Uncoupling of neurogenesis and differentiation during retinal development.

Conventionally, neuronal development is regarded to follow a stereotypic sequence of neurogenesis, migration, and differentiation. We demonstrate that this notion is not a general principle of neuronal development by documenting the timing of mitosis in relation to multiple differentiation events for bipolar cells (BCs) in the zebrafish retina using in vivo imaging. We found that BC progenitors undergo terminal neurogenic divisions while in markedly disparate stages of neuronal differentiation. Remarkably, the differentiation state of individual BC progenitors at mitosis is not arbitrary but matches the differentiation state of post-mitotic BCs in their surround. By experimentally shifting the relative timing of progenitor division and differentiation, we provide evidence that neurogenesis and differentiation can occur independently of each other. We propose that the uncoupling of neurogenesis and differentiation could provide neurogenic programs with flexibility, while allowing for synchronous neuronal development within a continuously expanding cell pool.

Introduction

The nervous system is indisputably the most complex structure assembled during vertebrate ontogenesis. Consequently, the developmental processes underlying nervous system assembly must be precisely orchestrated. Neuronal development in the central and peripheral nervous systems (CNS, PNS) is widely accepted to require three major steps: (i) neurogenesis, the birth of neurons by progenitor cell mitosis, (ii) migration, the relocation of post‐mitotic neurons from their birthplace in proliferative zones to specific locations, and (iii) neuronal differentiation, the acquisition of molecular and morphological features that permit the integration of newly generated neurons into synaptic circuits. To date, the prevailing view is that these ontogenetic events are discrete steps along a stereotypic sequence, beginning with neurogenesis, followed by migration and concluding with neuronal differentiation. There is however, particularly in the PNS, evidence that some hallmarks of neuronal differentiation might already occur in progenitors (Rothman et al, 1980; Rohrer & Thoenen, 1987; DiCicco‐Bloom et al, 1990; Godinho et al, 2007; Attardo et al, 2008). These observations call into question the invariant developmental sequence of neurogenesis, migration, and differentiation, and raise the unresolved question: How stereotypic is the developmental program of a defined progenitor population in vivo?

Here, we examined the developmental fate of a molecularly defined CNS progenitor population (expressing visual homeobox gene 1, vsx1) that gives rise to the vast majority of a specific interneuron cell type (bipolar cells, BCs) in the zebrafish retina by terminal symmetric divisions (He et al, 2012; Weber et al, 2014). The swift development as well as the genetic and optical accessibility of zebrafish permitted us to follow the entire developmental program of vsx1 progenitors, with single cell precision, in vivo. We used molecular, morphological, and cell biological markers of neuronal differentiation, in conjunction with chronic in vivo time‐lapse imaging, to determine the timing of mitosis in relation to a battery of developmental events. We discovered that for BCs, neurogenesis and multiple hallmarks of neuronal differentiation (such as somal positioning, neuronal marker expression, or neurite elaboration) are timed independently of each other. In other words, rather than dividing at a stereotypic point in their developmental trajectory, vsx1 progenitors of BCs undergo terminal mitosis at markedly disparate stages of differentiation, suggesting that differentiation is not time‐locked to mitosis. However, the state of differentiation of a vsx1 progenitor at mitosis is not arbitrary, but matches that of the post‐mitotic vsx1 BCs in its vicinity.

Results

Bipolar cell progenitor mitoses occur over an extended time‐period and relocate to non‐apical sites

In common with many parts of the developing vertebrate CNS, the retina begins as a pseudostratified neuroepithelium with spindle‐shaped progenitors that span its apico‐basal extent and undergo interkinetic nuclear migration, an oscillatory nuclear movement linked to specific cell cycle phases (Sauer, 1935; Baye & Link, 2008). At distinct but overlapping times, cells destined for different fates exit the cell cycle. Because mitotic divisions generally occur at the apical surface, newborn cells need to migrate varying distances to occupy their definitive locations within one of the emerging cellular laminae. Thus, while ganglion cells migrate furthest to occupy positions in the basal most part of the neuroepithelium, BCs have a shorter distance to relocate, and photoreceptors remain in situ at the apical surface. BCs, which are ultimately localized to the inner nuclear layer (INL) and confine their dendritic and axonal processes to the outer and inner plexiform layers (OPL, IPL), respectively, are generated over a protracted period, between 2 and 3 days post‐fertilization (dpf) in the zebrafish (He et al, 2012). Thus, early‐born cohorts of BCs are generated when the retinal neuroepithelium is not yet laminated, while later‐born cohorts are generated when the three cellular laminae are emerging.

To investigate the relationship between BC neurogenesis and differentiation, we examined the expression of vsx1, a transcription factor important for BC development (Passini et al, 1997; Chow et al, 2001; Vitorino et al, 2009; Shi et al, 2011), using a vsx1:GFP transgenic line (Kimura et al, 2008). In the zebrafish retina, vsx1 is expressed at low levels in the majority of committed, terminally dividing BC progenitors, up‐regulated during differentiation, and maintained at high levels in mature BCs (Vitorino et al, 2009). This developmental expression profile is faithfully reproduced in the bacterial artificial chromosome (BAC) vsx1:GFP transgenic line (Vitorino et al, 2009), so that we could follow nascent BCs from birth to maturity. Further, we took advantage of the zebrafish retina's gradient of development (Hu & Easter, 1999), to concurrently visualize immature (unlaminated) and more mature (laminated) parts of the retina and thus compare multiple states of BC differentiation in a single field of view.

