Embryonic origin and lineage hierarchies of the neural progenitor subtypes building the zebrafish adult midbrain.

Neurogenesis in the post-embryonic vertebrate brain varies in extent and efficiency between species and brain territories. Distinct neurogenesis modes may account for this diversity, and several neural progenitor subtypes, radial glial cells (RG) and neuroepithelial progenitors (NE), have been identified in the adult zebrafish brain. The neurogenic sequences issued from these progenitors, and their contribution to brain construction, remain incompletely understood. Here we use genetic tracing techniques based on conditional Cre recombination and Tet-On neuronal birthdating to unravel the neurogenic sequence operating from NE progenitors in the zebrafish post-embryonic optic tectum. We reveal that a subpopulation of her5-positive NE cells of the posterior midbrain layer stands at the top of a neurogenic hierarchy involving, in order, the amplification pool of the tectal proliferation zone (TPZ), followed by her4-positive RG cells with transient neurogenic activity. We further demonstrate that the adult her5-positive NE pool is issued in lineage from an identically located NE pool expressing the same gene in the embryonic neural tube. Finally, we show that these features are reminiscent of the neurogenic sequence and embryonic origin of the her9-positive progenitor NE pool involved in the construction of the lateral pallium at post-embryonic stages. Together, our results highlight the shared recruitment of an identical neurogenic strategy by two remote brain territories, where long-lasting NE pools serve both as a growth zone and as the life-long source of young neurogenic RG cells. Copyright Â

Introduction

Adult neurogenesis, ie. the de novo generation of neurons from neural progenitors maintained inside the adult brain, has been evidenced in all vertebrate species studied. It is an important component of brain plasticity, for example permitting song learning in songbirds, or the formation of new olfactory or spatial memories in rodents (Bond et al., 2015). In teleost fish such as the zebrafish, active neurogenic zones correlate with adult brain growth and the activation of specific sensory modalities, for example in the telencephalon and midbrain (Lindsey and Tropepe, 2014, Lindsey et al., 2014).

A striking feature of adult neurogenesis is its highly different efficacy between species and brain territories (Grandel and Brand, 2013; Chapouton et al., 2007). While a primary component of this difference relates to neuronal maturation and survival, several studies also hint to huge variations in neural progenitor cell properties, such as their propensity for recruitment and their lineage progression (Alunni and Bally-Cuif, 2016). Overall, it remains an intense subject of investigation to understand which lineages can underlie successful neurogenesis in the adult brain.

Teleost fish, which harbor constitutive and efficient neurogenic niches in all brain subdivisions (Adolf et al., 2006), are unique models to approach this question. Our recent work pointed to the coexistence of two different neural progenitor subtypes, radial glial cells (RG) and neuroepithelial progenitors (NE), across the germinal (ventricular) zone of the dorsal telencephalon (pallium) in the adult zebrafish (Dirian et al., 2014). RG cells, covering most of the pallial ventricle and generally quiescent, self-renewing and multipotent (generating neurons and other RG cells) (Chapouton et al., 2010, Rothenaigner et al., 2011), are the bona fide counterparts of rodent telencephalic neural stem cells (NSCs). RG cells express the bHLH transcription factor Her4 (orthologous to mammalian Hes5). In addition, a small pool of her4-negative, NE cells, inherited from embryogenesis, continuously feeds the ventricular zone with freshly born RG that later enter into quiescence and serve as neurogenic NSCs. These RG and NE progenitors drive differential growth rates and RG age across the pallium, and are endowed with different reconstruction potential of the pallial NSC population following chemical NSC ablation (reviewed in Than-Trong and Bally-Cuif (2015)). Assessing whether and which distinct progenitor subtypes and lineages operate in other neurogenic niches will reveal the general strategies that can balance NSC maintenance and neurogenesis at the population and individual cell levels in the adult vertebrate brain.

Towards this aim and as a counterpart to the pallium, intense work is focusing on the optic tectum (TeO), a major visual centre of the zebrafish brain. This domain, homologous to the superior colliculus of mammals, is composed of a deep granular layer of periventricular interneurons, projection neurons (periventricular grey zone, PGZ) and ventricular RG cells, and a superficial neuropile layer receiving input from retinal ganglion cells (Corbo et al., 2012). One key feature of this structure is to continuously expand from a growth zone located at its posterior edge (tectal proliferation zone, TPZ) (Fig. 1A,B) (Deves and Bourrat, 2012), thus managing the de novo generation of tectal neurons and glial cells and the long-lasting maintenance of progenitors in the adult. Experiments assessing the retention after chase of a BrdU label, incorporated into the DNA of dividing cells during S phase, identified a small niche of slow-proliferating, long-lasting progenitors within the TPZ in the adult zebrafish and medaka (Alunni et al., 2010, Ito et al., 2010, Lindsey et al., 2014) (Fig. 1A). These cells display apicobasal polarity, do not express glial markers, and were interpreted as a NE population of NSCs. BrdU lineage tracing further showed that the TPZ contributes to both neurons and RG in the tectum, arranged in transversal columns of increasing age as one moves away from the TPZ (Alunni et al., 2010, Ito et al., 2010). In addition, a minute population of progenitor cells with NE morphology was identified in a tela-like extension, the “peripheral midbrain layer”/“posterior midbrain lamina” (PML) bridging the TPZ with the more ventrally located torus semicircularis (TSc) and its associated isthmic proliferation zone (IPZ) (Grandel et al., 2006) (Fig. 1A). Some PML cells express the bHLH transcription factor Her5 (orthologous to mammalian Hes7) and proliferate slowly (Chapouton et al., 2006). To date, the fate of her5-positive cells was transiently assessed in adults using as a tracer the stability of GFP in the Tg(her5PAC:egfp) transgenic zebrafish (Tg(her5:egfp)) (Tallafuss and Bally-Cuif, 2003). This demonstrated their contribution to the generation of neurons and oligodendrocytes populating the tegmentum, in the basal midbrain (Chapouton et al., 2006). In the embryo, the PML region also in part expresses her5 (Tallafuss and Bally-Cuif, 2003), and time-lapse imaging identified individual PML cells generating neurons in the PGZ, or in the TSc (Recher et al., 2013). Thus, the PML serves as a progenitor source for tectal and torus neurons in the embryo.

