Emi2 enables centriole amplification during multiciliated cell differentiation

Abstract

Massive centriole amplification during multiciliated cell (MCC) differentiation is a notable example of organelle biogenesis. This process is thought to be enabled by a derived cell cycle state, but the key cell cycle components required for centriole amplification in MCC progenitors remain poorly defined. Here, we show that emi2 (fbxo43) expression is up-regulated and acts in MCC progenitors after cell cycle exit to transiently inhibit anaphase-promoting complex/cyclosome (APC/C)^cdh1 activity. We find that this inhibition is required for the phosphorylation and activation of a key cell cycle kinase, plk1, which acts, in turn, to promote different steps required for centriole amplification and basal body formation, including centriole disengagement, apical migration, and maturation into basal bodies. This emi2-APC/C-plk1 axis is also required to down-regulate gene expression essential for centriole amplification after differentiation is complete. These results identify an emi2-APC/C-plk1 axis that promotes and then terminates centriole assembly and basal body formation during MCC differentiation.

Centriole amplification and basal body production during MCC differentiation are regulated by an emi2-APC/C-plk1 axis.

INTRODUCTION

The multiciliated cell (MCC) is a specialized vertebrate cell type that uses hundreds of motile cilia to propel fluid along luminal surfaces in various organ systems (1). To form these cilia, MCC progenitors differentiate by producing hundreds of new centrioles that become the basal bodies required to extend multiple motile cilia. This process shares similarities to the centrosome cycle during cell division (2), when a sequence of events, including procentriole assembly, disengagement, and maturation, is regulated by the cell cycle machinery (3). In a similar manner, MCC progenitors undergo centriole amplification and form hundreds of basal bodies by exploiting links to the cell cycle at both the transcriptional and posttranscriptional levels.

Two small, related, coiled-coil proteins, called Gemc1 and Multicilin, initiate MCC differentiation transcriptionally by forming and acting in ternary complexes with E2f4 or 5 and Dp1. These E2F proteins normally function in postmitotic cells to repress gene expression required for cell cycle progression (4, 5) but are co-opted by Multicilin and Gemc1 in MCC progenitors to selectively activate gene expression required for centriole amplification and basal body formation (but not other cell cycle functions) (6–8). At the same time, Multicilin and Gemc1 up-regulate the expression of additional downstream transcription factors, notably Foxj1, which activate gene expression used to build a motile axoneme (4, 9). Multicilin and Gemc1, and many of their targets, are transiently expressed in differentiating MCCs, while Foxj1 and its targets continue to be expressed in fully differentiated cells. How this pattern of differential gene expression in MCCs progenitors is regulated is not known.

The process that promotes centriole amplification and basal body formation during MCC differentiation also appears to be regulated by the kinases and ubiquitin ligases that govern different phases of the cell cycle (10). MCC differentiation in cultured progenitors is disrupted by the application of small-molecule inhibitors to Cdk1, Plk1, and Cdk2, resulting in a block in centriole amplification and basal body formation (10–12). Differentiating MCCs also inappropriately display mitotic markers and mitotic chromatin after treatment with chemicals that disinhibit Cdk1 or inhibit the activity of the anaphase-promoting complex/cyclosome (APC/C), a ubiquitin ligase complex that degrades cyclins and other substrates during the cell cycle (12). Last, two cyclin genes, Ccna1 and Ccno, are markedly up-regulated transcriptionally during MCC differentiation and are required for centriole amplification and basal body formation (11, 13, 14). Thus, cell cycle kinases and ubiquitin ligases become active in MCC progenitors in a manner that enables centriole amplification while avoiding other cell cycle behaviors. The key cell cycle players required to produce this derived cell cycle state in MCC progenitors are not fully known.

Here, we examine the link between MCC differentiation and the cell cycle, by focusing on emi2 (Fbox43), a potent inhibitor of the APC/C ubiquitin ligase that acts during multiple points in the cell cycle (15, 16). The results presented here lead to a model in which emi2, markedly up-regulated by Multicilin, functions to dampen APC/C^cdh1 activity at early stages of MCC differentiation. We further show that the inhibition of APC/C^cdh1 is required to promote the phosphorylation and activation of the key cell cycle kinase, plk1, which acts, in turn, to promote the different steps involved in centriole amplification and basal body formation. Unexpectedly, many of the structural and regulatory genes required for centriole amplification and basal body formation are transiently expressed during early MCC differentiation, initially up-regulated by Multicilin, but then down-regulated by the emi2-APC/C-plk1 axis. These results provide key mechanistic insights into the link between gene expression and the key cell cycle regulators required for MCC differentiation.

RESULTS

emi2 is required for MCC differentiation

The larval skin of Xenopus embryos contains MCCs identical to those in the mammalian lung that synchronously arise when endogenous skin progenitors exit the cell cycle at around Nieuwkoop and Faber (NF) stage 11 (17), migrate into the outer epithelium (referred to as outer cells), and form MCCs over a ~12-hour period (Fig. 1A). Essentially all of the cells in the developing skin, including the outer cells in the superficial epithelium, can also be converted into ectopic MCC progenitors, by expressing an inducible form of Multicilin that contains the ligand binding domain from the human growth hormone receptor, called Multicilin-HGR (18). Both endogenous and ectopic MCC progenitors undergo the same synchronous program of centriole amplification with similar kinetics: Numerous nascent procentrioles appear at basal assembly sites, procentrioles migrate apically, and centrioles disengage and disperse (Fig. 1A). Last, these centrioles mature into basal bodies that dock at the apical surface to initiate ciliogenesis (Fig. 1A).

Fig. 1. Defective MCC differentiation in emi2 mutant embryos.

(A) Depiction of MCC differentiation in Xenopus skin progenitors, according to developmental staging by Nieuwkoop and Faber (17). (B to E) Confocal images of MCCs in a stage 26 wild-type (B and C) and an emi2^gRNA1 mutant (D and E) embryo, expressing cnt4-GFP (green, centrioles) and mRFP (blue, cell boundaries), and stained with anti-acetylated tubulin antibody (red, cilia). (F) Quantification of different cell types in the skin within a 203-μm² field (see Materials and Methods). Each point represents data from a different embryo. (G) Quantification of phenotypes among MCCs (n = 15 to 18) within a 203-μm² field from 10 different embryos (see the main text for criteria). (H and I) Confocal images of ectopic MCCs as in (B) and (D). (J) Quantification of phenotypes among ectopic MCCs, as in (G). A small fraction (~5%) of outer cells failed to respond to Multicilin-HGR in both wild-type and mutant embryos. (K) MCC differentiation was scored as described above at the indicated developmental stage. Scale bars, 5 μm. Error bars = SD. All values obtained with wild-type and emi2^gRNA1 embryos were significantly different (P < 0.0001), based on comparison using an unpaired t test, except those indicated as ns (not significant) or **(P < 0.01).

RNA sequencing (RNA-seq) analysis of synchronized Xenopus MCC progenitors (both endogenous and ectopic) previously revealed that emi2 (fbxo43) is markedly up-regulated at stages when centriole amplification occurs, in a Multicilin-dependent fashion (fig. S1A). Single-cell RNA-seq of Xenopus tropicalis embryos also reported that emi2 expression is specifically and transiently expressed in MCC progenitors (19). Emi2 has not been previously implicated directly in centriole amplification but has been shown to regulate the cell cycle during meiosis and early embryonic cell division (16, 20) as a potent inhibitor of the ubiquitin ligase, APC/C. To determine the function of emi2 and, by extension, APC/C during MCC differentiation, we generated F0 mutants by injecting one of two independent emi2 guide RNAs (gRNAs; fig. S1, B and K), along with Cas9 protein into Xenopus embryos, and assessed phenotypes that arise within the MCCs of the larval skin.