At 2 dpf, we found low levels of GFP expression in the unlaminated part of the retina in which vsx1:GFP+ (henceforth referred to as vsx1 +) cells span the entire apico‐basal axis. These vsx1 + cells represent both BC progenitors and post‐mitotic, undifferentiated BCs (Morgan et al, 2006; Randlett et al, 2013). By contrast, vsx1 + cells in the more mature, laminated part of the retina showed high levels of GFP expression and confined their processes to the IPL and OPL, suggestive of post‐mitotic, differentiated BCs (Fig 1A). To our surprise, immunostaining of vsx1:GFP retinas for phosphorylated histone H3 (pH3), a late G2/M‐phase marker (Hendzel et al, 1997), revealed vsx1 + pH3+ cells not only in the unlaminated retina but also in the INL of the laminated retina. This suggests that not only post‐mitotic BCs but also vsx1 + progenitors reside within the INL (Fig 1A–D). Notably, while vsx1 + progenitors in the unlaminated retina undergo mitosis at the apical surface, like “classical” progenitors in many parts of the CNS (Fig 1B), vsx1 + progenitors in the laminated retina undergo mitosis in the INL (Fig 1C), akin to previously described non‐apical progenitors (Godinho et al, 2007; Weber et al, 2014). Given the precocious expression of some neuronal characteristics in such non‐apical progenitors, we asked how similar vsx1 + progenitors were to the post‐mitotic vsx1 + BCs in their immediate surround.

Figure 1

Vsx1 + progenitors undergo mitosis in different proliferative zones and match the expression of molecular markers of post‐mitotic BCs in their vicinity

Confocal images of a coronal cryostat section from a 2 dpf vsx1:GFP retina with immature, neuroepithelial (“unlaminated”, orange) and mature, laminated regions (“laminated”, cyan). Left panel, vsx1:GFP; right panel, vsx1:GFP shown in conjunction with pH3 antibody staining to label cells in late G2/M‐phase. Scale bar: 10 μm.

High magnification images of (B) an apically dividing vsx1 + progenitor (orange arrowhead) in an unlaminated region where cells span the retina and express GFP weakly and (C) a non‐apically dividing vsx1 + progenitor (cyan arrowhead) in a laminated region where cells confine their processes to the OPL and IPL (dashed lines) and express high levels of GFP. Cellular membranes are labeled with BODIPY methyl ester. Scale bar: 10 μm.

Quantification of vsx1 + progenitor mitoses at apical (lightly shaded) and non‐apical (darkly shaded) locations in the unlaminated (“un”, orange) and laminated (“lam”, cyan) retina. Data are presented as mean ± SEM, 1,391 mitotic divisions, 80 sections from at least 14 fish.

Quantification of vsx1:GFP fluorescence intensity in dividing vsx1 + progenitors in the unlaminated (“un”, orange) and laminated (“lam”, cyan) retina. Data are presented as mean ± SEM, 86 progenitors, 43 sections from at least 13 eyes, Mann–Whitney U‐test, ***P ≤ 0.0001.

Correlation of vsx1:GFP fluorescence intensity in progenitors and their surrounding cells in the unlaminated (orange circles) and laminated (cyan circles) regions of the retina. Eighty‐six progenitors from 43 sections from at least 13 eyes, r 2 = 0.78. Inset: Analysis for all pH3+ cells (n = 13) from a single section.

Confocal images of a cryostat section from a 2 dpf vsx1:GFP retina immunostained with antibodies against pH3 and Crx. Vsx1 + progenitors (GFP+ pH3+) are Crx negative (orange arrowheads) in the unlaminated retina (left panels, orange bars above panels) and are Crx positive (cyan arrowhead) in the laminated region (right panels, cyan bars above panels). Scale bar: 10 μm.

Correlation of Crx antibody staining intensity between mitotic BC progenitors and the vsx1 + cells in their vicinity. Thirty‐eight progenitors, 10 sections from at least five eyes, r 2 = 0.87; orange circles, apical mitoses; cyan circles, non‐apical mitoses.

Marker expression in progenitors matches the surrounding post‐mitotic bipolar cells

GFP levels during mitotic division in vsx1 + pH3+ progenitors showed a striking (3.8‐fold) increase between the unlaminated and laminated regions of the developing retina (Fig 1E), which correlated well with the GFP levels in surrounding vsx1 pH3− cells (Fig 1F; the surrounding vsx1 + cells in laminated regions are expected to be ~90% post‐mitotic BCs on average; for an estimate of this number see Appendix Supplementary Materials and Methods). Furthermore, comparison of vsx1:GFP levels in pH3+ progenitors and pH3− surrounding cells across the entire developmental gradient in single retinas revealed a linear increase along the gradient (Fig 1F, inset). Hence, with regard to vsx1 expression, progenitors in the laminated retina are more similar to their BC neighbors than to their early dividing peers and form a continuum with regard to vsx1 promoter activity in lock‐step with surrounding BC differentiation. Direct time‐lapse observation of vsx1:GFP levels confirmed a parallel increase of fluorescence levels in BC progenitors and surrounding post‐mitotic BCs in vivo (Fig EV1). Moreover, based on the decay of GFP in a vsx2:GFP BAC line (Vitorino et al, 2009), vsx2 down‐regulation was similarly linked to the progression of differentiation along the retinal gradient independent of mitotic status.

Figure EV1

In vivo time‐lapse recording of dividing BC progenitors in a vsx1:GFP retina

Confocal in vivo image of a 2 dpf vsx1:GFP retina. Expression of GFP, depicted using a “Fire” look‐up table (LUT), follows the retinal differentiation gradient, appearing first in the ventro‐nasal patch and subsequently in the nasal, dorsal, and temporal parts of the retina. Scale bar: 10 μm.

Time‐lapse images of the boxed region in (A). Left panels represent the raw levels of GFP fluorescence (Fire LUT). Right panels depict GFP fluorescence adjusted for better visualization (gray). Dashed circles mark mitotic events. Scale bar: 10 μm.

