Atypical cadherin Dachsous1b interacts with Ttc28 and Aurora B to control microtubule dynamics in embryonic cleavages

Summary

Atypical cadherin Dachsous (Dchs) is a conserved regulator of planar cell polarity, morphogenesis, and tissue growth during animal development. Dchs functions in part by regulating microtubules by unknown molecular mechanisms. Here we show that maternal zygotic (MZ) dchs1b zebrafish mutants exhibit cleavage furrow progression defects and impaired midzone microtubule assembly associated with decreased microtubule turnover. Mechanistically, Dchs1b interacts via a conserved motif in its intracellular domain with the TPR motifs of Ttc28 and regulates its subcellular distribution. Excess Ttc28 impairs cleavages and decreases microtubule turnover, while ttc28 inactivation increases turnover. Moreover, ttc28 deficiency in dchs1b mutants suppresses the microtubule dynamics and midzone microtubule assembly defects. Dchs1b also binds to Aurora B, a known regulator of cleavages and microtubules. Embryonic cleavages in MZdchs1b mutants exhibit increased and in MZttc28 mutants decreased sensitivity to Aurora B inhibition. Thus, Dchs1b regulates microtubule dynamics and embryonic cleavages by interacting with Ttc28 and Aurora B.

Introduction

Dachsous (Dchs in vertebrates/Ds in Drosophila) is a single-pass transmembrane protein with 27 extracellular cadherin repeats and a less-characterized intracellular domain (Clark et al., 1995; Halbleib and Nelson, 2006). It was first shown in Drosophila to influence organ size and planar polarity through heterophilic interactions with its binding partner, another giant cadherin Fat (Adler et al., 1998; Clark et al., 1995; Yang et al., 2002). In Drosophila, a Golgi-associated kinase, Four-jointed (Fj), is involved in the modulation of Ds-Fat interactions to regulate tissue polarity (Brittle et al., 2010; Cho and Irvine, 2004; Simon, 2004; Simon et al., 2010). Current evidence places the Fat/Ds/Fj module upstream and/or in parallel to the core planar cell polarity (PCP) signaling pathway in regulation of tissue polarity (Adler et al., 1998; Casal et al., 2006; Donoughe and DiNardo, 2011; Ma et al., 2003; Matakatsu and Blair, 2004; Matis et al., 2014; Yang et al., 2002).

dachsous is evolutionarily conserved, with two homologs in mammals and three in zebrafish (Li-Villarreal et al., 2015; Mao et al., 2011). Dchs1 neonatal mutant mice manifest abnormalities in multiple organs, such as kidney and skeletal morphogenesis defects (Bagherie-Lachidan et al., 2015; Kuta et al., 2016; Mao et al., 2015; Mao et al., 2016; Mao et al., 2011). Zebrafish mutants lacking both maternal and zygotic function of dchs1b (MZdchs1b) display numerous embryogenesis defects, including abnormal cleavages and gastrulation movements (Li-Villarreal et al., 2015), revealing conserved and unique functions of Dchs in vertebrate development. Mutations in human DCHS1 homolog have been implicated in Van Maldergem syndrome that includes periventricular neuronal heterotopia. Knockdown of Dchs1 in mouse embryonic brain leads to abnormal localization of neural progenitor cells, reminiscent of the human periventricular neuronal heterotopia phenotype (Cappello et al., 2013). Moreover, recent studies also link mitral valve prolapse, a cardiac valve disease, to DCHS1 mutations (Durst et al., 2015). The broad involvement of Dchs cadherins in vertebrate development and human disease underscores the need to understand the underlying cellular and molecular mechanisms.

Several molecules have been identified in Drosophila that bind to the intracellular domain (ICD) of Ds and affect downstream events. Lowfat, a cytoplasmic protein, binds to Ds ICD and functions to modulate Ds/Fat expression levels (Mao et al., 2009), while the WD40-repeat protein Riquiqui has been shown to interact with Ds ICD to control the Hippo signaling pathway (Degoutin et al., 2013). Recently, the SH3 domain protein Vamana/Dlish has emerged as a novel interactor of Ds ICD to influence the downstream effector, unconventional myosin Dachs (Misra and Irvine, 2016; Zhang et al., 2016). Yet, neither Vamana/Dlish nor Dachs homologs have been identified in vertebrates.

Current studies also suggest that Dchs regulates cellular processes through microtubule cytoskeleton. In the Drosophila developing wing epithelium, Ds/Fat are thought to provide directional cues to the core PCP components by regulating polarization of apical microtubules (Harumoto et al., 2010; Matis et al., 2014). In zebrafish, defective organization and bundling of the microtubule network are thought to underlie a subset of defects in MZdchs1b mutants, including dorsal determinant transport and epiboly (Li-Villarreal et al., 2015). However, the molecular mechanisms through which transmembrane protein Dchs influences the microtubule cytoskeleton remain unknown.

Here we report that MZdchs1b zebrafish mutants exhibit cleavage furrow progression defects and impaired assembly of midzone microtubules during cytokinesis. By studying microtubules in vivo, we provide evidence that these defects are associated with decreased turnover of microtubules during embryonic cleavages, implying Dchs1b promotes microtubule dynamics. We identify Ttc28, a tetratricopeptide (TPR) motif-enriched cytoplasmic protein, as a molecular link between Dchs1b and microtubule dynamics. We show that Dchs1b interacts via a conserved motif within its ICD with the TPR motif of Ttc28 and regulates its subcellular distribution in zebrafish blastulae. Our experiments attribute the microtubule turnover and embryonic cleavage defects in MZdchs1b mutants to mis-localized Ttc28 during cleavage stages. Overexpression of Ttc28 disrupts embryonic cleavages and decreases microtubule turnover, whereas loss of Ttc28 increases microtubule dynamics. Moreover, genetic inactivation of ttc28 in MZdchs1b mutants suppresses the microtubule dynamics and midzone microtubule defects, without normalizing embryonic cleavages. We further discover that Dchs1b and Ttc28 influence microtubule dynamics and embryonic cleavages in part via Aurora B, a conserved regulator of cell division and microtubules (Basant et al., 2015; Douglas et al., 2010; Gruneberg et al., 2004; Nunes Bastos et al., 2013; Yabe et al., 2009). MZ cellular island (cei) mutants lacking Aurora B function or embryos treated with Aurora B inhibitor, exhibit severe cleavage defects (Yabe et al., 2009). We demonstrate that embryos treated with Aurora B inhibitor, exhibit reduced microtubule turnover similar to that in MZdchs1b mutants. Consistent with Dchs1b and Ttc28 having opposing effects on microtubule dynamics and cleavages, embryonic cleavages in MZdchs1b mutants are sensitized while MZttc28 mutants are less sensitive to Aurora B inhibition. Taken together, our studies identify Ttc28 and Aurora B as novel interactors of Dchs1b that control microtubule dynamics during embryonic cell divisions.

Results

Loss of dchs1b results in cleavage furrow progression defects

We first investigated the mechanisms through which Dchs1b regulates early cleavages in zebrafish by employing the Tg[βactin2:GCaMP6s]^stl351 transgenic line (Chen et al., 2017), which allows detection of calcium signaling associated with cleavage furrow initiation, propagation, and deepening (Webb et al., 1997). In contrast to the stereotypical calcium-signaling pattern associated with synchronous cleavages that occur at about 15-minute intervals in wild-type (WT) embryos (Kimmel et al., 1995), time-lapse confocal imaging revealed aberrant calcium activities in cleavage-stage MZdchs1b^fh275/fh275 mutants that harbor a strong/null nonsense mutation in this gene (Li-Villarreal et al., 2015) (Figure 1A; Movie S1). Whereas the calcium signals marking furrow initiation were usually positioned perpendicular to the preceding cleavage furrow in WT embryos, they were positioned at various angles in MZdchs1b mutants (Figures 1B and S1A), indicating a cytokinetic furrow positioning defect. In MZdchs1b mutant embryos, furrow calcium signaling sometimes failed to extend laterally during the propagation phase (Figure 1C). In addition, the narrowing of the calcium signal associated with furrow deepening in WT embryos, was delayed in MZdchs1b mutants (Figure 1D). These data indicate cytokinesis defects in MZdchs1b mutant embryos.

Figure 1. Disrupted embryonic cleavages in MZdchs1b.

(A) Representative time-lapse still images of cleavage furrow calcium signaling indicated by Tg[(3actin2:GCaMP6s]^stl351 during cleavage stages in WT and MZdchs1b. Scale bar, 300 um.