Fig. 1

Adult her5-positive cells of the PML express the characteristics of neuroepithelial progenitors. A, B: Schematic cross-section of one tectal hemisphere from an adult zebrafish (level indicated in B, dorsal up, midline indicated by the vertical broken line) showing the different relevant cell populations (color-coded, see keys on figure). Arrows indicate the known lineages, ie: green arrow: generation of PGZ neurons and radial glia from TPZ proliferating progenitors (Alunni et al., 2010, Ito et al., 2010); dark red arrow: generation of tegmental neurons by her5-positive cells of the PML (Chapouton et al., 2006). C: Cross section of an adult optic tectum at the level indicated in B, immunostained for mCherry and PCNA (color-coded) and counterstained with DAPI, position of the PML (arrowheads) and the Her5-mCherry-positive domain compared to the TPZ and IPZ. C1: low magnification showing one hemisphere, C2-C4: high magnifications of the white boxed area in C1 (C2: red and green channels with DAPI, C3: red channel only, C4: green channel only). Arrowheads to the PML on panel C2. D-F: compared expression of Her5-mCherry with markers for progenitors (Sox2), glial cells (GS) and tight junctions in neuroepithelial cells (ZO1, green arrowheads) on high magnifications of the PML area on cross sections of an adult optic tectum (C1-F3: High magnification of the yellow boxed area in C1, panels 1: red and green channels with DAPI, panels 2: red mCherry channel only, panels 3: green channel only). White arrowheads to the PML on panels 1. Scale bars: C 50 µm, D-F 10 µm. Abbreviations: CCe: crista cerebellaris, IPZ: isthmic proliferation zone, PGZ: periventricular grey zone, PML: peripheral midbrain layer/posterior midbrain lamina, TeO: tectum opticum, TPZ: tectal proliferation zone, TSc: torus semi-circularis.

Together, these studies identified distinct progenitor subtypes and partial lineages in the posterior midbrain, but leave important questions open as to the exact hierarchical cell type progressions that build the adult optic tectum. Namely: (i) is there a lineage relationship between embryonic and post-embryonic (including adult) her5-positive progenitors, (ii) do post-embryonic her5-positive progenitors also contribute neurons and RG to dorsal midbrain structures such as the TeO, (iii) if yes, do they do so via the TPZ as an intermediate, and (iv) are adult tectal RG neurogenic (or in other words, does a sequence “NE to RG” contribute to adult tectal neurogenesis), as is the case in the adult telencephalon. Here we use a combination of Cre-lox and Tet-on genetic tracing of her5- and her4-positive cells to answer these questions. Our results place the her5-positive cell pool at the top of a hierarchy building the post-embryonic optic tectum via proliferating NE progenitors then neurogenic RG, in a manner that recapitulates the neurogenic sequence operating in the lateral pallium. However, in contrast to the latter territory, tectal RG appear only transiently neurogenic under physiological conditions.

Materials and methods

Zebrafish Lines

Genomic zebrafish BAC (N° CH73-271E23) was modified by homologous recombination to insert nls-mCherry (plasmid N°233, Tol2 kit) (Kwan et al., 2007) or ERT2-Cre-ERT2 in frame after the second amino acid of the second exon of the her5 gene. The PME:ERT2CreERT2 plasmid was obtained by subcloning the Kpn1-BamHI fragment from the pCDNA:ERT2CreERT2 (Casanova et al., 2002). For efficient lambda-red recombination, a sequence including 354 base pairs (bp) upstream of the fusion site and a 479 bp sequence starting at the her5 TGA stop codon were used as homologous fragments encompassing both mCherry and ERT2-Cre-ERT2. A bacterial Zeocin cassette with a unique MaubI restriction site at each end was introduced upstream of the 479 bp fragment to ease recombinant-BACs selection on Zeocin-containing plates. Both assembled full-length sequences were isolated and used in recombination with the CH73-271E23 BAC. Upon selection and characterization of recombined clones, the Zeocin cassette was then removed by MaubI digestion and re-ligation.

Tol2 was obtained by modifying the iTol2-Amp plasmid (gift from U. Strähle) to insert the CMLC2:GFP sequences (plasmid N°395, Tol2 kit), the whole sequence was then inserted at the loxP511 site of pTARBAC2.1. Two fragments of 104 bp and 182 bp, located upstream and downstream loxP511 respectively, were used to direct lambda-red homologous recombination. These fragments were first cloned to surround the Tol2 construct. The assembled sequence was then excised and purified before use to target the loxP511 site. Recombinant BACs were selected on Ampicillin-containing plates and then fully characterized by restriction analysis and sequencing of the recombined regions (Agate bioservices, Bagard, France).

The transgenic lines Tg(her5:nls mCherry) and Tg(her5:ERT2CreERT2) were made by coinjecting 1-cell embryos with a mix containing 15 ng/µl of this plasmid and 15 ng/µl of transposase capped RNA. F0 adults were screened for transmission by crossing into wildtype fish and recording fluorescence in F1 embryos.

The Tg(her4:rtTA, GFP:cmlc2) and the Tg(GFP:biTRE:H2amCherry, crist:Venus) transgenic lines were obtained with the gateway strategy using plasmids developed in Jensen lab (Campbell et al., 2012) (a full description of these lines will be provided in Furlan et al. (in preparation)).

Embryos/larvae up to 5 dpf were maintained and staged as described (Kimmel et al., 1995). Adult zebrafish 3–6 months were maintained using standard fish-keeping protocols. All experiments were carried out in accordance to the official regulatory standards of the Department of Essonne (agreement number A 91–577 to L.B.-C).

Other transgenic lines used in this study are as follows: Tg(her5:eGFP) (Tallafuss and Bally-Cuif, 2003), Tg(βactin:lox-stop-lox-hmgb1-mcherry) (Wang et al., 2011), Tg(gfap:eGFP) (Bernardos and Raymond, 2006), Tg(her4:eGFP) (Yeo et al., 2007), Tg(her4:dRFP) (Yeo et al., 2007), Tg(her4:ERT2CreERT2) (Boniface et al., 2009).

4-OHT treatments, bromodeoxyuridine and 9TB treatments

4-Hydroxytamoxifen (4-OHT, T176, Sigma) treatments were performed as previously described (Mosimann et al., 2011); see Table 1 for optimal recombination conditions. Fish were then washed four times, transferred into fresh embryo medium, and grown as usual.

Table 1

Optimal 4-OHT concentrations for her5:ERT2CreERT2 and her4:ERT2CreERT2 recombinations.

Recombination stage	4-OHT concentration	Time
80%	5 µM	4 h
1 dpf	5 µM	6 h
2 dpf	5 µM	6 h
5 dpf	5 µM	30 h
1.5mpf	5 µM	100 h

9-t-butyl doxycycline (9-TB, B-0801, Tebu-bio) was dissolved in water at a final stock concentration of 5 mg/ml, then diluted in EM prior to use (see Table 2 for optimal induction conditions adjusted depending of the stage). 9TB treatments were performed in the dark at 28 °C.

Table 2

Optimal 9-TB concentrations for optimal her4:rtTA inductions.

Induction stage	9-TB concentration	Time
≤2 dpf	2.5 µg/ml	6 h
5–15 dpf	10 µg/ml	6 h
1.5 mpf	10 µg/ml	96 h

Bromodeoxyuridine (BrdU) labelling at 5 and 15 dpf was performed as previously described (Coolen et al., 2012).

Immunohistochemistry and in situ hybridization

Immunohistochemistry and in situ hybridization were performed as described previously (Bosco et al., 2013, Chapouton et al., 2010); see Table 3, Table 4 for the list of antibodies and probes used in this study.

Table 3

List of antibodies used in this study.