Individual mitotic events marked in (B) at progressively later times during development. GFP fluorescence levels of the mitotic progenitors and that of six cells in their surround. Orange circles, apical mitotic divisions; cyan circles, non‐apical mitotic divisions; gray squares, individual surrounding cells; black squares, average fluorescence of surrounding cells.

To further establish this similarity in molecular differentiation of vsx1 + progenitors and their post‐mitotic neighbors, we examined two additional molecular markers of BC differentiation: cone‐rod homeobox (Crx) and Ribeye a. Crx is a transcription factor expressed in mature photoreceptors and BCs (Liu et al, 2001; Shen & Raymond, 2004). In the unlaminated retina, antibody staining for Crx (combined with pH3) revealed little or no expression in vsx1 progenitors or the vsx1 + cells in their vicinity, whereas in laminated regions, high levels of Crx expression were found in virtually all vsx1 progenitors and surrounding post‐mitotic vsx1 + BCs (Fig 1G). The levels of Crx antibody staining in mitotic vsx1 + progenitors along the differentiation gradient strongly correlated with that of the vsx1 + cells in their immediate surround (Fig 1H). Moreover, time‐lapse imaging of a crx:mCFP transgenic line (Suzuki et al, 2013), in which Crx+ cells are faithfully labeled (Fig EV2A), regularly revealed non‐apical crx:mCFP+ mitotic divisions in the laminated retina (Fig EV2B and C). Finally, we analyzed the expression of Ribeye a, a structural protein of ribbon synapses in photoreceptors and BCs (Wan et al, 2005). As expected, using fluorescence in situ hybridization, we found ribeye a mRNA only in the laminated retina, where post‐mitotic cells predominate (Fig EV2D and E). Notably, we also observed ribeye a mRNA‐containing cells that were pH3+ (Fig EV2F). The fact that these cells were located in the INL suggests they are BC progenitors. Using a transgenic line designed to report ribeye a expression in BCs (ctbp2:mEGFP; Odermatt et al, 2012), we observed ctbp2:mEGFP+ cells dividing at non‐apical locations, giving rise to BCs (Fig EV2G). Hence, even with regard to a marker linked to synaptic structures unique to BCs in the inner retina, we find that BC progenitors co‐differentiate with post‐mitotic BCs in the surround. We next asked whether this similarity extended beyond molecular markers to cellular morphology and dynamics.

Figure EV2

BC progenitors dividing in the laminated retina express Crx and ribeye a

Confocal image of a cryostat section of a 2 dpf crx:mCFP retina immunostained with antibodies against Crx and pH3. Co‐labeled cells are indicated by magenta dots. Scale bars: 10 μm.

In vivo image of a 2 dpf crx:mCFP retina. Expression of mCFP in the INL is largely limited to the “laminated” region (cyan bar above the figure panel) of the retina while only photoreceptors are labeled in the unlaminated parts of the retina (orange bar above panel). One crx:mCFP+ progenitor (cyan arrowhead) can be seen in an early stage of division, which judged by transgene expression levels, belongs to the most differentiated cells in this intermediate part of the INL. Scale bar: 10 μm.

In vivo time‐lapse images of a crx:mCFP+ progenitor undergoing mitotic division (pseudo‐colored cyan) in the INL of a 2 dpf embryo. Ninety‐seven such divisions were observed in four time‐lapse recordings of a total of 32.1 h. Scale bar: 10 μm.

Contrast inverted confocal images of an eye from a 2 dpf ctbp2:mEGFP transgenic fish immunostained to visualize GFP (left panel) and processed for fluorescence in situ hybridization to detect expression of a ribeye a specific exon (right panel). Scale bar: 10 μm.

High magnification of boxed area in (D). Expression of the ctbp2:mEGFP transgene and endogenous ribeye a mRNA is restricted to the INL in the laminated region of the retina (cyan bar over figure panel). Dashed line indicates onset of expression. Scale bar: 10 μm.

Confocal images of a 2 dpf ctbp2:mEGFP retina labeled to detect GFP, pH3, and ribeye a mRNA. A ribeye a + pH3+ non‐apically dividing progenitor (magenta arrowheads, middle panel) can be seen in a laminated region of the retina (cyan bars above panels). Scale bar: 10 μm.

ctbp2:mEGFP+ progenitor cell undergoing mitotic division (pseudo‐colored cyan) in the INL during an in vivo time‐lapse recording of a 2 dpf retina. Eighty‐seven such divisions were observed in two time‐lapse recordings totaling 32.8 h. Scale bar: 10 μm.

Progenitor morphology and cell biology correspond to the surrounding post‐mitotic bipolar cells

