Summary
Centrioles are composed of long-lived microtubules arranged in nine triplets. In unicellular eukaryotes, loss of the noncanonical tubulins, delta-tubulin and epsilon tubulin, result in loss of the triplet microtubule structure. However, the contribution of triplet microtubules to mammalian centriole formation and stability is unknown. Here, we report the first characterization of delta-tubulin and epsilon-tubulin null human cells. Centrioles in cells lacking either delta-tubulin or epsilon-tubulin lack triplet microtubules and fail to undergo centriole maturation. These aberrant centrioles are formed de novo each cell cycle, but are unstable and do not persist to the next cell cycle, leading to a futile cycle of centriole formation and disintegration. Disintegration can be suppressed by paclitaxel treatment. Delta-tubulin and epsilon-tubulin physically interact, indicating that these tubulins act together to maintain triplet microtubules and that these are necessary for inheritance of centrioles from one cell cycle to the next.
Introduction
The major microtubule organizing center of mammalian cells, the centrosome, is composed of a pair of centrioles with associated appendages and pericentriolar material. The centrioles have a nine-fold symmetry and are formed, in part, of long-lived microtubules, which persist through multiple cell divisions (Kochanski and Borisy, 1990; Balestra et al., 2015). In most organisms, including humans, the centriolar microtubules have a triplet structure, found only in centrioles. This structure consists of a complete A-tubule and associated partial B-tubule attached to the A-tubule wall, and a partial C-tubule attached to the B-tubule wall.
The molecular mechanisms involved in making triplet microtubules are not well-understood, even in the well-characterized somatic centriole cycle of mammalian cells. In these cells centrioles duplicate once per cycle, such that daughter cells receive exactly one pair of centrioles. Centriole duplication is initiated at the G1-S transition when the kinase PLK4 localizes to a single focus on the mother centriole (Sonnen et al., 2012). Subsequently, the cartwheel, formed by SASS6 oligomerization, assembles to template the 9-fold symmetry of the centriole (Guichard et al., 2017; Hilbert et al., 2016). Microtubules are added to the cartwheel underneath a cap of CP110 (Kleylein-Sohn et al., 2007). By G2-M, the triplet microtubules are completely formed (Vorobjev and Chentsov, 1982). Subsequently, the A- and B-tubules elongate to the full ~500 nm length of the centriole, forming a distal compartment with doublet microtubules and marked by POC5 (Azimzadeh et al., 2009). In mitosis, the cartwheel is lost, the newly-formed centriole becomes disengaged from its mother, and acquires pericentriolar material (Vorobjev and Chentsov, 1980; Vorobjev and Chentsov, 1982; Khodjakov and Rieder, 1999; Tsou and Stearns, 2006; Tsou et al., 2009). In G2-M of the following cell cycle, the centriole acquires appendages, marking its maturation into a centriole that can nucleate a cilium (Graser et al., 2007; Guarguaglini et al., 2005).
Members of the tubulin superfamily are critical for centriole formation and function. All eukaryotes have alpha-, beta- and gamma-tubulin, but the tubulin superfamily also includes three less-studied members, delta-tubulin, epsilon-tubulin, and zeta-tubulin. Recent work has shown that these noncanonical tubulins are evolutionarily co-conserved, making up the ZED tubulin module (Turk et al., 2015). In the unicellular eukaryotes Chlamydomonas, Tetrahymena, Paramecium and Trypanosoma, mutations in delta-tubulin or epsilon-tubulin result in centrioles that lack triplet microtubules (Dupuis-Williams et al., 2002; Dutcher and Trabuco, 1998; Dutcher et al., 2002; Gadelha et al., 2006; Garreau de Loubresse et al., 2001; Goodenough and StClair, 1975; Ross et al., 2013). Humans and other placental mammals have delta-tubulin and epsilon-tubulin, but lack zeta-tubulin (Findeisen et al., 2014; Turk et al., 2015). Here, we show that human cells lacking delta-tubulin or epsilon-tubulin also lack triplets, that this results in unstable centrioles and initiation of a futile cycle of centriole formation and disintegration, and identify an interaction between delta-tubulin and epsilon-tubulin.