The cell-type composition of the skin was similar in F0 emi2 mutant embryos and wild-type controls, indicating that zygotic emi2 is unlikely to have an early role in embryonic cell division or cell-type specification (Fig. 1, B to F) (16). By contrast, MCC differentiation in the emi2 mutants was severely altered in a notably variable manner (Fig. 1G). At one extreme, MCCs in emi2^gRNA1 mutant embryos extended just one or two cilia (≤2 cilia/cell), largely lacked apical basal bodies, but contained aggregates of centriolar material located deep in the cytoplasm, indicating a differentiation arrest at stage II or earlier (Fig. 1, A and D). The few cilia extended by these MCCs likely nucleate from basal bodies derived from the preexisting centrioles, as has been shown to occur when these cells express just Foxj1 (21). In less extreme cases, the MCCs in emi2^gRNA1 mutant embryos contained more centriolar material near the apical surface, and more of these centrioles docked and extended cilia (albeit many fewer than normal) (Fig. 1, E and G; partially differentiated). The two emi2 gRNAs produced a similar range of defective MCC phenotypes, indicating specificity, but varied in their efficiency, with ~10% and ~50% normal MCCs with emi2^gRNA1 and emi2^gRNA2, respectively (Fig. 1G and fig. S1, C to F). The same MCC phenotypes, in approximately the same frequency and proportions, also occurred in ectopic MCCs induced in emi2 mutants using Multicilin-HGR (Fig. 1, H to J, and fig. S1, G to J). Last, the prevalence of the different MCC phenotypes remained unchanged in emi2^gRNA1 mutants out to stage 34, approximately 24 hours after the completion of differentiation (Fig. 1K), indicating that the loss of emi2 permanently arrested MCC differentiation at various points, rather than simply slowed MCC differentiation. Thus, emi2 is required for the differentiation of both endogenous and ectopic MCC progenitors, apparently acting downstream of Multicilin.

Super-resolution imaging of the emi2 mutant phenotypes

To determine where the block in MCC differentiation occurs in emi2 mutants, we used super-resolution microscopy to examine centriole amplification, by imaging fluorescent protein–tagged centriolar markers, including cep135, cnt4, cep152, and scored basal body formation, by imaging cep164 immunofluorescence, a maturation marker. At the early stages of MCC differentiation (stage 16), procentriole initiation at the basal assembly sites revealed by these markers appeared to be qualitatively similar in emi2^gRNA1 mutant and wild-type embryos (fig. S2, A to F) (22). However, these basal assembly sites disappear in wild-type MCCs after stage 16 as new centrioles migrate apically, but they persist in the emi2^gRNA1 mutants at least through stage 26, appearing even larger and more elaborate with time (fig. S2, O versus F). Moreover, when wild-type MCCs were imaged at stage 26, their centrioles were docked apically and extended cilia (Fig. 2A), were fully disengaged on the basis of a one-to-one labeling with cnt4–green fluorescent protein (GFP) and cep135-tomato (Fig. 2G), contained a ring of cep164 protein, a marker of distal appendages and maturation (Fig. 2, H, L, and P), and labeled weakly with cep152-GFP, a centriolar marker associated with assembly sites in differentiating MCCs (fig. S2A), but known to be lost in mature basal bodies (fig. S2H) (23). In contrast, the pattern of these markers in emi2^gRNA1 mutant MCCs indicated defects in centriole assembly and basal body formation at multiple levels. First, the total number of centrioles (based on cnt4-GFP foci) in emi2^gRNA1 mutants were significantly reduced relative to that in wild type (P < 0.0001), suggesting that procentriole initiation was partially affected (Fig. 2F). Second, a large fraction of the centrioles in emi2^gRNA1 mutant MCCs showed significant disengagement defects based on the presence of rosettes, in which multiple cnt4-GFP foci were associated with a single focus of cep135-tomato (Fig. 2, A to D and G, and fig. S2, V to W). Emi2 mutant MCCs typically contain two large rosettes (Fig. 2, B and D; M1 and M2) likely corresponding to procentriole initiation at the base of the preexisting centrioles (mother centriole pathway) as well as numerous centrioles typically arrayed in rosettes (Fig. 2C and fig. S2), likely corresponding to procentriole initiation at de novo sites. Third, the centrioles in severely affected emi2^gRNA1 mutant MCCs were located basally, suggesting that apical migration was largely blocked (fig. S2, P to R and X). In partially differentiated cells, more centrioles were located apically (fig. S2X), but many of these were in rosette structures, indicating that apical migration and disengagement are distinct events. Fourth, essentially all centrioles in severely defective emi2^gRNA1 mutant MCCs lacked cep164 staining, except for the two basal bodies likely derived from the preexisting centrioles that extend cilia (Fig. 2, J, N, and Q). More cep164 staining occurred on centrioles in partially differentiated cells, but this often appeared incomplete rather than ring-shaped (Fig. 2, K, O, and Q). Last, cep152-GFP continued to localize at high levels with cep135-tomato in both partially and severely affected emi2^gRNA1 mutant MCCs, a further indication that basal body maturation failed to occur (fig. S2, J to O) (3). Thus, emi2 mutant MCCs display defects at multiple points in MCC differentiation, a small effect on procentriole initiation, and more severe defects in centriole apical migration, disengagement, and maturation.

Fig. 2. Centriole amplification and basal body formation in emi2 and stil mutants.

(A to E) Super-resolution images of MCCs expressing cep135-tomato and cnt4-GFP in wild-type and mutant embryos as indicated in the lower left of each panel. Boxed regions in (B) are shown at higher magnification in (C) and (D), while the boxed region in (E) is shown at higher magnification in the inset. (F) Quantification of cnt4-GFP foci located within 6 μm of the apical surface in wild-type and mutant MCCs. (G) The fraction of engaged versus disengaged centrioles within each MCC based on the number of cnt4-GFP foci per cep135-tomato focus (see fig. S2W) in 7 to 10 wild-type and mutant embryos. (H to O) Super-resolution images of MCCs expressing cnt4-GFP (green) and stained for cep164 (red) in wild-type and mutant embryos as indicated in the lower left of each panel. For emi2^gRNA1, both a severely affected MCC (≤2 cilia: J and N) and partially differentiated MCCs (partial: K and O) are shown. Insets show boxed areas at higher magnification. (P to R) Plots showing the number of cnt4-GFP foci on the x axis, and ring-shaped cep164 staining on the y axis for a given MCC in wild-type and mutant embryo. Regression line with slope is shown. Scale bars, 2 μm; Error bars = SD. Values in mutant embryos differ significantly from wild-type embryos for all measurements (P < 0.0001) based on comparisons using unpaired t tests.

For comparison, we also disrupted stil, a gene required for initiating new centriole assembly in cycling cells (24) that is also markedly up-regulated in MCC progenitors by Multicilin (~10-fold) (7). MCCs in stil^gRNA1 mutants contained many fewer cnt4-GFP foci compared to those in wild-type or emi2^gRNA1 mutants (P < 0.0001; Fig. 2, E and F), indicating a marked decrease in procentriole initiation. However, those generated were largely disengaged (Fig. 2G), apically located (fig. S2, S to U), and showed ring-shaped cep164 staining (Fig. 2, I, M, and R), in contrast to the immature procentriole rosettes observed basally in the emi2 mutants. Thus, the stil and emi2 phenotypes show a complementary disruption in the process of procentriole initiation and the subsequent steps of centriole amplification and basal body formation.