To examine individual cells of the vsx1 + BC lineage, we generated a transgenic Gal4‐driver line (referred to as Q26) that was selected to label a sparse subset of vsx1 + cells. As Gal4 expression in Q26 is restricted to vsx1 BCs and their progenitors in laminated parts of the retinal gradient (Appendix Fig S1), we almost exclusively observed non‐apically dividing Q26 + progenitors, which we followed by time‐lapse imaging as they terminally divided to produce BCs (Fig 2A). We examined the cleavage plane in dividing Q26 + progenitors (n = 65) and found divisions along the circumferential, apico‐basal, and centro‐peripheral axes with no preference for a particular orientation, in line with other reports of non‐apically dividing progenitors (Kimura et al, 2008; Weber et al, 2014). During mitosis, the vast majority of Q26 + progenitors had limited their processes to the plexiform layers (Fig 2A at 0′, Fig 2B), as is characteristic for BCs in the mouse and zebrafish retina (Morgan et al, 2006; Randlett et al, 2013) but not for apically dividing progenitors (Das et al, 2003; Miyata et al, 2004; Noctor et al, 2004). Prior to mitotic division however, the processes of non‐apically dividing progenitors extended beyond the OPL or IPL (see Fig 2A at −89′) and over time remodeled to become restricted to the synaptic layers. We asked whether this morphological remodeling occurred in a fixed time‐window relative to mitotic division and focused our analyses, for technical reasons (see Appendix Supplementary Materials and Methods) on the retraction of processes from the apical surface. We found that apical process retraction occurred over an extended period of time (18 min to > 9 h) prior to mitotic division (Fig 2C). In contrast, for apically dividing BC progenitors in the unlaminated retina, apical process retraction to the OPL was a post‐mitotic event (as observed in vsx1:GFP), but again occurred over an extended time span following mitotic division (a few min to > 8 h). Thus, a single differentiation step, the remodeling of the apical process, occurs both pre‐ and post‐mitotically, and over a time span of more than 17 h relative to mitosis. Notably however, our time‐lapse recordings in the Q26 line suggested that apical process remodeling is locally coordinated. When we identified progenitors that had just undergone apical process retraction to the OPL and asked whether post‐mitotic BCs in the immediate vicinity had also done the same (Fig 2D), we found that, on the population level, apical process remodeling occurred concurrently (Fig 2E). Moreover, once pruned, the apical and basal processes of pre‐mitotic progenitors could form lateral arbors. While these arbors regressed during mitosis, at earlier time points we could not distinguish them from the dendritic and axonal arbors of surrounding post‐mitotic BCs (brackets in Fig 3A and E; Fig 4D at −645′), suggesting that the morphological processes of BC differentiation proceed independent of progenitor mitosis. In accordance with observations from other systems and species (Miyata et al, 2001; Das et al, 2003; Saito et al, 2003; Kosodo et al, 2008) the basally directed process of dividing Q26 + progenitors exhibited either splitting or asymmetric inheritance by one daughter cell followed by new outgrowth by the other daughter (Appendix Fig S2). Remarkably, when new process outgrowth was observed, it exhibited directed targeting of the IPL without overshooting beyond it.

Figure 2

Morphological rearrangement of vsx1 progenitors matches surrounding BCs

Confocal in vivo time‐lapse recording of a retina from the Q26 transgenic line (crossed to a UAS:memYFP reporter) showing a non‐apically dividing vsx1 + progenitor (pseudo‐colored cyan) with processes restricted to the IPL and OPL (dashed lines) during mitosis (0′). The last time point at which an apical process (open arrowhead) is detected is 89 min prior to mitosis. Scale bar: 10 μm.

Quantification of the distinct morphologies adopted by non‐apically dividing vsx1 + progenitors at M‐phase entry (122 progenitors, 17 fish). Open arrowheads indicate cytoplasmic processes extending beyond the synaptic layers (OPL and IPL, dashed lines).

Quantification of the time interval between retraction of the apical process (triangle) and mitosis (cyan circle) of non‐apically dividing vsx1 + progenitors shows a broad range from 18 min to more than 540 min. As only mitosis, but not process retraction was observed for the progenitor depicted with small cyan dots, the movie length of 540 min is an underestimate. 18 progenitors from 11 fish.

Schematic of apical process retraction (triangle) in a non‐apical vsx1 + progenitor (cyan soma) and the presence or absence of apical processes in the surrounding, post‐mitotic BCs.

Quantification of the percentage of surrounding post‐mitotic BCs without an apical process at the time when pre‐mitotic vsx1 + progenitors undergo apical process retraction (triangle, 60.3 ± 6.6%) and at the time when these progenitors undergo mitosis (cyan circle, 95.1 ± 1.7%). Data are presented as mean ± SEM, 17 progenitors, 10 fish.

Figure 3

Vsx1 progenitors in the unlaminated and laminated region exhibit distinct cell biological behaviors

Concurrent visualization of nucleokinesis in a vsx1 + progenitor in the laminated retina (Q26; UAS:memYFP, upper panels) and the cell cycle stage (mOrange2‐PCNA mRNA, lower panels) by in vivo time‐lapse confocal imaging. Progenitor soma is pseudo‐colored cyan in YFP channel and outlined in cyan in mOrange channel. Scale bar: 10 μm.

Nuclear movement of vsx1 + progenitors in the unlaminated region (vsx1:GFP) of the retina prior to mitosis at the apical surface (orange circle). Seven cells, five fish. Black line represents the average position of all vsx1 + progenitors prior to mitosis in unlaminated regions.

Nuclear trajectories of vsx1 + progenitors in the laminated region that undergo mitosis at the INL/OPL interface (cyan circle, left panel) and post‐mitotic BCs in their vicinity (right panel). In the time interval during which trajectories were tracked, all 12 vsx1 + progenitors move to the INL/OPL interface prior to mitosis; 8 of 12 post‐mitotic surrounding BCs do so (five fish). Black lines represent the average position of all vsx1 + progenitors prior to mitosis in laminated regions (left panel) or of the post‐mitotic BCs (right panel).

Centrosomes (centrin4‐YFP mRNA, grayscale) of a vsx1 + progenitor (magenta mask, Q19) and post‐mitotic BCs in the laminated region concurrently translocate to the OPL and remain clustered there. The centrosome of the highlighted vsx1 + progenitor is pseudo‐colored green. Scale bar: 10 μm.

Continuation of the time‐lapse from (D) (centrosomes not depicted). The vsx1 + progenitor (magenta mask) shows transient lateral arborizations in the synaptic layers as wide as its soma (see brackets at −210′, −165′). A further example can be seen in (A) (see brackets at −67′).

Figure 4

The differentiation status of late‐born bipolar cells and earlier‐born bipolar cells in their vicinity are similar

Fluorescence intensity of a non‐apically dividing vsx1 + progenitor before it undergoes mitotic division (cyan trace), at the time of mitotic division (cyan circle) and of the two BC daughter cells (green ovals) over the course of 10 h (green traces).

