Expression of a Barhl1a reporter in subsets of retinal ganglion cells and commissural neurons of the developing zebrafish brain

Promoting the regeneration or survival of retinal ganglion cells (RGCs) is one focus of regenerative medicine. Homeobox Barhl transcription factors might be instrumental in these processes. In mammals, only barhl2 is expressed in the retina and is required for both subtype identity acquisition of amacrine cells and for the survival of RGCs downstream of Atoh7, a transcription factor necessary for RGC genesis. The underlying mechanisms of this dual role of Barhl2 in mammals have remained elusive. Whole genome duplication in the teleost lineage generated the barhl1a and barhl2 paralogues. In the Zebrafish retina, Barhl2 functions as a determinant of subsets of amacrine cells lineally related to RGCs independently of Atoh7. In contrast, barhl1a expression depends on Atoh7 but its expression dynamics and function have not been studied. Here we describe for the first time a Barhl1a reporter line in vivo showing that barhl1a turns on exclusively in subsets of RGCs and their post-mitotic precursors. We also show transient expression of barhl1a:GFP in diencephalic neurons extending their axonal projections as part of the post-optic commissure, at the time of optic chiasm formation. This work sets the ground for future studies on RGC subtype identity, axonal projections and genetic specification of Barhl1a-positive RGCs and commissural neurons.

BarH-like homeodomain (BARHL) transcription factors play crucial roles in the control of neural cell fate specification, migration, subtype identity acquisition and survival during development of the retina and the brain [1][2][3][4][5][6][7] . Studies have also implicated Barhl in neurodegenerative and neoplastic disorders 8,9 . Previous barhl studies have suggested that whole genome duplication (WGD) during vertebrate evolution generated two homologs of barhl: barhl1 and barhl2 10 . In addition, studies in Xenopus and mouse suggest that barhl2 (previously named MBH1 and XBH1) is the only family member expressed in the retina of tetrapods 11,12 . Specifically, barhl2 expression is found in both amacrine cells and retinal ganglion cells (RGCs) of the developing and mature retina 13,14 . Furthermore, Barhl2 is both sufficient and essential for determining the subtype specific identity of amacrine cells as well as to promote the maturation and survival of RGCs. Lastly, the expression of barhl2 in RGCs appears to depend on Atoh7 (also known as Ath5) -a bHLH transcription factor required for the specification of RGCs in vertebrates [13][14][15][16][17][18][19] .
In Zebrafish, three barhl paralogs were found 4,7,10 . Studies based on both Barhl protein sequence analysis and conserved gene synteny between barhl genes suggest that they likely arise from the additional round of WGD in teleosts followed by the loss of one barhl paralog 10 . These studies also have revealed that barhl1b (previously called barhl1 or barhl1.1) and barhl1a (previously called barhl1.2) are more closely related to the mammalian barhl1. Conversely, the barhl2 paralogue is more closely related to the mammalian barhl2 10 . Concordantly, similarly to its barhl1 counterpart in tetrapods, barhl1b lost its retinal expression, maybe due to a redundant function with barhl1a and relaxed evolutionary pressure in its locus 4,10 . On the contrary, retinal expression of barhl1a and barhl2 is retained, but these two paralogs appear to have diversified their function in retinal lineages 10 . In favour of this hypothesis, studies have shown that, similarly to the mammalian counterpart, Barhl2 is an amacrine cell subtype identity-biasing factor downstream of the transcription factor Ptf1a 20 . Furthermore, taking advantage of the zebrafish transgenesis combined with accessibility to 3D time-lapse imaging [21][22][23][24][25] we have previously shown that barhl2-expressing amacrine subtypes consistently arise within the lineage of Atoh7 upon reproducible asymmetrical divisions of RGC progenitors 20 (see also Fig. 9). Interestingly, while barhl2 turns on exclusively in amacrine cells under the control of Ptf1a 20 , the expression of barhl1a depends on Atoh7 and appears therefore to be restricted to the ganglion cell layer (GCL) 10 . The generation pattern of individual barhl1a-expressing cells within the Atoh7 cell lineage, the lineage relationship among barhl1a and barhl2 -expressing cells in vivo as well as the role played by Barhl1a in RGC genesis have so far remained unknown.