Transient decrease in APC/C activity during MCC differentiation

The results above suggest that APC/C activity needs to be inhibited by emi2 in MCC progenitors in order for centriole amplification and basal body formation to occur. To assess this possibility further, we measured APC/C activity during MCC differentiation, by imaging a well-characterized fluorescent reporter containing an APC/C degron from hGeminin (1-110) fused to monomeric azami green (mAG) (referred to as mAG-GEM) (25, 26). RNA encoding mAG-GEM was injected into embryos along with an RNA encoding Histone2B-Cherry (fig. S3, A to C) as a normalization control. We found that the levels of normalized, nuclear mAG-GEM fluorescence varied in dividing outer cells based on the known changes in APC/C activity during cell cycle progression (Fig. 3, A and C, and fig. S3, C and D); reporter fluorescence was essentially lost as cells entered mitosis (Fig. 3A, asterisks), and thus low in G₁-G₀ when APC/C activity is known to be at its highest and then gradually increased in levels as cells reentered the cell cycle, reaching its highest levels in G₂ (Fig 3, A and C). mAG-GEM reporter activity is known to increase upon reentry into the cell cycle because APC/C^cdh1 activity is abolished by the emi2 paralog, emi1, at the G₁-S transition (15, 27). The levels of normalized mAG-GEM fluorescence in dividing outer cells also increased significantly (although still cycled) when these cells express emi2 RNA, consistent with the ability of emi2 to inhibit APC/C activity (Fig. 3, B and C).

Fig. 3. Emi2 inhibits APC/C activity during MCC differentiation.

(A and B) Confocal images of the skin of wild-type (A) and an emi2 RNA–injected (B) embryo expressing mAG-GEM (green), mRFP (red), and H2b-Cherry (red). Scale bars, 10 μm. (C) Normalized pixel intensity of mAG-GEM for individual nuclei is plotted versus nucleus size, showing data from one field (203 μm², n = 40 to 50 nuclei) from three wild-type embryos (black) or three embryos misexpressing emi2 RNA (red). (D to F) Normalized mAG-GEM pixel intensity in skin nuclei measured within one confocal field (203 μm², n = 25 to 50 cells) in three different embryos (e1 to e3) at the indicated stages, for wild-type outer cells (OCs, black), ectopic MCCs (red), or endogenous MCCs (blue, only scorable at stage 15 or later). Median normalized pixel density for each group is indicated. (G to J) Normalized mAG-GEM pixel intensity in skin nuclei, as in (D) to (F), measured in wild-type outer cells (OCs, black), ectopic MCCs in wild-type (WT, blue), or emi2^gRNA1 mutant (red) embryos. Median normalized pixel density for each group is indicated. (K) Diagram summarizing the observed changes in APC/C activity mediated by emi2 during early MCC differentiation. Statistical comparisons shown are P values obtained using unpaired t tests. *P < 0.05; **P < 0.01; ***P < 0.001; ****P <0.0001; ns, not significant.

We predicted that the levels of normalized mAG-GEM fluorescence should markedly decrease as MCC progenitors in the skin leave the cell cycle when APC/C^cdh1 activity is known to increase to its maximal level. Indeed, in ectopic MCC progenitors, a marked fraction of cells lose mAG-GEM fluorescence around stages 11 and 12, or 2 to 3 hours after inducing Multicilin activity, and cell cycle exit occurs (Fig. 3D, red versus black) (18). We also observed that differentiating MCCs exit the cell cycle based on a marked increase in the levels of a Cdt1-based fluorescent reporter (25), indicating extremely low levels of a second critical cell cycle ubiquitin ligase, SCF^skp2 (fig. S3, I and J). Despite their postmitotic status, however, both endogenous and ectopic MCC progenitors went on to display moderate levels of the mAG-GEM reporter at stage 15 (Fig. 3E, blue and red dots, respectively), consistent with idea that APC/C activity is attenuated during early differentiation. At stage 18, when centriole amplification is largely complete, expression of mAG-GEM reporter was again lost (Fig. 3F), suggesting that APC/C activity returned to the high levels expected for postmitotic cells.

We next asked whether emi2 mediates the observed attenuation of APC/C activity during early MCC differentiation. The pattern of APC/C activity in dividing outer cells in both wild-type and emi2^gRNA1 mutant embryos was similar (fig. S3, E to H), as was the pattern in ectopic MCC progenitors up to stage 12 (Fig. 3G). However, by stage 14, when the levels of APC/C activity decreased in ectopic MCC progenitors (and the levels of the mAG-GEM reporter increased; Fig. 3H, blue), those in the emi2^gRNA1 mutants remained high (low mAG-GEM levels; Fig. 3H, red). The attenuation of APC/C activity in MCC progenitors was also apparent at stage 16 (moderate mAG-GEM levels; Fig. 3I, blue), before returning to high levels at stage 20 (low mAG-GEM levels; Fig. 3J, blue), while those in the emi2^gRNA1 mutants remained high at all stages (Fig. 3, H to J, red). Together, these results indicate that high levels of APC/C activity arise in postmitotic MCC progenitors upon cell cycle exit, that these levels are then attenuated by emi2 at early stages of differentiation, and that high levels of APC/C reappear once differentiation is complete (Fig. 3K).

emi2 misexpression in MCC progenitors leads to cell cycle reentry

The results above suggest that emi2 is required to promote basal body production in postmitotic MCC progenitors by inhibiting APC/C at early stages. However, this function needs to reconcile with the fact that the inhibition of APC/C in G₁ (by emi1) is a critical event promoting cell cycle reentry (15, 27). To examine this issue further, we injected emi2 RNA into emi2^gRNA1 mutant and wild-type embryos, using levels that slightly increased the levels of emi2 RNA but not affecting cell division or MCC differentiation (fig. S4, A to C). In emi2^gRNA1 mutants, emi2 RNA misexpression was sufficient to rescue the MCC differentiation phenotypes observed in both endogenous (Fig. 4, A to C) and ectopic MCCs (fig. S4, D to F), leading to largely normal basal body number and organization (Fig. 4D and fig. S4E). Moreover, in wild-type embryos, misexpression of emi2 RNA in MCC progenitors markedly increased the levels of the APC/C reporter at early stages, indicating that APC/C levels were hypersuppressed to levels normally seen in G₂ just before mitosis in dividing outer cells (Fig. 4, E to G). Although this hypersuppression did not perturb MCC differentiation (Fig. 4D and fig. S4B), it did cause MCC progenitors to engage in inappropriate cell cycle behavior. For example, wild-type MCCs do not incorporate EdU at stage 14 or later, while those misexpressing emi2 RNA continue to incorporate EdU at stage 16 (Fig. 4, H to J). In addition, MCC progenitors in emi2 RNA–injected embryos inappropriately entered into mitosis (albeit infrequently), based on a low incidence of mitotic figures as late as stage 22 (Fig. 4, K to M). These mitotic figures were disorganized due to the presence of the supernumerary centrioles generated as MCCs differentiate (Fig. 4L). Thus, these results indicate that the levels of emi2 are calibrated to ensure that APC/C activity is reduced in postmitotic MCC progenitors, when centriole amplification and basal body formation occur. Higher levels of emi2, however, lead to a hypersuppression of APC/C that causes inappropriate cell cycle reentry (15, 27).

Fig. 4. Emi2 misexpression hypersuppresses APC/C activity in MCC progenitors.

