The interplay of atoh1 genes in the lower rhombic lip during hindbrain morphogenesis

The Lower Rhombic Lip (LRL) is a transient neuroepithelial structure of the dorsal hindbrain, which expands from r2 to r7, and gives rise to deep nuclei of the brainstem, such as the vestibular and auditory nuclei and most posteriorly the precerebellar nuclei. Although there is information about the contribution of specific proneural-progenitor populations to specific deep nuclei, and the distinct rhombomeric contribution, little is known about how progenitor cells from the LRL behave during neurogenesis and how their transition into differentiation is regulated. In this work, we investigated the atoh1 gene regulatory network operating in the specification of LRL cells, and the kinetics of cell proliferation and behavior of atoh1a-derivatives by using complementary strategies in the zebrafish embryo. We unveiled that atoh1a is necessary and sufficient for specification of LRL cells by activating atoh1b, which worked as a differentiation gene to transition progenitor cells towards neuron differentiation in a Notch-dependent manner. This cell state transition involved the release of atoh1a-derivatives from the LRL: atoh1a progenitors contributed first to atoh1b cells, which are committed non-proliferative precursors, and to the lhx2b-neuronal lineage as demonstrated by cell fate studies and functional analyses. Using in vivo cell lineage approaches we revealed that the proliferative cell capacity, as well as the mode of division, relied on the position of the atoh1a progenitors within the dorsoventral axis. We showed that atoh1a may behave as the cell fate selector gene, whereas atoh1b functions as a neuronal differentiation gene, contributing to the lhx2b neuronal population. atoh1a-progenitor cell dynamics (cell proliferation, cell differentiation, and neuronal migration) relies on their position, demonstrating the challenges that progenitor cells face in computing positional information from a dynamic two-dimensional grid in order to generate the stereotyped neuronal structures in the embryonic hindbrain.


Introduction
The assembly of functional neural circuits requires the specification of neuronal identities and the execution of developmental programs that establish precise neural network wiring. The generation of such cell diversity happens during embryogenesis, at the same time that the brain undergoes a dramatic transformation from a simple tubular structure, the neural tube, to PLOS ONE | https://doi.org/10.1371/journal.pone.0228225 February 3, 2020 1 / 23 a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 a highly convoluted structure-the brain-, resulting in changes in the position of neuronal progenitors and their derivatives upon time. Thus, the coordination of progenitor proliferation and cell fate specification is central to tissue growth and maintenance. The comprehension of how neuronal heterogeneity is achieved implies the understanding of how the neurogenic capacity is acquired, how the number of progenitors vs. differentiated neurons is balanced, and how their relative spatial distribution changes upon morphogenesis. Neurogenesis is initiated by proneural genes, which trigger the specification of neuronal lineages and commit progenitors to neuronal differentiation by promoting cell cycle exit and activating a downstream cascade of differentiation genes [1]. Once neuronal progenitors are committed, the first step towards achieving the diversity observed in adults occurs early in development with the division of neuronal progenitor cells into distinct domains along dorsoventral (DV) axis, which will give rise to different types of neurons in response to morphogen signals emanating from local organizing centers [2]. The next level of complexity arises with the interpretation of the two-dimensional grid, along the DV and anteroposterior (AP) axes, of molecularly distinct progenitor regions that will control the final neuronal fate.
The hindbrain undergoes a segmentation process along the AP axis leading to the formation of seven metameres named rhombomeres (r1-r7) that constitute developmental units of gene expression and cell lineage compartments [3][4][5]. This compartmentalization involves the formation of a cellular interface between segments called the hindbrain boundary [6], which exhibit distinct features such as specific gene expression [7] and biological functions [8][9][10][11]. The hindbrain is the most conserved brain vesicle along evolution [12,13], and in all vertebrates the dorsal part of the hindbrain gives rise to a transient neuroepithelial structure, the rhombic lip (RL). RL progenitors will generate different neuronal lineages according to their position along the AP axis. The most anterior region of the RL, which coincides with the dorsal pole of r1, is known as Upper Rhombic Lip (URL) and produces all granule cells of the external and internal granular layers of the cerebellum [14,15]. The rest of the RL, which expands from r2 to r7, is known as Lower Rhombic Lip (LRL) and gives rise to deep nuclei of the brainstem, such as the vestibular and auditory nuclei and most posteriorly the precerebellar nuclei [16,17]. The genetic program for cerebellum development is largely conserved among vertebrates [16]; as an example, zebrafish and mouse use similar mechanisms to control cerebellar neurogenesis with a crucial role of atoh1 and ptf1 genes [17,18]. For the LRL, we know both the contribution of ptf1a/atoh1a proneural progenitor populations to specific deep nuclei [19], and the distinct rhombomeric identity [20]. However, little is known about how progenitor cells from the LRL behave during neurogenesis and how their transition into differentiation is regulated, in order to balance the rate of differentiation and proliferation to produce the proper neuronal numbers.