Quantification of the fluorescence intensity of four pairs of BC daughters (green ovals) derived from non‐apically dividing vsx1 + progenitors (cyan circle) 10 h after division. The fluorescence intensity of each of the BC pairs was normalized to the mitotic cell from which they were derived. BC pairs depicted are derived from two fish. The different overall rates of GFP fluorescence increase can be explained by where along the differentiation gradient cells originate.

Comparison of the vsx1 expression levels of apical progenitors, non‐apical progenitors, and post‐mitotic BCs derived from non‐apically dividing progenitors 10 h following division. The fluorescence expression levels of cells surrounding each of the three cell categories are also quantified. Apically dividing vsx1 + progenitors (orange circle) and surrounding cells (dull orange bar, mean, and standard deviation, SD). Non‐apically dividing vsx1 + progenitors (cyan circle) and surrounding cells (dull cyan bar, mean, and SD). In total, 43 apically dividing progenitors and 524 cells in their surround, and 43 non‐apically dividing progenitors and 521 cells in their surround from at least 13 eyes were used for this analysis. Post‐mitotic BCs (green oval) and earlier‐born surrounding cells (dull green bar, mean, and standard deviation). Eleven post‐mitotic BCs derived from seven mitotic divisions and 105 cells in their surround from three fish were used for this analysis.

Confocal in vivo time‐lapse recording of the emergence of ribeye a puncta (pseudo‐colored magenta, ctbp2:mCherry‐ctbp2) in a BC derived from a non‐apically dividing vsx1 + progenitor (pseudo‐colored cyan, Q26; UAS:memYFP), 7 h following cell cycle exit. Insets of the magenta‐boxed region (at −118′ and 0′) of the axon terminal of the BC reveal the emergence of discernible ribeye a puncta at the 0′ time point (white arrowheads). Scale bars: 10 μm.

Quantification of the time interval between mitosis and clustering of ribeye puncta (triangle) in the axon terminals of the ensuing BC daughters. Eight BCs derived from seven non‐apically dividing progenitors (cyan circles) and eight BCs derived from apically dividing vsx1 + progenitors (orange circles) from three fish were used for this analysis. In seven of the eight BCs derived from apical progenitors (depicted as small orange dots), we observed clustering but not the mitotic division, thus the time interval in these cases is an underestimate.

Quantification of the axon terminal arbor width of BCs derived from non‐apical vsx1 + progenitors and the cells in their surround. At 10 h post‐division, mean arbor width for BCs was 4.34 ± 1.05 μm, SD; the mean arbor width for cells in their surround was 4.29 ± 1.24 μm, SD. In total, nine BCs derived from seven non‐apical divisions and 56 surround cells from six fish were used for analysis.

Next, we asked whether vsx1 + progenitors undergo interkinetic nuclear migration. We used an mRNA construct encoding a fusion protein between mOrange2 and proliferating cell nuclear antigen (PCNA; Fig 3A and Appendix Fig S3) that allows determination of progenitor cell cycle phase by distinct nuclear localization patterns (Leonhardt et al, 2000; Leung et al, 2011). Similar to classical progenitors, vsx1 + progenitors in the unlaminated region exhibited rapid apically directed interkinetic nuclear migration during G2 before mitosis at the apical surface (Fig 3B and Appendix Fig S3). This movement was absent in non‐apically dividing vsx1 + progenitors, making their nuclear movements largely indistinguishable from the post‐mitotic BCs in their immediate vicinity (Fig 3A and C). However, the non‐apically dividing vsx1 + progenitors always translocated their nuclei to the INL/OPL interface prior to mitosis. This movement can be explained by the location of the centrosome. By expressing fluorescently tagged centrin4, we found that BC centrosomes actively relocated from the apical surface to the INL/OPL interface (Fig 3D and E), resulting in a dendritic, rather than a somatic location at maturity (Fig EV3). Like apical process retraction, centrosome relocation occurred at the same time for progenitors and surrounding, post‐mitotic BCs (Figs 3D and EV3). Thus, nucleokinesis and centrosome relocation represent further differentiation steps for which vsx1 progenitors are time‐locked to their post‐mitotic neighbors.

Figure EV3

Centrosomes of BCs are clustered at the OPL

In vivo confocal images of a 2 dpf retina in which centrosomes (centrin4 mRNA) and cellular membranes (BODIPY‐Texas Red) are labeled. The outer part of the INL is largely devoid of centrosomes. The emergence of the OPL (region to the right of the arrowhead) coincides with the clustering of centrosomes in this location. Solid gray line indicates the apical surface. Orange and cyan bars above the images represent unlaminated and laminated parts of the retina, respectively. Scale bar: 10 μm.

In vivo confocal images of a 2 dpf retina in which centrosomes (centrin4 mRNA) and a subset of BCs (Q19) are labeled. BC somata are devoid of centrosomes. The relocation of BC centrosomes from the apical surface (solid gray line) to the OPL coincides with the retraction of the apical process of BCs (arrowhead). Orange and cyan bars above the images represent unlaminated and laminated parts of the retina, respectively. Scale bar: 10 μm.

Confocal in vivo images of a 3 dpf retina with an isolated BC, in which the centrosome and cellular membranes are labeled with YFP and cerulean, respectively (plasmid encoding UAS:centrin4‐YFP/UAS:memCerulean was injected into Q26). The centrosome (green) is localized to the dendritic tuft at the OPL rather than in the soma. Scale bar: 5 μm.