(A and B) Confocal images of the skin, labeled as in Fig. 1B. Scale bars, 10 μm. (C) Quantification of phenotypes among MCCs (n = 15 to 18) within a 203-μm² field from 10 different embryos as in Fig. 1. (D) Basal body number/MCC scored from 10 different embryos. Mean values are shown. (E to G) Normalized mAG-GEM pixel intensity in skin nuclei as shown in Fig. 3, in wild-type outer cells (OCs, black) or in ectopic, emi2^gRNA1 mutant MCCs, without (red) or with emi2 RNA injection (blue), at the indicated stage. Median pixel density for each group is indicated. Median values are shown and statistical comparisons using unpaired t tests. **P < 0.01; ***P < 0.001; ****P < 0.0001. (H and I) Confocal images of MCCs stained for cilia at the apical surface (H, asterisks) and EdU uptake (I, asterisks). Scale bars, 5 μm. (J) Fraction of ectopic or endogenous MCCs in a 203-μm² field from eight different embryos that incorporated EdU applied at the indicated stage without (WT) or with emi2 RNA injection. (K and L) Mitotic figures (mf) detected using H2b-Cherry (red) in the skin of emi2 RNA–injected embryo at stage 22. Inset in (K) shown at higher magnification in (L). Scale bar, 10 μm. (M) The number of mitotic figures detected in ectopic MCCs within a 203-μm² field at the indicated stage in different embryos. Error bars = SD. Emi2 RNA injection significantly (P < 0.0001) rescued the MCC phenotypes, basal body number, and led to significant defects in EdU uptake and mitotic figures, based on comparisons using unpaired t tests.

The emi2 mutant phenotype is exacerbated by cdh1 misexpression

The results above indicate that emi2 acts in MCC progenitors after cell cycle exit when cdh1 is known to be the main activator of APC/C. To examine this possibility further, we injected embryos with cdh1 RNA, using low levels that did not significantly perturb early cell cycle progression (fig. S5, C and D). MCC differentiation in cdh1 RNA–injected embryos was indistinguishable from wild-type embryos (Fig 5, H and I, and fig. S5, A, B, and E to H). In contrast, when the same levels of cdh1 RNA were introduced into emi2^gRNA1 mutant embryos, most MCCs failed to extend any cilia (Fig. 5, C and D), thus significantly increasing the number of MCCs with the most severe phenotype (≤2 cilia; P < 0.0001; Fig. 5, H and I). The MCCs in these embryos also contained many fewer centrioles near the apical surface (Fig. 5, G versus F), and most centrioles failed to stain with cep164 (P < 0.0001; Fig. 5, J to O). These results indicate that the levels of APC/C^Cdh1 vary in emi2 mutants, allowing some MCCs to partially differentiate. Upon cdh1 RNA misexpression, the levels of APC/C^Cdh1 presumably rise, causing a complete block in centriole amplification and basal body formation. These results also suggest that emi2 and APC/C^cdh1 likely interact to compensate for changes in component levels.

Fig. 5. APC/C^cdh1 activity antagonizes centriole amplification and leads to a loss of activated plk1.

(A to D) Confocal images of endogenous MCCs labeled as in Fig. 1B in embryos treated as indicated in the lower left of each panel. An apical (C) and basal slice (D) of the same cell are shown. Scale bars, 5 μm. (E to G) The number of cnt4-GFP foci located with 1 μm of the apical surface was scored in a given MCC, where each column represents MCCs scored within the same field of a given embryo. Data in (F) differ significantly from that in (E) (P = 0.03), while that in (G) differ significantly from that in (F) or (E) (P < 0.0001). (H and I) Quantification of phenotypes in endogenous and ectopic MCCs as in Fig. 1 (G and J). Injecting cdh1 RNA did not statistically change the MCC phenotypes in wild-type embryos, but significantly (P < 0.0001) exacerbated them in emi2^gRNA1 mutants. (J to L) Super-resolution images of an MCC expressing cnt4-GFP (green) and stained for cep164 (red) in embryos treated as indicated in the lower left of each panel. Scale bars, 2 μm. (M to O) Plot showing the number of cnt4-GFP foci on the x axis, and ring-shaped cep164 staining on the y axis per MCC treated as indicated. Regression line with slope is indicated in each case. (P) Western blot analysis with the indicated antibodies of total protein extracts prepared from control (st14) or Multicilin-HGR–injected skin progenitors, treated as indicated, and analyzed at different stages. Error bars = SD.

Targets of APC/C^Cdh1during early MCC differentiation

The results above indicate that APC/C^Cdh1 acts to inhibit the different steps involved in basal body production in MCC and that emi2 is required to block this inhibition. This finding further suggests that there are critical targets of APC/C^Cdh1 in postmitotic MCC progenitors that mediate basal body production but are degraded when emi2 is absent. To explore this possibility, we carried out Western analysis on several candidate proteins using total extracts prepared from Xenopus skin progenitors induced to undergo MCC differentiation using Multicilin-HGR (Fig. 5P).

Cyclins are well-known targets of APC/C^Cdh1 in dividing cells. We initially focused on ccna1 since it is the cell cycle cyclin most prominently up-regulated during MCC differentiation (7, 28), and Cyclin A1 mutant mice extend fewer cilia than normal in the MCCs of the respiratory airways (11). Consistent with these published data, ccna1 levels in MCC progenitors increased early and peaked at the time of centriole amplification, before decreasing after this stage is complete (Fig. 5P). Significantly, ccna1 levels are largely the same in emi2^gRNA1 mutant as in wild-type MCCs, even in the presence of exogenous cdh1 RNA (Fig. 5P).

We further examined the expression of the mitotic cyclins, based on the proposal that these cyclins are also active during MCC differentiation (12). The levels of ccnb1 or 2 RNA are largely unchanged when MCC differentiation is induced with Multicilin (table S1). Consistent with this RNA expression, proteins in MCC progenitor extracts recognized by antibody directed against Xenopus cyclinB1 and 2 are constant in levels during MCC differentiation and, moreover, are not significantly changed in emi2^gRNA1 mutant MCC progenitors, even in the presence of cdh1 RNA.

Last, we examined the levels of plk1, a key cell cycle kinase, which has been extensively associated with the centrosome cycle in dividing cells (29–32), is a target of APC/C during mitosis (33), and is required for ependymal MCC differentiation (12). Plk1 RNA is not up-regulated by Multicilin (7) and plk1 protein levels do not change significantly at early stages of MCC differentiation (Fig. 5P), nor were the levels of plk1 protein markedly changed in emi2^gRNA1 mutant MCCs, even in the presence of cdh1 RNA (Fig. 5P). Together, these results indicate that many of the obvious targets of APC/C^cdh1 during the cell cycle are not detectably altered when the levels of APC/C ^cdh1 activity increase in the context of MCC progenitors.

Plk1 activity promotes different steps of centriole amplification and basal body formation during MCC differentiation

We also analyzed MCC progenitors using a phospho-specific antibody that detects plk1 phosphorylation on Thr²⁰¹ (plk1^pT201), a key event in increasing plk1 kinase activity during the cell cycle (34). Notably, the levels of plk1^pT201 markedly increased in MCC progenitors to high levels by early stage 18, compared to cycling progenitors, and this increase was markedly reduced in emi2^gRNA1 mutants and essentially abolished when cdh1 RNA was also present (Fig. 5P). This result indicates that plk1 activity is strongly up-regulated in MCC progenitors during early differentiation in an emi2-dependent manner.

We next used constitutively activated forms of plk1 to determine whether the differentiation arrest in emi2^gRNA1 mutant MCCs is indeed caused by the loss of activated plk1. Xenopus plk1 kinase activity can be constitutively activated by substituting an aspartic acid (D) for Ser¹²⁸ (S128D) or Thr²⁰¹ (T201D) (35–37). Single mutants in Xenopus plk1 show a small increase in kinase activity, but the two mutations synergize to cause full activation (35–37).