In this work, we sought to understand the role of atoh1 genes in the generation of the neuronal derivatives of LRL. We used complementary strategies in the zebrafish embryos to provide information about the gene regulatory network operating in the specification of LRL cells, and the kinetics of cell proliferation and behavior of atoh1a-derivatives. We unveiled that atoh1a is necessary and sufficient for specification of LRL cells by activating atoh1b, which worked as a differentiation gene to transition progenitor cells towards neuronal differentiation in a Notch-dependent manner. This cell state transition involved the release of atoh1a-derivatives from the LRL: atoh1a progenitors contributed first to atoh1b cells, which are committed non-proliferative precursors, and to the lhx2b-neuronal lineage as demonstrated by cell fate studies and functional analyses. Using in vivo cell lineage approaches we showed that the proliferative cell as well as their mode of division, relied on the position of the atoh1a progenitors within the dorsoventral axis.

Zebrafish lines and genotyping
Zebrafish (Dario rerio) were treated according to the Spanish/European regulations for the handling of animals in research. All protocols were approved by the Institutional Animal Care and Use Ethic Committee (Comitè Etica en Experimentació Animal, PRBB) and the Generalitat of Catalonia (Departament de Territori i Sostenibilitat), and they were implemented according to European regulations. Experiments were carried out in accordance with the principles of the 3Rs. Embryos were obtained by mating of adult fish using standard methods. All zebrafish strains were maintained individually as inbred lines. The transgenic line Mu4127 carries the KalTA4-UAS-mCherry cassette into the 1.5Kb region downstream of egr2a/krx20 gene, and was used for targeting UAS-constructs to rhombomeres 3 and 5, or as landmark of these regions [21]. Tg[ßactin:HRAS-EGFP] line, called Tg[CAAX:GFP] in the manuscript, displays GFP in the plasma membrane and was used to label the cell contours [22]. Tg[tp1:d2GFP] line is a readout of cells displaying Notch-activity [23] in which cells with active Notch express GFP. The Tg[HuC:GFP] line labels differentiated neurons [24]. atoh1a fh282 mutant line in the Tg[atoh1a:GFP] background, which carried a missense mutation within the DNA-binding domain, was previously described in [18]. Embryos were phenotyped blind and later genotyped by PCR using the following primers: Fw primer 5 0 -ATGGA TGGAATGAGCACGGA-3' and Rv primer 5 0 -GTCGTTGTCAAAGGCTGGGA-3'. Amplified PCR products underwent digestion with AvaI (New England Biolabs), which generated two bands: 195 bp + 180 bp for the WT allele and 195 bp + 258 bp for the mutant allele. Since the atoh1a fh282 mutant allele only caused a deleterious phenotype in homozygosity, wild type and heterozygous conditions showed identical phenotypes and they were displayed in all our experiments as a single wild type condition. Invitrogen). Either Draq5 TM (1:2000; Biostatus, DR50200) or DAPI were used to label nuclei. After staining, embryos were either flat-mounted and imaged under a Leica DM6000B fluorescence microscope, or whole-mounted in agarose and imaged under a SP8 Leica confocal microscope.

BrdU staining and TUNEL analysis
Cells in S-phase were detected by BrdU-incorporation (Roche). Briefly, embryos were dechorionated and incubated in 10mM BrdU diluted in 5%DMSO 30min at RT. Embryos were washed with fresh water, fixed in 4%PFA at RT, and dehydrated in MetOH. After progressive rehydration, embryos were permeabilized with Proteinase K (Invitrogen) at 10 μg/ml 15min at RT, fixed 20min in 4%PFA, and washed 3x10min in PBS before immunostaining with anti-BrdU (1:50, Becton Dickinson).
Distribution of apoptotic cells was determined by TdT-mediated dUTP nick-end labeling of the fragmented DNA (TUNEL, Roche). Briefly, whole embryos at 30hpf were fixed in 4% PFA and dehydrated in 100% MetOH were permeabilized with Proteinase K at 25 μg/ml, and preincubated with TUNEL mixture during 60 min at 37˚C according to the manufacturer's instructions. DAPI (1:500; Molecular Probes) was used to label nuclei.

Quantification of the phenotypes
For quantifying the number of differentiated neurons in atoh1a WT Tg[atoh1a:GFP] and ato-h1a fh282 Tg[atoh1a:GFP] embryos, confocal MIP of ventral stacks were used and all cells present in the r4/r5 and r5/r6 domain were counted (see Table 1 for numbers and statistics).
In order to quantify the number of proliferating LRL-cells in atoh1a WT and atoh1a fh282 embryos in the Tg[atoh1a:GFP] background, the number of mitotic figures within the atoh1a: GFP progenitor domain was assessed (see Table 2 for numbers and statistics).
For the quantification of the total number of LRL atoh1a:cells in atoh1a WT and atoh1a fh282 embryos in the Tg[atoh1a:GFP] background, embryos at 24hpf were stained with Draq5 and the total number of nuclei of atoh1a:GFP cells was assessed in r5 (see Table 2 for numbers and statistics).