Antibodies	Species	Dilution	References	Antibody registry
anti-mCherry	rabbit	1/300	Clontech (632496)	AB_10013483
anti-GFP	chicken	1/1000	Aves lab (gpf-1020)	AB_10000240
anti-PCNA	mouse (IgG2a)	1/250	Dako (M0879)	AB_2160651
anti-Sox2	mouse	1/200	Abcam (ab171380)	not registered
anti-ZO1	mouse (IgG1)	1/100	Invitrogen (A339100)	AB_2533147
anti-Glutamine Synthetase	mouse (IgG2a)	1/1000	Millipore (MAB302)	AB_2110656
anti-HuC	mouse (IgG2b)	1/250	Invitrogen (A21271)	AB_10562207
anti-BrdU	rat	1/250	Abcam (ab6326)	AB_305426

Table 4

List of probes used in this study.

Gene	Probe references
cre	(Dirian et al., 2014)
her4.1	(Takke et al., 1999)
her5	(Geling et al., 2003)
her9	SourceBioscience, BC079516

Image acquisition

Images taken using a confocal microscope (Zeiss LSM700) were processed with ZEN 2011 software (Carl Zeiss MicroImaging) and Photoshop CS6. All confocal views are single planes (thickness 0.8 mm) except panels in Fig. 4 A2-A4 and B2-B4, which are maximal projections of 3 µm. Dorsal whole-mount views of the midbrain were taken using a Nikon macrozoom.

Fig. 4

her4-positive RG exhibit neurogenic activity in the post-embryonic TeO. Compared fate in the 5mpf TeO of RG cells expressing her4 at 1.5mpf, assessed by Cre-mediated tracing (A) and rtTA-mediated birthdating (B). Panels 1. Cross-sections of a TeO hemisphere immunostained for mCherry and GS and counterstained with DAPI and observed in confocal microscopy. Panels 2–4. High magnifications of the areas boxed in panels 1 (single and merged channels as indicated, maximum projection of 4 sections over 3 µm). In both cases, a single stripe of mCherry-positive progeny cells is observed, identically located along the antero-posterior and dorso-ventral axes and comprising both neurons and RG. Scale bars: A1 and B1 100 µm, A2-A4 and B2-B4 50 µm.

Results

An organized arrangement of NE progenitor subtypes and RG in the caudal midbrain

As a preliminary step towards identifying neurogenic lineages in the TeO, we aimed to precisely characterize the cell subtypes of the caudal midbrain and their relative spatial arrangement in a single comparative study. GFP stability in Tg(her5:egfp) fish can result in overestimating the her5-positive zone (Chapouton et al., 2006), and her5-positive cells are very few and difficult to detect at adult stage by in situ hybridization. Thus, we first used BAC recombination to build a novel transgenic line expressing the less stable nls-mCherry protein under control of the her5 regulatory elements (Tg(her5BAC:nlsmCherry), hereafter referred to as Tg(her5:mCherry)) (Fig. S1A).

In the adult midbrain, mCherry-positive cells were restricted to a ventral subset of the adult PML cells (Fig. 1B-C4, S1C), also expressing her5 RNA in transgenic fish (Fig. S1E, F), and comprised within the eGFP-positive domain in our previous Tg(her5:egfp) line (Fig. S1B). Tg(her5:mCherry)-positive cells of the PML are mosaically positive for the Proliferating Cell Nuclear Antigen (PCNA, a marker of G1 and subsequent phases of the cell cycle) (Fig. 1C, S1H, I). As previously described, this territory intervenes the TPZ and IPZ, which are strongly and uniformly PCNA-positive (Fig. 1C, S1I and not shown), and expresses the progenitor marker Sox2 (Fig. 1D). The PML does not contain HuC-positive neurons (Fig. S1J) and Tg(her5:mCherry)-positive cells also do not express Glutamine Synthase (GS) and Glial Fibrillary Acidic Protein (GFAP) (Fig. 1E, S1K), which label most astroglial cells in the adult zebrafish brain (Than-Trong and Bally-Cuif, 2015)). However, and in agreement with a NE progenitor nature, mCherry-positive cells display polarized expression of the apical marker ZO-1 (Fig. 1F). Compared to our previous data using stable eGFP tracing, these results identify her5-expressing cells as a PML subpopulation of progenitors of NE identity. Two NE progenitor populations are thus juxtaposed in the caudal midbrain: PML cells (some of which are her5-positive) and label-retaining progenitors located along the ventricle at the posterior edge of the TPZ (Ito et al., 2010) (Fig. 1A, dark green).

We next aimed to position these NE progenitors and the TPZ compared to tectal her4-positive cells and RG. In the adult TeO, mature RG have their cell bodies in the PGZ aligned along the tectal ventricle, extend basal processes that reach the pial surface across the superficial neuropile, and express GS and GFAP (Ito et al., 2010) (Fig. 1A and not shown). The her4:egfp and her4:dRFP reporters, expressing eGFP or dRFP under control of the same her4 regulatory element in transgenic fish (Yeo et al., 2007), are RG markers in the adult telencephalon. We found that these reporters also faithfully recapitulate endogenous her4 expression (Fig. S1D, G and not shown) and label cells with RG morphology in the TeO. Most of these cells co-express GS and GFAP (Fig. 2A). In addition, a small population of her4-positive RG cells extends beyond the GS/GFAP posterior limit to reach into the most anterior aspect of the TPZ (Fig. 2A, B, red arrows). The latter area displays weaker PCNA staining, possibly indicative of progressive cell cycle exit (Fig. 2B3).

Fig. 2

The her5-positive, TPZ, her4-positive and mature radial glia domains are spatially ordered in the posterior TeO. All views are confocal images. A, B. Compared expression of Her4-RFP, GFAP-GFP and GS (A) or PCNA (B) on cross-sections of the adult midbrain in double transgenic animals (high magnifications of the area boxed in the schematic). Panels 1 are RFP, GFP and GS merge views with DAPI counterstaining, Panels 2–4 are single channel views (color-coded). Note that Her4-RFP-positive glial cells (red arrows) extend into a more lateral ventricular domain than mature RG (A 1–4, B3, green or blue arrows point to the lateral limits of GFAP-GFP and GS expression, respectively) and reach into the PCNA-positive domain (B4, white arrow). C-F. Compared position of the Her5-mCherry (red, red arrows) and Her4-GFP (green, green arrows) positive territories in the posterior midbrain at increasing stages from embryo to adult, observed on cross sections (levels and stages indicated on the schematics). C′-F′. Interpretative drawings of C-F highlighting in red the her5-positive zone and in green the area covered by the cell bodies of her4-positive RG. Scale bars: A,B 100 µm, C-F 50 µm. Abbreviations: Cb: cerebellum, TeO: tectum opticum, TPZ: tectal proliferation zone, TSc: torus semi-circularis, V: ventricle.

Together, these observations combined with previous data (Alunni et al., 2010, Ito et al., 2010) highlight a linear arrangement of potential progenitor cell subtypes in the adult posterior midbrain and TeO, with, starting from posterior: NE cells of the PML (a subset of which expresses her5 and/or PCNA), label-retaining NE cells of the TPZ, amplifying cells of the TPZ, her4-positive RG, and finally GFAP/GS/her4-positive RG, which then extend as a uniform population up to the anterior TeO. This ordered, adult-like arrangement is present from embryonic stages onwards, as visible for example by the mutual exclusion of the Tg(her5:mCherry) and Tg(her4:eGFP) signals at all stages analysed starting at 2 dpf, with a progressive restriction of the Tg(her5:mCherry)-positive domain within the PML as the TeO grows and the her4-positive RG layer expands (Fig. 2C-F′).

