Asymmetric inheritance of the apical domain and self-renewal of retinal ganglion cell progenitors depend on Anillin function

During retinal development, asymmetric self-renewing divisions of progenitor cells prolong the temporal progression of neurogenesis (Cayouette et al., 2006; Cepko, 2014; Parameswaran et al., 2014). The asymmetric inheritance of apical components has been suggested to provide instructive information for the specification of cell fate (Clark et al., 2012; Huttner and Brand, 1997; Kechad et al., 2012). The complex cellular and molecular interactions linking this inheritance to daughter cell self-renewal and differentiation in vivo remain elusive. Recent studies reported cleavage furrow components as potential signalling sources for instructing cell fate (Chen and Zhang, 2013; Dubreuil et al., 2007; Ettinger et al., 2011; Kosodo and Huttner, 2009; Pollarolo et al., 2011; Singh and Pohl, 2014). For example, newborn sensory neurons of the Drosophila notum inherit asymmetrically distributed cytokinesis remnants of the last mitotic cleavage (Pollarolo et al., 2011). Other studies have suggested that the extent of release of the post-mitotic midbody – a transient intercellular bridge from the cleavage furrow just before abscission – couples cell division to cell fate in neuroepithelial, stem and cancer cells (Dubreuil et al., 2007; Ettinger et al., 2011). It remains unknown whether cleavage furrow components also instruct asymmetric cell division in vivo in the developing vertebrate neuroepithelium.

Here, we combined in vivo time-lapse confocal imaging with cell biology and functional analyses to examine the role of the F-actin-binding protein Anillin – a crucial component of the cleavage furrow and midbody (Field and Alberts, 1995; Kechad et al., 2012; Rincon and Paoletti, 2012; Ronkainen et al., 2011) – in the occurence of asymmetric divisions in the developing zebrafish retina. We report novel evidence supporting that Anillin is important not only for progenitor cell self-renewal, but also for balancing symmetric and asymmetric outcomes essential for correct retinal neurogenesis. This study further provides novel insights into the interplay between cytokinesis machinery components and the retinal ganglion cell (RGC) fate determinant Ath5 (Atoh7) during asymmetric neurogenic cell divisions in vivo.

To investigate the role of Anillin during retinal neurogenesis we established an anillin:anillin-eGFP transgenic line that recapitulates anillin expression in vivo (Fig. 1A,B). All cells in the retinal neuroepithelium are proliferative prior to 28 hours post fertilisation (hpf) (Li et al., 2000), strongly expressing anillin (Fig. 1B). The ath5:gap43-RFP transgene, in which membrane-tethered RFP (Gap43-RFP) is expressed under the control of the ath5 promoter (Zolessi et al., 2006), highlights the cell cycle exit of RGCs – the first-born neurons in the vertebrate retina. Time-lapse imaging in double-transgenic anillin:anillin-eGFP;ath5:gap43-RFP embryos revealed that anillin:anillin-eGFP is downregulated in the differentiating Gap43-RFP-positive cells of the RGC layer, suggesting anillin downregulation by Ath5 (Fig. 1C,C′). Consistently, retinae from lakritz mutant (ath5^−/−) embryos, in which progenitors undergo extra rounds of proliferative divisions instead of differentiating into RGCs (Jusuf et al., 2012; Kay et al., 2001; Kelsh et al., 1996), failed to downregulate anillin mRNA (supplementary material Fig. S1A). The presence of a well-characterised Ath5 consensus binding site (CACCTG) (Del Bene et al., 2007) in an active enhancer region upstream of anillin [highlighted by histone H3 lysine 4 mono-methylation (H3K4me1) mark] (supplementary material Fig. S1B) further suggests that Ath5 directly regulates anillin.