Results and Discussion
To determine the roles of delta-tubulin and epsilon-tubulin in the mammalian centriole cycle, null mutations in TUBD1 and TUBE1 were made using CRISPR/Cas9 genome editing in hTERT RPE-1 human cells. Recent work has established that loss of centrioles in mammalian cells results in a p53-dependent cell-cycle arrest (Bazzi and Anderson, 2014; Lambrus et al., 2015; Wong et al., 2015). We found that homozygous null mutations of delta-tubulin or epsilon-tubulin could only be isolated in TP53 -/- cells, thus all subsequent experiments use RPE1 TP53 -/- cells as the control.
Three TUBD1 -/- and two TUBE1 -/- cell lines were generated (Figure 1A and Figure 3 – supplemental figure 1). The TUBD1 -/- lines are all compound heterozygotes bearing, proximal to the cut site, small deletions of less than 20 base pairs on one chromosome and insertion of one base pair on the other, resulting in frameshift and premature stop mutations. The two TUBE1 -/- lines are compound heterozygotes bearing large deletions surrounding the cut site, that in each case remove an entire exon and surrounding DNA, including the ATG start site. In all cases, the next available ATG is not in-frame. We conclude that these alleles are likely to be null, or strong loss-of function mutations.
We next assessed the phenotype of TUBD1 -/- and TUBE1 -/- cells stably expressing GFP-centrin as a marker of centrioles. Many cells in an asynchronous population had multiple, unpaired centrin foci (Fig. 1B). These foci also labeled with the centriolar proteins CP110 and SASS6 (see Figs 2 and 3). To determine whether these foci are centrioles, and to assess their ultrastructure, we analyzed them using correlative light-electron microscopy. In serial sections of interphase TUBE1 -/- (Fig 1 B) and TUBD1 -/- (Fig 1C) cells, some of the centrin-positive foci corresponded to structures that resemble centrioles, but were narrower and shorter than typical centrioles and lack appendages. They contained a ~80 nm wide central lumen, with luminal content resembling cartwheel architecture that extends throughout the centriole’s length (Fig 1 – supplement 1).
Centrioles in TUBD1 -/- and TUBE1 -/- cells were of similar diameter: 165 nm +/-15 nm in TUBD1 -/- cells (n = 19 centrioles), 164 nm +/-13 nm in TUBE1 -/- cells (n = 11 centrioles), compared to 220 nm diameter of the proximal end of typical mammalian cell centrioles (Loncarek et al., 2008; Wang et al., 2015). The reduced diameter of these aberrant centrioles is consistent with the presence of only singlet microtubules. Indeed, only singlet microtubules were identified in the two cross-sections observed, from TUBD1 -/- cells (Figure 1C). These results demonstrate that cells lacking either deltatubulin or epsilon-tubulin form defective centrioles that lack normal triplet microtubules. This is similar to the defects reported for delta-tubulin and epsilon-tubulin mutants in unicellular eukaryotes.
Centrioles in both tubulin mutants were also shorter than typical, mature centrioles: 230 nm +/-45 nm in in TUBD1 -/- cells (n = 14 centrioles), and 271 nm +/-43 nm in in TUBE1 -/- cells (n = 11 centrioles), compared to approximately 500 nm for typical human cell centrioles (Paintrand et al., 1992). Newly-formed mammalian centrioles, or procentrioles, reach their full length by elongation in G2-M, creating a distal compartment that is a feature of centrioles in some, but not all, organisms. We sought to determine whether the aberrant centrioles in TUBD1 -/- and TUBE1 -/- cells are capable of elongation and formation of the distal compartment. We analyzed the ultrastructure of centrioles in a TUBE1 -/- prometaphase cell using correlative light-electron microscopy (Fig. 2A). These aberrant centrioles (n = 3) exhibited a striking morphological phenotype, consisting of two electron-dense segments, one of ~50 nm and the other of ~200 nm, connected by singlet microtubules spanning a gap of ~250 nm. The total length (~500 nm) of these structures approximates that of typical mature mammalian centrioles.