To address these questions, direct and dynamic visualization of barhl1a-expressing cells is needed. We used Tol2 transposase-mediated BAC transgenesis to generate a Barhl1a reporter line where GFP expression is driven by barhl1a regulatory regions. This reporter recapitulates barhl1a mRNA expression and helped us determine that barhl1a turns on exclusively in a sub-population of RGCs after the cell cycle exit of atoh7-expressing progenitors. In addition, the reporter allows us to visualize axonal connectivity derived from barhl1a-expressing neurons. Furthermore, we provide the first description of the expression dynamic of barhl1a in commissural neurons of the forebrain. This study thereby provides the foundation for further investigations of the role of Barhl1a transcription factors in nerve cell subtype identity acquisition and maintenance in this in vivo model system.

Results
Barhl1a in the retina is up-regulated in a subset of atoh7-expressing post-mitotic RGC precursors.
In order to visualize barhl1a-expressing cells in vivo we generated a Tg(barhl1a:GFP) transgenic line expressing the reporter eGFP under barhl1a regulatory genomic elements. To this end, a BAC spanning the barhl1a genomic locus was used to perform Tol2 transposon-mediated BAC transgenesis replacing eGFP with the barhl1a coding sequence (see Methods). The newly generated Tg(barhl1a:GFP) transgenic embryos displayed the distribution pattern of GFP as expected from barhl1a expression in the posterior thalamus and zona limitans intratalamica (ZLI) 4 ( Supplementary Fig. 1A). To further assess the reliability of the Tg(barhl1a:GFP) line, we compared the spatial expression pattern of barhl1a mRNA and GFP mRNA in Tg(barhl1a:GFP) embryos by double fluorescent in situ hybridization (FISH). Transcripts of both genes can be consistently found as shown, for example, at 35 and 40 hours post-fertilization (hpf), in the developing GCL as well as in the optic tectum ( Fig. 1).
To directly investigate the dynamics of the appearance of individual Barhl1a:GFP-positive cells in the retina, we used 3D time-lapse imaging on Tg(barhl1a:GFP) embryos. Imaging of the developing retina of Tg(barhl1a:GFP) embryos for at least 20 hours revealed that GFP positive cells become first visible in the ventral retina at ~ 29-30 hpf and spread subsequently to the nasal, dorsal and temporal retina in a temporal and spatial fashion predictive of the wave of RGC differentiation ( Fig. 2A-D and Supplementary Movie 1) 26 . Individual cells turning on barhl1a:GFP can be seen located either in the basal half or in the apical half of the neuroepithelium, from where they subsequently migrate basally until they reach the GCL ( Fig. 2A-D and Supplementary Movie 1). We never observed Barhl1a:GFP cells migrating to the apical surface to undergo division. This suggests that barhl1a:GFP turns on in RGC precursors after the terminal division of atoh7-expressing neuroepithelial cells. Immuno-labelling of Tg(barhl1a:GFP) embryos with an antibody to Cxcr4b, a marker of post-mitotic RGCs 27 further supports our notion that barhl1a:GFP is expressed in RGCs and their post-mitotic precursors ( Fig. 2E-G').
To further investigate expression of Tg(barhl1a:GFP) with respect to atoh7 we generated a Tg(barhl1a:GFP ;atoh7:gap43-RFP) double transgenic line in which the Atoh7:gap43-RFP labels all atoh7-expressing progenitors and their derived retinal cell types, including barhl2-expressing amacrine cells 20,28 . Studies have shown that Atoh7:gap43-RFP becomes first visible in the anterior ventral retina already at 27-28 hpf, underlying the onset of retinal neurogenesis 20,28,29 . Consistently, GFP expression from the Barhl1a reporter could be seen appearing at ~30 hpf, hence with two or three hours delay after the onset of atoh7:gap43-RFP expression ( Fig. 3A-A"). Likewise, consistently with the idea of Barhl1a:GFP being restricted to post-mitotic RGCs, GFP expression was always seen in the basal half of the neuroepithelium whilst it remains excluded from the retinal apical surface where mitosis takes place ( Imaging in the retina of Tg(barhl1a:GFP;atoh7:gap43-RFP) embryos also revealed that not all Atoh7:gap43-RFP cells in the GCL appear as Barhl1a:GFP positive, as it would be expected if Barhl1a labels a subpopulation of RGCs (see asterisks in Figs. 2G' , 4A-A' ,C-C"). To assess this, Barhl1a:GFP positive cells were counted in the central retina of 50 hpf, Tg(barhl1a:GFP;atoh7:gap43-RFP) embryos immunostained against GFP and Zn5 (also known as Alcama/DM-GRASP/Neurolin/Zn8 30 ), a cell adhesion molecule that is transiently found on the entire surface of differentiated RGCs, and is therefore a suitable marker for labeling all RGCs 30 . Quantification of the Barhl1a:GFP nuclei shows that, on average, only about 54% of the RGC population (1391 cells, n = 7 retinae) is also Barhl1a:GFP positive (Fig. 4B and Supplementary Table 1).