(B) Quantification of cleavage furrow orientation by monitoring calcium signaling in WT and MZdchs1b at 2-4 cell stage. Directional distributions of the positioning angle θ is shown as a rose diagram of 5° bins. Each bin represents a percentag e of the total population with the circular line indicating the percentile. N, number of embryos. P<0.005, Mardia-Watson-Wheeler test.

(C) Quantification of calcium signaling dynamics during furrow propagation in WT and MZdchs1b at 2-4 cell stage. The Y-axis shows the percentage of furrow calcium signaling width d at each time point divided by the total cell width d'. N, number of embryos.

(D) Quantification of the ratio of the first furrow deepening calcium-signaling width at the onset of second cleavage in WT and MZdchs1b. The Y-axis shows the ratio of the initial cleavage calcium signaling width w divided by the total cell width w'. N, number of embryos. *** P<0.001, Student's unpaired t-test. Error bars represent standard deviation.

(E) Representative images of mitotic and cytokinesis events in the blastomeres stained using γ-Tubulin (green) antibody and counterstained with DAPI (blue). Scale bar, 30 μm. In the “merge”images, arrowheads indicate two or more nuclei in one cell, asterisk indicates abnormal chromatinbridge, and arrow indicates micronuclei.

(F) Quantification of mitotic events in WT and MZdchs1b embryos at 2.5 hpf. N, number of embryos.n, number of cells. See also Figure S1, Movie S1.

To test whether cytokinesis anomalies could lead to aberrant furrow calcium activities, we treated WT Tg[βactin2:GCaMP6s]^stl351 zygotes with Aurora B inhibitor ZM447439. Consistent with the role of zebrafish Aurora B in cytokinetic furrow establishment and extension (Yabe et al., 2009), we found that cleavage furrow calcium signaling was positioned normally but failed to propagate laterally after ZM447439 treatment (Figures S1C-S1E; Movie S2). These observations further support the conclusion that MZdchs1b mutants exhibit cytokinesis defects during embryonic cleavages. Moreover, using confocal imaging for γ-Tubulin and DNA staining, we detected mitosis and cytokinesis defects in MZdchs1b mutants at 2.5 hours post fertilization (hpf), including polyploidy, chromatin bridges, multipolar spindles, and micronuclei (Figures 1E, 1F and S1B). Collectively, these data indicate that dchs1b is required for normal cell divisions during early zebrafish embryogenesis.

Dchs1b promotes microtubule dynamics during embryonic cleavages

The regulation of microtubule dynamics is critical during mitosis and cytokinesis, and both astral and spindle midzone microtubules contribute to the specification of a cleavage furrow (Bringmann and Hyman, 2005; Eggert et al., 2006). As our earlier work revealed abnormal microtubule organization in MZdchs1b mutants (Li-Villarreal et al., 2015), we hypothesized that the observed mitotic and cytokinesis defects in MZdchs1b mutants are due to mis-regulated microtubules. To explore dynamic microtubule organization during mitosis and cytokinesis in MZdchs1b mutants, we performed confocal time-lapse imaging during cleavage stages using Tg[ef1α:dclk-GFP] transgenic line in which microtubules are labeled with green fluorescent protein (GFP) (Tran et al., 2012). We found that midzone microtubule assembly during cytokinesis was significantly delayed in MZdchs1b mutants compared to WT (Figures 2A and 2B), and strikingly, in some severely affected mutant blastomeres the midzone microtubules failed to coalesce into a midbody, resulting in cytokinesis failure (Figure 2A; Movie S3). The above data demonstrate that irregular midzone microtubule assembly can contribute to the cytokinesis defects in MZdchs1b mutants.

Figure 2. Midzone microtubule assembly and YCL microtubule dynamics defects in MZdchs1b.

(A) Representative time-lapse still images of microtubule dynamics during cytokinesis in WT and MZdchs1b embryos using Tg[ef1a:dclk-GFP]. Arrowheads denote the initial midzone microtubule width. Scale bar, 30 μm.

(B) Quantification of the midzone microtubule assembly activities in WT and MZdchs1b. The Y-axis shows the ratio of midzone microtubule width d at each time point divided by the initial width d'. N, number of embryos. **** P<0.0001, paired student's two-tailed t-test was used to determine the difference between control and mutants across the time course. Error bars represent SEM.

(C) Representative still time-lapse images of YCL microtubules marked with Tg[ef1a:dclk-GFP] during cleavage stages in WT and MZdchs1b. To the right a cartoon of 8-celled embryo with the black box illustrating the analyzed region; A, animal pole; V, vegetal pole. Scale bar, 30 μm.

(D) Quantification of YCL microtubule density during individual cell cycles from the time-lapse movies in WT and MZdchs1b. N, number of embryos. Error bars represent SEM.

(E) Quantification of EB3-GFP track speed in the YCL in WT and MZdchs1b. The Y-axis shows the relative frequency of all quantified tracks in corresponding speed bins. The right panel shows a representative speed heatmap in the region of interest. A, animal pole; V, vegetal pole. N, number ofembryos; n, number of EB3 tracks.

(F) Directional distribution of EB3-GFP track angles in WT and MZdchs1b. N, number of embryos; n,number of EB3 tracks. See also Figure S6, Movie S3 and S4.

To test whether the midzone microtubule defects in MZdchs1b mutants were associated with altered microtubule dynamics, we examined microtubules by time-lapse confocal microscopy in the yolk cytoplasmic layer (YCL), a thin layer in the yolk cortex. During cleavage stages, microtubules originate from the marginal blastomeres and extend along the animal-vegetal embryonic axis into the YCL (Solnica-Krezel and Driever, 1994). However, the dynamic changes of microtubule network during cleavages have not yet been characterized with live imaging. We found that the YCL microtubules depolymerize and polymerize from the animal toward the vegetal pole in a cell cycle-dependent manner (Movie S4, 2-2.5 hpf), giving rise during consecutive cell cycles to a zone of low microtubule density between the trailing edge of depolymerizing microtubules and the leading edge of polymerizing microtubules (Figure 2C). By quantifying the microtubule density from confocal time-lapse imaging, we found that compared to WT, YCL microtubules in MZdchs1b mutants exhibited slower turnover rate and often failed to form a microtubule low-density zone (Figures 2C and 2D; Movie S4). On average, the microtubule density in MZdchs1b mutants was decreased to 49.4±9.3% of its maximal density, in contrast to 16.3±2.6% in WT embryos (Figure 2D). Similarly reduced microtubule dynamics in MZdchs1b mutants were also observed using a different microtubule-labeling transgenic line Tg[βactin2:EMTB-3xGFP] (Wuhr et al., 2010) (Figures S6A and S6B). Based on these results, we propose that Dchs1b promotes microtubule turnover during embryonic cleavages.

To investigate how Dchs1b promotes microtubule dynamics, we injected synthetic RNA encoding EB3-GFP, a plus end microtubule marker (Stepanova et al., 2003), into WT and MZdchs1b mutant zygotes and quantified the microtubule plus end EB3-GFP comet speed at 4-4.5 hpf. We found that a larger fraction of EB3-GFP comets in MZdchs1b mutant embryos exhibited a faster speed compared to WT embryos (Figure 2E) and longer track displacement (Figure S6C), while comparable track duration (Figure S6D), suggesting a faster microtubule polymerization rate in MZdchs1b mutants. Moreover, by monitoring EB3-GFP comet trajectories, we observed defective microtubule polarization angles in MZdchs1b mutants in comparison with WT, in which a significant alignment with the animal-vegetal axis of the embryo was observed (Figure 2F). These results are in agreement with defective microtubule polarity previously associated with Dachsous mutations in zebrafish (Li-Villarreal et al., 2015) and Drosophila (Harumoto et al., 2010; Matis et al., 2014). Moreover, they are consistent with lower microtubule turnover in MZdchs1b mutants (Figure 2D) and they further suggest Dchs1b promotes microtubule turnover in part by negatively regulating microtubule polymerization speed.