Scheme illustrating distinct developmental steps (nucleokinesis, centrosome relocation, apical process retraction, transient arborizations of the apical process in the OPL and mitosis at the INL/OPL interface) for the vsx1 + progenitor from Fig 3D and E, observed by time‐lapse imaging. Nucleokinesis trajectories of the cell are color coded to indicate the local cytoarchitecture of the retina when these events took place (orange: unlaminated; cyan: laminated).

The post‐mitotic differentiation status of late‐born bipolar cells is similar to the early‐born bipolar cell population in their vicinity

We assessed the rate of differentiation of the BC progeny resulting from non‐apically dividing vsx1 progenitors in the laminated region of the retina. First, we monitored vsx1‐driven GFP expression levels in post‐mitotic BCs over the course of 10 h after their exit from the cell cycle and found a steady increase in fluorescence intensity (Fig 4A). Ten hours post‐division, GFP expression levels in BC sibling pairs were remarkably similar to each other (Fig 4A and B) and remained similar to the cells in their surround, which largely comprised of earlier‐born post‐mitotic BCs (Fig 4C).

Next, we used an established transgenic line, ctbp2:mCherry‐ctbp2 (Pelassa et al, 2014), to examine the emergence of ribbon synapses in BC axon terminals. We assessed the time interval between exit from the cell cycle and the appearance of ribeye a clusters (marker of ribbon synapses) as an indication of presynaptic differentiation in vsx1 BCs (Fig 4D). For late‐born BCs derived from non‐apically dividing vsx1 + progenitors, the average time interval was 6.5 h. By contrast, for early‐born BCs this interval was longer than 13 h on average (Fig 4E). Thus, late‐born BCs acquire features of presynaptic differentiation with greater speed than their earlier‐born counter‐parts. Finally, we measured the lateral extent of the terminal axonal arbors of BCs derived from non‐apical vsx1 + progenitors 10 h after they exited the cell cycle, and found this to be remarkably similar to that of the earlier‐born post‐mitotic BCs in their vicinity (Fig 4F). Taken together, late‐born BCs continue to differentiate in lock‐step with earlier‐born BCs in their vicinity, including the elaboration of markers of axonal and synaptic differentiation, thus contributing to synchrony in local neuronal development.

Experimentally delaying cell division does not delay vsx1 progenitor differentiation

Our experiments have established that along the developmental gradient, BC progenitors blend into the differentiation landscape that surrounds them with regard to their morphological, cell biological, and molecular characteristics. Moreover, the “head‐start” gained by pre‐mitotic differentiation in late‐dividing progenitors continues in their BC progeny, so that the lag in differentiation between late‐born and early‐born BCs is remarkably low. Two potential scenarios could explain these observations: (i) Multiple, “fixed” vsx1 progenitors exist, each of which undergoes mitosis at a stereotypic time point in the cell's differentiation trajectory. (ii) Alternatively, neurogenesis and differentiation could be uncoupled from each other, so that mitotic divisions could occur at various points in any given vsx1 cell's differentiation program. To distinguish between these two possibilities, we delayed BC progenitor divisions using hydroxyurea and aphidicolin (HUA). HUA treatment rapidly reduced the number of cells entering the G2/M‐phase (Appendix Fig S4), but a small number of progenitors continued to divide, albeit with a prominent delay. We could now ask whether during delayed progenitor mitosis, differentiation stalled (as implied by the existence of “fixed” progenitors) or whether it continued and remained in synchrony with surrounding post‐mitotic BC differentiation (as predicted by the “uncoupling” scenario; Fig 5A). We tracked 10 vsx1 progenitors in HUA‐treated embryos from late S‐phase until mitosis. Knowing that progenitors, in control experiments, divided on average 142 ± 3.9 min after the onset of late S‐phase, we could determine when a HUA‐treated cell should have divided (“expected” mitosis) and measured the delay with which the division actually occurred (“observed” mitosis, delay range approximately 3.5–9 h; Figs 5B and EV4). Because we could predict with a high degree of accuracy (86.4%), whether progenitors, in control conditions, would undergo mitosis at the apical surface or in the INL well before the divisions occurred (see Appendix Supplementary Materials and Methods), we could ask whether the HUA‐induced delay of mitosis would shift divisions from apical to non‐apical locations (as an “uncoupling” scenario would imply). Nine of the 10 HUA‐treated cells fulfilled criteria that identified them to be destined to divide apically. However, all 10 cells instead underwent mitosis in the INL after process remodeling, suggesting that they had been shifted from an apical to a non‐apical phenotype simply by delaying mitosis (Fig 5C). Furthermore, vsx1:GFP expression levels in the 10 delayed progenitors increased from the point of expected mitosis to observed mitosis (Fig 5D). Notably, even hours after the delayed mitosis occurred, progenitors still matched the fluorescence levels of cells in their immediate surround (Fig 5E), suggesting uncoupling of cell division and differentiation. Together, these findings support the “uncoupling” scenario laid out above and hence argue for an independence of neurogenesis and differentiation programs during BC development.

Figure 5

Neurogenesis and differentiation of vsx1 + progenitors are independent of each other

Schematic representation of expected outcomes if the immature retina (progenitors, P; neurons, N) is treated with HUA to delay the cell cycle. Upper panel: If there are multiple “fixed” progenitors, a block of cell division should stall progenitors at the differentiation state in which they normally would have undergone mitosis. The result would be a “salt‐and‐pepper” pattern of undifferentiated (light green) and differentiating progenitors (dark green). Lower panel: If cell cycle and differentiation are independent, all progenitors should homogenously differentiate. Open arrowheads indicate cytoplasmic processes not confined to the OPL and IPL, filled arrowheads indicate cytoplasmic processes confined to the synaptic layers.