Anillin is an established midbody component, and asymmetric inheritance of the midbody during stem cell division has been reported in multiple systems, including mouse radial glial progenitors (Dubreuil et al., 2007; Ettinger et al., 2011; Hesse et al., 2012; Kosodo et al., 2004). This led us to investigate Anillin-eGFP dynamics in dividing retinal progenitor cells. Live imaging confirmed the localisation of Anillin-eGFP in the gap43-RFP-expressing progenitor cell nuclei at interphase (asterisk, Fig. 1D, t=0′), to the contractile ring during cytokinesis (arrow, Fig. 1D, t=0′) and at the daughter cell apical interface consistent with the formation of a midbody at the end of cytokinesis (Fig. 1D, t=5′ and t=10′). Interestingly, 60% of analysed divisions (n=9 out of 15) displayed asymmetric positioning of the Anillin-eGFP spot to one daughter cell apical domain at the end of cytokinesis (Fig. 1E-F). Consistent with a correlation between Anillin-eGFP inheritance and cell fate, we observed that the daughter cell inheriting the Anillin-eGFP spot retracts its apical process and migrates basally towards the differentiating RGC layer, while the sibling cell invariably migrates back towards the apical surface (n=5 out of 5, Fig. 1G; supplementary material Movie 1). In addition to the observed dynamics of anillin expression downstream of Ath5, these data are consistent with a role for Anillin in asymmetric neurogenic cell divisions in vivo.

To test the role of Anillin in retinal neurogenesis we performed a loss-of-function analysis. Anillin knockdown causes cytokinesis defects that affect early vertebrate development (Gbadegesin et al., 2014; Reyes et al., 2014) (supplementary material Movie 2). We therefore generated anillin hypomorphic conditions (supplementary material Fig. S2A-C) and performed all analyses in mosaic zebrafish embryos as previously described (Jusuf et al., 2012). Live imaging of dividing transplanted progenitor cells labelled with H2B-RFP (nuclei, red) and LifeAct-Venus (F-actin, yellow; Riedl et al., 2008) highlighted the appearance of an F-actin-enriched spot at the basal side of the cell body before anaphase (Kosodo et al., 2008) and its basal-apical displacement during cytokinesis (supplementary material Fig. S2D,E), indicating that vertical furrowing progresses normally in the mononucleated anillin hypomorphic cells.

The midbody has been reported as a positional cue for apical F-actin accumulation at the daughter cell interface before the formation of new cell-cell junctions (Herszterg et al., 2013; Morais-de-Sa and Sunkel, 2013; Singh and Pohl, 2014). Expression of the anillin-eGFP reporter together with a red lifeact-Ruby construct revealed that midbody formation at the end of cytokinesis indeed coincides with the transient accumulation of F-actin around it (arrow, Fig. 2A; supplementary material Movie 3). Consistent with the apical Anillin-eGFP spot, quantification of LifeAct-Venus intensity revealed that more than half of the analysed divisions (57%, n=8 out of 14) distribute the apical F-actin-rich domain asymmetrically to one daughter (Fig. 2B,C; supplementary material Movie 4). This asymmetric distribution is drastically reduced in anillin hypomorphic conditions (15%, n=2 out of 13) (Fig. 2B,C; supplementary material Movie 5), indicating that Anillin is required for proper apical domain segregation in dividing retinal progenitor cells.

The evolutionarily conserved apical polarity protein Par3 (Pard3) colocalises with F­actin at the level of junctional belts in neuroepithelial cells and its inheritance during cell division has been linked to neuronal fate (Alexandre et al., 2010; Bultje et al., 2009; Chen and Zhang, 2013; Dong et al., 2012; Shi et al., 2003; Takekuni, 2003; Zolessi et al., 2006). We investigated whether Par3 asymmetric inheritance depends on Anillin. Time-lapse imaging of a Par3-GFP protein revealed that, whereas in control clones most divisions distribute Par3 asymmetrically (80%, n=12 out of 15, Fig. 2D; supplementary material Movie 6), anillin hypomorphic conditions favour symmetric Par3 distribution (Fig. 2D, n=15 out of 15; supplementary material Movie 7). Thus, Anillin is required for the correct apical distribution of F-actin and Par3, which might be essential for asymmetric cell fate outcome.

We next assessed whether the observed effect on Par3 and F-actin reflects changes in the mode of cell division. Hypomorphic anillin conditions produce an overall increase in ath5:GFP (Poggi et al., 2005) or ath5:gap43-RFP transgene expression (Fig. 3A-C; supplementary material Movies 8 and 9) and an expansion of the RGC layer (Fig. 3C), which is efficiently rescued by co-injection of anillin-eGFP mRNA (Fig. 3D). Although quantification of phospho-histone H3 (pH3)-positive cells at 30 hpf revealed no significant difference between the control and anillin hypomorphic conditions (Fig. 3E,F), the number of pH3-positive cells within the ath5:GFP population was significantly lower in anillin hypomorphic conditions (Fig. 3E,G). Thus, retinal progenitor divisions are more likely to generate ath5-expressing daughters, and ath5-expressing cells are less likely to undergo further rounds of cell division.