Dchs1b physically interacts with Ttc28

The above observations support a model whereby Dchs1b influences microtubule dynamics and/or polarity to contribute to the regulation of embryonic cleavages. However, it is still unknown mechanistically how Dchs/Ds could affect microtubule polarity or dynamics in vertebrates/Drosophila. To identify a molecular link between Dchs and microtubules, we performed affinity purification-mass spectrometry (AP-MS) experiments using the intracellular domain of rat Dchs1 (Dchs1-ICD) in HEK293 cells (Figure S2A). These experiments identified Lix1L, a known interactor of Dchs (Mao et al., 2009), validating our approach. One of the top hits from these experiments was Ttc28/TPRBK, a tetratricopeptide (TPR) motif-enriched cytoplasmic protein. Studies in mammalian cells implicated Ttc28 in the regulation of mitosis and cytokinesis via binding to Aurora B, a known regulator of microtubules during cell divisions (Izumiyama et al., 2012). The zebrafish ttc28 homolog is highly conserved in vertebrates (Figure S2B) and is ubiquitously expressed during cleavage and blastula stages, with its transcripts becoming enriched in neural tissues at 1 day post fertilization (dpf) (Figures 3A and 3B). This novel interaction between mammalian Dchs1 and Ttc28 is also conserved in zebrafish, as co-immunoprecipitation (Co-IP) experiments performed in HEK293 cells demonstrated that zebrafish Ttc28 interacted with both full length zebrafish Dchs1b and Dchs1b-ICD (Figures 3C, 3D, S3A and S3B). Subsequent experiments showed that HA-Ttc28-N and HA-Ttc28-N-1 but not HA-Ttc28-ΔN-1 co-immunoprecipitated with Dchs1b-ICD, indicating that Dchs1b-ICD interacted with the N-terminal TPR domain and specifically with the first 15 TPR motif repeats of Ttc28 (Figures 3E, 3F and S3C). To define the region of Dchs1b-ICD involved in this interaction, we carried out deletion and truncation mapping of Dchs1b-ICD in Co-IP experiments. These studies revealed that CM2, one of the three Dchs1b-ICD conserved motifs (CM) (Hulpiau and van Roy, 2009), was necessary and sufficient to mediate the interaction with Ttc28-N (Figures 3G, 3H and 3J). Focusing on CM2, we observed that deleting CM2-N, the first half of CM2, abolished the binding between Dchs1b-ICD and Ttc28-N (Figures 3I and 3J). However, substitutions of every three contiguous residues within CM2-N to Alanine could not completely abolish the interaction with Ttc28-N (Figure S3D), suggesting Dchs1b-ICD interacts with Ttc28-N through several residues in the conserved CM2-N motif. It is noteworthy that the evolutionarily conserved 356-residue Dchs1b-ICD is predicted to be almost entirely disordered except for a near 40-residue region comprising the CM2 motif (Figure 3J)(Oates et al., 2013). This length of amino-acid sequence is typically insufficient to support a self-contained higher order structure, but it can form an oligomerization domain or assume a highly ordered conformation upon interacting with another protein (Yegambaram et al., 2013). Cumulatively, these data indicate that the interaction between Ttc28 and Dchs1b is conserved from zebrafish to mammals and requires the N-terminal TPR motifs of Ttc28 and the CM2-N motif of Dchs1b-ICD.

Figure 3. Dchs1b interacts with the N-terminal TPR motifs of Ttc28 via its ICD CM2-N motif.

(A) RT-PCR assays to compare the expression levels of ttc28 at different developmental stages in zebrafish. gapdh housekeeping gene as a reference control.

(B) The expression patterns of ttc28 revealed by whole mount in situ hybridization at 8-cell stage (1.15 hpf), 4.3 hpf, and 1 dpf zebrafish. ttc28 sense probe was used as a negative control.

(C) Co-IP assays of HA-Ttc28 with full-length Dchs1b-sfGFP or membrane-GFP.

(D) Co-IP assays of Flag-Dchs1b-ICD with HA-Ttc28.

(E) Co-IP assays of Flag-Dchs1 b-ICD with HA-Ttc28-N, HA-Ttc28-N-1, or HA-Ttc28-ΔN-1.

(F) Schematic of Ttc28 truncation constructs and the Co-IP results of Ttc28 deletion mapping.

(G) Co-IP assays of HA-Ttc28-N with Flag-Dchs1b-ICDΔCM1, Flag-Dchs1b-ICDΔCM2, or Flag-Dchs1b-ICDΔCM3.

(H) Co-IP assays of HA-Ttc28-N with Flag-Dchs1b-ICD-N1, Flag-Dchs1b-ICD-N2, or Flag-Dchs1b-ICD-N3.

(I) Co-IP assays of HA-Ttc28-N with Flag-Dchs1 b-ICD, Flag-Dchs1b-ICDΔCM2-N, or Flag-Dchs1b-ICDΔCM2-C.

(J) Schematic of Dchs1 b-ICD deletion and truncation constructs and sequence alignment of Dchs CM2 motifs from different species. Disorder consensus bar, intensity of green corresponds to the strength of the consensus disorder agreement among 8 prediction algorithms (Oates et al., 2013). Red line, residues of zebrafish Dchs1 b involved in Ttc28 interaction. See also Figure S2 and S3.

Dchs1b regulates Ttc28 subcellular distribution

Next, we sought to understand whether the interaction between Dchs1b and Ttc28 could influence Ttc28 subcellular distribution in vivo. To address this, we injected synthetic RNA encoding HA-Ttc28-mCherry (600 or 900 pg) into one-celled WT zygotes and used confocal microscopy to examine the resulting blastulae at 3-4 hpf. We observed that HA-Ttc28-mCherry was enriched in the perinuclear region, including the centrosomes, and at the plasma membrane (Figures 4A, S2C, S2D, and S4E). Interestingly, when Dchs1b-sfGFP was co-expressed with HA-Ttc28-mCherry in WT, the accumulation of HA-Ttc28-mCherry at the centrosomes was reduced and it mostly co-localized with Dchs1b-sfGFP at the plasma membrane (Figures 4A, 4B and S4A). By mosaically expressing Dchs1b-sfGFP in WT blastulae ubiquitously expressing HA-Ttc28-mCherry, we also observed a strong reduction of HA-Ttc28-mCherry signal at the centrosomes in cells expressing Dchs1b-sfGFP (Figure 4C). Further, we observed that HA-Ttc28-N-mCherry, but not HA-Ttc28-ΔN-1-mCherry, was recruited to the plasma membrane by Dchs1b-sfGFP (Figures S4B and S4C), and the accumulation of HA-Ttc28-mCherry at the centrosomes was reduced less when co-expressed in WT embryos with Dchs1b-ΔCM2-sfGFP compared to Dch1b-sfGFP (Figure S4E), consistent with the Co-IP results. Conversely, in MZdchs1b mutant blastulae, the plasma membrane localization of HA-Ttc28-mCherry was strongly reduced (Figures 4A, 4B and S4A). We interpret these data to mean that Dchs1b is sufficient and required to recruit Ttc28 to the plasma membrane, and consequently that Dchs1b regulates subcellular Ttc28 distribution.

Figure 4. Dchs1 b regulates Ttc28 subcellular localization, and Ttc28 is sufficient to influence YCL microtubule dynamics and embryonic cleavages.

(A) Representative images showing the subcellular localizations of HA-Ttc28-mCherry (magenta) with membrane GFP (green) or Dchs1b-sfGFP (green) in the WT and MZdchs1b at 3-4 hpf in embryos injected with synthetic RNAs encoding HA-Ttc28-mCherry (600 pg), membrane GFP (25 pg), and Dchs1b-sfGFP (600 pg). Scale bar, 30 μm.

(B) Quantification of the relative HA-Ttc28-mCherry intensity at the plasma membrane and centrosome normalized to the cytoplasmic signal in the same cell in the various injection conditions. n, number of cells quantified. ns, not significant. **** P<0.0001, Tukey's multiple comparisons test.

(C) Cartoon of the experimental design and confocal images of mosaic analyses of HA-Ttc28-mCherry and Dchs1b-sfGFP interaction in vivo. HA-ttc28-mCherry RNA was injected at 1 -cell stage and dchs1b-sfGFP RNA was subsequently injected into one of the blastomeres at 8-cell stage. Confocal imaging was performed at 3-4 hpf. Scale bar, 30 μm.

(D) Representative time-lapse still images of YCL microtubules in mCherry and HA-Ttc28-mCherry RNA (900 pg) injected Tg[ef1α:dclk-GFP] WT embryos at 1.75-2.5 hpf. Scale bar, 30 μm.

(E) Quantification of YCL microtubule density in mCherry control and HA-Ttc28-mCherryoverexpressing embryos at 1.75-2.5 hpf. Error bars represent SEM. N, number of embryos.

(F) Representative images of blastomeres stained with γ-Tubulin (green) and counterstained with DAPI (blue) in WT and MZdchs1b embryos at 4-5 hpf injected with either mCherry or HA-ttc28-mCherry RNA (900 pg) at 1 -cell stage. Arrows indicate multiple nuclei, and arrowheads indicate chromatin bridges. Scale bar, 30 μm.

(G) Quantification of mitotic events at 4-5 hpf in WT and MZdchs1b blastulae injected with mCherry or HA-ttc28-mCherry RNA (900 pg) at 1-cell stage. N, number of embryos; n, number of cells. See also Figure S2, S4, and Movie S5.