Besides RGCs, the GCL also contains displaced amacrine cells. These displaced amacrine cells are however present in very low numbers, which would not largely affect the percentage of Barhl1a:GFP RGCs. We aimed nevertheless to rule out the possibility that some of the GFP-negative cells in the GCL may be displaced amacrine cells. To this end we generated a Tg(barhl1a:GFP;ptf1a:dsRed) double transgenic line, in which the ptf1a:dsRed transgene labels all amacrine and horizontal cells [31][32][33] (Fig. 5). Analysis of the retinae of Tg(barhl1a:GFP;ptf1a:dsRed) embryos at 50 hpf revealed that expression of the two reporter genes are mutually exclusive, with all displaced amacrine cells expressing the Ptf1a reporter being Barhl1a:GFP-negative (Fig. 5). This further confirms that expression of barhl1a is restricted to a RGC population.
We conclude that barhl1a turns on in about half the RGC precursors. The other half of the atoh7-expressing RGC precursors would then differentiate without the direct contribution of Barhl1a in the retina. The Barhl1a reporter line offers the advantage to visualize not only the development of neuron cell bodies in the retina and brain but also of their axonal processes in vivo (Fig. 6). Imaging analysis of the brain of Tg(barhl1a:GFP) embryos indeed revealed, besides the optic chiasm, other two early commissural tracts deriving from barhl1a-expressing cells. A first bundle of GFP positive fibers become visible at around 20-22 hpf, extending on both sides of the ventral-anterior diencephalon towards the midline, and forming a commissure around the presumptive optic tract (Fig. 6B  axons can be clearly identified; which exited the retina and crossed the ventral midline to form the optic chiasm 45 (Fig. 6C' ,D"). Conversely, only very few of these gap43-RFP -positive axons are also Barhl1a:GFP-positive ( Fig. 6C-C"), further suggesting that the birth and/or maturation of Barhl1a:GFP RGCs is delayed with respect to the first wave of RGC genesis. Rotating the reconstructed 3D volume further highlighted the two contiguous but distinct commissures identified by the Atoh7 and Barhl1a reporters, respectively ( Fig. 6D-D"). Lastly, Tg(barhl1 a:GFP;atoh7:gap43-RFP) embryos were co-labeled immuno-histochemically with antibodies against acetylated Tubulin and GFP. The GFP signal overlaps, albeit only partially with that of acetylated Tubulin in the POC fibers ( Fig. 7), further indicating that the Barhl1a:GFP cell projections contribute to this diencephalic commissure.
Notably, another bundle of Barhl1a:GFP positive axons could be seen at around 27 hpf, which extend dorsally and cross the midline at the presumptive rostral border of the developing tectum ( Fig. 6B,C" and Supplementary Movie 6). Based on their temporal appearance and anatomical position, these commissural fibers are likely to grow along the path pioneered by the posterior commissure (PC) and might be therefore residing in the ventral midbrain or pretectal nuclei (Supplementary Movie 7) 36,38,42 .