We injected RNA encoding wild type or the constitutively activated forms of plk1 into embryos at levels that did not appear to lead to alterations in the cell cycle since the density of outer cells or MCCs in the skin was unchanged (fig. S6, L and M). In addition, RNA injection produced levels of wild-type or mutant plk1 similar to those present endogenously (fig. S6I). Injecting these RNAs did not disrupt MCC differentiation (fig. S6L), although basal body number was progressively reduced as plk1 activity increased (fig. S6K; WT:160 ± 26; plk1 RNA:173 ± 36; plk1^S128D RNA:129 ± 24; plk1^S128D/T201D RNA:91 ± 25). The severity of the MCC mutant phenotype (fig. S6J), the number of apical cnt4-GFP (fig. S6, F versus E), or the ratio of cep164 staining (Fig. 6, C and H, versus Fig. 6, B and G) in emi2^gRNA1 mutant embryos was not significantly changed upon injecting RNA encoding wild-type plk1. However, all of these parameters of MCC differentiation were partially and significantly rescued (P < 0.0001) when emi2^gRNA1 mutant embryos were also injected with RNA encoding plk1^S128D (fig. S6J; fig. S6, G versus E; and Fig. 6, D and I, versus Fig. 6, B and G) and were fully rescued by injecting RNA encoding plk1^S128D/T201D (fig. S6J; fig. S6, H versus E; Fig. 6, E and J, versus Fig. 6, B and G).

Fig. 6. Activated Plk1 promotes centriole amplification and basal body production.

(A to E) Super-resolution images of MCCs expressing cnt4-GFP (green) and stained for cep164 in embryos treated as indicated in the lower left of each panel. Insets (upper right) show boxed areas at higher magnification. Scale bars, 2 μm. (F to J) Plots showing the number of cnt4-GFP foci on the x axis and ring-shaped cep164 staining on the y axis per MCC in embryos treated as indicated. Regression line with slope is shown in each case. (K to P) Super-resolution images of MCCs expressing cep135-tomato and cnt4-GFP in embryos treated as indicated in the lower left of each panel. Insets (upper right) show boxed areas at higher magnification. Scale bars, 2 μm. (Q) The fraction of engaged (red) versus disengaged (black) centrioles for each MCC based on the number of cnt4-GFP foci per cep135-tomato focus in 7 to 10 embryos, treated as indicated. (R to W) Plots showing the number of cnt4-GFP foci on the x axis and ring-shaped cep164 staining on the y axis per MCC, in embryos treated as indicated. Regression line with slope is shown in each case. Error bars = SD. Statistical analysis used unpaired t tests. *P < 0.05; ****P < 0.0001; ns, not significant.

Similar results were obtained when the different plk1 forms were expressed in emi2^gRNA1 mutants also injected with cdh1 RNA. Both the disengagement defect (Fig. 6Q) and maturation defect in emi2^gRNA1 mutant embryos (Fig. 6S) were exacerbated by cdh1 RNA injection (Fig. 6, Q and T). These phenotypes were not significantly changed by coinjecting wild-type plk1 RNA (Fig. 6, N, Q, and U), but were partially rescued by injecting plk1^S128D RNA (Fig. 6, O, Q, and V), and fully rescued by injecting plk1^S128D/T201D RNA (Fig. 6, P, Q, and W). Again, all parameters of MCC differentiation were significantly rescued (P < 0.0001) to a larger degree by plk1^S128D/T201D compared to the partial rescue obtained with plk1^S128D.

Thus, these results are consistent with the model in which the levels of activated plk1 are severely compromised in emi2 mutants expressing cdh1 and increasing the levels of activated plk1 progressively rescues centriole amplification and basal body formation during MCC differentiation. Lower levels of plk1 kinase activity appear to be required for apical migration from basal assembly sites, while higher levels are required for full disengagement and maturation.

Loss-of-function analysis of plk1

We next asked whether a reduction of plk1 kinase activity would cause similar MCC differentiation phenotypes as seen in emi2^gRNA1 mutants. Two different gRNAs against plk1 were injected along with Cas9 protein into embryos. This approach produced only a modest decrease in the levels of plk1 or plk1^pT201 proteins in MCC progenitors (fig. S7A) and did not produce any detectable phenotype in MCC differentiation (Fig. 7, B, F, and I). This result and the fact that plk1 RNA expression is not up-regulated by Multicilin (7) suggest that the levels of plk1 required for MCC differentiation in early Xenopus embryos are largely met by a maternal component. However, injecting plk1^gRNA1 along with emi2^gRNA1 led to a further reduction in the levels of plk1^pT201 (Fig. 7J and fig. S7A) and greatly exacerbated (P < 0.0001) the emi2^gRNA1mutant phenotype (Fig. 7, D and H, versus Fig. 7, C and G; Fig. 7I), in a manner similar to that of cdh1 RNA injection shown above. These results provide further support for the model that emi2 is required during MCC differentiation to inhibit APC/C^cdh1, thus increasing the levels of activated plk1.

Fig. 7. Loss-of-function analysis of plk1.

(A to D) Confocal images of MCCs labeled as in Fig. 1B, in embryos treated as indicated in the lower left of each panel. (E to H) The number of cnt4-GFP foci located with 1 μm of the apical surface was scored in a given MCC, where each column represents MCCs within a field of a given embryo. Values in (E) versus (F) are not significantly different, while those in (E) versus (G), and those in (G) versus (H) differ significantly (P < 0.0001), using unpaired t tests. (I) Quantification of phenotypes among MCCs (n = 15 to 18) within a 203-μm² field from 10 different wild-type and mutant embryos, as indicated. Values do not differ significantly in wild-type versus plk1^gRNA1 embryos but do differ significantly in emi2^gRNA1 single mutants versus plk1^gRNA1/emi2^gRNA1 double mutants (P < 0.0001: Partial and ≤2 cilia; P = 0.04: Normal). (J) Western blot analysis with the indicated antibodies of total protein extracts prepared from skin progenitors that were isolated from embryos treated as indicated. These data are quantified further in fig. S7A. Scale bars, 5 μm, Error bars = SD, Statistical analysis used unpaired t tests.

The emi2-APC/C-plk1 axis down-regulates the expression of genes required for centriole amplification

One conceivable component of the emi2^gRNA1 mutant phenotype is an alteration in gene expression required for MCC differentiation. To examine this possibility, we took advantage of the fact that the differentiation of Xenopus skin progenitors is relatively synchronized, thus providing a temporal readout of how gene expression changes during MCC differentiation (7). Previous RNA-seq analysis of this population revealed a marked increase in approximately 1000 genes in response to Multicilin by stage 17 in both endogenous and ectopic MCCs (peak of centriole amplification), including downstream transcription factors (FoxJ1, Foxn4, and Myb), and many of the structural and regulatory factors associated with centriole amplification, basal body formation, and motile ciliogenesis (6, 7). When RNA-seq was used to compare MCC progenitors at stage 17 wild-type and emi2^gRNA1 mutant embryos, essentially no differences were found (Fig. 8A and table S2), indicating that the phenotypes in emi2^gRNA1 mutant embryos are not caused by a failure to activate the appropriate genes.

Fig. 8. The emi2-APC/C-plk1 axis regulates MCC gene expression.

(A and B) Volcano plot of RNA transcripts significantly changed in emi2^gRNA1 mutant versus wild-type MCCs at stage 17 (A) (red dots) compared to all transcripts that change in response to Multicilin-HGR at stage 17 (gray dots) (7) and at stage 26 (B) where red points represent core MCC genes [defined by Quigley and Kintner (7)]. (C) Plot showing the log₂ fold change in RNA levels of the indicated gene in isolated skin progenitors at stages 17 and 26, either wild type or after inducing MCC differentiation using Multicilin-HGR. Each QPCR value is normalized to a housekeeping control (Odc, ΔCt) and then compared to the levels detected in unmanipulated wild-type progenitors at stage 17 (ΔΔCt). (D) Analysis as in (C) applied to wild-type and stil^gRNA1 mutant MCC progenitors. (E) Analysis as in (C) applied to wild-type and emi2^gRNA1 mutant MCC progenitors that also express cdh1 RNA. (F) Analysis as in (C) applied to wild-type and emi2^gRNA1 mutant MCC progenitors that also express plk1^S128D/T201D RNA. Error bars = SD.