For the quantification of the delamination time of atoh1a:cells in atoh1a WT and atoh1a fh282 embryos in the Tg[atoh1a:GFP] background, we kept track of the time of division of a given cell (t0) and the time of delamination of the resulting cells (tf) and calculated the difference between tf and t0. Table 1. Quantification of differentiated cells in atoh1a WT and atoh1a fh282 embryos at 24hpf and 36hpf with the t-test values (Fig 4M and 4N

Conditional overexpression
The full-length coding sequences of zebrafish atoh1a-and atoh1b [27] were cloned into the MCS of a custom dual vector that expressed Citrine from one side of 5xUAS sequence and the cDNA of interest from the opposite side [32]. Mu4127 embryos (expressing KalT4 in r3 and r5) were injected either with H2B-citrine:UAS, H2B-citrine:UAS:atoh1a or H2B-citrine:UAS: atoh1b constructs at the one-cell stage, grown at 28.5˚C and analyzed at 24hpf for atoh1a/b and lhxb2 in situ hybridization and Citrine expression.

Pharmacological treatments
atoh1a WT Tg[atoh1a:GFP] and atoh1a fh282 Tg[atoh1a:GFP] sibling embryos were treated either with 10 μM of the gamma-secretase inhibitor LY411575 (Stemgent) or DMSO for control. The treatment was applied into the swimming water at 28.5˚C from 24hpf to 30hpf. After treatment, embryos were fixed in 4%PFA for further analysis.

Expression of proneural genes within the zebrafish hindbrain
We first analyzed the formation of molecularly distinct neural progenitor domains, each of them able to generate particular neuronal cell types, during hindbrain embryonic development. We performed a comprehensive spatiotemporal analysis of the expression of distinct proneural genes along the anteroposterior (AP) and dorsoventral (DV) axes within the hindbrain and defined the DV order of proneural gene expression. The expression profiles of atoh1a, ptf1a, ascl1a, ascl1b, and neurog1 indicated that their onset of expression differed along the AP axis (S1 Fig). The dorsal most progenitor cells expressed atoh1a all along the AP axis from 18hpf onwards, which remained expressed there until at least 48hpf (S1A-S1C Fig;  Fig 1A-1E). ptf1a expression started in rhombomere 3 (r3) at 18hpf and from 21hpf onwards it expanded anteriorly towards r1 and r2 (S1D and S1E Fig), ending up expressed all along the AP axis of the hindbrain with different intensities (S1F Fig; [17]). These two proneural genes were the most dorsally expressed as shown by transverse sections (S1A'-S1F' Fig). ascl1a and ascl1b displayed overlapping expression profiles along the AP axis in a rhombomeric restricted manner with slightly different intensities (S1G and S1J Fig). Nevertheless, their DV expression differed: ascl1a expression was adjacently dorsal to ascl1b and constituted a smaller territory (S1G'-S1I', S1J'-S1L' and S1R Fig). Indeed, ascl1a and ptf1a mainly overlapped along the DV axis occupying the region in between atoh1a and ascl1b (S1P-S1R Fig). Although by 24hpf ascl1a-cells seemed to be more laterally located than ascl1b-cells (compare S1I with S1L Fig), this just reflected the lateral displacement of the dorsal part of the neural tube upon hindbrain ventricle opening: the hindbrain at early stages was a closed neural tube resembling the spinal cord (S1 Fig, 18-21hpf stages), whereas at late stages all progenitor cells were in the ventricular zone facing the brain ventricle after lumen expansion (S1C Fig, 24hpf; compare S2A', S2B', S2E' and S2F' with S2C', S2D', S2G' and S2H' Fig). At 24hpf, ascl1a/b expression was restricted to rhombomeres, and by 42hpf their expression was clearly confined to the rhombomeric domains that flank the hindbrain boundaries (S2A-S2D Fig) as previously shown in [32,33]. Finally, neurog1 was expressed in a more ventral position (S1M-S1O and S1M'-S1O' Fig atoh1a and atoh1b were sequentially expressed in partially overlapping domains The three atoh1 paralogs -atoh1a, atoh1b and atoh1c-were shown to be expressed within the hindbrain and to contribute to the development of the cerebellum, with the expression of atoh1c restricted to the upper rhombic lip [17,18]. Since our main interest was understanding the development of the lower rhombic lip (LRL), we focused on the study of atoh1a and atoh1b and compared their onset of expression. atoh1a preceded the expression of atoh1b in the most dorsal progenitor cells of the hindbrain at 14hpf (Fig 1A and 1A'). This was in contrast with the onset in the otic epithelium, where atoh1b was expressed earlier than atoh1a (see magenta in the otic placode in Fig 1A; [27]). At 18hpf, atoh1a expression remained in the dorsal most cells, whereas atoh1b expression domain was more lateral, overlapping with atoh1acells and mostly contained within this expression domain ( Fig 1B, 1B', 1C and 1C'). Upon the opening of the neural tube, the atoh1a/b domains were laterally displaced and atoh1a remained medial whereas atoh1b positioned lateral ( Fig 1D and 1D'), and by 42hpf -when the fourth ventricle was already formed-atoh1b expression was completely lateral, and atoh1a remained dorsal and medial ( Fig 1E and 1E'). Thus, atoh1a and atoh1b were dorsally expressed but they differed in their mediolateral (apicobasal) position. To demonstrate that they were kept as progenitor cells, we stained Tg[HuC:GFP] embryos with atoh1a/b and observed that neither atoh1a nor atoh1b were expressed in differentiated neurons ( Fig 1F-H and 1F'-1H'). Their differential apicobasal distribution and the fact that progenitor cell divisions always happened in the apical domains, suggested that atoh1b-progenitor cells might have experienced a basal displacement of their cell body before undergoing differentiation. To demonstrate this, we stained embryos with atoh1a/b and anti-pH3, a marker for mitotic figures, and observed that more atoh1a than atoh1b cells seemed to undergo mitosis ( Fig 1I, 1I', 1J and 1J'). In this same line, analyses of single mitotic cells in the transgenic Tg[atoh1a:GFP] fish line that labeled atoh1a-expressing cells and their derivatives [18], showed that mitotic atoh1a:GFP cells were always located in the ventricular domain ( Fig 1K-1K"; see black asterisks in Fig 1K' and 1K"), whereas the ones that did not divide were laterally displaced just above the neuronal differentiation domain (see white asterisks in Fig 1K' and 1K") as atoh1b cells. To demonstrate that indeed basal atoh1b did not proliferate, embryos were incubated with BrdU and assayed for atoh1b expression (Fig 1L-1L"). We observed that indeed atoh1b cells did not incorporate BrdU, and therefore did not undergo S-phase (see white asterisks in Fig 1L-1L"). Thus, atoh1b cells may derive from atoh1a progenitors that diminished their proliferative capacity and behaved as committed progenitors transitioning towards differentiation.
atoh1a progenitors gave rise to atoh1b cells and lhx2b neurons Next, we sought to unravel whether atoh1b cells derived from atoh1a progenitors and to which neuronal derivatives the atoh1a progenitors gave rise. For this we used the same Tg[atoh1a: GFP] fish line than before [18], which allows to label the cell derivatives of atoh1a progenitors due the stability of GFP, and combined in situ hybridization experiments with immunostaining, using atoh1 probes and specific neuronal differentiation genes such as lhx2b, lhx1a, and pan-neuronal differentiation markers such as HuC (Fig 2; S3 Fig). Although neuronal progenitors expressing atoh1a were restricted to the dorsal most region of the hindbrain, their derivatives were allocated in more ventral domains already at early stages of neuronal differentiation (Fig 2A and 2A', compare magenta and green domains). atoh1b cells, located more laterally than atoh1a cells, expressed GFP ( Fig 2B and 2B', see white arrowhead in B' pointing to magenta/white cells in the green territory) indicating that indeed, they derived from atoh1a progenitors and according to their position they were transitioning towards differentiation. At this stage in which neuronal differentiation just started, ventral atoh1a derivatives constituted a lateral subgroup of differentiated neurons expressing the terminal factor lhx2b (see white asterisks indicating magenta/white cells in Fig 2C and 2C'). Note that the more medial lhx2b neurons in r4 did not arise from atoh1a cells (

Reconstruction of the atoh1a lineage
Next question was to address how the rate of differentiation and proliferation of atoh1a cells was balanced to achieve the needed cell diversity. For this, we used genetic lineages that allowed to delineate cell types arising from atoh1a subsets. To trace the atoh1a neuronal lineages we used a transgenic line that expressed the H2A-mCherry fluorescent reporter protein under the control of enhancer elements of the atoh1a. Tg[atoh1a:H2A-mCherry] fish were crossed with Tg[CAAX:GFP] -to have the contour of the cells-and embryos at 24hpf were imaged over 14h. Information about plasma membrane, cell fate and position was simultaneously recorded every 7min (Fig 3A as an example). We monitored the atoh1a progenies and studied their behavior according to their position along the DV axis to (Fig 3B-3E). We tracked 40 atoh1a-cells, 22 dorsal most (see cells encircled in orange in Fig 3B) and 20 adjacently ventral (see cells encircled in white in Fig 3C), and analyzed their trajectories, when and how many times they divided during the 14h that they were imaged (Fig 3D), and by which mode of division they did so (Fig 3E) attending to their morphology and location: symmetrically giving rise to two progenitor cells (PP) or two neurons (NN), or asymmetrically generating one progenitor cell and one neuron (NP). Of the 22 tracked dorsal most cells (Fig 3B and  3D), only 59% of them divided, and they did so only once (Fig 3D,

atoh1a is necessary and sufficient for neuronal specification