Post-embryonic her5-positive cells are at the origin of the adult PML and progressively build the TeO in an anterior to posterior sequence

With the aim to identify the lineages building the adult TeO, we first addressed the fate of her5-positive PML cells. Because Tg(her5:mCherry) animals faithfully recapitulated her5 expression, we recombined a construct encoding a conditional version of the Cre recombinase, ERT2CreERT2, into the same BAC, and generated Tg(her5BAC:ERT2CreERT2) transgenic animals (thereafter called Tg(her5:ERT2CreERT2)) (Fig. S1A). We verified that cre and her5 expression were identical at all relevant stages (5 embryos analysed, Fig. S2A-D and not shown). eGFP stability in Tg(her5:egfp) animals had previously allowed to show that her5 expression is dynamic and that its earliest domain in the embryonic neural plate (80% epiboly) was the presumptive territory for the midbrain and anterior rhombencephalon (including the cerebellum and rhombomeres 1 and 2) (Tallafuss and Bally-Cuif, 2003). Thus, to further validate the Tg(her5:ERT2CreERT2) line, we crossed it into the ubiquitous floxed reporter line Tg(actb2:lox-stop-lox-hmgb1mCherry) (Wang et al., 2011) and applied tamoxifen to double transgenic embryos at 80% epiboly (animals thereafter referred to as her5). In 3mpf adults, the midbrain and cerebellum were mCherry-positive (100% of brains analysed, n=3, Fig. 3A1). In particular, close inspection demonstrated that, as expected, virtually all TeO cells were stained (Fig. 3A2, A3).

Fig. 3

Post-embryonic her5-positive cells contribute to the generation of the posterior TeO via a TPZ intermediate. A-D. Adult fate of her5-positive cells assessed using Cre-lox recombination at the indicated stages (top schemes) and analysed in whole-mount brains (panels 1, dorsal views, anterior left, dotted line to surround the TeO) or on horizontal sections (panels 2 and 3, confocal views, with high magnifications of boxed areas in panels 3). The anterior limit of the TeO domain issued from recombined cells is indicated by yellow arrowheads in panels 1 and 2. Small red arrow in panel D2 point to transversal mCherry-positive polyclones (see high magnification in E). White arrowheads in panels 3 point to the PML, containing mCherry-positive cells (red arrowhead in D3) (in average, 15% ±4% cells per PML, n=2 fish and 6 sections analysed). E, F. mCherry-positive polyclones induced by a mosaic recombination in her5 fish at 5dpf and analysed at adult stage. Confocal views of triple-immunostained horizontal sections (areas boxed in scheme), merged panel (E1) includes DAPI counterstaining, single channel panels (E2-4) are color-coded. E. Compared adult expression of mCherry with the RG marker GS (green) and the neuronal marker HuC/D (blue) in her5 fish. The mCherry polyclones include ventricular RG (small green arrows point to RG cell bodies) and neurons. F. Compared expression of mCherry with PCNA-positive proliferating progenitors, showing that mCherry polyclones contribute to the TPZ (yellow arrows to the TPZ, white arrowheads to the PML) with high magnifications of the boxed area in panel F1. Scale bars: A2-D2 100 µm, A3-D3 50 µm, E,F 20 µm. Abbreviations: Cb: cerebellum, TeO: tectum opticum, TPZ: tectal proliferation zone, TsC: torus semicircularis.

her5 expression is restricted to the PML from at least 2 dpf onwards (Fig. 2C), and we therefore assessed the fate of her5-positive cells past that stage, after verifying that tamoxifen application indeed drove selective mCherry expression in PML cells (100% of cases, 7 embryos analysed, Fig. S2E, compare with Fig. 2C, and not shown). We found that recombination at 5dpf, when the her5-positive territory is already limited to a discrete PML subdomain (Fig. 2D), labelled a large posterior half of the TeO in adults (100% of cases, 12 whole-mount brains analysed to determine the anterior boundary of the mCherry domain, 3 brains sectioned, Figs. 3D1–3). These results demonstrate that post-embryonic her5-positive PML cells are TeO progenitors. Further, we found that the anterior and ventral limits of the recombined territory within the adult TeO recede as recombination is induced at increasing post-embryonic time points (Fig. 3B1-D1, B2-D2, yellow arrowheads, at least 5 brains analysed per condition -except for T(5dpf): 12 brains-). Thus, post-embryonic tectal growth is driven by her5-positive PML progenitors in a ventro-anterior to posterior schedule, which resembles the anterior to posterior mode of TeO generation that was described from the TPZ (Alunni et al., 2010, Ito et al., 2010).

Based on these results, we further probed the parallels between the progenitor activities of her5-positive PML cells and the TPZ. BrdU pulses applied at post-embryonic stages label TPZ progenitors and generate after chase transversal cell clones spanning the width of the PGZ and comprising RG cells and neurons (Alunni et al., 2010, Ito et al., 2010). To test whether the post-embryonic her5-positive PML could also generate such clones, we made use of mosaic recombinations at 5dpf to follow the fate and organization of polyclones specifically generated from her5-positive cells. We found that TeO mCherry-positive cells in her5 adults were organized in transversally oriented polyclones that crossed the depth of the PGZ in columns and included both RG cells and neurons (Fig. 3E1-4, green arrows), resembling the clones issued from the TPZ. An economical interpretation of these results would be that her5-positive PML cells give rise to TPZ cells. To formally assess this hypothesis, we stained her5 adults for PCNA. We found indeed that some TPZ (PCNA-positive) cells expressed mCherry, demonstrating that the adult TPZ, at least in part, derives from the post-embryonic her5-positive pool (yellow arrows to mCherry-positive TPZ cells, Fig. 3F1-4). The pattern of polyclones appears different from fish to fish (Fig. S2G-I), in agreement with a mosaic and random recombination in her5 fish and with the generation from the her5 pool of most TPZ cells rather than a specific subpopulation.

The progenitor activity of the her5-positive PML domain likely persists in adults

To determine whether the PML could still serve as a progenitor source in adults, we first carefully analysed mCherry-positive polyclones generated upon mosaic recombination in her5 fish. We observed polyclones located close to the TeO posterior border in 3mpf adults (Fig. 3D3, in average, 15% ±4% mCherry-positive cells per PML (total number of PML cells per section: 20±3, n=2 fish and 6 sections analysed), and in particular within the TPZ (white arrowheads to the TPZ, Fig. 3F), indicating that the generation of mCherry-positive cells continues at time points close to the time of analysis. In the same animals, we could document scattered mCherry-positive cells in the PML (Fig. 3D3, red arrowhead). Together, these observations suggest that the adult PML could serve as a long-lasting cell pool feeding mCherry-positive cells into the TPZ. Recombination in her5 fish was unfortunately inefficient past 5dpf, probably due to the low intensity of Cre expression (not shown), and it was not possible to directly validate this interpretation. Alternatively, the polyclones found adjacent to the TPZ at late stages could derive, with a delay, from long-lasting mCherry-positive TPZ progenitors generated from the her5-positive pool after the 5dpf stage.

her4-positive tectal RG can generate neurons at post-embryonic stages

Having assessed the sequence “her5-positive PML>TPZ” upstream of the post-embryonic generation of TeO neurons and RG, we wondered about the neurogenic contribution of tectal RG to TeO formation. Although GFAP- or GS-positive tectal RG are never found to express PCNA, we observed neuronal clones juxtaposed to mCherry-positive RG in her5 adults (100% of cases, 3 brains analysed, Fig. 3E). These could be interpreted as originating, in part, via mCherry-positive RG. Mosaic recombination at other stages led to similar observations (8 embryos analysed, Fig. S2F).