Excess Ttc28 disrupts embryonic cleavages and reduces microtubule turnover

We reasoned that Ttc28 mis-localized from the plasma membrane in MZdchs1b mutant blastomeres could contribute to the abnormal embryonic cleavages. Consistent with this hypothesis, injection of increasing amounts of HA-Ttc28-mCherry but not mCherry control RNA (900 pg) into WT zygotes, led to reduced YCL microtubule turnover during early cleavages (Figures 4D and 4E; Movie S5). Overexpression of HA-Ttc28-mCherry or HA-Ttc28 also resulted in mitosis and cytokinesis defects at 4-5 hpf (Figures 4F, 4G and S4D), similar to those observed in MZdchs1b mutants (Figures 1E and 1F). However, overexpression of HA-Ttc28-mCherry in MZdchs1b mutants only slightly enhanced the mitotic defects (Figure 4G). In addition, both HA-Ttc28-N-mCherry and HA-Ttc28-ΔN-1-mCherry mis-expression in WT embryos led to cell division defects, albeit less frequently compared to HA-Ttc28-mCherry (Figure S4D). These results imply that Ttc28 is sufficient to influence microtubule dynamics during embryonic cleavages. Moreover, the TPR domain of Ttc28 can induce embryonic cleavage defects, and its interaction with Dchs1b is not essential for this activity.

ttc28 genetically interacts with dchs1b to regulate microtubule dynamics

To investigate further Ttc28 function during zebrafish embryogenesis, we generated ttc28 zebrafish mutants using the CRISPR/Cas9 approach (Jinek et al., 2012). The resulting frame-shifting indel ttc28^stl362 and ttc28^stl363 mutations greatly reduced level of ttc28 transcripts indicating that they represent strong/null alleles (Figures 5A and 5B). Although no overt morphological or cleavage anomalies were observed in zygotic or MZttc28^{stl363/stl363} mutants (thereafter MZttc28 mutants) (Figures S5A-S5C and data not shown), we observed increased YCL microtubule turnover in MZttc28 mutants during early cleavages for both alleles (Figures 5C, 5D, S6A, and S6B; Movie S6). Furthermore, the increased YCL microtubule dynamics in MZttc28 mutants could be rescued by overexpressing HA-Ttc28 or HA-Ttc28-mCherry (Figures 5C and 5D). Consistently, midzone microtubules assembled faster in MZttc28 compared to WT blastulae (Figures 5E and 5F; Movie S7), supporting the notion that Ttc28 limits microtubule dynamics.

Figure 5. Ttc28 is required to regulate microtubule dynamics.

(A) Schematic of ttc28 sgRNA targeting site and the sequences of ttc28^stl362 and ttc28^stl363 mutant alleles.

(B) qPCR analyses of ttc28 transcript levels in WT, MZttc28^{stl362/stl362}, and MZttc28^{stl363/stl363} mutants at 2-cell stage. Two different sets of primers were used and normalized to β-actin. Error bars represent standard deviation. N=3 biological repeats.

(C) Representative time-lapse still images of YCL microtubules at 1.75-2.5 hpf in WT, MZttc28^{stl363/stl363}, and HA-Ttc28-mCherry RNA-injected (900 pg) MZttc28s^tl363/stl363 Tg[ef1α:dclk-GFP] embryos. Scale bar, 30 um.

(D) Quantification of YCL microtubule density from time-lapse movies in WT, MZttc28^{stl363/stl363}, and HA-Ttc28 or HA-Ttc28-mCherry RNA-injected (900 pg) MZttc28^{stl363/stl363} embryos at 1.75-2.5 hpf. Error bars represent SEM. N, number of embryos.

(E) Representative time-lapse still images of microtubule dynamics during cytokinesis in WT andMZttc28^{stl363/stl363} embryos using Tg[ef1α:dclk-GFP]. Scale bar, 30 μm.

(F) Quantification of midzone microtubule assembly rates in WT and MZttc28^{stl363/stl363} blastulae at 1.75-2.5 hpf. **** P<0.0001, paired student's two-tailed t-test. Error bars represent SEM. N, number of embryos. See also Figure S5, S6, Movie S6 and S7.

To test whether ttc28 genetically interacts with dchs1b, we generated MZdchs1b; MZttc28 compound mutants and examined the YCL microtubule dynamics in comparison with MZdchs1b single mutants. Time-lapse analyses revealed that the reduced YCL microtubule turnover observed in MZdchs1b single mutants (Figures 2C and D) was partially suppressed in MZdchs1b; MZttc28 compound mutants (Figures 6A and 6B). Likewise, the midzone microtubule assembly delay we observed in MZdchs1b single mutants (Figures 2A and 2B) was also suppressed in MZdchs1b; MZttc28 compound mutants (Figures 6C and 6D; Movie S8), which exhibited phenotype more similar to MZttc28 single mutants (Figure 5F). Taking into consideration these MZdchs1b; MZttc28 double mutant phenotypes and that overexpression of HA-Ttc28-mCherry in MZdchs1b mutants did not substantially increase abnormal cleavages (Figure 4G), we conclude that Ttc28 acts largely downstream of Dchs1b to influence microtubule dynamics during early cleavages, although roles parallel to Dchs1b are also plausible. We note that mitosis and cytokinesis defects were still detected in MZdchs1b; MZttc28 mutants at a frequency comparable to that observed in MZdchs1b single mutants (Figure 6E), implying the suppression of microtubule dynamics and the midzone microtubule assembly defects alone is insufficient to normalize cleavages. Given the presence of additional abnormalities in MZdchs1b mutants compared to MZttc28 mutants, such as cleavage furrow mis-positioning and propagation defects (Figures 1A-1D and S1A), Dchs1b likely engages in interactions with additional molecules to mediate other aspects of embryonic cleavages.

Figure 6.

ttc28 genetically interacts with dchs1b to regulate microtubule dynamics (A) Representative time-lapse still images of YCL microtubules in MZdchs1b single and MZdchs1b; MZttc28^{stl363/stl363} compound mutants in Tg[ef1α:dclk-GFP] transgenic background at 1.75-2.5 hpf. Scale bar, 30 μm.

(B) Quantification and comparison of YCL microtubule density in WT, MZdchs1b single and MZdchs1b; MZttc28^{stl363/stl363} compound mutant at 1.75-2.5 hpf. Error bars represent SEM. N, number of embryos.

(C) Representative time-lapse still images of microtubule dynamics during cytokinesis in MZdchs1b single and MZdchs1b; MZttc28^{stl363/stl363} compound mutant embryos using Tg[ef1α:dclk-GFP] Scale bar, 30 μm.

(D) Quantification of midzone microtubule assembly rates in WT, MZdchs1b single and MZdchs1b; MZttc28^{stl363/stl363} compound mutant blastulae at 1.75-2.5 hpf. The Y-axis shows the ratio of midzone microtubule width d at each time point divided by the initial width d'. **** P<0.0001, paired student's two-tailed t-test. Error bars represent SEM. N, number of embryos.

(E) Comparison of mitotic events in WT, MZdchs1b single and MZdchs1b; MZttc28^{stl363/stl363} compound mutant embryos at 3 hpf. N, number of embryos; n, number of cells. See also Movie S8.

Functional interactions between Dchs1b, Ttc28, and Aurora B

In our search for additional Dchs1b-interacting proteins during embryonic cleavages, we turned to Aurora B, as it was previously shown that mammalian TTC28 binds to Aurora B/Aurkb to regulate mitosis and cytokinesis (Izumiyama et al., 2012). Moreover, zebrafish MZcei mutants lacking Aurora B function or embryos treated with ZM447439 Aurora B inhibitor, exhibit severe cleavage defects (Yabe et al., 2009)(Figure S1C-E). Our co-IP experiments failed to detect biochemical interactions between HA-Ttc28 and EGFP-Aurkb expressed in HEK293 cells (not shown). However, we found that both HA-Ttc28 and EGFP-Aurkb were co-pulled down by Flag-Dchs1b-ICD (Figure 7A). Notably, we also detected a weak interaction between Flag-Dchs1b-ICD and EGFP-Aurkb in the absence of HA-Ttc28 (Figure 7A), suggesting Dchs1b-ICD can bind to Aurora B independent of Ttc28.

Figure 7. Functional interactions between Dchs1b, Ttc28, and Aurora B in embryonic cleavages.

(A) Co-IP assays of Flag-Dchs1b-ICD with EGFP-Aurkb and HA-Ttc28.

(B) Quantification of cleavage defects in WT, MZttc28^{stl363/stl363}, and MZdchs1b embryos at 3-3.5 hpf following DMSO control or 75 μM ZM447439 treatment. The right panel shows the representative severity of cleavage defects. N, number of embryos.