Our observations suggested that proliferating atoh1a progenitors gave rise to post-mitotic atoh1b precursors and lhx2b neurons in a sequential manner. However, in order to elucidate the hierarchy between these factors and cellular types, we analyzed the effect of atoh1a mutation on the neuronal differentiation domain (Fig 4). We made use of the available atoh1a fh282 mutant fish in the Tg[atoh1a:GFP] background, which carried a missense mutation within the DNA-binding domain [18]. First, we observed that mutation of atoh1a resulted in a complete loss of atoh1b expression within the hindbrain (Fig 4A, 4A', 4D, 4D', 4G, 4G', 4J and 4J'), suggesting that atoh1a was necessary for atoh1b expression and supporting the previous result that atoh1b cells derived from atoh1a progenitors. This phenotype was accompanied with the loss of the most lateral lhx2b-neuronal population (see white asterisk in Fig 4B, 4B', 4E, 4E', 4H, 4H', 4K and 4K'), but not of the lhx2b-medial column in r4 that remained unaffected (see white arrowhead in Fig 4B, 4B', 4E, 4E', 4H, 4H', 4K and 4K'), as it was anticipated since this specific population of lhx2b neurons did not derive from the atoh1a cells (Fig 2D). Although the overall pattern of neuronal atoh1a:GFP cells was not dramatically changed ( Fig 4C, 4C', 4F, 4F', 4I, 4I', 4L and 4L'), when the number of neurons at different AP positions was assessed we could observe a clear decrease in the number of differentiated atoh1a neurons in the ato-h1a fh282 mutant embryos at both the onset and progression of neuronal differentiation ( Fig  4M and 4N, quantification of green dashed inserts in Fig 4C, 4F, 4I and 4L; Table 1).
To address the possibility that the decrease in the number of neurons in atoh1a fh282 mutants was the result of a smaller number of atoh1a progenitor cells, we quantified the number of LRL atoh1a:GFP cells undergoing mitosis (Fig 5A), and the overall number of atoh1a:GFP cells (Fig 5B), both in atoh1a WT and atoh1a fh282 embryos. No significative differences were observed, suggesting that loss of atoh1a function did not affect the original number of LRL progenitors (Fig 5A; LRL atoh1a:GFP cells displaying PH3-staining: atoh1a WT 17.9 ± 3.6 cells n = 15 vs. atoh1a fh282 15.9 ± 3.1 cells, n = 8; Fig 5B; total atoh1a:GFP cells: atoh1a WT 69.5 ± 6.4 cells n = 15 vs. atoh1a fh282 68.4 ± 7.5 cells, n = 8; see Table 2). Next, we investigated whether atoh1a mutation resulted in an increase of apoptotic cells by TUNEL assay (Fig 5C and 5D). The pattern of cell death was the same sparse staining in the wild type and atoh1a fh282 sibling  Table 2), suggesting that mutation of atoh1a did not result in a substantial increase of apoptosis. Since the domains of neural bHLH gene expression are established and/or maintained by cross-repression resulting in the control of specific neuronal populations [1], we sought whether this neuronal loss was due to a change in cell fate rather than to a reduction of the number of progenitor cells. Thus, we analyzed proneural gene expression changes both in wild type and mutant context (Fig 5E-5J; atoh1a WT n = 8, ato-h1a fh282 n = 10). We observed that upon atoh1a mutation, atoh1a expression dramatically increased as previously reported [18] (compare Fig 5E and 5E' with 5H and 5H') and the GFPexpressing progenitor cells did not die (Fig 5F-5F", 5I and 5I'). In addition, these cells remained in an intermediate domain since they did not completely migrate towards their final ventral destination as they did in atoh1a WT embryos (compare Fig 5F' and 5I'; see white arrow in Fig 5H'-5J'). When we analyzed their possible cell fate switch, by assessing whether the GFP-expressing progenitor cells in the mutant context acquired the expression of the adjacent proneural gene ptf1a, atoh1a:GFP progenitors in the atoh1a fh282 embryos did not display ptf1a expression (compare Fig 5G, 5G', 5J and 5J', see white arrow in J'). These observations indicated that in the absence of atoh1a function cells remained as post-mitotic but undifferentiated progenitors, and the LRL domain was properly specified since no changes in the number of cells was observed. Loss of atoh1a function resulted in accumulation of atoh1a:GFP progenitors unable to migrate and finally differentiate. In order to demonstrate that these committed precursors arrested, we performed high-resolution time-lapse imaging of both atoh1a WT and atoh1a fh282 Note that in wild type embryos, the cell delaminates and migrates towards ventral, allocating in the corresponding neuronal differentiation zone (see the first three dorsal frames and then the following ventral ones), whereas in atoh1a fh282 embryos the indicated cell remains within the dorsal epithelium (see that there are four dorsal and two medial frames because the cell never reaches ventral). M) Box-plot indicating the time of delamination from the LRL of atoh1a:GFP cells in atoh1a WT and atoh1a fh282 embryos. Note that cells from wild type embryos exit the LRL much earlier than the cells from mutant siblings. Since the atoh1a fh282 mutant allele only caused a deleterious phenotype in homozygosity, wild type and heterozygous conditions showed identical phenotypes and they were displayed as single wild type condition. nt; neural tube lumen; ov, otic vesicle. Scale bars correspond to 50 μm. ns, non-statistically significant; ��� p<0.001.