To directly assess the fate of tectal RG, we made use of the Tg(her4:ERT2CreERT2) line (Boniface et al., 2009), expressing Cre under the same regulatory elements as Tg(her4:egfp) and Tg(her4:drfp), and leading to faithful recombination in RG cells of the embryonic and post-embryonic telencephalon (Dirian et al., 2014). We verified that cre expression in this line recapitulates her4 expression in the embryonic and post-embryonic midbrain (Fig. S3A-F and not shown, 100% of cases, at least 5 embryos per condition), and drives faithful recombination in post-embryonic midbrain RG cells (100% of cases, 7 embryos analysed, Fig. S3G). We recombined her4 animals at different stages from embryo to 1.5mpf and analysed the location and nature of mCherry-positive progeny cells in 5mpf adults, specifically looking for the generation of neurons. In all cases, mCherry-positive cells organized, in each tectal hemisphere, as a single transversal stripe across the PGZ (100% of cases, 3 brains analysed, Figs. 4A, 5A). This stripe generally covered the whole width of the TeO, extending from the RG layer into the entire PGZ and including isolated neurons in the neuropile (Fig. 4A and not shown). These results indicate that at least some her4-positive RG are neurogenic progenitors in the post-embryonic to adult TeO.

The shape of her4-generated polyclones, which appear of equal width across the TeO thickness, suggests that the neurogenic process initiated by her4-positive progenitors does not involve major further amplification steps and lies at the end of the neurogenic cascade building the TeO. To test this hypothesis, we used a birthdating strategy (Campbell et al., 2012) making use of the Tet-On system driven by the her4 promoter (Knopf et al., 2010) and driving transient induction of H2a-mCherry in her4-expressing cells upon a pulse of 9-t-butyl doxycycline (9-TB) (to be reported elsewhere, Furlan et al., in prep.). The incorporation of H2a-mCherry into the chromatin backbone stably labels cells at the time of induction, and will permanently label post-mitotic cells generated within a few divisions post-induction, while further divisions will dilute the label (Furutachi et al., 2015). In control experiments, we validated the reliability of this method by applying a transient pulse of 9-TB to transgenic animals at relevant t time points (animals thereafter referred to as her4 fish) and assessing the coincidence of H2a-mCherry expression with her4 and the RG marker GS in the TeO (Fig. S3H, I, 100% of cases, 4 brains analysed). To determine whether neurons were generated from her4-positive RG after few divisions during post-embryonic TeO construction, we applied 9-TB at successive time points spanning the period 2 dpf to 1.5mpf. We found that such inductions, at all stages, generated transversal TeO stripes of H2a-mCherry-positive RG and overlying neurons (100% of the cases, 3 brains analysed per condition, Fig. 4B) that resembled in all aspects the Cre recombination-induced profiles (Fig. 4A). We conclude that the neurogenic period of RG does not exhibit an amplification delay after the onset of her4 expression and is constant thereafter until her4-positive RG loose neurogenic competence.

The neurogenic activity of her4-positive tectal RG corresponds to a brief window of competence linked with their location at or close to the TPZ

The location of the mCherry-positive stripes in the adult tectum of her4 and her4 animals was not random along the gradient of tectal growth: their position appeared to recede towards posterior when Cre recombination or 9TB inductions were triggered at increasing ages (Fig. 5A1, A2 and 5D). These results suggest that the neurogenic period of her4-positive progenitors is linked with the existence of a narrow window of competence associated with their age.

Fig. 5

Post-embryonic her4-positive tectal RG are neurogenic during a brief window of competence linked with their position close to the TPZ. A-D. Compared distributions in the adult TeO of the progeny of her4- (A1, A2) and her5- (B1, B2) positive cells, traced from 1dpf (A1, B1) or 5dpf (A2, B2) by means of Cre-mediated recombination in her4 and her5 animals. Confocal images of horizontal sections at equivalent dorso-ventral positions, anterior left, immunostained for mCherry expression and counterstained with DAPI. Red arrows in A1, A2 point to the posterior limit of the her4-derived mCherry-positive stripe. Pink arrows in B1, B2 point to the anterior limit of the her5-derived mCherry-positive domain. Note that the position of these limits recedes towards posterior as recombination is induced at later stages, and that the her4-derived stripe is consistently anterior to the her5-derived domain. C1 and C2 are interpretative drawing (color-coded, see key) of superimposed her4 and her5 recombinations at 1dpf and 5dpf, respectively. D. Top panel: 9TB induction and analysis time points for the her4 animals photographed in D1-D4. Bottom panels: Whole-mount dorsal views of TeOs in adult her4 fish induced between 2 dpf and 1.5mpf, observed for H2a-mCherry expression under a fluorescence binocular. TeO surrounded by dotted line, anterior left, arrows to the H2a-mCherry-positive stripes. Note the receding position of these stripes with later induction time points. E. Compared antero-posterior positions in the adult TeO of an -recombined stripe (red) and BrdU-positive cells pulsed at 5dpf (green). Yellow arrow to the identical location of the two stripes. Horizontal section observed by confocal microscopy, immunostained for mCherry and BrdU, and counterstained with DAPI. F. Antero-posterior position in the adult TeO of BrdU-positive cells pulsed at 15dpf (green). Horizontal section observed by confocal microscopy, immunostained for BrdU and counterstained with DAPI. Note the coincidence with the anterior limit of the domain (pink arrow in panel B2). G. High magnification of a her4 -recombined stripe in the adult TeO, co-immunostained for mCherry and GS and counterstained with DAPI. Confocal views of a horizontal section, anterior left, yellow arrows to the mCherry-positive stripe. G4 is an interpretative drawing of G1 with identical color code. From anterior to posterior: anterior RG cells (yellow asterisk) are not overlaid with mCherry-positive neurons (black asterisk); RG at the level of the stripe (yellow asterisk, yellow arrows in G1) are overlaid with mCherry-positive neurons (red stripe); RG located posterior to the stripe are not recombined (green asterisk). Scale bars: A,B,E,F 100 µm, G 20 µm.