(C) Representative time-lapse still images of YCL microtubules in Tg[ef1 α:dclk-GFP] transgenic WT embryos after DMSO or 400 μM ZM447439 treatment. Scale bar, 30 μm.

(D) Quantification of YCL microtubule density in DMSO control and ZM447439 (400 μM) –treated embryos. Error bars represent SEM. N, number of embryos.

(E) Models of Dchs1 b interacting with Ttc28 and Aurora B to control microtubule dynamics and embryonic cleavages in zebrafish blastomeres of WT and MZdchs1b mutant embryos. Dchs1b, Ttc28 and Aurora B, during cleavages. Dchs1 b binds to Ttc28 at the membrane to limit its microtubule dynamics reducing activity in the cytoplasm. Potential direct or indirect effects of Ttc28 on Aurora Bare also illustrated. Binding of Dchs1 b to Aurora B promotes its activity stimulating microtubuledynamics. Aurora B has additional effects on early cleavages. See also Movie S2.

To further test functional interaction between these three proteins, we treated WT, MZttc28, and MZdchs1b embryos with a subthreshold dosage of ZM447439 (75 μM) to partially inhibit Aurora B activity and examined embryonic cleavage defects at 3-3.5 hpf. We found that, in comparison with WT embryos, MZttc28 embryos treated with the same dosage of ZM447439 exhibited a milder spectrum of cleavage defects (Figures 7B, ttc28^stl363 allele, and S5D, ttc28^stl362 allele). In contrast, MZdchs1b embryos displayed more severe cleavage defects upon treatment with the same concentration of ZM447439 compared to WT embryos (Figure 7B). These data suggest Dchs1b and Aurora B act similarly to regulate embryonic cleavages, while Ttc28 antagonizes Aurora B activity.

Given the above results, we reasoned Aurora B regulates microtubule dynamics similarly to Dchs1b and opposite to that of Ttc28. To test this, we treated WT Tg[ef1α:dclk-GFP] zygotes with DMSO control or Aurora B inhibitor ZM447439 (400 μM) to examine the YCL microtubule dynamics. We found that inhibition of Aurora B activity led to reduced YCL microtubule turnover (Figures 7C and 7D), phenocopying MZdchs1b single mutants (Figures 2C and 2D) and overexpression of Ttc28 (Figures 4D and 4E) and contrasting the increased microtubule turnover observed in MZttc28 mutants (Figures 5C and 5D). Taken together, we propose functional interactions between Dchs1b, Ttc28, and Aurora B in the control of microtubule dynamics and embryonic cleavages: Dchs1b promotes microtubule turnover and normal embryonic cleavages in part by interacting with Aurora B and in part by sequestering at the membrane and inhibiting Ttc28, which limits microtubule turnover thus antagonizing Aurora B function (Figures 7E).

Discussion

Dachsous cadherins play essential roles in diverse developmental processes, from egg activation, early cleavages, embryo patterning, gastrulation movements, neuronal migration, through tissue polarity and growth during organogenesis, in Drosophila, zebrafish and mammals. Whereas recent studies in Drosophila and zebrafish proposed that Dchs functions through regulation of actin and microtubule cytoskeletons (Harumoto et al., 2010; Li-Villarreal et al., 2015; Matis et al., 2014), the molecular links between Dchs and microtubules remain to be identified. Our current findings indicate that Dchs1b promotes microtubule turnover and midzone microtubule assembly and define a molecular mechanism through which Dchs1b regulates microtubules during embryonic cleavages. Through studies in mammalian cells and zebrafish, we demonstrate that Dchs ICD interacts with Ttc28 protein, which has been previously implicated in regulation of mammalian cell division through its interaction with Aurora B (Izumiyama et al., 2012). Whereas we did not detect biochemical interaction between zebrafish Ttc28 and Aurora B, we uncovered that Dchs1b binds Aurora B. We show that Dchs1b interacts with Ttc28 via its ICD CM2-N motif to regulate Ttc28 subcellular distribution in zebrafish embryos. Whereas MZttc28 mutants showed increased microtubule turnover during early cleavages, its overexpression decreased microtubule turnover and resulted in mitotic defects, thus phenocopying MZdchs1b mutants. Moreover, inactivation of ttc28 in MZdchs1b mutants largely normalized microtubule dynamics and spindle midzone microtubule assembly, but did not suppress cell division defects, implying additional interactions are involved. In line with Dchs1b and Aurora B functionally interacting to promote microtubule dynamics and normal cleavages, we demonstrate that embryos treated with Aurora B inhibitor, exhibited reduced microtubule turnover similar to those in MZdchs1b mutants and embryonic cleavages in MZdchs1b mutants were sensitized to treatments with Aurora B inhibitor. In support of Dchs1b/Aurora B and Ttc28 having opposing effects on microtubule dynamics and cleavages, MZttc28 mutants exhibited increased microtubule turnover and were less sensitive to Aurora B inhibition. Based on these biochemical and genetic interaction experiments, we propose that during embryonic cleavages in zebrafish, Dchs1b binds to and restrains Ttc28 activity to limit microtubule dynamics and to ensure proper midzone microtubule assembly, and these functions are in part mediated through interactions with Aurora B (Figure 7E).

Our previous studies demonstrated embryonic cleavage abnormalities in MZdchs1b mutants, but how loss of Dchs1b function leads to these defects was unknown (Li-Villarreal et al., 2015). Here we provide evidence that impeded cytokinetic furrow progression during cleavage stages is responsible for the embryonic cleavage defects in MZdchs1b mutants (Figure 1). Moreover, these cytokinesis defects are associated with and likely caused by aberrant midzone microtubule assembly (Figure 2A). Consistent with this model, extensive studies in different model organisms and mammalian cultured cells have shown that perturbing midzone microtubules or proteins involved in microtubule regulation impairs or blocks cytokinesis (D'Avino et al., 2005; Eggert et al., 2006).

Previously, Ds has been suggested to control microtubule organization and polarity in Drosophila pupal wing epithelium (Gomez et al., 2016; Harumoto et al., 2010; Matis et al., 2014). Our work indicates that Dchs1b can also regulate microtubule turnover. Taking advantage of the unique thin yolk cytoplasmic layer (YCL) of the early zebrafish embryo that contains dynamic and elaborate microtubule arrays (Solnica-Krezel and Driever, 1994; Strahle and Jesuthasan, 1993), we observed cyclical assembly and disassembly of microtubule arrays during early cleavages, allowing us to examine microtubule dynamics in vivo (Figure 2C). These analyses revealed that the YCL microtubules display reduced turnover in MZdchs1b mutants using two different transgenic lines in which microtubules are marked with different GFP fusion proteins (Tran et al., 2012; Wuhr et al., 2010) (Figures 2C, 2D, S6A, and S6B). Monitoring microtubule polymerization by tracking EB3-marked plus ends, further suggests that Dchs1b promotes microtubule turnover by limiting microtubule polymerization speed (Figure 2E). During cytokinesis, astral microtubules elongating from the two spindle poles are thought to contact at the equator to initiate furrow specification and ingression, and subsequently a signal originating from the midzone is required to promote the completion of furrow ingression (Bringmann and Hyman, 2005). We reasoned that reduced microtubule dynamics in MZdchs1b mutants could compromise this process. Consistent with this notion, stabilizing microtubules by taxol in mammalian cells delays cytokinetic furrow onset and prevents its completion (Shannon et al., 2005).

Dchs is mostly thought to function as a ligand for Fat to control Fat-dependent signaling activity (Matakatsu and Blair, 2004; Simon et al., 2010). In the Drosophila eye, epistasis experiments suggest that Ds acts upstream of Fat to regulate PCP (Yang et al., 2002). Accordingly, driving expression of a Ds construct without the ICD is sufficient to rescue the PCP phenotype in the Drosophila wing (Matakatsu and Blair, 2006). However, Ds has been shown to mediate Drosophila wing disc growth independent of Fat and PCP signaling (Matakatsu and Blair, 2006), and Fat4 is thought to signal through Dchs1 to regulate nephron progenitors during mouse kidney growth (Bagherie-Lachidan et al., 2015; Mao et al., 2015). In addition, expression of Dchs1b ICD is sufficient to rescue the bundled microtubule phenotype in the zebrafish YCL during epiboly, also supporting an autonomous function of Dchs (Li-Villarreal et al., 2015). In this study, we uncover Ttc28 and Aurora B as novel interactors of Dchs1b that specifically bind to its ICD, consistent with a model where Dchs directly influences intracellular processes (Figure 7E).