We hypothesized that the aberrant centrioles formed in TUBD1 -/- and TUBE1 -/- cells elongate in G2-M, but that only the A-tubule is present and thus able to elongate as a singlet. In this model, the shorter, distal density might correspond to the CP110 cap, under which the centriolar microtubules elongate (Kleylein-Sohn et al., 2007). The longer, 200 nm density corresponds to the proximal centriole end containing the cartwheel, as observed above in interphase cells. A prediction of this model is that the distance between CP110 and the SASS6 fluorescent labels would increase by about 200 nm in mitosis. We found that in TUBD1 -/- and TUBE1 -/- interphase cells, similar to control TP53 -/- cells, the centroids of CP110 and SASS6 foci were separated by a mean distance of 0.3 μm, whereas in mitotic cells the foci were separated by a mean distance of 0.5 μm (Fig 2B and 2C). Thus, centrioles in TUBD1 -/- and TUBE1 -/- cells elongate at the appropriate time in the cell cycle, and have a cap and proximal end typical of newly-formed centrioles. The lack of electron-dense structure between the cartwheel and cap might be due to a failure to recruit distal compartment components. Consistent with this, we found that the distal compartment component POC5 is absent from these aberrant centrioles (Fig 2D).
Together, these results indicate that the primary centriolar defect in cells lacking delta tubulin or epsilon-tubulin is the absence of triplet microtubules. To determine the consequences of loss of triplet microtubules on the centriole cycle and centrosome formation, we first determined the distribution of centrioles in asynchronously dividing cell populations, as determined by staining for the established centriole proteins centrin and CP110. TP53 -/- control cells had a typical centriole number distribution, with approximately 50% of cells having two centrioles, corresponding to cells in G1 phase, and 40% having three to four centrioles, corresponding to cells in S through M phases. In contrast, in TUBD1 -/- and TUBE1 -/- cells, approximately 50% of cells had 5 or more centriole foci, whereas 50% of cells had no detectable foci positive for both centrin and CP110 (Fig. 3A and 3B). Similar centriole distributions were found in other, independently derived, TUBD1 -/- and TUBE1 -/- cell lines. In addition, this phenotype could be rescued by expression of delta-tubulin and epsilon-tubulin, respectively (Fig 3 – supplement 1A – 1C).
We reasoned that a possible explanation for the centriole distribution in TUBD1 -/- and TUBE1 -/- cells is that the centriole structures we observed by EM are produced de novo in each cell cycle, and that these aberrant centrioles are unstable and do not persist into the next cell cycle. This hypothesis predicts that the aberrant centrioles in TUBD1 -/- and TUBE1 -/- cells would 1) not be paired, since de novo centrioles only form in the absence of an existing centriole, 2) lack markers of maturation such as distal appendages, since they would not persist to the point of acquiring such proteins, 3) fail to recruit substantial pericentriolar material, since the centriole-centrosome conversion occurs at entry to the next cell cycle, and 4) would be formed in S phase, and be lost at some point prior to the subsequent S phase.
In agreement with this hypothesis, the centrioles, as visualized by centrin and CP110 were never observed to be closely apposed, as is typical of wild-type cells (Fig. 3A). Rather, in interphase they appeared to be distributed within the central region of the cell (Fig. 3A). The centrioles in asynchronous TUBD1 -/- and TUBE1 -/- cells all lacked Cep164, a component of the centriolar distal appendage and marker of mature centrioles that have progressed through at least one cell cycle (Fig. 3C), whereas approximately 40% of all centrioles were positive for Cep164 in asynchronous control cells, consistent with the cycle of distal appendage acquisition (Nigg and Stearns, 2011). Lastly, most of the centrioles in TUBD1 -/- and TUBE1 -/- cells lacked detectable gamma-tubulin (Fig. 3C), and those that stained positive had less than centrioles in control cells (Fig 3 – supplement 1D). In addition, we noted that SASS6, the cartwheel protein that is present in nascent and recently-formed centrioles, but is lost from centrioles at the mitosis-interphase transition in human cells, was present in most of the centrioles in TUBD1 -/- and TUBE1 -/- cells, consistent with these centrioles originating in the observed cell cycle, but not having successfully persisted into the subsequent cell cycle.