To more precisely stage this competence window, we carefully analysed the position of the transversal mCherry-positive stripe in her4 animals. We observed that, in all cases, all tectal RG located anterior to the stripe were also mCherry-positive, although not overlaid by mCherry-positive neurons (Fig. 5G). Thus, these RG were her4-positive at the time of recombination and were properly recombined but were not neurogenic during the chase time. In contrast, all neurons and RG cells located more posteriorly were mCherry-negative, indicating that they come from cells that were not her4-positive at the time of recombination (Fig. 5G). These observations indicate that the her4-positive subpopulation of RG exhibiting neurogenic competence is systematically located in the most posterior aspect of the her4 expression domain at the time of recombination. To analyse whether this neurogenesis-competence window coincided with a location of her4-positive RG close to the TPZ, we applied a BrdU pulse at the time of 4-OHT administration in her4 animals followed by long enough a chase to obtain an identifiable BrdU stripe away from the TPZ. We found that the BrdU-positive and the recombined mCherry-positive stripe were coincident (Fig. 5E). Together, these results demonstrate that her4-positive tectal RG are neurogenic during a transient window of competence that approximately correlates in time with the moment of their generation from the TPZ.

Amplification characteristics and timing of the her4-dependent neurogenesis process in the post-embryonic TeO

The genetic tracing tools developed here further offer the possibility to roughly estimate the time needed for progenitor cells to transit from the her5-positive to the neurogenic her4-positive status. To directly compare the fate of her5- and her4-positive cells in the adult TeO, we recombined her5 and her4 animals in parallel at the same time points. As previously observed (Fig. 3B), her5 and her5 animals displayed a wide distribution of mCherry-positive cells throughout the adult TeO, except for a dorso-anterior component, which we interpret as being generated from cells switching off her5 expression before 1dpf and 5dpf respectively (Fig. 5B1, B2,) (100% of cases, 5 and 12 brains analysed, respectively). In comparison, her4 stripes appeared consistently located anterior to the her5 domain generated by recombination at the same stage (100% of cases, 3 her4 brains analysed per conditions, compare Fig. 5A with 5B, summary in 5C). We interpret the distance between these two mCherry-positive domains as reflecting the time needed by the progeny of her5-positive cells to turn on her4 expression. As a proxy to translate this physical distance into time, we compared the position of BrdU stripes after varying chase times with the anterior limit of the her5-induced mCherry-positive domain. We found that, in the adult brain, a BrdU stripe pulsed at 15dpf is positioned at or close to the anterior limit of the mCherry-positive domain of her5 animals (compare Fig. 5F with Fig. 5B2) (100% of cases, 3 BrdU brains analysed). Thus, at these early juvenile stages, the time needed from her5- to her4-positivity approximates 10 days.

Evidence for an additional, although minor, “direct” route of neuronal generation in the TeO not involving her4-positive progenitors

The results above are in agreement with the existence of the following multistep neurogenic sequence in the post-embryonic TeO: her5-positive PML>TPZ>her4-positive RG>neurons. We next aimed to determine whether a parallel direct neurogenic route from TPZ progenitors, ie. without an intermediate her4-positive state, might also be involved in TeO construction.

As a first, indirect approach, we mapped the connection point of her5/PML-derived polyclones to the TPZ in adult her5 animals. We could capture polyclones just as they exited the TPZ in a few her5 adult animals (n=5 clones) and all were in connection with the ventral most aspect of the TPZ (Fig. 6A), next to the transition domain giving rise to her4-positive RG (Fig. 2A-B).

Fig. 6

Evidence for a minor direct route of TeO neurons generation independent of her4-positive RG progenitors. A. Organization of polyclones at the TPZ border in her4 adults, co-immunostained for mCherry and PCNA and counterstained with DAPI. High magnification of a horizontal section, confocal microscopy. A1: merge; A2, 3: single channels. Note that polyclones extending across the entire PGZ are attached to the most ventricular domain of the TPZ. B, C. Compared fate of her4-positive and BrdU-labelled progenitors in the adult TeO. Top panel: experimental scheme. B. Low magnification of a horizontal section showing an entire TeO hemisphere in a double labelled adult, co-immunostained for mCherry, BrdU and GS, and counterstained with DAPI. C. High magnification views of the area boxed in B, located anterior to the labelled stripe. C1: merge, C1-C3: single channels. Confocal views. Note that recombination of the RG layer anterior to the stripe is complete. D. High magnification views of the stripe area, as boxed in B. D1: merge, D2-D4: single channels. Confocal views. Yellow arrows point to double mCherry and BrdU-positive neurons, green arrow to neurons positive for BrdU only. Scale bars: A, B 100 µm, C, D 50 µm.

We next attempted a direct approach tracing TPZ cells in comparison with her4 progeny cells. her4-derived polyclones coincide with TPZ-derived BrdU-labelled stripes, when Cre recombination and BrdU labelling are concomitant (Fig. 5E). Thus, if an RG-independent neurogenic route also exists, it must take place during an identical temporal window than the neurogenic sequence involving RG progenitors. Thus, to probe the existence of this route, we designed an experiment where all derivatives of her4-positive progenitors are labelled during a long temporal window while a brief (20-min) BrdU pulse is applied during this window to trace TPZ fate (Fig. 6B). In her4 animals analysed as adults at 3mpf, mCherry-positive her4-derivatives distribute, as expected, over a broad transversal stripe across the PGZ. Importantly, all RG within this stripe appear mCherry-positive, which attests of a complete or nearly complete recombination efficiency (Fig. 6C1-4). A brief BrdU pulse was applied at 6dpf, half way through the recombination period. In agreement with the restricted duration of this pulse, BrdU-positive RG and neurons distributed at 3mpf in a narrow transversal stripe included within the mCherry-positive domain (Fig. 6C). We expect neurons formed from a direct, her4-independent route in such as scheme to carry the BrdU label yet be mCherry-negative. Careful analysis of her4, BrdU-pulsed animals revealed that such cells were a minority (Fig. 6D, green arrows) (35.5% ±5% of BrdU-positive cells are mCherry-negative within polyclones, 2 brains analysed, from a total of 199 BrdU-positive cells counted on 9 sections). One needs to keep in mind, in addition, that the possible presence of some individual non-recombined her4-positive RG within the recombined domain will lead to overestimate the her4-independent route.

These results together suggest the existence of a direct route of neuronal generation from the TPZ, but highlight that this route contributes to a minority of TeO neurons.

The neurogenic sequence involved in post-embryonic tectal growth matches the sequence operating in the lateral pallium

We recently described that neurogenic RG NSCs of the lateral pallium originate from a NE progenitor pool located at the lateral edge of the pallial ventricular zone and activated at post-embryonic stages (Dirian et al., 2014). We therefore wondered whether the neurogenic sequence model of the TeO could be extended to interpret lateral pallium construction. Pallial NE progenitors express the e(spl) gene her9 (Dirian et al., 2014) (closely related to mammalian Hes1), and pallial RG express her4 (Chapouton et al., 2010). We thus spatially compared the location of her9- or her4-expressing progenitors at the lateral pallial edge with intermediate markers of the tectal neurogenic sequence, ie. PCNA (identifying a proliferation zone between NE and RG progenitors) and GS (identifying mature RG). Combined triple labelling at adult stage demonstrated that the her9-positive domain partly overlapped a PCNA-positive domain, itself partly overlapping the her4-positive domain, in a distal to proximal sequence (Fig. 7A, B). Thus, an “her9-positive NE pool>proliferation zone>her4-positive RG>neurons” sequence, reminiscent of post-embryonic tectal neurogenesis, operates at the post-embryonic pallial edge (Fig. 7C). Further, in the lateral pallium like in the tectum, the GS- and her4-positive RG domains appeared coincident except for youngest RG cells, which are her4-positive only.