We next compared MCC progenitors from emi2^gRNA1 mutant and wild-type embryos at stage 26, when MCC differentiation is largely complete (fig. S7B). Unexpectedly, the major difference was a significant marked up-regulation of ~100 genes in the emi2^gRNA1 mutant MCCs (Fig. 8B and table S3). These emi2-regulated genes are strongly induced by Multicilin in MCC progenitors at stage 17 and normally down-regulated in MCCs by stage 26 (fig. S7C and table S3), making up a significant fraction of the genes transiently expressed during MCC differentiation (fig. S7D). In addition, a vast majority of these emi2-regulated genes encode factors required to regulate centriole amplification (ccno, cdk1, cdc20b, aurka, bora, and emi2 itself), initiate procentriole formation (stil, cep152, sass6, plk4, cep135, ccdc67, etc.), build centrioles (cnt3, cep76, cep44, cep57l1, Poc1b, etc.), or mature basal bodies (odf1, odf2, cep83, etc.; table S3). Thus, genes involved in both centriole amplification and basal body formation undergo an on-off cycle during MCC differentiation in an emi2-dependent manner. Furthermore, this increase in gene expression can potentially account for the elaborate assembly of structures detected basally in emi2^gRNA1 mutant MCCs at later stages (fig. S2).

The RNA-seq results were validated with quantitative polymerase chain reactions (QPCR) to assess the expression of mcidas and emi2 (early regulators), ccno and deup1 (centriole amplification), and foxj1 (motile cilia). All five genes were confirmed to be induced by Multicilin in skin progenitors at a stage 17 equivalent but only the expression of foxj1 remained elevated at stage 26, and unchanged in emi2^gRNA1 mutant versus wild-type MCCs (Fig. 8C, blue versus black bars). By contrast, the two genes associated with centriole amplification (ccno and deup1) as well as mcidas and emi2 are strongly down-regulated by stage 26 (Fig. 8C, red bars at stage 26 versus 17) in an emi2-dependent manner (Fig. 8C, blue bars at stage 26), consistent with the RNA-seq data.

The results above raise the possibility that there is a direct link between centriole amplification and the on-off cycle of centriolar gene expression during MCC differentiation. To examine this possibility, we used QPCR to analyze gene expression in the stil mutant MCCs, where centriole assembly is effectively blocked at an early stage (Fig. 2). We first determined that ectopic MCC progenitors do indeed display the same phenotype as endogenous MCCs in stil^gRNA1 mutants (fig. S7, E to H). We then carried out QPCR analysis of ectopic MCC progenitors, finding similar levels of ccno, deup1, mcidas, emi2, and foxj1 RNA in wild-type and stil mutant MCCs at both stage 17 and stage 26 (Fig. 8D). Thus, centriole gene expression during MCC differentiation is not regulated per se by the levels of centriole amplification but specifically by emi2 activity.

To determine whether this misregulation of late MCC gene expression in emi2 mutants was exacerbated by cdh1 or rescued by plk1^S128D/T201D, we carried out QPCR analysis on wild-type or emi2 mutant MCC progenitors that also misexpress cdh1 or plk1^S128D/T201D RNA. While misexpressing cdh1 further increased the gene expression that is up-regulated in emi2 mutants at late stages (Fig. 8E), misexpressing plk1^S128D/T201D RNA markedly reduced this expression back to wild-type levels (Fig. 8F). Thus, the regulation of centriole amplification by the emi2-APC/C-plk1 axis also acts to regulate the on-off cycle of gene transcription that drives this process during MCC differentiation.

DISCUSSION

Multicilin along with the E2F proteins is sufficient to markedly up-regulate the expression of structural genes involved in centriole amplification and basal body formation, thus setting the stage for MCC differentiation (6, 7, 18). However, centriole amplification is normally a process that is coupled to different phases of the cell cycle, suggesting that Multicilin also promotes this process by establishing a novel cell cycle state. Here, we show that MCC progenitors display a novel cell cycle state defined by low SCF^Skp2 activity indicative of a postmitotic cell but moderate APC/C^cdh1 levels indicative of entry into the cell cycle. We find that Multicilin promotes this state in Xenopus skin progenitors by up-regulating the expression of the APC/C inhibitor, emi2. We further show that emi2 inhibits APC/C^cdh1 to enable the phosphorylation and activation of plk1, which, in turn, acts in a graded manner to promote the different steps required for centriole amplification and basal body formation. This emi2-APC/C-plk1 axis also mediates a negative feedback loop that reduces the expression of centriole amplification genes once differentiation is complete.

This study extends on previous observations showing that artificially low APC/C activity during MCC differentiation causes progenitors to initiate inappropriate cell cycle behavior (12). Specifically, we show that the opposite is also true: High levels of APC/C^cdh1 activity arise as progenitors exit the cell cycle to initiate MCC differentiation and, if not constrained, block centriole amplification and basal body formation at multiple points, including centriole disengagement, apical migration, and maturation. We further find that MCC progenitors have co-opted emi2 to inhibit this APC/C^cdh1 activity, thus carrying out a function that is similar to that of its paralog, emi1, during cell cycle reentry. In the latter case, emi1 is up-regulated by E2f1 in G₁ and triggers entry into S-phase by inhibiting APC/C^cdh1 as part of a bistable switch (15, 27). In a similar manner, if the levels of emi2 are increased in MCC progenitors by RNA misexpression, they inappropriately undergo DNA synthesis and mitosis. Thus, Multicilin co-opts the E2Fs to up-regulate emi2 activity, thus reducing APC/C^cdh1 activity and enabling centriole amplification and basal body formation in MCC progenitors, while avoiding a threshold level that would trigger cell cycle reentry (Fig. 9).

Fig. 9. Model for an emi2-APC/C-plk1 axis operating during MCC differentiation.

Diagram illustrating the role of emi2 in modifying APC/C^cdh1 during MCC differentiation, thus allowing plk1 activation, based on phenotypes that are generated in both loss- and gain-of-function experiments.

The activation of APC/C^cdh1 in emi2 mutant MCCs presumably leads to a loss of critical players required for centriole amplification and basal body maturation to move forward. This observation was exploited here to identify these critical players based on their sensitivity to APC/C^cdh1 activity, and on whether they restore centriole amplification and basal body formation in rescue experiments. This approach led to the finding that plk1 kinase activity, as indicated by the levels of plk1^pT201, is markedly up-regulated at early stages of MCC differentiation, is extremely sensitive to APC/C^cdh1 activity, and is sufficient to rescue centriole amplification and basal body formation in emi2 mutant MCCs, even those with high levels of APC/C^cdh1activity.

Significantly, the phenotype of the stil mutant MCCs indicates that the multiple steps mediating centriole amplification can occur efficiently even if the first step, procentriole initiation, is markedly reduced. In addition, the variable nature of the emi2 mutant phenotype and the differential rescue obtained with plk1 mutants with different levels of kinase activity indicate that the different steps required for centriole amplification all require plk1 activity but to different extents. Thus, apical migration of centrioles from basal assembly sites appears to occur more readily when plk1 activity is low, while centriole disengagement and basal body formation appear to occur more readily with high levels of plk1 activity. We also note that the activated forms of plk1 reduce the final number of basal bodies that form, suggesting that the levels of plk1 activity also need to be tightly regulated for centriole amplification to occur at a normal pace. Together, these results lead to the model (Fig. 9) in which emi2 is required during MCC differentiation to inhibit APC/C^cdh1, leading to the phosphorylation and activation of plk1, which, in turn, acts in a graded manner to promote multiple steps in centriole amplification and basal body formation. This model is consistent with previous studies implicating plk1 in MCC differentiation (38) and with the role of plk1 in multiple steps of the centrosome cycle in dividing cells (29–31, 39).