To assess this hypothesis we performed in vivo time-lapse analysis of daughter cell fate in clones transplanted from ath5:GFP transgenic embryos. We considered symmetric versus asymmetric divisions based on two criteria: (1) Ath5:GFP positive (A) versus Ath5:GFP negative (self-renewing progenitor, P) in sibling daughters from divisions of H2B-RFP-labelled cells; (2) neuronal fate (N) versus self-renewing progenitor (P) in sibling daughters from divisions of H2B-RFP/Ath5:GFP-positive cells (Fig. 4) (Poggi et al., 2005). We found that 56% of divisions of H2B-RFP-positive cells were asymmetric, generating one GFP-positive and one GFP-negative daughter (A/P) (n=5 out of 9) (Fig. 4A,D; supplementary material Movie 10). By contrast, all divisions analysed in anillin hypomorphic clones were symmetric, entailing GFP in both daughter cells (Fig. 4B-D; supplementary material Movie 11) (n=10). This is consistent with the observed increase in Ath5:GFP-positive cells. Within the Ath5:GFP-positive control clones most analysed divisions (82%, n=9 out of 11) were asymmetric, generating one post-mitotic daughter, which retracts the apical process and migrates to the RGC layer, and one self-renewing progenitor (N/P) (Fig. 4A,D; supplementary material Movie 12) (Jusuf et al., 2012). The concomitant appearance of an extending axon from the cell body or expression of the cxcr4b:cxcr4b-GFP transgene (Donà et al., 2013) further indicated differentiation into an RGC (data not shown; Fig. 4C,C′) (Pujic et al., 2006). Conversely, most divisions in anillin hypomorphic clones were symmetric, generating two Cxcr4b-GFP-positive daughters (N/N; 70%, n=7 out of 10) (Fig. 4B-D; supplementary material Movie 13). This is consistent with the decrease in pH3-positive cells among the Ath5:GFP population. Hence, when Anillin function is impaired, retinal progenitors undergo symmetric rather than asymmetric divisions, in which they are more likely to upregulate ath5, become post-mitotic and adopt a neuronal (RGC) fate (Fig. 4E).

Our study provides compelling evidence that Anillin contributes to the regulation of asymmetric neurogenic cell divisions in the retinal neuroepithelium – favouring the maintenance of proliferating progenitors while also being required for the proper distribution of F-actin and Par3 at the daughter cell apical domain. Our time-lapse analysis of asymmetric cell divisions of RGC progenitors suggests that the differentiating (RGC) daughter transiently inherits the Anillin-eGFP-labelled midbody. This is in agreement with the idea that instructive information accompanying this inheritance is linked to neural fate commitment (Alexandre et al., 2010; Pollarolo et al., 2011; Zolessi et al., 2006).

This study establishes a framework for future mechanistic investigations to address how apical Anillin, F-actin and associated signalling molecules contribute to long-term fate decisions in asymmetric neurogenic cell divisions (Galvagni et al., 2012; Parameswaran et al., 2014). The nature of this information (e.g. cell cycle exit, cell polarity or neuronal fate acquisition) and the underlying signalling pathways have yet to be determined, but the evidence provided here suggests that they are likely to integrate feedback regulatory loops between ath5 and anillin. Anillin expression correlates with the proliferation and metastatic potential of human tumours from many different tissue origins (Hall et al., 2005). Therefore, it is tempting to speculate that similar mechanisms, whereby components of the cytokinesis machinery contribute to maintaining the intrinsic asymmetry of proliferative divisions, might be common to all proliferating stem and progenitor cells.

Zebrafish (Danio rerio) breeding/raising followed standard protocols (Kimmel et al., 1995). Fish care is under the supervision of animal welfare agencies and in accordance with local (Tierschutzgesetz 111 Abs. 1 Nr. 1) and European Union animal welfare guidelines. For imaging, embryos were treated with 0.0045% 1-phenyl-2-thiourea (Sigma) to delay pigment formation. Lines used are indicated in the supplementary Materials and Methods.