Confocal images of a 2 dpf retina from a vsx1:GFP embryo injected with a p53 morpholino and mOrange2‐PCNA mRNA. A vsx1 + progenitor (dashed magenta outline) is shown before HUA treatment (left panel), at the time when it would have been “expected” to undergo mitosis (middle panel, orange diamond) and when it actually underwent mitosis (right panel, cyan circle). The retina and the vsx1 + progenitor continue to mature after the “expected” mitosis (retraction of cytoplasmic processes, mitosis at non‐apical location, and up‐regulation of GFP). Open arrowheads indicate cytoplasmic processes not confined to the OPL and IPL, filled arrowheads indicate cytoplasmic processes confined to the synaptic layers. Dotted lines indicate extent of vsx1 + cell somata across retinal thickness. Scale bar: 10 μm.

Progenitors that were expected to divide at the apical surface (exp), divided non‐apically (obs). 10 progenitors, four fish.

Quantification of vsx1:GFP fluorescence intensity of progenitors at the time when they were expected to undergo mitosis (exp, orange) and when they underwent mitotic division (obs, cyan). Data are presented as mean ± SEM, 10 progenitors, four fish. Mann–Whitney U‐test, ***P = 0.0002.

The fluorescence intensity of HUA‐treated vsx1:GFP progenitors at the time when they were expected to undergo mitosis (diamonds), and when they were observed to undergo mitosis (circles), plotted against the intensity of the surrounding cells (10 progenitors, four fish). One cell (cyan diamond) was expected to divide non‐apically. Green line indicates the fluorescence change of the lowest expressing progenitor for clarity.

Figure EV4

mOrange2‐PCNA allows for studying HUA‐induced delay of mitosis

Time‐lapse imaging of vsx1:GFP retinas expressing mOrange2‐PCNA was used to determine the time interval between the beginning of late S‐phase and the beginning of M‐phase for progenitors. In control fish the average time for this interval was 142 ± 48 min (mean ± 2 standard deviations, SD, 38 cells from six fish, orange circles represent single cells). HUA‐treated progenitors were significantly delayed compared to controls, with an average time interval of 534 ± 32 min (10 cells from four fish, P < 0.0001, Mann–Whitney U‐test, cyan circles). The delay with which each of the 10 HUA‐treated progenitors reached M‐phase was calculated by subtracting the average “expected” time interval between S‐ and M‐phase (obtained from control cells, orange diamond) from the observed time interval between S‐ and M‐phase for each HUA‐treated progenitor (cyan circles). An example of this calculation for the cell tracked in (B) is shown.

Time‐lapse confocal images of a HUA‐treated vsx1:GFP retina expressing mOrange2‐PCNA. Based on control experiments, the outlined cell (for which the delay in mitosis is depicted in A, cyan cell with a pink outline) would have been “expected” to undergo mitotic division around the 148 min time point (orange diamond; closest to the calculated mean “expected” time of 142 min), but entered M‐phase at 693 min (cyan circle). Scale bar: 10 μm.

Discussion

Our study is the first, to our knowledge, to address the relative timing of neurogenesis, migration, and differentiation for a molecularly defined CNS population in vivo and to elucidate the effects of such timing on progenitor characteristics. In contrast to the widely held view, we found that developing neurons did not adopt a stereotypic sequence of neurogenesis followed by migration and subsequent differentiation. Rather, to our surprise, the developmental trajectories that progenitors adopted were variable and accommodated remarkable flexibility. This resulted in vsx1 + BC progenitors with a wide variety of molecular, morphological, and cell biological characteristics. Importantly, rather than representing many distinct populations, the differentiation status of these progenitors formed a continuum in lock‐step with the differentiation of surrounding post‐mitotic BCs along the developmental gradient of the retina. Accordingly, cells dividing in the laminated parts of the retina were more BC‐like than their early dividing counter‐parts. Indeed, without the aid of time‐lapse imaging or cell cycle markers it would have been impossible to distinguish between pre‐mitotic progenitors and post‐mitotic cells. Our results support the conclusion that a stereotypical and fixed sequence of ontogenetic events is not essential during neuronal development. Thus, at least for terminally dividing progenitors, which generate a substantial part of the CNS neuronal population (Nakashima et al, 2015), mitosis does not have to occur before neuronal differentiation is initiated (or at any specific step thereafter), but rather can be flexibly intercalated between other developmental steps.

Our findings could explain two previous intriguing observations: First, that neuronal progenitors in different parts of the nervous system show precocious signs of differentiation prior to cell cycle exit (Rothman et al, 1980; Rohrer & Thoenen, 1987; DiCicco‐Bloom et al, 1990; Miyata et al, 2004; Godinho et al, 2007; Attardo et al, 2008; Prasov & Glaser, 2012); and second, that blocking mitosis does not halt neuronal differentiation in many parts of the Xenopus CNS including the retina, spinal cord, and brain stem (Harris & Hartenstein, 1991). Our results now offer a unifying explanation for these previous observations, suggesting that they might simply result from a fundamental uncoupling of cell cycle and neuronal differentiation during normal development. Indeed, the increasingly recognized prevalence of precocious progenitors in many parts of the nervous system of a range of species [e.g., basal progenitors in the neocortex (Haubensak et al, 2004; Miyata et al, 2004; Noctor et al, 2004); neural crest‐derived PNS progenitors (Rothman et al, 1980; Rohrer & Thoenen, 1987; DiCicco‐Bloom et al, 1990); retinal progenitors (Godinho et al, 2007; Prasov & Glaser, 2012)], implies that such uncoupling may be a general principle of neural development.

