced  using the sgRNA against the tbb-1 gene by a selection marker (Pmyo-2::mCherry: a pharyngeal promoter-driven mCherry amplified from pCFJ90 (Addgene #19327) (Frøkjær-Jensen et al., 2008) that contains synthetic sgRNA sites at both ends. For the second editing, the replaced ORF region of tbb-1 was re-replaced with the gfp::tbb-1 sequence by using the sgRNA against the synthetic sgRNA site. Primers for homologous repair templates are listed in Table S3. The obtained allele was named tj30[gfp::tbb-1].

The gfp::tbb-2 strain was constructed by replacing the tbb-2(gk129)ts deletion allele to the gfp::tbb-2 sequence with sgRNA against the gk129 allele-specific sequence. The homologous repair template contained the wild-type tbb-2 genomic sequence that spans the deleted region of gk129 (but lacks the 3′-terminus of the tbb-2 ORF) to which the gfp coding sequence was inserted at the 5′-terminus of the tbb-2 ORF. Primers for homologous repair templates are listed in Table S3. The obtained allele was named tj26[gfp::tbb-2].

The [β1;β1] (tbb-2-to-tbb-1 substitution) strain was constructed by two-step genome editing. For the first editing, the tbb-2(gk129)ts deletion allele was replaced by the gfp::tbb-2[tbb-1 substitution] sequence by using the sgRNA against the 3′-terminus of tbb-2 ORF. For the second editing, the gfp-coding sequence was removed from the gfp::tbb-2[tbb-1 substitution] strain by using two sgRNAs against the 5′ and 3′ of the gfp sequence. The gfp::tbb-2[tbb-1 substitution] DNA fragment was constructed by modifying the 14 codons of tbb-2 ORF to encode the TBB-1 amino acid sequence. Primers for homologous repair templates are listed in Table S3. The obtained allele was named tj41[tbb-2[tbb-1 substitution]].

The [β2;β2] (tbb-1-to-tbb-2 substitution) strain was constructed by replacing the tbb-1(gk207) deletion allele to the tbb-1[tbb-2 substitution] sequence by using the sgRNA against the 3′-terminus of tbb-1 ORF. The tbb-1[tbb-2 substitution] DNA fragment was constructed by modifying the 12 codons of tbb-1 ORF to encode TBB-2 amino acid sequences. The homologous repair template contained the tbb-1[tbb-2 substitution] sequence with homologous sequences of both 5′ and 3′ of tbb-1 genomic loci that span the deleted region of gk207. Primers for homologous repair templates are listed in Table S3. The obtained allele was named tj59[tbb-1[tbb-2 substitution]].

All CRISPR/Cas9-mediated alleles were confirmed by sequencing of the corresponding genomic regions.

For live imaging, embryos were collected from dissected hermaphrodite adults and mounted on 2% agarose pads with egg buffer (Edgar, 1995). Images were taken with an Axioplan 2 imaging microscope (Carl Zeiss, Jena, Germany) using a C-Apochromat 63×/1.2 NA water corr objective lens (Carl Zeiss) for differential interference contrast (DIC) microscopy, or a CSU-X1 spinning-disk confocal system (Yokogawa Electric, Musashino, Japan) mounted on an IX71 inverted microscope (Olympus, Tokyo, Japan) with a UPlanSApo 60×/1.30 NA silicone objective lens (Olympus) or a UPlanSApo 100×/1.40 NA oil objective lens (Olympus) for fluorescent microscopy, under the control of MetaMorph software (Molecular Devices, Sunnyvale, CA). All images were taken by Orca-R2 12-bit/16-bit cooled CCD camera (Hamamatsu Photonics, Hamamatsu, Japan). To analyze the spindle oscillation by DIC microscopy, images were acquired every second at a single z-section by using 50 ms exposure time with camera gain 0 and without binning. To quantify the amount of MTs by fluorescent microscopy, images were acquired by using a 60× lens at 17 z-sections with 1 µm steps using 150 ms exposure time with camera gain 150 for GFP, and 500 ms exposure time with camera gain 255 for mCherry without binning. To quantify the EBP-2::GFP signal, images were acquired by using a 100× lens every 700 ms at a single z-section using streaming for 60 frames (total acquisition time was 42 s), with camera gain 150 and binning 2.

Images were processed and analyzed by ImageJ/Fiji software (National Institutes of Health, Bethesda, MD). To quantify the amount of the GFP-tagged tubulin in a whole embryo, z-sectioned image stacks were projected using the sum intensity algorithm and the area of an embryo was determined by Threshold function. Centrosomal and cytosolic signal of gfp-tagged tubulin was quantified as follows. Using sum-intensity-projected stacks, signal intensities within 50×50 pixel circles around the anterior- and the posterior-centrosomes were measured, and their average value was used as the centrosomal signal. Signal intensities within a 50×50 pixel circle in cytosol were used as the cytosolic signal.