Morpholinos (1 ng/µl) were: CtrlMO, 5′-CCTCTTACCTCAGTTACAAT-TTATA-3′ (Gene Tools); ATGMO, 5′-GTACGCTACAAGCTGAAAG-TAAAGT-3′; and SpMO, 5′-TTTCACAAAAAGCTCTCACCTCGGT-3′. Controls and rescue experiments are described in the supplementary Materials and Methods.

The Cell Counter plugin from Fiji software was used to determine the average ratio of pH3-positive cells among total cells (DAPI staining) or among Ath5:GFP cells in every 2 µm of a 16 µm-thick z-projection.

Six plasmids were used: BAC/anillin:anillin-eGFP and CMV:anln-eGFP_pDestTol2 (see supplementary Materials and Methods for cloning details and BAC transgenesis); pCS2+:H2B-RFP (Jusuf et al., 2012); pCS2+:Lifeact-Venus (kindly provided by R. Köster, TU Braunschweig, Germany); pDestTol2/CMVSP6:lifeact-Ruby vector (generated by I. Weisswange and kindly provided by J. Wittbrodt, University of Heidelberg, Germany); and pCS2+:pard3-GFP (kindly provided by P. Alexandre and J. Clarke, King's College London, UK).

Immunostaining was performed as described (Jusuf et al., 2012; supplementary Materials and Methods). Primary antibodies: rabbit anti-phospho-histone H3 (Millipore, 06-570; 1:500), mouse anti-Zn5 (ZIRC Zn-5; 1:200) and chicken anti-GFP (Life Technologies, A10262; 1:200). Secondary antibodies: goat anti-rabbit IgG (H+L) conjugated to Cy5 (Invitrogen, A10523; 1:500), goat anti-mouse IgG (H+L) conjugated to Alexa fluor 546 (Invitrogen, A11003; 1:500) and donkey anti-chicken IgG (H+L) conjugated to Alexa fluor 488 (Jackson, 703-546-155; 1:250). Nuclei were counterstained with DAPI (Sigma; 1:1000).

For ISH, antisense probes for anillin and ath5 were generated as previously described (Schuhmacher et al., 2011). See supplementary Materials and Methods for cloning and ISH protocol details.

A Leica SpE confocal laser scanning microscope was used with a Leica 40×1.15 NA oil-immersion or Leica 63×1.2 NA water-immersion objective and Leica Application Suite (LAS) software (see supplementary Materials and Methods for further details of imaging). Volocity software (version 5.3, PerkinElmer) or Fiji was used for image processing and analyses. For analysis of F-actin spot displacement (supplementary material Fig. S2), coordinates (x, y, z) of the spot at the basal side of the cell body at anaphase and at the apical surface at telophase were determined. Distance covered and velocity were calculated between the two points. For lineage analysis, mosaic clones were randomly selected in the time-lapse series and daughter cells tracked manually in 4D after division. For A/P and A/A divisions (Fig. 4A,B), the ratio GFP intensity/background was assessed with Fiji. Analysis of Anillin and F-actin distributions is described in supplementary material Fig. S3 legend. Par3-GFP distribution was assessed on frontal z-sections across the centre of the dividing daughters (Alexandre et al., 2010).

Statistical tests were performed using Prism 5.0 (GraphPad), Microsoft Excel and R (Fisher's exact test, Student's t-test and Wilcoxon Mann–Whitney test) with P<0.05 considered significant. Classification of the distance between the daughter cell interface and Anillin/F-actin centre of mass was performed using R. For comparison between CtrlMO and anlnMO we used unpaired homoscedastic two-tailed Student's t-test [lineage analysis (GFP-positive daughters), rescue of the RGC layer, F-actin spot displacement], Wilcoxon Mann–Whitney test (F-actin distribution in CtrlMO versus anlnMO conditions) and two-tailed Fisher's exact test (Par3 inheritance and lineage studies). All data represent three or more independent experiments.

We thank R. Köster, J. Wittbrodt, D. Gilmour, P. Alexandre and J. Clarke for constructs and transgenics; J. L. M. Cerdan for statistics; S. Bouffard for cloning; B. Wittbrodt, E. Leist, A. Saraceno, M. Majewski, B. Seiferling, T. Kellner for fish maintenance and technical assistance; M. Carl, W. A. Harris, C. Norden, N. Foulkes and D. Gilmour for constructive inputs; J. McCall for editing; J. Wittbrodt and K. Schumacher for sharing equipment and space.