What would be the advantages of uncoupling neurogenesis and neuronal differentiation? By comparison with the orderly sequence of developmental events in the classical model, we suggest that the uncoupled model provides two advantages, namely speed and synchrony (Fig EV5). In the classical model, because differentiation can only be initiated following mitosis, the time required for the majority of cells in a defined population to differentiate is dictated by the delay in their birth dates. By contrast, in the uncoupled model, differentiation steps can already occur at the progenitor stage, permitting differentiation across the population to be faster. Furthermore, despite the extended time span over which mitotic divisions can occur, cells in a given population differentiate in relative synchrony. The uncoupling of mitosis and differentiation could thus be particularly pertinent for the assembly of essential neural circuits where swift maturation is paramount for survival (Nikolaou & Meyer, 2015).

Figure EV5

Schematic comparison of “classical” and “uncoupled” models of neuronal development

An idealized representation of the developmental trajectories of four progenitors tracked over time until they differentiate and reach maturity. The period of neurogenesis is the same for cells in the “classical” and “uncoupled” models. As compared to the “classical” model in which differentiation is initiated only after cell cycle exit, “uncoupling” neurogenesis and differentiation permits a cell population to reach maturity faster and in relative synchrony.

Materials and Methods

Animals

All experiments were performed according to regulations as approved by the local regulatory bodies. Zebrafish were maintained, mated, and raised as described in Mullins et al (1994). Embryos were kept in 0.3× Danieau's solution at 28.5°C and staged as previously described (Kimmel et al, 1995). Fish were either in an AB wild‐type or roy orbison (Ren et al, 2002) background. The transgenic lines used are listed in Appendix Table S1. We generated Tg(vsx1:Gal4)q26 (Q26) and Tg(14xUAS:memTagRFP‐T) by Tol2 mediated insertion (Kawakami, 2004). For details of constructs used to make transgenic fish and for transient injections see Appendix Supplementary Materials and Methods.

mRNA synthesis and injection

Plasmids were linearized (PCNA: NotI, centrin4: ApaI). Capped mRNA was produced using the Ambion mMESSAGE mMACHINE kit (Applied Biosystems) according to the manufacturer's instructions. mRNA was injected at 100 ng/μl into one‐ or two‐cell stage embryos.

Immunohistochemistry and in situ hybridization

Immunostaining to detect pH3, Crx, and GFP was performed using adaptations of previously published protocols either on cryosections (Williams et al, 2010) or on whole‐mount embryos (Hunter et al, 2011). In situ hybridization to visualize ribeye a mRNA was performed on whole‐mount embryos using a digoxigenin‐labeled riboprobe and Fast Red TR/naphthol AS‐MX (Sigma) to detect alkaline phosphatase activity. For a detailed description see Appendix Supplementary Materials and Methods.

In vivo imaging

Embryos were prepared for imaging as described previously (Godinho, 2011; Engerer et al, 2016). Between 10 and 18 h post‐fertilization (hpf), embryos were transferred to 0.3× Danieau's solution containing 0.003% 1‐phenyl‐2‐thiourea (PTU, Sigma) to inhibit melanin formation. At 2.25 days post‐fertilization (dpf), manually dechorionated embryos were anesthetized using 0.02% tricaine (PharmaQ) in medium containing PTU and embedded laying on their side in low‐melting agarose (0.7–0.8%, Sigma).

Fish were imaged starting at 2 dpf on an Olympus FV1000 confocal/2‐photon and an Olympus FVMPE‐RS 2‐photon microscope using water‐immersion objectives (Olympus 20×/NA 0.95, Olympus 25×/NA 1.05, Zeiss 40×/NA 1.0, Nikon 25×/NA 1.1, and Nikon 40×/NA 0.8) or a silicon‐immersion objective (Olympus 30×/NA 1.05). Embryos were maintained at 28.5°C during all in vivo recordings. At each time point z‐stacks were acquired of the peripheral retina, encompassing its entire circumference.

Image processing

Images were viewed and processed using open‐source ImageJ/Fiji software (http://fiji.sc). The StackReg function was used for drift compensation in xy. Image panels were assembled in Photoshop CS5 (Adobe) and combined into figures using Illustrator CS5 (Adobe). The “Gaussian blur” function was used to filter noise for clarity. Gamma was not adjusted.

Hydroxyurea‐aphidicolin treatment

Hydroxyurea (Sigma) and aphidicolin (BioViotica) were used at a final concentration of 20 mM and 150 μM, respectively, in 0.3× Danieau's containing 1.0–1.7% DMSO. Embryos were injected with a p53 morpholino (0.5–1 mM, Gene Tools) at the one‐ or two‐cell stage to ameliorate HUA‐induced apoptosis (Girdler et al, 2013). At 2 dpf embryos were mounted in agarose as described for in vivo imaging above, but leaving the tail fin un‐embedded for better drug access. Retinas were imaged for one time point prior to HUA administration. The 0.3× Danieau's medium was replaced with HUA‐containing medium, and time‐lapse recording was immediately resumed. Recordings were generally limited to < 16 h after HUA addition as high levels of cell death were observed thereafter.

Data analysis

For details, see Appendix Supplementary Materials and Methods.

Statistics

Mean values and standard error of the mean (SEM) were calculated using Microsoft Excel. We used the Mann–Whitney U‐test to compare datasets with GraphPad Prism 5. Data are presented as mean ± SEM unless indicated otherwise. P‐values < 0.05 are denoted with “*”, P < 0.01 with “**”, and P < 0.001 with “***”.

Author contributions

PE, PRW, TM, and LG conceived of the project and designed experiments. TM and LG supervised the project. PE performed all in vivo imaging experiments, with some help from NO. PE and PC performed in situ hybridization. PE, SCS, TY, NO, and BO generated new constructs and transgenic lines for this research. PE, PRW, TM, and LG wrote the paper with input from all authors.

Conflict of interest

The authors declare that they have no conflict of interest.

Appendix

Click here for additional data file.

Expanded View Figures PDF

Click here for additional data file.

Review Process File

Click here for additional data file.