To investigate the fate of newly-formed centrioles in TUBD1 -/- and TUBE1 -/- cells, we next tested the cell cycle-dependence of the formation and loss of aberrant centrioles in TUBD1 -/- and TUBE1 -/- cells (Fig 4A). As in previous experiments, about 50% of TUBD1 -/- and TUBE1 -/- cells in an asynchronous population had centrin and CP110- positive foci corresponding to aberrant centrioles. TUBD1 -/- and TUBE1 -/- cells were analyzed in different cell cycle stages as follows: G0/G1 – synchronized by serum withdrawal, S phase – identified from asynchronous culture by PCNA labeling, G2 – synchronized by the CDK1 inhibitor RO-3306, and M – identified from asynchronous culture by presence of condensed chromatin (Fig. 4A). TUBD1 -/- and TUBE1 -/- cells in G0/G1 mostly lacked centriole structures, whereas cells in S-phase, G2 and mitosis had them. These results indicate that in TUBD1 -/- and TUBE1 -/- cells, aberrant centrioles are formed in S-phase, persist into mitosis, and are absent in G1. We note that this loss of centriole structure is likely due to a specific event that occurs at the mitosis-interphase transition, rather than simply time since formation, since cells were arrested in G2 for 24 h, which is substantially longer than the normal progression through mitosis to G1, yet the centriole structures persisted (Fig 4A).
To more finely determine the timing of centriole loss in the mitosis-interphase transition, control or TUBE1 -/- cells were synchronized by mitotic shakeoff, and the presence of centriole foci was assessed over time as cells entered G1 (Fig 4B). In control cells, the number of centrioles follows the pattern expected from the centriole duplication cycle. In TUBE1 -/- cells, the majority of mitotic cells had centrioles. By 1 h after shakeoff, the fraction of interphase cells without centrioles had increased to 50%, and this fraction continued to increase at 2 h and 3 h after shakeoff. By 12 h after shakeoff, 56 +/-12% of cells had entered S-phase, and centriole structures began to appear, consistent with de novo centriole formation. Thus, delta-tubulin and epsilon-tubulin are not required to initiate centriole formation in human cells, but the aberrant centrioles that form in their absence are unstable and disintegrate during progression from M phase to the subsequent G1 phase.
Centrioles formed de novo can persist to form fully mature centrioles (Lambrus et al., 2015; Wong et al., 2015), but have also been reported to be structurally defective (Wang et al., 2015). We tested whether the phenotype we observed is specific to loss of delta-tubulin and epsilon-tubulin, rather than a property of de novo centrioles in general, by assessing whether de novo centrioles formed in the presence of delta-tubulin and epsilon-tubulin would also disintegrate upon cell cycle progression. RPE-1 TP53 -/- cells were treated with centrinone to inhibit PLK4 (Wong et al., 2015), a kinase required for centriole duplication, for more than 2 weeks to obtain acentriolar cells. Centrinone was then washed out from mitotic cells; by 12 h after shakeoff, 36% of cells had entered S-phase, and centriole structures began to appear, consistent with de novo centriole formation. However, in contrast to TUBD1 -/- and TUBE1 -/- cells, these de novo centrioles persisted through the subsequent G1 (Fig 4C). We conclude that centriole instability in TUBD1 -/- and TUBE1 -/- cells is due to a specific defect in their structure, and is not a general feature of de novo centrioles, similar to previous reports (La Terra et al., 2005).
We hypothesized that centriole disintegration may result from instability of the centriolar microtubules, perhaps as a result of elongation in G2-M phase. To test this, microtubules were stabilized in G2-M stage TUBE1 -/- cells by addition of the microtubule-stabilizing drug paclitaxel. This treatment did not inhibit centriole elongation (Fig 4 – supplement 1B). Cells were allowed to enter mitosis in the presence of paclitaxel, and subsequently forced out of mitosis using the CDK inhibitor RO-3306. This treatment was sufficient to stabilize centrioles from mutant cells in G1, compared with cells that had not been treated with paclitaxel (Fig 4D and Fig 4 – supplement 1). These stabilized centrioles lose their SASS6 cartwheel and fail to recruit detectable gamma-tubulin (Fig 4 – supplement 1). We conclude that stabilization of the centriolar microtubules in TUBE1 -/- cells stabilizes the centriole structure.
One striking observation of this work is that the phenotypes of delta-tubulin and epsilon tubulin null mutants are similar. This strongly suggests that the proteins work together to accomplish their function. To test this hypothesis, we assessed the ability of delta tubulin and epsilon-tubulin to interact by co-expression in human HEK293T cells. Epsilon-tubulin could be immunoprecipitated with delta-tubulin from co-expressing cells, and not from control cells (Fig 5A).
