Sox1a mediates the ability of the parapineal to impart habenular left-right asymmetry

Left-right asymmetries in the zebrafish habenular nuclei are dependent upon the formation of the parapineal, a unilateral group of neurons that arise from the medially positioned pineal complex. In this study, we show that both the left and right habenula are competent to adopt left-type molecular character and efferent connectivity upon the presence of only a few parapineal cells. This ability to impart left-sided character is lost in parapineal cells lacking Sox1a function, despite the normal specification of the parapineal itself. Precisely timed laser ablation experiments demonstrate that the parapineal influences neurogenesis in the left habenula at early developmental stages as well as neurotransmitter phenotype and efferent connectivity during subsequent stages of habenular differentiation. These results reveal a tight coordination between the formation of the unilateral parapineal nucleus and emergence of asymmetric habenulae, ensuring that appropriate lateralised character is propagated within left and right-sided circuitry.


Introduction
Although once considered to be a mark of cognitive superiority of the human cortex, it is 15 now clear that left-right asymmetries are a consistent feature of all vertebrate brains studied, as well as invertebrate nervous systems (Alqadah et al., 2018;Concha et al., 2012;Duboc et al., 2015;Frasnelli, 2013;Frasnelli et al., 2012;Rogers, 2014). Lateralisation of brain function has many potential advantages, such as sparing energetically expensive brain tissue, decreasing reaction time by avoiding eliciting incompatible responses, providing an 20 advantage in motor learning and facilitating coordinated behaviour in social animals (Concha et al., 2012;Rogers, 2014;Rogers, 2013;Vallortigara and Rogers, 2005). Not only does evolutionary conservation of brain asymmetries emphasise the importance of hemispheric lateralisation, but it also allows comparative developmental and behavioural studies between species. 25 Zebrafish (Danio rerio) have become an advantageous model organism in studying brain asymmetries owing to their rapid embryonic development, amenability to genetic manipulation, as well conveniently small size and transparency for developmental imaging and behavioural analysis. With respect to CNS lateralisation, the focus has long been on the epithalamus which displays overt left-right asymmetries in structure and function not only 30 in zebrafish but in a large number of vertebrates, albeit the extent and laterality of these asymmetries varies greatly between different groups (Aizawa et al., 2011;Concha and Wilson, 2001).
The epithalamus is a dorsal subdivision of the diencephalon constituted by bilateral habenular nuclei and a medially positioned pineal complex. The habenula (Hb) is a 35 phylogenetically old brain structure, which functions as a relay station conveying information from the limbic forebrain and sensory systems to the ventral midbrain (Aizawa et al., 2011;Bianco and Wilson, 2009), whereas the pineal has a conserved role in melatonin release and regulation of circadian rhythms (Ekstrom and Meissl, 2003;Sapede and Cau, 2013). The pineal complex also contains a photoreceptive accessory nucleus in some 40 species: a frontal organ in anuran amphibians, a parietal eye in some species of lizards and a parapineal nucleus in jawless and teleost fish (Concha and Wilson, 2001).
It has been hypothesised that the asymmetric connectivity of the dHb in fishes might reflect 65 the division in processing sensory versus forebrain contextual input, whereas in mammals such division is lost due to lack of direct sensory input to the epithalamus (Stephenson-Jones et al., 2012). The light-sensitive parapineal nucleus is one such example of an asymmetric sensory functions in the epithalamus of teleost fish and lampreys (Borg, 1983;Yanez et al., 1999). 70 In addition to its possible photosensory function, the left-sided parapineal is also essential for the development of most left-right asymmetries in the zebrafish habenulae.