Fig. 7

The sequence of progenitor subtypes at the lateral pallial edge is compatible with a reiteration of the TeO sequence. Compared expression of Her4-GFP, GS and PCNA (A) and her9, Her4-GFP and PCNA (B) in triple labelled cross-sections of the adult pallium, counterstained with DAPI. her9 expression is revealed by fluorescent in situ hybridization, and all other markers by immunocytochemistry. Confocal views. Panels 2–5 are high magnifications views of the boxed areas in panels 1. Panels 1 is the merged view, panels 3–5 are single channels. Red and green arrows in A3, A4 point to the lateral limit of the GS- and Her4-GFP-positive zones, respectively. The red arrowhead in B3 points to the medial limit of the her9-positive domain. Scale bars: 50 µm. C. Interpretative drawing showing the relative positions of the her9, PCNA, her4 and GS-positive zones. This organization is reminiscent of the TeO neurogenic sequence. Progression from a lateral pool to RG was shown in Dirian et al., (2014).

Discussion

This work uses genetic lineage tracing methods to further extend our understanding of the neurogenic sequence building the post-embryonic and adult zebrafish optic tectum. Previous studies in the adult zebrafish and medaka, relying on BrdU incorporation, had highlighted a major proliferation zone, the TPZ, located at the posterior tectal edge, and had shown that the TPZ was at the origin of tectal neurons and RG (Alunni et al., 2010, Ito et al., 2010). The TPZ itself could be subdivided into a ventricular, posterior subpopulation of slower dividing cells with NE characteristics, at the origin of a more anterior growth zone of active proliferation (Alunni et al., 2010, Ito et al., 2010). We now evidence, in the post-embryonic brain, a series of spatially organized progenitor cell type transitions, which includes and brackets the previously identified TPZ: we reveal that her5-positive NE progenitors of the PML constitute the most upstream progenitor pool in this sequence and is at the origin of the TPZ, and that transiently neurogenic her4-positive RG act downstream of the TPZ as a major source of neurons. Such a sequence, shared between the post-embryonic tectum and lateral pallium, is reminiscent of the progenitor transitions occurring during embryonic neurogenesis in vertebrates. The present data and our previous work (Dirian et al., 2014) further demonstrate that, in both the post-embryonic tectum and lateral pallium, the upstream NE progenitor pool is issued in lineage from a restricted embryonic NE pool of similar molecular and cellular characteristics. We propose that the tectum and lateral pallium highlight a shared strategy of post-embryonic neurogenesis that relies on the long-lasting maintenance of NE progenitors and recapitulates an embryonic program of neuronal generation.

The neurogenic sequence of the post-embryonic optic tectum involves several progenitor subtypes and recapitulates an embryonic progenitor progression

Our work identifies her5-positive NE cells as the most upstream cell type in the neurogenic hierarchy of the post-embryonic tectum at least until 5dpf. Two main sets of arguments, combined, support the proposed hierarchy: the first is the generation of tectal RG (and neurons) in her5 animals, and the second the generation of neurons from her4-positive RG in her4 animals. The polyclones obtained in her5 fish vary in their position and extent between samples (Fig. S2G-I), strongly supporting that they result from random mosaicism in Cre recombination, rather than identifying a specific subset of RG. Because most RG appear neurogenic at some time point in their life, the data together largely support a her5 >her4 > neuron lineage. One should keep in mind however that all analyses were conducted at the population level, hence one cannot formally exclude the existence of individual her5-positive progenitors that would fail to give rise to neurogenic RG, or of individual RG that would fail to give rise to neurons. Because her5 expression weakens with age, and possibly also due to the low proliferative activity of these cells, Tg(her5:Cre2ERT2)-driven recombination was very inefficient past 5dpf, and we could not directly assess the progenitor potential of this population at late juvenile or adult stages. We could show however that some Tg(her5:mCherry)-positive cells of the adult PML co-express the proliferation marker PCNA, highlighting their progenitor capacity. In addition, sparsely recombined her5 adults display mCherry-positive neuronal polyclones attached to the TPZ, ie. generated around the analysis time point. In the same brains, a mosaic distribution of mCherry-positive cells can be seen in the adult PML. These results together suggest the maintenance of a late contribution of her5-positive PML cells to tectal construction. An alternative possibility, which we cannot formally exclude, is that the direct contribution of her5-positive PML cells stops on the way to adulthood and is taken over by long-lasting progenitors of the TPZ (Ito et al., 2010), themselves mCherry-positive as issued from the PML.

The contribution of her4-positive RG cells to tectal neurogenesis was previously unreported, and is a major novel finding of this paper. We unambiguously demonstrate this point here using two different genetic ways, Cre-mediated lineage tracing and Tet-on-mediated neuronal birthdating. Perfect overlap of the recombined/induced cells with her4 expression or RG markers validated both techniques. The neurogenic activity of post-natal tectal RG was unexpected, given that this population is virtually never expressing proliferation markers at post-embryonic stages, in contrast to the TPZ, where the bulk of proliferation is observed. Understandably, the limited time window of neurogenic competence of her4-positive tectal RG, and its close physical association with the leading edge of the TPZ, rendered its detection difficult without genetic tracers. At this point, the RG nature of her4-positive cells is based on morphological observations in the her4:gfp and her4:drfp lines, which highlight their overt radial shape with an identifiable basal process. When these cells exit the TPZ, they do not express yet mature RG markers such as GS or GFAP, leaving morphology as only available criterion to categorize them (Than-Trong and Bally-Cuif, 2015). Clearly, her4 expression does not simply correspond to a commitment time point that all TPZ progenitors would transit through: our demonstration of an alternative neurogenic route from the TPZ not involving her4-positive progenitors indeed reinforces the notion that her4 expression highlights a specific, RG-dependent neurogenic process.

Together, our results conclusively implicate her5-positive and her4-positive progenitors in post-embryonic TeO neurogenesis, and demonstrate their lineage relationship. her5-positive cells display overt apico-basal polarity, express progenitor markers (eg. Sox2, Musashi) (this paper and (Chapouton et al., 2006)) but are negative for RG markers such as GFAP or GS (or her4), and are arranged as a polarized monolayer of cuboidal cells linked by tight junctions. Together, these are characteristics of NE progenitors ((Than-Trong and Bally-Cuif, 2015) and references therein). During embryonic development in vertebrates, progenitors transit over time from NE to RG cells, each being associated with a phase of neuronal production: NE progenitors amplify and generate early neurons building a primary scaffold, while later NE-derived RG produce most embryonic neurons (Chapouton and Bally-Cuif, 2004, Gotz and Huttner, 2005). Likewise, we show that, in the post-embryonic tectum, her5-positive NE cells are at the origin, via the TPZ, of two neurogenesis routes: a major route involves her4-positive RG, while a more minor route is direct and independent of her4-positive RG. Thus, post-embryonic tectal neurogenesis recapitulates embryonic sequences of progenitor transitions. Because tectal neurogenesis is continuous from NE progenitors, however, and because neurogenesis occurs from a discrete NE source without overt cell migrations, NE and RG cells are concomitantly present in the post-embryonic tectum and their transition is ordered in space. It will remain to be determined whether the neurons generated from both the RG-dependent and direct routes are of different subtypes or endowed with different functions. Strikingly, as revealed by the coincident BrdU- and Her4-mCherry-positive stripes in our double tracing experiments, they exit the TPZ at or near the same time to integrate into identical transversal columns across the PGZ.