https://doi.org/10.1371/journal.pone.0228225.g005 embryos from 24hpf onwards and followed the birth and migration of these atoh1a:GFP progenitors (Fig 5K and 5L). Before migrating, atoh1a progenitors in the wild type context, extended their apical and basal feet along the mediolateral axis of the neuroepithelium (dorsal stacks in Fig 5K; white asterisk indicating the tracked cell), and then moved away from the dorsal epithelium towards the mantle zone where they resided as differentiated neurons (see ventral stacks in Fig 5K; white asterisk indicating the tracked cell). This transition was accomplished in an average period of 4.5h (Fig 5K and 5M; t = 275min ± 102; n = 28 tracked cells). In contrast, atoh1a fh282 progenitors failed to transition and detach (see dorsal stacks in Fig 5L; white asterisk indicating the tracked cell) to barely migrate basally (see medial stacks in Fig 5L; white asterisk indicating the tracked cell). Indeed, after 9.5h of imaging most of atoh1a fh282 cells still remained in the dorsomedial epithelial region (Fig 5L and 5M; t = 569min ± 180; n = 9/12 tracked cells). Thus, our observations revealed that atoh1a was necessary for initial steps of neuronal differentiation (apical abscission and migration).
To further demonstrate the requirement of atoh1a in atoh1b expression and lhx2b neuronal differentiation, and to better dissect the proneural gene hierarchy, we performed conditional gain of function experiments. We injected Mu4127 embryos expressing Gal4 in r3 and r5 with H2B-citrine:UAS vectors carrying either atoh1a or atoh1b genes, and analyzed the effects in atoh1 genes and lhx2b neurons (Fig 6, Table 3). The atoh1a transgene proved successful, as atoh1a expression was spread along the DV axis, where it induced the expression of atoh1b (compare Fig 6A', 6B', 6D' and 6E') as well as ectopic lhx2b neurons in r5 (compare Fig 6C' and 6F'), a rhombomere usually devoid of these neurons at this stage. This was a cell autonomous effect, since all cells expressing atoh1b or lhx2b ectopically expressed Citrine, and therefore atoh1a (compare green cells in Fig 6E-6H with magenta cells in E'-H'). On the other hand, although atoh1b expression resulted in ectopic lhx2b induction (Fig 6H' and 6I') it did not activate atoh1a expression (Fig 5G'), demonstrating that atoh1b and atoh1a were not interchangeable, and atoh1a was upstream atoh1b. Overall, our results proved that atoh1a progenitors activated atoh1b, which allowed them to transition towards differentiation and contribute to the lhx2b neuronal population. Moreover, these experiments demonstrated the neurogenic potential of atoh1b, and importantly, its role in assigning a neuronal identity subtype.

Notch-signaling regulates the transition of atoh1a cycling progenitors towards atoh1b committed cells
We showed that atoh1a cycling cells gave rise to atoh1b post-mitotic committed precursors. Since this commitment is suspected to be irreversible and leading towards neuronal differentiation, we thought the Notch signaling pathway as a reasonable candidate to be regulating this transition. Thus, we explored the Notch activity within the LRL to understand how atoh1b expression was restricted to a given atoh1a-domain in the neural tube. First, we assessed Notch activity by the use of the Tg[tp1:d2GFP] transgenic line, which is a readout of Notchactive cells [23]. Indeed, Notch-activity was restricted to the most dorsomedial atoh1a cell population (Fig 7A and 7A'), whereas the more laterally located atoh1b cells were devoid of it ( Fig  7B and 7B'). This suggested that Notch activity was responsible of preventing atoh1a progenitor cells to transition to atoh1b and therefore modulating neuronal differentiation. To demonstrate this, we conditionally inhibited Notch activity by incubating Tg[atoh1a:GFP] embryos with the gamma-secretase inhibitor LY411575, and asked whether atoh1a/b expression domains were altered. Upon inhibition of Notch activity, there was an increase of atoh1bexpression at expense of atoh1a (Fig 7C, 7D, 7F and 7G): atoh1b expression was expanded more medially, and atoh1a expression dramatically decreased (compare the border of the atoh1b expression in Fig 7D' with 7G'). As expected, the atoh1b cells did not arise de novo but derived from atoh1a:GFP progenitors (Fig 7E, 7E', 7H and 7H'), supporting the hypothesis that Notch-pathway regulated either the transition from neural stem cells to neuronal progenitors, or the transition of atoh1a progenitors towards differentiation. To respond to this question, we conditionally inhibited the Notch-pathway in embryos where atoh1a was mutated, and therefore no cells could be transitioning towards differentiation. Upon LY-treatment, ato-h1a fh282 embryos displayed a similar phenotype than non-treated mutant embryos (compare  Table 3 for numbers of analyzed embryos. r, rhombomere. Scale bars correspond to 50 μm. https://doi.org/10.1371/journal.pone.0228225.g006 Table 3. Analysis of the phenotypes in gain-of-function experiments (Fig 6).  , namely: atoh1a expression increased (Fig 7I and 7I'; [18]), atoh1b expression was highly diminished (Fig 7J and 7J'), and GFP-expressing progenitor cells failed to reach the neuronal differentiation domain (Fig 7K and 7K'). Thus, even though inhibition of N-activity triggered the neurogenic program, lack of atoh1a function impeded the LRL-progenitors to proceed towards differentiation, supporting the hypothesis that the transition of atoh1a progenitors towards differentiation depends on atoh1a function and is regulated by Notch.