In the pseudostratified neuroepithelium of the neural tube, post-mitotic neurons and glia are generated from progenitor cells that span the entire thickness of the neuroepithelium, from the basal to the apical surface. The nuclei of these progenitor cells undergo interkinetic nuclear migration, with mitotic nuclei being located at the apical surface whilst differentiating cells migrate to more basal locations 46,47 . Since Barhl1a:GFP-positive cells span the entire thickness of the diencephalic neuroepithelium, we asked whether these cells would be found at different developmental phases of neurogenesis. Tg(barhl1a:GFP) neurons were firstly labelled immune-histochemically with a HuC/Elav antibody labelling the cell body of differentiated neurons 48 . Overlap of HuC/Elav and GFP occurs in cells located further away from the apical (ventricular) surface, indicating that they are post-mitotic cells (Fig. 8A,A'). Conversely, GFP positive cells immediately adjacent to the apical ventricular surface expressed the mitotic marker phospho-histone 3 (pH3) 49 , indicating that they are mitotic precursors, as well as could be seen undergoing mitotic divisions ( Fig. 8B-B'). Neuronal progenitors in the zebrafish neuroepithelium have been reported to express markers of radial glia, such as the glial acidic fibrillary protein (GFAP) 48,50 . Mosaic expression of an gfap:lss-kate construct, where the promoter of gfap has been cloned upstream of the lss-mKate1 gene, highlighted GFAP:Lss-mKate1/Barhl1a:GFP double positive cells at different apical-basal distances within the thickness of the neuroepithelium (Fig. 8C). In summary, the observed spatial-temporal dynamics of Barhl1a:GFP are consistent with Barhl1a being a subtype identity factor for populations of commissural neurons in the brain as well as in the retina. The Barhl1a reporter indeed specifically labels particular neuronal subsets in the forebrain and their extending axons, which form commissures across brain hemispheres. These barhl1a:GFP-expressing axons follow pre-existing embryonic commissures: the POC, the PC and the optic chiasm in a stereotypic pattern that is reminiscent of the previously described developmental sequence of the formation of archetypal tracts 36,37,51,52 .

Discussion
Barhl genes have a crucial role in the control of neuronal subtype acquisition and maintenance during development and in the adult 1,2,5,6,8,15,53 . Loss of Barhl2 protein in the postnatal mammalian retina causes programmed cell death of more than 50% of RGCs 14 . Therefore, a better understanding of the Barhl function is essential to prevent RGC death and/or enable RGCs and axon regeneration. However, the mammalian Barhl2 is concomitantly expressed in amacrine cells and it is required for their subtype identity acquisition 14,15 . Untangling these two concurrent functions of Barhl2 might therefore be challenging in mammals. Barhl proteins play evolutionarily conserved roles in retinal cell type maturation 13,20 . Furthermore, the two zebrafish barhl paralogues genes, We here started to explore whether it could be possible to disentangle the dual function of the single mammalian Barhl2 gene in the retina, by analyzing its zebrafish orthologous genes. With the generation of the Interestingly, these Barhl1a:GFP RGCs represent only about 54% of the RGCs population, which appear to display distinct retinotectal projections. There is a delay in the appearance of Barhl1a:GFP-positive axons compared to the appearance of Atoh7:gap43-RFP axons. One explanation for this delay might be that Barhl1a:GFP cells comprise a population of later-born RGCs, which grow their axons along the optic tract pioneered by the early-born Atoh7-positive/Barhl2-negative RGCs 52 . Alternatively, the delayed onset of GFP could also reflect the requirement of barhl1a expression for the late maturation and/or maintenance of RGCs. Conclusions cannot be drawn here and future studies using the Tg(barhl1a:GFP) line and barhl1a-inducible Gal4 constructs may be used to investigate the distinct molecular signature, morphology, dendritic patterns and retinotopic targets of Barhl1a RGCs.
We have previously shown that about 58% of amacrine cells express barhl2 in the Zebrafish retina, all of which seem to arise from the asymmetric divisions of atoh7-expressing progenitors. Moreover, Zebrafish Barhl2 is both necessary and sufficient for subtype identity acquisition of these amacrine cells downstream of Ptf1a 20 . Given the remarkably similar temporal and subtype-restricted expression pattern of barhl1a, it will be interesting in future studies to examine if and how Barhl1a determines particular RGC subtype identities and retinotopic projections, and whether these subtypes tend to have recurrent clonal relationships and connectivity patterns with their lineally-related Barhl2 amacrine subtypes (Fig. 9). Lastly, the fact that Barhl1a RGCs and Barhl2 amacrine cells are lineally related (that is: they arise from the same type of progenitor) suggests that these retinal cells share gene expression profiles and/or epigenetic ancestry. It will be exciting to explore whether such insights gained from a genome duplication in zebrafish will allow to narrow down the shared and distinct gene networks that are  www.nature.com/scientificreports www.nature.com/scientificreports/ relevant for the role of Barhl in the maturation of RGCs and amacrines as well as in the survival of RGCs in the mammalian retina. Understanding this could be instrumental, for instance, for the identification of key factors whereby reprogramming of lineally-related retinal cell types, such as Barhl-dependent amacrines and RGCs, could be achieved to prevent RGC death or encourage their regeneration from amacrine cells 54 .