For the analysis of spindle oscillation during one-cell anaphase stage, L4 worms were transferred to 24.5°C and after 12–24 h embryos were collected from young adult worms. Quantification of spindle oscillation was performed using ImageJ with a plug in ‘Automated centrosome tracking’ (Cluet et al., 2014) or by manually tracking centrosomes. For manual tracking, centrosomes were tracked using ImageJ/Fiji software and the ‘point tool’ for each time point, and their distances from the midline along the anterior-posterior axis of embryos were calculated.

EBP-2::GFP comets were manually tracked using ImageJ/Fiji software. To quantify growth rates and lifetimes, a circle with radius of 5 µm was drawn around the anterior centrosome and comets emanating from its circumference were measured. Kymographs of each detected comet were drawn by using the ‘make montage tool’ and growth rates were calculated from the angle of trajectory. The lifetimes of EBP-2 comets were manually measured from the movie.

For sample preparation, ten adult worms were collected in 7.2 µl of DDW and flash-frozen in liquid nitrogen. After thawing, 0.8 µl of 10% NP40 and 2 µl of 5× SDS-PAGE sample buffer were added, followed by boiling for 5 min. The lysate (10 µl) was separated by 5–20% gradient SDS-PAGE gel (Wako, Osaka, Japan). The primary antibodies used were as follows: mouse anti-β-tubulin antibody E7 in ascitic fluid (1:10; Developmental Studies Hybridoma Bank, Iowa City, IA), mouse anti-α-tubulin antibody DM1A (1:5000; T9026, Sigma-Aldrich, St Louis, MO), rabbit anti-Histone H3 antibody (1:5000; ab1791, Abcam, Cambridge, UK) and rat anti-GFP antibody (1:10,000; Medical and Biological Laboratories, Nagoya, Japan). The secondary antibodies used were as follows: horseradish peroxidase (HRP)-conjugated goat anti-mouse (1:50,000; cat. no. 115-035-166, Jackson ImmunoResearch Laboratories, West Grove, PA), HRP-conjugated donkey anti-rabbit (1:200,000; cat. no. 711-035-152, Jackson ImmunoResearch Laboratories) and HRP-conjugated donkey anti-rat (1:50,000; cat. no. 712-035-153, Jackson ImmunoResearch Laboratories). Signals were detected in chemiluminescence assays (Chemi-lumi One Super, Nakarai Tesque, Kyoto, Japan).

Rat anti-GFP antibody had been generated at the Medical and Biological Laboratories (Nagoya, Japan) by using affinity-purified 6×His-tagged GFP as the antigen.

RNA interference (RNAi) was carried out as described previously (Maeda et al., 2001). To synthesize double-stranded RNA (dsRNA) for tba-1, tba-2, tbb-1 and tbb-2, the 3′ sequence of the coding region and the subsequent 3′ untranslated region (UTR) of each tubulin gene was amplified from the following fosmids or cDNA clones: WRM0629aH08 for tbb-1, WRM0625dB09 for tbb-2, yk1467h08 for tba-1 and yk1536b11 for tba-2 (cDNA clones were gifts from Yuji Kohara, National Institute of Genetics, Mishima, Japan). Template DNA fragments were PCR-amplified using primer sets containing a promoter sequence of T7 polymerase and each tubulin-gene-specific sequence that was designed according the description by Wright and Hunter, 2003 (i.e. primer sequences were designed so that exact nucleotide matches to other tubulin genes were less than 14 bp to reduce cross-interference). Primers for dsRNA are listed in Table S4. dsRNA was transcribed in vitro with RiboMAX T7 Express System (Promega, Madison, WI) and purified by phenol-chloroform extraction. L4 worms were soaked in dsRNA solution of 2 mg/ml, incubated at 24.5°C for 24 h and then used for imaging or immunoblot for 12-24 h after recovery from the dsRNA solution.

Statistical analysis was performed using R software (R Foundation for Statistical Computing, Vienna, Austria) or MEPHAS (http://www.gen-info.osaka-u.ac.jp/MEPHAS/). For data sets of embryonic lethality or frequencies of lifetime of EBP-2 comets, approximate P-values were calculated according to Fisher's Exact test with simulated P-value (based on 1e+07 replicates) using R software. For other data sets, Steel-Dwass test was used.

We thank the Caenorhabditis Genetic Center, which is funded by the National Institutes of Health Center for Research Resources, for providing strains; Yuji Kohara (National Institutes of Genetics, Mishima, Japan) for providing cDNA clones and the members of Sugimoto lab for discussions.