Discussion
Progenitor cell populations undergo important changes in their relative spatial distribution upon morphogenesis, which need to be precisely coordinated with the balance between progenitor cells vs. differentiated neurons. Here, we have defined the role of atoh1 genes along the development of the LRL population, and how this progenitor cell population behaves during the early neurogenic phase. The spatiotemporal activation of proneural genes in the hindbrain shows that the neurogenic capacity is regionalized along the AP axis, such as that hindbrain boundaries and rhombomere centers remain devoid of neurogenesis [33]. This is valid for most of proneural genes except for atoh1 genes, because these are expressed all along the AP axis in the dorsal most hindbrain; however, RL derivatives delaminate from the dorsal epithelium, migrate and transitorily locate in the boundary regions. Interestingly, our results demonstrate that the function of different atoh1 genes depends on the context. In the inner ear, atoh1a and atoh1b cross-regulate each other but are differentially required during distinct developmental periods: atoh1b activates atoh1a early, whereas in a late phase atoh1a maintains atoh1b [27]. In the URL, atoh1a and atoh1c have equivalent function in the generation of granular cells progenitors [18], whereas we argue that in the LRL atoh1a and atoh1b are not interchangeable, since they work directionally and have distinct functions. Although in the URL atoh1a activates the expression of neurod1 in intermediate, non-proliferative precursors [35], neurod1 expression is not detected in the zebrafish LRL before the 48hpf, implying that atoh1b is the one defining LRL intermediate precursors rather than neurod1 during early LRL-derived neurogenesis.
Zebrafish has three atoh1 genes, atoh1a, atoh1b and atoh1c, which are expressed in overlapping but distinct progenitor domains within the rhombic lip [17,18]. Although atoh1a and atoh1c specify different, non-overlapping pools of progenitors within the URL, in the LRL while atoh1b largely overlaps with atoh1a it defines a cellular state rather than a progenitor lineage. atoh1b is expressed in a cell population that derives from atoh1a progenitors, and it has diminished its proliferative capacity; thus, atoh1b cells experienced a basal displacement of their cell body behaving as committed progenitors transitioning towards differentiation. This observation implies that atoh1 gene duplication in teleosts resulted in a gene sub-functionalization: atoh1a may behave as the cell fate selector gene, whereas atoh1b functions as a neuronal differentiation gene maintaining the transcriptional program initiated by atoh1a. In our conditional functional experiments, atoh1a ectopic expression was rapidly downregulated, whereas ectopic atoh1b remained active at later stages, highlighting the different roles of atoh1a and atoh1b in initiating vs. maintaining the differentiation program, and that atoh1a and atoh1b are not interchangeable. Interestingly, atoh1a/b/c proteins are conserved in the basic region, characterized by being arginine-rich, and in the two helixes but not in the loop, which is known to be variable. This conserved region, the core of bHLH proteins, is located in the center of the three proteins. The N-and C-terminal regions are highly divergent except for certain amino acids such as serine and threonine, predicted to be phosphorylation sites that may modulate the function of the distinct atoh1 proteins (S5 Fig). Interestingly, first-born neurons from the LRL delaminate and migrate towards medio-ventral positions to allocate in rhombomeric boundaries. Later-born LRL neurons follow the same trajectory, pile up with them and settle more laterally generating what we call neuronal arch-like structures. We think that this pattern of neuronal organization responds to some kind of chemo-attractant signal derived from boundary cells, as first atoh1a derivatives have a tendency to allocate within rhombomeric boundaries independently from their AP position upon differentiation. Many of such signalling pathways have been described for LRL migrating cells in the mouse embryo [36]; however, signals participating in this particular context are unknown. Nonetheless, boundary cells are signalling centres instructing the neuronal allocation in the neighbouring tissue [9]; thus, one plausible hypothesis is that boundary cells might dictate the allocation of newly-differentiated neurons.