Unlike in the retina, where Barhl1a:GFP is restricted to post-mitotic neurons, barhl1a:GFP in the brain is turned on in mitotic progenitors and retained in their post-mitotic daughters. Using time-lapse analysis in the brain of Tg(barhl1a:GFP) embryos we discovered for the first time that Barhl1a:GFP in the diencephalon labels small populations of cells contributing to the POC and the PC tracts. We also found partial overlapping expressions of barhl1a and lmx1b1, a regulator of dopaminergic neurons 44 . It will be intriguing to assess in future studies whether these neurons co-expressing lmx1b1 and barhl1a are the ones from which such commissures arise. Also interestingly is the finding that the onset of Barhl1a:GFP-labeled POC fibers coincides with the onset of RGC axon maturation and is transient, i.e. it disappears when the optic tract is fully differentiated. Lastly, it was also intriguing to find that, whilst the Barhl1a:GFP labelled POC neurons also express GFAP, a marker of progenitors and glial cells 48,50 , the Barhl1a:GFP labelled POC fibers only partially co-localize with the POC axonal marker acetylated tubulin. Future studies will assess the precise (glial and/or neuronal) identity of Barhl1a:GFP commissural fibers; they will also assess whether the remarkably harmonized developmental pattern between the Barhl1a:GFP-labeled POC and optic chiasm may underlie a molecular cross-talk required for the proper formation of the optic tract.
In conclusion, our novel experimental tools and insights into the spatial-temporal dynamics of Barhl1a:GFP in the retina and brain provide a fundamental framework for further investigation of Barhl1a-specific RGCs and their retinotectal RGC projections as well as of Barhl1a-commissural projection neurons in vivo. Such investigations are likely to be relevant to dissect BarHl functions in mammals and to understand and ameliorate the pathophysiology of BarHl-linked diseases.

Methods
Animals and ethic statement. Fish were maintained at 26-28 °C and embryos raised at 28.5 °C and staged as described previously 55,56 . Fish were housed in two facilities: Fish facility at COS, University of Heidelberg, Germany; fish facility at CIBIO, University of Trento, Italy. Each facility is under supervision of and in accordance with local animal welfare agencies and European Union animal welfare guidelines (Tierschutzgesetz 111, Abs. 1, Nr. 1; Regierungspräsidium Karlsruhe and the Italian Ministry of Health -permit no.: 151/2019-PR). As zebrafish sex cannot be determined until they have reached age of reproduction, embryos of either sex were used exclusively before free-feeding stages. Embryos used for whole-mount imaging were treated with 0.0045% 1-phenyl-2-thiourea (Sigma) to delay pigment formation. Lines used in this study were generated in the zebrafish wild type background (AB/AB or AB/WIK). The fish facility is under the supervision of the local representative of the Animal Welfare Agency.

Fish lines and constructs.
For the generation of the Tg(barhl1a:GFP) line, the BAC clone CH120-215H7 (CHORI, BACPAC, Oakland, CA, USA) was modified using the flpe recombinase system in EL250 cells 57 . The coding sequence of eGFP with a SV40 polyA site was inserted within the first exon of the barhl1.2 genomic sequence and the tol2-AmpR cassette was inserted in the TARBAC2.1 backbone by PCR. The modified BAC was injected together with tol2 mRNA into one cell-stage zebrafish embryos to generate a stable transgenic line.