A similar arrangement of her4-positive, then GS/GFAP-positive RG can be observed in the torus semicircularis (homologous to the mammalian inferior colliculus) as one moves away from the her5-positive zone and PML (Fig. 1A, not shown). Although we have not studied neurogenesis in this territory, mCherry-positive RG and neurons were obvious on our preparations following her5-driven Cre recombination at post-embryonic stages (see Fig. 3A). Likewise, we had previously shown that post-embryonic her5 progenitors generated neurons in the tegmentum (Chapouton et al., 2006). We therefore propose that the post-embryonic her5-positive NE pool serves as a common source for the construction of the TeO, torus semicircularis and tegmentum, acting as a hub that orchestrates in time and space the construction of the midbrain. This appears similar to the function of the embryonic PML (Recher et al., 2013). Whether similar progenitor hierarchies are used in all domains remains to be demonstrated. It is to note that our data also demonstrate a contribution of early progenitor cells expressing her5 (until 1dpf) to the adult cerebellum (Fig. 3A,B), where NE cells have been described to serve as a source for granule neurons (Kaslin et al., 2009). NE progenitors of the adult cerebellum however do not maintain her5 expression and are directly neurogenic without a RG intermediate (Kaslin et al., 2009).

The optic tectum and lateral pallium highlight a shared mode of post-embryonic neurogenesis associated with tissue growth

Our results highlight striking similarities in the neurogenic sequences operating in the post-embryonic TeO and lateral pallium: in the TeO, our genetic tracing identifies a major “her5-positive NE>TPZ>her4-positive RG>neuron” sequence; in the lateral pallium, the compared expression analyses presented here, together with her4 tracing (Dirian et al., 2014), are suggestive of a “her9-positive NE>PCNA-positive domain>her4-positive RG>neuron” progression. The similarity in the neurogenesis process in these two domains extends to three further points. First, at the molecular level, the NE progenitor pools express specific e(spl) genes, her5 and her9, that were shown not to require Notch signalling for their expression in certain contexts (as opposed to her4, which is a systematic Notch target) (Bae et al., 2005, Geling et al., 2004, Radosevic et al., 2011, Stigloher et al., 2008). Whether these two territories are direct functional equivalents using distinct Her factors, or whether her5 versus her9 expression reflect some degree of heterogeneity within NE pools remains to be addressed. Second, at the lineage level, these pools are issued from a restricted, long-lasting embryonic progenitor population expressing the same molecular signature (this work, and (Dirian et al., 2014)). Finally, a striking feature of these post-embryonic NE pools is their extremely small size (a few hundreds of cells at most) compared to the huge territories that they generate. The latter observation is in line with their intermediate generation of large amplification domains (the TPZ at the posterior edge of the TeO, and a PCNA-positive domain partially overlapping the her9-positive zone in the lateral pallium). Together, these observations identify a distinct neurogenic strategy operating in the post-embryonic TeO and lateral pallium, which relies on small NE pools inherited from the embryonic neural tube and serving both as growth zones and long-lasting generators of young, adult-born RG progenitors (Fig. 8). The two domains differ however in the neurogenic activity of their RG progeny: while this activity under physiological conditions is very transient in the TeO, it is, at the population level, permanent in the lateral pallium, generating neurons throughout life. What accounts for this difference in the duration of the RG proliferation and neurogenic phases, and in particular whether it results from an intrinsic change in RG potential or is rather due to environmental cues, is unknown at present. In this context, it will be particularly interesting to molecularly compare pallial and tectal RG. An organization similar to that of the TeO is also found in the retina (Goldman, 2014, Lenkowski and Raymond, 2014), with a NE ciliary marginal zone generating a proliferation zone and her4-positive glia. Whether the latter cells serve as transient neurogenic intermediates, extending the model above, remains to be determined.

Fig. 8

Compared neurogenic sequences in the post-embryonic TeO and lateral pallium. Schematic representation of the similarities between the TeO and lateral pallium neurogenic sequences. The arrows indicate hierarchical relationships, and cell types and gene expression are color-coded. The indirect (i) and direct (ii) neurogenic routes evidenced for the TeO are indicated (in the lateral pallium, the existence of a direct route has not been studied). Both territories maintain NE progenitors, inherited from an embryonic NE pool expressing e(spl) genes, and that serve both as a growth zone and as a RG source. However, RG are transiently neurogenic in the TeO (this study) while they are constitutively neurogenic in the pallium (Dirian et al., 2014). In both cases, a pool of NE progenitor cells (red, left panels) inherited in lineage from an embryonic NE pool (right panels) is at the origin of neurogenic RG cells (green) via a transit amplification pool (blue). Post-embryonic RG of the lateral pallium generate neurons (purple) life long, while those of the TeO rapidly switch to a non-neurogenic state (green, dark grey surrounding). Horizontal grey lines represent the ventricular surface. Gene expression is color-coded (dotted: intermittent expression). Arrows indicate lineage relationship (demonstrated at the population level in the case of the TeO). Note that the embryonic NE pool of the lateral pallium expresses the two e(spl) genes her6 and her9 (Dirian et al., 2014). Abbreviations: PML: posterior midbrain layer; TPZ: tectal proliferation zone.

The neurogenic mode of the post-embryonic lateral pallium and TeO appears very different from the sequence at play in the post-embryonic dorsal pallium, where the population of RG NSCs is inherited from RG progenitors and amplifies by symmetric RG divisions. At this point, we may only speculate about the respective implications of these two neurogenesis modes. We previously demonstrated that the Notch signalling-independency of the lateral pallium NE pool allowed reconstitution of a RG NSC population following Notch inhibition at post-embryonic stages, while this failed in the dorsal pallium (Dirian et al., 2014). Whether this is also the case in the TeO remains to be tested. It is possible also that RG generated from NE progenitors at post-embryonic stages have younger characteristics than those originating from embryonic RG. Finally, in systems such as those studied here that minimally involve cell migration, post-embryonic NE pools permit the oriented growth of their dependent brain territories, as opposed to the relatively homogeneous growth observed in the dorso-medial pallium. This is likely to bear consequences on both the diversification of morphologies across brain territories and the age-dependent organization of neurons, and presumably of circuits, in the mature brain.

An important question will remain to assess whether the NE-dependent neurogenesis mode highlighted here, and shared between remote brain territories in zebrafish, is encountered in other species and notably in mammals. Along this line, we note that a Notch-independent territory has been described at the posterior edge of the mouse midbrain (Lutolf et al., 2002).