Balancing the rate of differentiation and proliferation in developing neural tube is essential for the production of appropriate numbers and achieving the needed cell diversity to form a functional central nervous system (CNS). This requires a finely tuned balance between the different modes of division that neural progenitor cells undergo [37]. Three distinct modes of divisions occur during vertebrate CNS development: self-expanding (symmetric proliferative, PP) divisions ensure the expansion of the progenitor pool by generating two daughter cells with identical progenitor potential, self-renewing (asymmetric, PN) divisions generate one daughter cell with the same developmental potential than the parental cell and another with a more restricted potential, and self-consuming (symmetric terminal neurogenic, NN) divisions generate two cells committed to differentiation, thereby depleting the progenitor pool [37,38]. Our in vivo cell lineage studies shed light into this specific question in respect to the atoh1a cell population. We reveal the importance of the initial allocation of atoh1a progenitors: dorsal most atoh1a progenitors display more neurogenic capacity than ventral ones, since they give rise only to NN divisions upon the early neurogenic phase, whereas atoh1a progenitors located just underneath undergo the three distinct modes of division ensuring the expansion of the atoh1a-pool and providing committed progenitors. Most probably, the originally located dorsal progenitors will quickly become atoh1b and transition towards differentiation allocating more laterally. Interestingly, in the amniote spinal cord the modes of progenitor division are coordinated over time [39], instead of space. Why such a difference? One explanation is that in the LRL, where the position of progenitor cells changes dramatically over time, the most efficient way to provide fast neuronal production without exhausting the pool of progenitors could be regionalising the proliferative capacity. On the other hand, in vivo experiments in the chick spinal cord showed that an endogenous gradient of SMAD1/5 activity dictated the mode of division of spinal interneuron progenitors, in such a way that high levels of SMAD1/5 signalling promoted PP divisions, whereas a reduction in SMAD1/5 activity forced spinal progenitors to reduce self-expanding divisions in favour of self-consuming divisions [40]. This would suggest that dorsal most atoh1a cells would respond less to BMP signalling than ventral atoh1a cells. However, during hindbrain morphogenesis there is an important change in the position of atoh1a progenitors, and therefore their relative position in respect to the gradient sources. Since morphogen gradients quickly decrease with distance [41,42], it is difficult to apply the same rationale here than in the spinal cord. Still very little is known about how these gradients are established within the hindbrain [43], and how hindbrain progenitors interpret the quantitative information encoded by the concentration and duration of exposure to gradients. An alternative explanation is that different E proteins may control the ability of atoh1a to instruct dorsal or ventral neural progenitor cells to produce specific, specialized neurons, and thus ensure that the distinct types of neurons are produced in appropriate amounts as it happens in the chick spinal cord [44].
The loss of atoh1a function clearly affects the formation of the lateral column of lhx2b differentiated neurons and decreases the number of overall differentiated neurons. But what are the derivatives of these atoh1a-derived lhx2b cells? It has been described that the hindbrain displays a striking organization into transmitter stripes reflecting a broad patterning of neurons by cell type, morphology, age, projections, cellular properties, and activity patterns [45]. According to this pattern, the lateral lhx2b column would correspond to glutamatergic neurons expressing the barhl2 transcription factor [46], which in turn is an atoh1a target [46,47]. Moreover, our observations revealed that atoh1a was necessary for initial steps of neuronal differentiation, such as apical abscission and migration. Interestingly, this phenotype resembled to the one of atoh1c fh367 mutants, in which the release of granule neuron progenitors from the URL required functional atoh1c [18], indicating that atoh1a replaced atoh1c function in this context.
Notch has been extensively studied as a regulator of proneural gene expression by a process called lateral inhibition, in which cells expressing higher levels of proneural genes are selected as "neuroblasts" for further commitment and differentiation, while concomitantly maintaining their neighbors as proliferating neural precursors available for a later round of neuroblast selection [48]. Indeed, in the LRL the transition atoh1a to atoh1b seems to be regulated by Notch-activity, since upon Notch-inhibition most of the atoh1a cells disappear and they become atoh1b, and therefore are ready to undergo differentiation. Thus, although atoh1a is the upstream factor in LRL cell specification, several mechanisms seem to be in place to precisely coordinate acquisition of the neurogenic capacity and progenitor vs. differentiation transitions. GFP cells corresponding to atoh1a-derivatives end up generating a neuronal arch-like structure (see white arrowheads) as development proceeds. ov, otic vesicle; r, rhombomere. Scale bars correspond to 50 μm. (TIF) S5 Fig. Amino acid sequence comparison of zebrafish atoh1 proteins. Comparison of zebrafish atoh1a, atoh1b and ato1hc proteins by Multiple Sequence Alignment CLUSTALW (MSA, EMBL-EBI). Sequence conservation (>70%) is displayed at the top as grey blocks with different hues. Amino acids highlighted in green correspond to those that match with the consensus sequence, which is displayed at the top in bold. Note how the three atoh1 proteins are conserved in the central regions and their sequence diverge in the N-and C-terminal domains. (TIF)