The barhl1a:gal4 bacterial artificial chromosome (BAC) construct was generated as follows: a BAC spanning the barhl1a genomic locus (CH211-215C18, BACPAC Resources Center) was used to perform Tol2 transposonmediated BAC transgenesis replacing gal4 with the barhl1a coding sequence. Transformation through electroporation of the pRedET plasmid was performed as described 59 . For Tol2 transposon-mediated BAC transgenesis, the iTol2-amp cassette 59 was amplified by PCR with the primer pair pTarBAC_HA1_iTol2_fw (5′-gcgta agcggggcacatttcattacctctttctccgcacccgacatag atCCCTGCTCGAGCCGGGCCCAAGTG-3′) and pTarBAC_HA1_ iTol2_rev (5′-gcggggcatgactattggcgcgccggatcgatccttaattaagtctactaATTATGATCCTCTAGATCAGCTC-3′), where the lower and upper cases indicate the pTarBAC2.1 sequences for homologous recombination and the iTol2-amp annealing sequences, respectively. Subsequently, the amplified iTol2-amp cassette was introduced into the backbone (pTarBAC2.1) of the Barhl1a-BAC. 500 ng of the PCR product (1 mL) were used for electroporation. www.nature.com/scientificreports www.nature.com/scientificreports/ 5′-caaaaccagtgtcataaaggacaaatgcacatttgatattgatttgactcGCCACCATGAAGCTACTGTCTTCTATCGAAC-3′ and 5′-ctgtgagaaagtatagactcgatcccaaagctcgagccgtttgatacctcCCGCGTGTAGGCTGGAGCTGCTTC-3′ primers were used to amplify and insert the gal4FF cassette into the BAC 60 . The lower and upper cases indicate the CH211-215C18 sequences for homologous recombination and the pGal4FF-FRT-Kan-FRT annealing sequences, respectively. 500 ng of the PCR product (1 mL) were used for electroporation in Barhl1a-iTol2-amp-BAC -containing cells. The BAC DNA preparation was performed using the HiPure Midiprep kit (Invitrogen), with modifications for BAC DNA isolation as described by the manufacturer. Tol2 transposase mRNA was prepared by in vitro transcription from XbaI-linearized pDB600 61 using the T3 mMessage mMachine kit (Ambion). RNA was purified using the RNeasy purification kit (Qiagen), diluted to a final concentration of 100 ng/μl for injection. At least 20 injected fish were backcrossed to wild type. Germline transmission was observed in the offspring from two of such crossings; which displayed the distribution pattern of GFP as expected from barhl1a expression in the posterior thalamus and zona limitans intratalamica (ZLI) (Supplementary Fig. 1A). Only one of the two F1 generations was used to carry out all of the following analyses.
Immunohistochemistry. The primary antibodies used in this study and their dilutions were the following:  , blocked using 10% blocking solution (10% heat-inactivated goat serum, 1% bovine serum albumin, 0.2% Triton X-100 in PBS) and incubated overnight at 4 degrees in 1% blocking solution (10% blocking solution in 1X PBS-Tween) in which primary antibodies were diluted. Primary antibodies were then washed out using 1X PBS-Tween and embryos were then incubated overnight at 4 degrees in 1% blocking solution (1% heat-inactivated goat serum, 1% bovine serum albumin, 0.2% Triton X-100 in PBS) in which secondary antibodies and 4′,6-diamidino-2-phenylindole (DAPI) were diluted. Prior to imaging, embryos were washed in 1X PBS-Tween and stored at 4 degrees.
For the immunohistochemistry on sections and RGC counting, 50 hpf Tg(barhl1a:GFP;atoh7:gap43-RFP) embryos were fixed in 4% paraformaldehyde (PFA) in 0.1 M phosphate saline buffer (PBS) overnight at 4 °C, rinsed and cryoprotected in 30% sucrose (w/v) overnight at 4 °C. Embryos were then mounted vertically (head down) in freezing medium (Jung Tissue Freezing Medium, Leica Microsystems), frozen in liquid nitrogen and cryosectioned immediately with Leica CM3050 S cryostat. The thickness of sections was 14 μm. Sections were collected on adhesion microscope slides (SuperFrost Plus, Menzel-Gläser) and left to dry overnight at 4 °C. For immunohistochemistry, microscope slides with cryosections were washed 3 times (each wash 15 min) in PTW, then covered with 10% goat blocking medium and incubated at room temperature for 1 h. The cryosections were then incubated overnight in primary antibodies: anti-chicken GFP (Life Technologies, A10262) and anti-mouse Zn5 (Zebrafish International Resource Center (ZIRC)) both diluted 1:500 in 1% goat blocking medium. Microscope slides were then washed again 3 times (each wash 15 min) in PTW and secondary antibodies were added: anti-chicken conjugated to Alexa Fluor 488 (from donkey; Jackson ImmunoResearch Laboratories, Inc., 703-545-155) and anti-mouse conjugated to Alexa Fluor 546 (from goat; Invitrogen, A-11030) both diluted 1:500 in 1% goat blocking medium. Cryosections were incubated with the secondary antibody mix for 1.5-2 h at 37 °C in the dark and then washed again in PTW (3 times 15 min) at room temperature. They were then incubated in DAPI (10 μg/ml in PTW) for 10-15 min and washed in PTW (3 times 15 min). Microscope slides were then dried from the back and edges with a paper tissue, 120 μL 60% glycerol was added and cryosections were covered with coverslips (24 × 60 mm, Carl Roth). Coverslips were sealed with nail polish and microscope slides were stored at 4 °C in the dark until imaging.  