Previous studies have also implicated the cyclin-dependent kinases (cdks) and cyclins in MCC differentiation (10–12). In light of our results, and the known interactions between the cdks and plk1 during the cell cycle, one likely possibility is that cyclin/cdk activity is required to promote the activation of plk1, and that in the context of these cells, this is sufficient to drive centriole amplification and basal body formation forward in a graded fashion while avoiding other cell cycle behavior. Current cell cycle models propose that cyclinA2/cdk activity drives the phosphorylation of plk1 on Thr²⁰¹ at the onset of mitosis, by phosphorylating bora, which, in turn, recruits aurkA to plk1 (40). Cyclin A1 potentially carries out a similar function during MCC differentiation, since it is one of the few cell cycle cyclins notably up-regulated transcriptionally during MCC differentiation in different species (table S1) (11), and Cyclin A1 mutant mice extend fewer cilia than normal in the MCCs of the respiratory airways (11). In addition, Ccno, an MCC-specific cyclin related to the cyclinA subfamily, could also be a critical factor in carrying out this function based on the disruption of basal body production observed in ccno mutants in human, mouse, and Xenopus (13, 14).

We emphasize that the levels of cyclin A1 at early differentiation stages are not appreciably altered in emi2^gRNA1 mutant MCCs (Fig. 5P), suggesting that additional targets of APC/C will need to be assessed to determine why emi2 mutant MCCs fail to generate phosphorylated plk1. Based on the model proposed above, players such as aurkA, phosphorylated bora, and/or ccno will need to be examined to determine what factors are rate-limiting and deficient when emi2 is mutant and APC/C^cdh1 levels are too high. Moreover, we cannot rule out the possibility that the phosphorylation and activation of plk1 are also mediated by additional players that are unique to MCC progenitors and sensitive to APC/C^cdh1 levels, particularly given the high levels of plk1^pT201 detected in differentiating cells in relation to those seen in cycling cells.

Our results also reveal the existence of regulatory feedback loops that likely ensure that the process of centriole amplification during MCC differentiation is robust. A regulatory feedback loop is evident in the cdh1 RNA–injected embryos, where wild-type MCCs compensate, but emi2 mutant MCCs are severely affected. We have also discovered a regulatory feedback loop that occurs between the genes transiently expressed during MCC differentiation and the emi2-APC/C-plk1 axis. Genes transiently expressed during MCC differentiation are likely to be direct targets of Multicilin and are highly enriched in those involved in centriole amplification and basal body formation (including emi2 and ccna1). Our data indicate that the emi2-APC/C-plk1 axis is required to shut off the RNA expression of this transient group, while genes regulated by downstream transcription factors, such as foxj1, stay on. This negative feedback loop would tend to balance the levels of gene expression required for organelle biogenesis and the activity of the emi2-APC/C-plk1 axis, providing robustness to this stage of MCC differentiation.

MATERIALS AND METHODS

Xenopus laevis embryos and microinjections

X. laevis embryos were prepared by in vitro fertilization using standard protocols (41). Twenty minutes after fertilization, embryos were dejellied using 3.0% cysteine (pH 8.2) and injected immediately with Cas9/gRNAs. At the first cleavage, additional injections were done to introduce synthetic mRNAs, typically targeting the animal blastomeres with four equidistant injection sites (typically 0.1 to 5.0 ng of RNA per embryo).

DNA constructs and RNA synthesis

Various tagged proteins were expressed in Xenopus embryos using injection of synthetic RNAs generated in vitro using SP6 polymerase to transcribe DNA templates based on the CS2. The open reading frame for various marker proteins or cell cycle components were inserted into different CS2-based vectors for tagging, including CS2-MT (six myc-epitopes), CS2-N1 (N-terminal GFP), and CS2-tomato (N-terminal). Previously described DNA templates to generate RNAs include those that encode cnt4-GFP (centriole/basal body marker), H2b-Cherry, GFP tagged with a nuclear localization signal (NLS-GFP), membrane-localized red fluorescent protein (mRFP), Multicilin-HGR, and cep152-RFP (centriole/basal body marker) (14, 18). New DNA templates were generated for expressing X. laevis cep135, cdh1 (fzr1), and plk1. PCR (Phusion) was used to amplify the corresponding open reading frame from a cDNA library of stage 17 X. laevis embryos and insertion into a CS2 vector using Gibson cloning (NEB). Constitutively activated forms of plk1 were generated in a CS2-MT-plk1 construct using the standard protocol for using PCR for site-directed mutagenesis using previously published primer sequences (table S6). While the wild-type and single mutant of plk1 contained six copies of the myc-epitope tag, that of the double mutant contained four copies. DNA templates for generating the FUCCI reporters were generated by amplifying sequences encoding mAG-hGEM(1-110) or mKO2-hCdt1(30-120) using PCR from pBOB-EF1-FastFUCCI-Puro (Addgene 86849) construct and inserting into the CS2 vector using Gibson cloning. DNA templates were linearized and used as templates to produce synthetic, capped mRNA in vitro with SP6 polymerase as described previously (21).

CRISPR mutagenesis

Coding sequences were selected for CRISPR targeting based on whether they are conserved in both the S and L gene variants, encode conserved protein domains required for protein function, and are considered optimized for CRISPR mutagenesis (https://chopchop.cbu.uib.no; table S6). Sixty picomoles of a synthetic gRNA (Synthego) was mixed with 12 pmol of NLS-Cas9 protein (PNA Bio) in a total volume of 5 μl and one injection of 20 nl was made into the animal pole of embryos, between 25 and 40 min after fertilization. Compound mutants were injected successively with different gRNA/Cas9 mixtures. At the two- to four-cell stage, embryos were injected with experimental or tracer RNAs.

CRISPR targeting of genes and the specificity of the mutant phenotypes were assessed in this study using different approaches. In some cases (emi2 and plk1), at least two gRNAs were tested to control for offsite mutations. In addition, efficient editing and indel production in embryos injected with Cas9/emi2^gRNA1 was assessed using the RNA-seq analysis, using reads that mapped to the L and S forms of the emi2 gene (fig. S1K). Last, in some cases, we assess protein levels (plk1) since F0 crispants potentially leave maternal RNA intact, which can potentially mask mutant phenotypes, especially in cases where the gene is not up-regulated during MCC differentiation.

RNA preparation, RNA-seq, and quantitative RT-PCR

Animal caps (skin progenitors) were isolated at stage 10 from wild-type, emi2^grna1, and stil^grna1 crispants that were injected at the two- to four-cell stage with RNA encoding Multicilin-HGR and in some experiments with cdh1 and plk1^S128D/T201D RNA. Isolated animal caps were cultured in 0.5× MMR (Marc’s Modified Ringer’s), treated with dexamethasone (1 μM) at stage 11 to induce Multicilin-HGR activity, and harvested at 9 and 20 hours (stages 17 and 26, respectively) with the proteinase K method, followed by phenol chloroform extractions, precipitation in 4 M LiCl, treatment with ribonuclease-free deoxyribonuclease, and a second series of phenol chloroform extractions and ethanol precipitation. RNA-seq libraries were constructed with Illumina Truseq RNA Sample Preparation kit v2 according to the manufacturer’s instructions and sequenced on a HiSeq 2500 at 1 × 50 base pairs to a depth of 20 million to 40 million reads. Each RNA-seq condition was performed in triplicate, using animal caps isolated from different females on different days.