It is poorly understood how vertebrate Dchs regulates intracellular events, as most of the Ds interactors identified in Drosophila, such as Dachs and Vamana, lack close homologs in vertebrates. Sequence alignment of Dchs among different species identified three conserved motifs (CM1-3) within its ICD that might be functionally relevant (Hulpiau and van Roy, 2009). We report here that a conserved region in Dchs1b ICD (CM2-N) is essential for its interaction with Ttc28, as deleting CM2-N completely abolished the interaction (Figures 3I and 3J). In addition, the Dchs1b-Ttc28 interaction via CM2-N is vital for the ability of Dchs1b to regulate Ttc28 subcellular distribution (Figures 4A-4C, and S4E). Given the largely disordered nature of Dchs ICD except the CM2 motif (Figure 3J), it is plausible that CM2 also mediates interactions with other proteins. We note that proteins with tandem repeats of TPR motifs are thought to serve as scaffold proteins (Zeytuni and Zarivach, 2012). Consistently, we show that the N-terminal region of TPR motifs in Ttc28 is sufficient to mediate the interaction with Dchs1b ICD, and its subcellular distribution could be influenced by Dchs1b (Figures 4A, 4B, S4B, and S4C). Drosophila Fat is thought to mediate PCP and Hippo signaling by altering the subcellular localization and activity of the unconventional myosin Dachs (Cho and Irvine, 2004; Mao et al., 2006). It is possible that Dchs1b regulates Ttc28 function in a similar manner, suggesting that binding via ICD and regulating subcellular distribution of cytoplasmic proteins is a key mechanism through which these giant cadherins regulate intracellular processes.

We provided several lines of evidence that by binding to Ttc28 at the membrane, Dchs1b regulates its subcellular distribution and microtubule dynamics during cleavage stages. Our experiments attribute the microtubule dynamics and embryonic cleavage defects in MZdchs1b mutants to Ttc28 being mis-localized from the membrane to cytoplasmic locations during cleavage stages (Figure 4A, B). Accordingly, overexpression of Ttc28 disrupted embryonic cleavages (Figure 4F, G) and decreased microtubule dynamics (Figure 4D, E), whereas loss of Ttc28 increased microtubule dynamics (Figure 5C, D). Moreover, genetic inactivation of ttc28 in MZdchs1b mutants largely suppressed the microtubule dynamics and midzone microtubule defects (Figure 6A-D), without normalizing embryonic cleavages (Figure 6E), implying additional interactions are involved.

We propose Aurora B is one such interactor, owing to its well-known function in direct regulation of microtubules and cell divisions (Basant et al., 2015; Douglas et al., 2010; Gruneberg et al., 2004; Murata-Hori and Wang, 2002; Nunes Bastos et al., 2013; Yabe et al., 2009). Moreover, the human TTC28 homolog was reported to bind to Aurora B and regulate cell divisions in mammalian cells (Izumiyama et al., 2012). Whereas, we were unable to detect this interaction between zebrafish proteins using Co-IP experiments (not shown), we obtained evidence that Dchs1b ICD binds to both Ttc28 and Aurora B, and its interaction with Aurora B can be independent of Ttc28 (Figure 7A). Consistent with Dchs1b and Aurora B interacting and having similar activity during cleavages, chemical inhibition of Aurora B activity resulted in reduced YCL microtubule dynamics (Figures 7C and 7D), phenocopying decreased microtubule turnover in MZdchs1b mutants (Figures 2C and 2D). Moreover, MZdchs1b mutant embryos were more sensitive to Aurora B inhibition than WT embryos (Figure 7B). Although we did not observe obvious embryonic cleavage defects in MZttc28 mutants, our observation that MZttc28 mutant embryos are less sensitive to Aurora B inhibition than WT embryos (Figures 7B and S5D), support a functional interaction between Ttc28 and Aurora B in this process. We speculate Ttc28 functions to influence embryonic cleavages either directly via antagonizing Aurora B activity or indirectly through limiting microtubule turnover (Figure 7E).

Critical roles of Ttc28 in development and disease are being increasingly recognized. We note that Ttc28 knock-out mouse exhibits vertebral fusion phenotypes (http://www.informatics.jax.org/marker/MGI:2140873), similar to what has been reported for Dchs1 mouse mutants (Kuta et al., 2016; Mao et al., 2016). In addition, TTC28 has been linked to several human diseases, including cancer, such as uveal melanoma and oesophageal squamous-cell carcinoma (Cancer Genome Atlas, 2012; Chang et al., 2017; Fujimoto et al., 2016), but the underlying mechanisms are not understood. Notably, current evidence suggests that microtubule dynamics is altered in cancer cell divisions and many effective chemotherapeutics are microtubule interfering drugs (Cirillo et al., 2017). Hence, our findings on Dchs regulating microtubule dynamics and cleavages through interactions with Ttc28 and Aurora B may provide molecular clues for further understanding how these proteins function during development and in human disease.

Star Methods

Contact For Reagent and Resource Sharing

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Lilianna Solnica-Krezel (solnical@wustl.edu).

Experimental Models and Subject Details

Zebrafish

Zebrafish WT AB*, dchs1b^fh275 (Li-Villarreal et al., 2015), Tg[βactin2:GCaMP6s]^stl351 (Chen et al., 2017), Tg[ef1α:dclk-GFP] (Tran et al., 2012), and Tg[βactin2:EMTB-3xGFP] (Wuhr et al., 2010) lines were used in this study. As described in previous studies (Li-Villarreal et al., 2015), dchs1b mutant phenotypes display an age-related decrease in penetrance and expressivity as reported for many early embryonic mutant phenotypes in zebrafish (Fekany et al., 1999; Li-Villarreal et al., 2015). MZ mutant embryos were generated by pairwise crossing of adult (3-12 months old) females and males homozygous for the studied mutations. In individual experiments, we used mutant and wild-type females of similar age. Sex cannot be determined at embryonic stages analyzed in this study.

All zebrafish experiments and procedures were performed in accordance with the Institutional Animal Use and Care Committee of the Washington University in St. Louis School of Medicine. Fish are maintained in 1.25, 3.0, or 8 L tanks with recirculating water purified by reverse osmosis and heated to 28.5°C. 10% of fresh water is added to the system daily. Water quality (pH and conductivity) is monitored by sensors built into the system and are adjusted automatically. Adult fish are housed at a density of <12/L. Zebrafish larvae from 5-15 dpf are nurtured using algae-fed rotifer suspension and subsequently according to their developmental stage using Tritone robots (Tecniplast) with combination of rotifer suspension (2-3 times daily) and dry food (Gemma 150) (6-9 times daily). Young fish approaching adulthood are fed 2 times with rotifer suspension and 10 times with dry food (Gemma 300). Adult fish are fed by Tritone robots, depending on the frequency of mating, with rotifer suspension (1 -2 times daily) and dry food (Gemma 300) (1 -5 times per day).

Method Details

Cloning and RNA synthesis

The pCR4-dchs1b-sfGFP plasmid was linearized with EcoRI to synthesize RNA for microinjection with a T7 promoter as described previously (Li-Villarreal et al., 2015). The open reading frame of dchs1b-sfGFP was cloned into pCMV-sport6.1 for HEK293 cell transfection. Flag-Dchs1b-ICD, Flag-Dchs1b-ICD-N1, Flag-Dchs1 b-ICD-N2, and Flag-Dchs1b-ICD-N3 were constructed by PCR and cloned into pcDNA3.1. Flag-Dchs1b-ICDΔCM1, Flag-Dchs1 b-ICDΔCM2, Flag-Dchs1 b-ICDΔCM3, Flag-Dchs1 b-ICDΔCM2-N, and Flag-Dchs1b-ICDΔCM2-C constructs were generated by overlap extension PCR. The full-length coding sequence of ttc28 was generated by PCR from zebrafish cDNA and subcloned into pT7Ts vector in frame with mCherry sequences by annealing extend PCR (Pont-Kingdon, 1997) for RNA synthesis. The ttc28 coding sequence was cloned into pCMV-sport6.1 with a N-terminal HA tag for cell transfection. HA-ttc28-N, HA-ttc28-N-1, HA-ttc28-ΔN-1, HA-ttc28-N-mCherry, and HA- ttc28-ΔN-1-mCherry were constructed by PCR. EGFP-Aurkb was constructed by PCR and inserted into pCS2 vector. All injected mRNAs were synthesized from linearized DNA plasmid templates mentioned above using SP6 or T7 mMessage mMachine Kit (Ambion).