www.nature.com/scientificreports www.nature.com/scientificreports/ Signal Amplification, as described by 10 . Barhl1a riboprobe was synthesized as previously reported 10 and gfp riboprobe was synthesized from a linearized pCS2:GFP plasmid with NotI (Fermentas or New England Biolabs) and transcribed with Sp6 (mMessage mMachine Sp6, Ambion). Incubation with the barhl1a probe was for 40 min, incubation with the gfp probe was for 30 min. Embryos were then kept in the dark for all following steps. For detection and staining of the antisense probes, embryos were washed 5 times 10 min with TNT (0.1 M Tris pH7.5, 0.15 M NaCl, 0.1% Tween-20), incubated with 1% H 2 O 2 in TNT for 20 min and washed again 5 times 10 min. Embryos were blocked in TNB (2% DIG Block in TNT) for 1 h at room temperature and afterwards incubated with Anti-Digoxigenin-POD Fab fragments diluted 1:100 in TNB. For signal detection, Fluorescein (FITC), Cyanine 3 (Cy3) or Cyanine 5 (Cy5) Fluorophore Tyramide by PerkinElmer was used. Embryos were then incubated in 1 × 4′,6-Diamidin-2-phenylindol (DAPI) in TNT overnight at 4 °C and washed several times in TNT the next day. The stained embryos were then used for imaging or kept in the dark at 4 °C until imaging.
For barhl1a/lmx1b1 double fluorescent in situ whole mount hybridization (FISH), antisense RNA probes for barhl1a and lmx1b1 were produced from full-length cDNA clones generated by 64 . Antisense probes were generated as described in 64  Imaging and quantification. For either imaging of fixed embryos or live imaging, embryos were embedded in 1% low-melting agarose (w/v, diluted in H 2 O; Biozym) in 35 mm Glass-bottom Microwell dishes (P35G-1.5-10-C, MatTek). They were oriented with truncated Microloader tips (Eppendorf) frontally (head touching the dish and body axis tilted about 45 °C) or laterally. After solidification, agarose was covered with a drop of water to avoid dehydration during imaging. Confocal microscopy was carried out with Leica SpE or Leica Sp5 confocal laser scanning microscope using a Leica 20× or 60 × 1.2 NA water-immersion objective and Leica Application Suite (LAS) software. In most cases, confocal images were taken sequentially from the whole head/retina, with a distance of 1 to 5 μm in z-direction. Optical stack of 40-60 μm for time-lapse and a maximum of 100μm for fixed embryos were imaged. Live imaging was performed as previously described [21][22][23] . Images were taken every 5 or 10 minutes for 24-42 h and the motorized XY stage was used to image multiple embryos. For all imaging experiments, laser power was minimized as much as possible to avoid bleaching and photo-toxicity. Sequential image acquisition was also performed using individual descanned Leica PMT detectors. All image processing and 3D reconstructions were done using either Volocity 6.0.1 (PerkinElmer) or Fiji (https://imagej.net/ImageJ). All data presented in this study derive from observations made in at least 10 embryos/experiment in the case of fixed samples and on at least 3 embryos/experiment in the case of live imaging. In order to quantify the number of Barhl1a:GFP cells in the GCL, 14 µm cryosections of 50 hpf (barhl1a:GFP;a-toh7:gap43-RFP) double transgenic embryos were immunostained against GFP and Zn5 (also known as Alcama/ DM-GRASP/Neurolin/Zn8), a cell adhesion molecule that is transiently found on the entire surface of differentiated RGCs 30 . Barhl1a:GFP positive cell nuclei were counterstained with DAPI. Sections were imaged with a Leica DM5000B compound microscope (20x objective). Three most central sections of each retina (n = 7 retinae from 6 embryos) were imaged. Both GFP positive and GFP negative nuclei were counted manually in Volocity Analysis version 5.3 (Improvision). To exclude cells in the periphery, where the CMZ continues to give rise to all retinal cell classes, only cells within the central quadrant of the retina were counted as previously described 20 . For every retina, the percentage of GFP-positive cells was calculated as a mean of the proportions observed in each individual section.