Together, our results show that delta-tubulin and epsilon-tubulin act together to create or stabilize structural features of centrioles. The most obvious such feature is the triplet microtubules, which define centrioles in most species, and are absent in delta-tubulin or epsilon-tubulin mutant cells in all organisms which have been examined. This suggests that delta-tubulin and epsilon-tubulin are required either to form the triplet microtubules, or to stabilize them against depolymerization. The former seems unlikely, since the presence of triplet centriolar microtubules is not strictly correlated with the presence of delta-tubulin and epsilon-tubulin in evolution (Fig 5B and Fig 5 - Supplemental Table 1). Among the organisms that completely lack the ZED tubulin module, C. elegans lacks triplet microtubules, but both Drosophila and the primitive plant Ginkgo biloba have triplet microtubules in their sperm cells. Since loss of the entire ZED tubulin module must have occurred independently in the dipteran insect and plant lineages, the most parsimonious interpretation is that triplet microtubule formation itself does not require delta-tubulin or epsilon-tubulin, rather than that these two lineages independently evolved mechanisms of triplet formation in their absence. Thus, we propose that delta-tubulin and epsilon-tubulin are required for stabilization of the centriolar triplets in most organisms, such that the centrioles can mature and recruit other proteins. We do not yet know the molecular basis of this differential requirement for delta-tubulin or epsilon-tubulins with respect to microtubule triplet stability. However, we note that the few centriole-bearing organisms that lack delta-tubulin and epsilon-tubulin have simpler centriole structures that lack distal appendages, and, to the extent it is possible to tell, lack a distal compartment that is typical of more complex centrioles.
Why do centrioles disintegrate in delta-tubulin and epsilon-tubulin mutant cells? We have shown that in TUBD1 -/- and TUBE1 -/- cells, aberrant centrioles with elongated singlet microtubules connecting the proximal and distal centriole segments become unstable as cells progress through mitosis. This is remarkably similar to the progressive loss of centrioles described in the original characterization of the epsilon-tubulin mutant bald-2 by Goodenough and St. Clair (Goodenough and StClair, 1975). In human cells, Izquierdo, et al. reported that centrioles in CEP295 -/- human cells also become unstable upon cell cycle progression, due to a failure of centrioles to recruit pericentriolar material that coincides with loss of the cartwheel during the centriole-to centrosome conversion at the end of mitosis (Izquierdo et al., 2014). Although the phenotypes are outwardly similar to the phenotypes we describe here, CEP295 is conserved in species lacking delta-tubulin and epsilon-tubulin (Fu et al., 2015), and centrioles in Chlamydomonas do not undergo centriole-to-centrosome conversion. We propose that the post-duplication centriole elongation that creates the distal compartment of the centriole is a critical time in centriole stability, and that the triplet microtubules, either directly or through proteins that associate with them, are required to prevent centriole disassembly subsequent to that step. One possible basis for the instability is that events at the distal end of the centriole associated with preparing it to serve as a basal body for a cilium in G1 expose the ends of the centriolar microtubules. The doublet microtubules normally present at the end would be resistant to depolymerization in this model, but the singlets found in delta-tubulin and epsilon-tubulin mutants might be unstable. In accordance with this possibility, stabilization of centriolar microtubules with paclitaxel was able to prevent centriole disintegration, even when both the SASS6 cartwheel and pericentriolar material are lost (Fig 4D and Fig 4 – supplement 1). Another possibility is that centrioles lacking the normal triplet structure would likely also lack the A-C linker, which is visible in EM as a bridge between the A- and C-tubules of adjoining triplets. Perhaps the A-C linker is most important for stability after the full elongation of the centriolar microtubules. No components of the A-C linker have been identified, but the poc1 mutant in Tetrahymena causes partial loss of this linker and results in instability of triplet microtubules (Meehl et al., 2016).
Here we have shown that delta-tubulin and epsilon-tubulin likely work together in a critical function for centriole structure and function, and that cells lacking delta-tubulin or epsilon-tubulin undergo a futile cycle of de novo centriole formation and disintegration. Our results show that in human cells, delta-tubulin and epsilon-tubulin act to stabilize centriole structures necessary for inheritance of centrioles from one cell cycle to the next, perhaps by stabilizing the main structural feature of centrioles, the triplet microtubules.