Furrow calcium imaging

Embryos were manually dechorionated with forceps and mounted in 0.3% low-melting agarose on a glass bottom dish (MatTek) within 40-50 minutes post fertilization (mpf). Time-lapse live imaging was carried out using a spinning-disk confocal microscope (Olympus IX81, Quorum) with a 28.5°C chamber. Z-stack was set up at 3 μm for a total of 50-55 slices, and the images were collected every 30 seconds from 2-cell to 16-32 cell stage.

YCL and midzone microtubule dynamics imaging

Embryos were mounted as described above at 8-16 cell stage. Time-lapse imaging was performed using a spinning-disk confocal microscope (Olympus IX81, Quorum) or Nikon spinning-disk confocal microscope with a 40× water lens at 28.5°C. Z-stack was set up for a total of 17-20 μm for YCL microtubule imaging or 45-50 μm for midzone microtubule imaging at 1 μm interval. Images were acquired every 15 seconds for YCL microtubules or every minute for midzone microtubules from 32-cell to 256-cell stage.

EB3-GFP dynamics imaging

Embryos were injected with 50pg of RNA encoding EB3-GFP prior to the one-cell stage. Injected embryos were mounted as described above at the 1,000-cell stage and image acquisition was performed between sphere and dome stages (4 – 4.5 hpf). Time-lapse imaging was performed using a Nikon spinning-disk confocal microscope with a 60× oil immersion lens at room temperature. Z-stack was set up for a total of 4-6 μm at 1 -1.5 μm interval. Images were acquired every 2 seconds.

Generation of ttc28 mutants in zebrafish

We used CRISPR/Cas9 system (Hwang et al., 2013; Jao et al., 2013) to generate ttc28 mutant zebrafish. In brief, one-celled WT zygotes were injected with 100 pg Cas9 synthetic RNA and 10 pg ttc28 sgRNA (5′-GGCATCAGTGGTGGTCCTGG-3′). Founders with germline transmission of ttc28 mutation(s) were outcrossed with WT fish to generate F1, and ttc28^stl362 containing 1 bp and ttc28^stl363 containing 10 bp deletions were recovered from the F1 generation by sequencing. Both ttc28^stl362 and ttc28^stl363 were used in the YCL microtubule dynamics imaging experiments. ttc28^stl363 was crossed to dchs1b^fh275 to test the genetic interaction between ttc28 and dchs1b.

Immunostaining

Embryos were fixed in sweet PFA (4% paraformaldehyde, 4% sucrose in PBS pH7.3) at 4°C overnight, and transferred into 100% methanol at -20°C for 48 hours. Samples were then rehydrated and incubated with blocking buffer (10% goat serum, 2%BSA, and 1% DMSO in 0.5-2.5% Triton X-100 PBSTr) at room temperature for 2 hours. Anti-γ-tubulin (T6557, Sigma) was used as primary antibody (1:500) and Alexa Fluor 488 as secondary antibody (1:500) with 4-6 × 30 minutes PBSTr washes. Samples were counter-stained with DAPI (4′,6-Diamidino-2-Phenylindole, Dihydrochloride, 0.4 μg/mL final concentration) for 30 minutes at room temperature and rinsed 3 times with PBS before imaging. Stained samples were mounted in 1.0% low-melting agarose on a glass bottom dish (MatTek) for imaging using a spinning-disk confocal microscope (Olympus IX81, Quorum) with a 60× water lens.

Co-immunoprecipitation and immunoblotting

Transfections in HEK293 cells were performed using Lipofectamine 2000 (Thermo Fisher), cultured for 24-36 hours and then lysed in TNTE buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM EDTA, 0.5% Triton X-100). Lysates were incubated with 15 μL prewashed Protein A/G beads and 0.5-1.0 μg anti-Flag (Sigma, F1804), anti-HA (Roche, clone 3F10), or anti-GFP (Torrey Pines, TP401) antibody for 2-4 hours or overnight at 4°C with agitation, followed by 3 washes in TNTE wash buffer (0.1% Triton X-100). Inputs and immunoprecipitates were resolved on SDS-PAGE and analyzed by immunoblotting. All Co-IP experiments were repeated as least 3 times.

Aurora B inhibitor treatments

Embryos were treated with pronase to remove chorion or manually dechorionated with forceps at 10-20 mpf, and subsequently incubated with DMSO or ZM447439 (20 mM stock solution in DMSO, Tocris Bioscience 2458) in 1X Danieau solution. Embryos were treated with 400 μM ZM447439 from 30 to 55 mpf before mounting in 0.3% low melting agarose for live confocal imaging. For Aurora B sensitivity experiments, dechorionated embryos were treated with 75 μM ZM447439 from 30 mpf to 3 hpf and observed under a dissecting microscope to analyze cleavage defects.

RT-PCR, qPCR, and in situ hybridization

cDNA from WT and ttc28 mutant embryos was prepared at various developmental stages using Trizol (Invitrogen) and iScript cDNA synthesis kit (Bio-Rad). qPCR was performed using SsoAdvanced SYBR Green Supermix (Bio-Rad) and CFX Connect Real-Time PCR Detection system (Bio-Rad). Whole mount in situ hybridization was carried out as previously described (Thisse and Thisse, 2008). Briefly, 4% PFA-fixed embryos were placed at -20°C in 100% methanol overnight. Samples were then rehydrated and prehybridized at 70°C in hybridizati on mix (50-65% formamide, 5 × SSC, 50 μg/ml Heparin, 500 μg/ml tRNA, and 0.1% Tween 20 in H2O) for at least 2 hours. ttc28 sense control and antisense RNA probes were synthesized using Digoxigenin RNA labeling kit (Sigma Aldrich), and used in the hybridization steps. Anti-Digoxigenin-AP antibody was used at 1:5000, and followed by standard Alkaline Phosphatase staining protocol.

Primers used in this study are shown in in the table below.

	Forward primer	Reverse primer
ttc28-rtF&R	GGCAAACAGGCCAACCGTCG	ACGCGCAGCACCTCGATGAT
ttc28-qF1&R1	GCGCTTCAGTCTCATCGCCGT	GGCAGACGAGGGTTTCCGACC
ttc28-qF2&R2	ACAGACCCCGGAGACCGACT	CCCACAACATTTTTAACAGCCCGGT
gapdh-rtF&R	GATACACGGAGCACCAGGTT	GCCATCAGGTCACATACACG
β-actin-qF&R	CGAGCTGTCTTCCCATCCA	TCACCAACGTAGCTGTCTTTCTG

Quantification and Statistical Analysis

Cleavage furrow calcium activity

Furrow positioning orientation was determined by measuring the angle between the preceding furrow calcium signal and the emerging furrow calcium signal from the bottom to the top at a range from 0° to 180° The rose diagram was generated with Rose.N et software. For furrow propagation dynamics, the time of the first frame when calcium signaling observed was determined as t=1, and the length of furrow calcium signaling was measured at each time point and divided by the total length of the dividing cell as a percentage. For furrow deepening process, the width of preceding furrow calcium signaling was measured at the time frame when succeeding division calcium signaling emerged, and divided by the total width of cell as a ratio; a higher ratio suggests a delay in furrow deepening.

Assembly of midzone microtubules

The time frame when anti-parallel midzone microtubule bundles initiated was defined as t=0, and the distance of the midzone microtubule bundle was measured as initial diameter d. Subsequent measurements of the midzone microtubule distance at 1 minute interval were defined as d' and divided by the initial diameter d as a ratio. In MZdchs1b mutants, mild phenotype was defined as midzone microtubule assembly could be readily detected, and severe phenotype was defined as those exhibited abnormal midzone activity as shown in Fig. 2A bottom panel. For quantification of midzone microtubule dynamic ratio, severe phenotype groups were not taken into account.

Yolk cytoplasmic layer microtubule turnover

The YCL microtubule density was analyzed using ImageJ Particle Analyzer plugin (https://imagej.net/Particle_Analysis). First, the region of interest (131 μm × 65.5 μm) was thresholded with ImageJ software to subtract the background signals, then particle analyzer plugin was used with the default setting to measure the percentage of white area, representing the empty area or microtubule-pixel unoccupied area in the region of interest, across the time-lapse movies. The values were then subtracted from 100% to give rise to percentages representing microtubule density at individual time frames. 100% microtubule density was defined as the whole region of interest occupied by microtubule pixels. For MZdchs1b mutants, severe group was defined as those embryos failing to show an obvious microtubule low-density gap, and the remaining embryos were grouped into the mild phenotype class. Both groups were taken into quantification.