RNA samples, as prepared above, were also used to measure gene expression using quantitative PCR as described previously (18). Briefly, RNA was converted in cDNA using reverse transcription and then assayed by quantitative PCR using the appropriate primer pairs (table S6) in triplicate with the ABI Prism 7900HT thermal cycler (Life Technology). Samples were normalized to the levels of ornithine decarboxylase RNA (odc) as an internal control.

Bioinformatics

Sequenced reads were quality-tested using FASTQC reads and mapped to v2 of the X. laevis 9.2 genome using RNA-STAR (42). Mapping was carried out using default parameters (up to 10 mismatches per read and up to 9 multimapping locations per read). The aligned reads were then counted, normalized, and tested for differential expression using the HOMER software package from UCSD and edgeR (43, 44). Differentially expressed genes were defined as having a false discovery rate <0.05 and a log₂ fold change >1 when comparing two experimental conditions. For visualization purposes, RNA-seq reads were also mapped to the X. laevis genome with bowtie2 (45) and loaded as tracks into the IGV browser (46).

Confocal and super-resolution imaging of centriole markers

Embryos expressing fluorescent proteins were processed for confocal imaging by fixing in quickfix {3.7% formalin and 0.3% glutaraldehyde in PBT [0.1% Triton X-100 in phosphate-buffered saline (PBS)]} for 10 min at room temperature and washing extensively with PBT. In some experiments, embryos were processed for cilia staining by incubating with an anti-acetylated tubulin, monoclonal antibody (Sigma-Aldrich; 1:2000), and a cy5-donkey-anti-mouse IgG. In embryos labeled by EdU injection, embryos were processed with Click-IT labeling, using the manufacturer’s recommendations (Invitrogen). For cep164 detection, embryos were fixed with 4% paraformaldehyde in PBS on ice for 1 hour, then gradually dehydrated with ethanol, and stored in −20°C for overnight. The embryos were gradually rehydrated with PBS, then pretreated with PBT for 10 min at room temperature (RT), and then incubated with cep164 antibody in 4°C overnight. The embryos were washed with PBT extensively and subsequently incubated with Alexa Fluor 568 (Invitrogen) for 45 min at RT and then washed with PBT extensively. All embryos were mounted between coverslips in Fluoromount-G (no. 0100-01, SouthernBiotech) for imaging.

Different imaging approaches were used to score different aspects of MCC differentiation. To quantify skin cell types (e.g., Fig. 1F), MCC differentiation (e.g., Fig. 1G), and apical centriole localization (e.g., Fig. 4E), we captured confocal images of randomly chosen 203-μm² fields from at least 10 embryos with an LSM700 microscope using a Plan Apochromat 63×/1.4 numerical aperture (NA) oil-immersion objective. Confocal images consisted of a stack of 0.5-μm slices, with the top slice capturing the apical surface and cilia projection, while the bottom slice (~6 μm) identified even poorly differentiated MCCs as these contained aggregates of cnt4-GFP deep in the cytoplasm.

Cell type was scored in these images by counting the number of MCCs (based on the cnt4-GFP) and outer cells (based on their large apical domain). The number of outer cells per field provides an indication of general cell cycle defects, since cell division at early development stages largely occurs without cell growth, thus leading to a marked reduction in cell size. Larger outer cells, and thus fewer cells per field, strongly indicate that cell division has slowed or has been blocked.

MCC differentiation was scored as follows: MCCs with a normal pattern of basal bodies at the apical surface with cilia extension (normal), MCCs with an abnormal pattern of basal bodies at the apical surface, many of which were aggregated in rosette-like structures, with partial ciliation (partially differentiated), or MCCs with few apical basal bodies, extending two or fewer cilia (≤2 cilia). The percentage of MCCs in each category was calculated for each field scored, usually from 10 different embryos, and plotted as a single point. Apical migration was scored by determining the number of cnt4-GFP foci localized within 1 μm of the apical surface. This number was determined for all of the MCCs within a given field and plotted separately for each embryo.

Scoring centriole disengagement (e.g., Fig. 2G) and centriole number/maturation (e.g., Fig. 2P) required super-resolution images taken on the LSM880 Airyscan system (Zeiss) with a Plan Apochromat 63×/1.4 NA oil-immersion objective and 2.5× or 4× zoom for each acquisition set. Embryos were scanned from apical to basal to image the apical domain where cnt4-GFP resides (~2.5 or ~6 μm depth for control and mutant embryos). Serial z-stack scanning was carried out at 120- or 200-nm intervals. Disengagement was scored based on the number of cnt4-GFP foci associated with a single cep135-tomato focus (e.g., Fig. 2, B and C). Centriole number and maturation was scored by determining the number of total cnt4-GFP foci and the number with ring-shaped cep164 staining within each MCCs of randomly chosen 52- or 33-μm² fields taken from at least 15 embryos in three separate experiments. Airyscan-processed images were analyzed using ZEN software or ImageJ. The linear regression analysis was performed using Prism software.

Analysis of APC/C activity using FUCCI reporters

Embryos, injected at the two-cell stage with RNA encoding mAG-hGEM and H2b-Cherry, were fixed in quickfix, mounted, and imaged in an LSM700 microscope using a Plan Apochromat 63×/1.4 NA oil-immersion objective. Images were captured at 0.5× zoom (203-μm² field), using the same settings for a given experiment and set to prevent signal saturation. The pixel intensity within the nucleus for the two fluorescent proteins was measured using ImageJ and corrected by subtracting background fluorescence in the cytoplasm. The pixel intensity of mAG-hGEM was normalized by dividing by the pixel intensity of H2b-Cherry for each nuclei and then adjusted by determining the maximum ratio of mAG-hGEM to H2b-Cherry in dividing outer cells at each time point, and setting this value to one in both wild-type and mutant MCCs.

Immunoblotting analysis using X. laevis whole-cell lysates

Western analysis of MCC progenitors was performed by generating mutant and RNA-injected embryos as described above and then isolating the skin progenitors at stage 10.5 by removing the animal cap, trimming away marginal zone tissue, and culturing in 0.5× MMR. Multicilin-HGR was activated by treating with dexamethasone at stage 11, which is sufficient to drive most skin progenitors (~95%) into MCC differentiation (18). At the indicated stage equivalent, the animal cap tissue was collected and directly lysed with rigorous pipetting and vortexing in 2× lithium docecyl sulfate (LDS) sample buffer (Thermo Fisher Scientific, NP0008), followed by boiling for 10 min. The sample proteins were subjected to SDS–polyacrylamide gel electrophoresis and transferred to polyvinylidene difluoride membranes that were then blocked with 0.1% casein blocker (#161-0783, Bio-Rad) in 0.5× PBS for 30 min followed by overnight incubation with the indicated primary antibodies at 4°C in 0.1% Tween-20 in blocking buffer. After extensive washing in 0.1% Tween-20 in PBS, the blots were incubated with Alexa Fluor 680– or 800–conjugated secondary antibodies (Thermo Fisher Scientific, A21058, A21076, and A32735) for 45 min at room temperature, washed with 0.1% Tween-20 in PBS, and imaged using Odyssey (LI-COR) instrument. The primary antibodies used in this study are listed in table S6.

Supplementary Materials

This PDF file includes:

Figs. S1 to S8

Other Supplementary Material for this manuscript includes the following:

Tables S1 to S6

View/request a protocol for this paper from Bio-protocol.