HA-Ttc28-mCherry intensity

To quantify the relative HA-Ttc28-mCherry intensity, mean pixel intensity at a square size of 1.24 μm by 1.24 μm (5×5 pixel) was measured on the plasma membrane, centrosome, or cytosol at individual z-plane. 6 spots for plasma membrane and cytosol, and 2 spots for centrosome were measured for each cell as shown in Fig. S4A and averaged for each group. The relative intensity of HA-Ttc28-mCherry on plasma membrane or centrosome was calculated by subtracting the cytosol signal intensity.

EB3-GFP tracks analyses

To quantify the speed, duration, and displacement parameters of individual EB3-GFP tracks, comets in time-lapses were converted to spots with a diameter of 5 pixels plotted at x,y positions corresponding to the centroid of each comet at each timepoint. These positions were tracked and analyzed using the TrackMate ImageJ plugin (Tinevez et al., 2017). The difference of Gaussians detector (DoG) algorithm was used to detect individual spots, and the linear motion Linear Assignment Problem (LAP) tracker was used to join these spots into tracks. To quantify trajectory angles of individual EB3-GFP tracks, angles were calculated between each comet's initial and final locations.

Statistical analyses

All statistical analyses were performed with PAST and Graphpad Prism 7 software. Mardia-Watson-Wheeler test was used to compare the furrow calcium signaling initiation position between WT and mutants. Furrow deepening calcium signaling widths were analyzed using Student's unpaired t-test. For midzone microtubule bundling experiments, paired Student's two-tailed t-test was used to determine the difference between control and mutants. Tukey's multiple comparisons test was performed to quantify the Ttc28-mCherry subcellular distribution experiments. See figure legends for additional information about number of embryos used in each experiment.

Supplementary Material

1

Figure S1, related to Figure 1. Cleavage defects in MZdchs1b mutants and abnormal cleavage furrow-associated calcium activities after Aurora B inhibition in WT (A) Quantification of cleavage plane orientations in WT and MZdchs1b embryos between 8-cell and 64-cell stages. (B) Representative images of WT and MZdchs1b blastomeres labeled with membrane-RFP (red) and H2B-GFP (green) at 3 hpf. Arrowheads denote the abnormal mitotic events. Scale bar, 30 μm. (C) Representative time-lapse still images of cleavage furrow calcium signaling in Tg[βactin2:GCaMP6s]^stl351 embryos following 400 μM Aurora B inhibitor ZM447439 treatment. Scale bar, 150 μm. (D) Quantification of furrow initiation orientation in DMSO control and ZM447439-treated Tg[/3actin2:GCaMP6s]^stl351 embryos. ns, not significant. (E) Quantification and comparison of the cleavage furrow-propagation calcium signaling in DMSO control and ZM447439-treated Tg[βactin2:GCaMP6s]^stl351 embryos at 2-4 cell stage. N, number of embryos.

Figure S2, related to Figure 3, 4. Identification of Ttc28 as a Dchs1b binding protein Heatmap of the raw spectral count data for two biological repeats of Dchs1-ICD pull-down in comparison with control from the AP-MS experiments. The arrow marks TTC28 AP-MS results. Schematic of TTC28 sequence identity and similarity from difference species. Representative still images of HA-Ttc28-mCherry (magenta) and H2B-GFP (green) during cell divisions at 2.5-3 hpf. Scale bar, 30 μm. Representative images showing the co-localization of HA-Ttc28-mCherry (magenta) and centrosome marker Xcentrin-GFP (green) in WT embryos at 2.5-3 hpf. Scale bar, 30 μm.

Figure S3, related to Figure 3. Deletion and mutation mapping to identify the binding regions of Dchs1 b and Ttc28 Co-IP assays of full-length Dchs1 b-sfGFP with full-length HA-Ttc28. a-HA antibody was used to immunoprecipitate the cell lysates, which were immunoblotted with a-GFP and a-HA antibodies. Co-IP assays of Flag-Dchs1b-ICD with full-length HA-Ttc28. α-Flag antibody was used to pull down the immunoprecipitates. Co-IP assays of HA-Ttc28 or HA-Ttc28-N with Flag-Dchs1b-ICD. Co-IP assays of HA-Ttc28-N with different Flag-Dchs1b-ICD CM2-N mutation constructs.

Figure S4, related to Figure 4. Ttc28 is recruited by CM2 motif in Dchs1 b ICD to the plasma membrane through its N-terminal TPR motifs Related to Figure 4A, quantification of HA-Ttc28-mCherry signal intensity at the cytoplasm, plasma membrane, and centrosome before normalization to the signal in the cytoplasm. Representative images of the subcellular localizations of HA-Ttc28-mCherry, HA-Ttc28-N-mCherry, and HA-Ttc28-ΔN-1-mCherry (magenta) together with membrane GFP/H2B-GFP (green) at 3-4 hpf. Scale bar, 30 μm. Representative images of the subcellular localizations of HA-Ttc28-mCherry, HA-Ttc28-N-mCherry, and HA-Ttc28-AN-1-mCherry (magenta) together with Dchs1 b-sfGFP/H2B-GFP (green) at 3-4 hpf. Scale bar, 30 μm. Quantification of mitotic defects in mCherry control, HA-Ttc28, HA-Ttc28-mCherry, HA-Ttc28-N-mCherry, and HA-Ttc28-ΔN-1-mCherry overexpressing embryos at 4-5 hpf. N, number of embryos. n, number of cells. (E) Representative images of the subcellular localizations of HA-Ttc28-mCherry at 3-4 hpf, upon injection of either 900 pg (left column) or 600 pg (three right columns) of HA-Ttc28-mCherry RNA into WT embryos co-expressing either membrane GFP (25 pg RNA), Dchs1 b-sfGFP (600 pg RNA) orDchs-ACM2-sfGFP (600 pg RNA). Scale bar, 30 μm.

Figure S5, related to Figure 5. ttc28 mutant embryos exhibit normal furrow calcium activities during cleavage stage, but are less sensitive to Aurora B inhibition (A-C) Quantification of cleavage furrow calcium activities in WT and MZttc28^{stl363/stl363} at 2-4 cell stage. N, number of embryos. ns, not significant. Error bars represent standard deviation. (D) Quantification of cleavage defects in WT and MZttc28^{stl362/stl362} embryos at 3-3.5 hpf after DMSO control or 75 μM ZM447439 treatment.

Figure S6, related to Figure 2, 5. Analyses of YCL microtubule dynamics in Tg[fiactin2:EMTB-3xGFP] transgenic background Representative time-lapse images of YCL microtubules in WT, MZdchs1b, and MZttc28^{stl362/stl362} embryos during cleavage stages using Tg[/3actin2:EMTB-3xGFP] at 1.75-2.5 hpf. Scale bar, 30 um. Quantification and comparison of YCL microtubule density in WT, MZdchs1b and MZttc28^{stl362/stl362} embryos at 1.75-2.5 hpf. N, number of embryos. (C, D) Quantification of EB3-GFP track displacement and duration data in the YCL in WT and MZdchs1b. N, number of embryos.

Movie S1. Related to Figure 1. Movie shows cleavage furrow calcium activities during cleavage stages in WT and MZdchs1b embryos from the animal pole view.

Movie S2. Related to Figure 7. Movie shows the cleavage furrow calcium activities in Tg[(3actin2:GCaMP6s]^st embryo after ZM447439 treatment.

Movie S3. Related to Figure 2. Movie shows microtubule dynamics during cell divisions in the WT and MZdchs1b blastomeres at 2-2.5 hpf from the animal pole view.

Movie S4. Related to Figure 2. Movie shows that the YCL microtubules depolymerize and repolymerize from the animal toward the vegetal pole during cleavage stages in WT, mild MZdchs1b, and severe MZdchs1b mutants at 2-2.5 hpf (lateral view).

Movie S5. Related to Figure 4. Movie shows the YCL microtubule dynamics in HA-Ttc28-mCherry-overexpressing WT embryos.

Movie S6. Related to Figure 5. Movie shows the YCL microtubule dynamics in MZttc28^{stl363/stl363} mutants.

Movie S7. Related to Figure 5. Movie shows representative midzone microtubule assembly activities in the WT and MZttc28^{stl363/stl363} blastomeres during cleavage stages.

Movie S8. Related to Figure 6. Movie shows representative midzone microtubule assembly activities in MZdchs1b single mutant and MZdchs1b; MZttc28^{stl363/stl363} compound mutant blastomeres during cleavage stages.

Highlights.

• Zebrafish Dachsous1b promotes microtubule turnover and midzone microtubule assembly

• Dachsous1b intracellular domain binds Ttc28 and Aurora B

• Dachsous1b and Aurora B promote and Ttc28 limits microtubule turnover

• Dachsous1b, Aurora B and Ttc28 functionally interact to control embryonic cleavages