Hence, mutants in which the parapineal is not properly specified (Clanton et al., 2013;Regan et al., 2009;Snelson et al., 2008) or in experimental setups where the parapineal is laser-ablated at early developmental stages (Aizawa et al., 2005;Bianco et al., 2008;Concha 75 et al., 2003;Gamse et al., 2005;Gamse et al., 2003), left dHb characteristics fail to develop and the habenulae exhibit right isomerism (a double-right phenotype). One of the mechanisms possibly influenced by the parapineal is the differential timing of neurogenesis in the left and right dHb. As shown by 5-bromo-2-deoxyuridine (BrdU) birth-date analysis, neurogenesis peaks at 32 hpf in the dHbL (more prominent on the left) and at 50 hpf in the 80 dHbM (more prominent on the right) (Aizawa et al., 2007). However, the early onset of asymmetric neurogenesis, marked by expression of the neuronal marker cxcr4b specifically in the left dHb, can already be detected at 28 hpf and requires left-sided epithalamic Nodal signalling (Roussigne et al., 2009). Around that time, left-sided Nodal signalling also determines the direction of parapineal migration -in the case of absent or bilateral 85 epithalamic Nodal signalling, parapineal migration is randomised and consequently habenular asymmetries are reversed in 50% of the embryos (Aizawa et al., 2005;Concha et al., 2000). Since the asymmetries in Nodal-dependent habenular neurogenesis are very subtle, biasing the migration of the parapineal to the left side might provide a mechanism to further enhance left dHb neurogenesis. 90 In this study, we address the role of the Sox family transcription factor encoding gene sox1a in mediating the ability of the parapineal to influence habenular development.
Zebrafish sox1a and sox1b have arisen from an ancestral vertebrate Sox1 gene during teleost genome duplication (Bowles et al., 2000) and show largely overlapping expression at early stages from 21 somites to 25 hours post fertilisation (hpf) in the telencephalon, 95 hypothalamus, eye field, early lateral line and otic vesicle primordia, trigeminal placode, lens and spinal cord interneurons (Gerber et al., 2019;Okuda et al., 2006). However, sox1aspecific expression has been detected in the lateral line primordium at 24 hpf (Gerber et al., 2019) and in the parapineal from 26-28 hpf onwards, but not the pineal anlage from which the parapineal arises (Clanton et al., 2013). Hence, Sox1a is a candidate transcription factor 100 for being involved in parapineal specification and/or the role of the parapineal in imparting habenular asymmetry.
Through analysis of the role of sox1a in epithalamic development, we find that the parapineal forms and migrates normally in sox1a -/mutant zebrafish larvae but the habenulae exhibit right isomerism. Furthermore, transplants of a few wild-type parapineal 105 cells are able to rescue epithalamic asymmetries in sox1a -/embryos and induce left-dHb characteristics in both left and right habenula. A time-course of parapineal ablations reveals a previously unsuspected step-wise regulation of habenula development by the parapineal.
Our results highlight the essential role of the parapineal and of Sox1a in asymmetric development of adjacent habenula. 110 Results sox1a is expressed in the developing parapineal from the onset of its formation Whole mount in situ hybridisation analysis showed that sox1a is expressed in the parapineal from the onset of its formation between 26-28 hpf ( Figure 1A-D") (Clanton et al., 2013).
Fluorescent in situ labelling of sox1a mRNA in embryos expressing the Tg(foxD3:GFP) and 115 Tg(flh:eGFP) transgenes in the whole pineal complex (Concha et al., 2003) revealed that sox1a is first expressed in a few cells located on the left side of the forming parapineal at 28 hpf, and thereafter in all parapineal cells as they undergo collective migration to the left side of the epithalamus ( Figure 2A). Additionally, sox1a is expressed in other areas such as the lens vesicle, anterior forebrain, ventral diencephalon, hindbrain and pharyngeal arches 120 ( Figure 1A-D'), as has also been described previously (Gerber et al., 2019;Okuda et al., 2006;Thisse, 2004).
The parapineal forms in sox1a -/mutants Using CRISPR/Cas9 genome editing (Auer et al., 2014a;Auer et al., 2014b;Gagnon et al., 2014;Talbot and Amacher, 2014), we generated two sox1a mutant lines (Figure 2 -figure  130 supplement 1). The sox1a ups8 allele (hereafter referred to as sox1a -/-), has an 11 bp deletion in the single exon of the sox1a gene, which leads to a premature stop codon at amino acid 62. As a result, the mutant Sox1a protein lacks the HMG DNA binding domain and is predicted to be non-functional. Indeed, no sox1a mRNA was detected in the parapineal of mutant embryos ( Figure 2B) suggesting nonsense mediated decay of the mutant transcript. 135 The second sox1a u5039 allele has a 10 bp deletion leading to a premature stop at amino acid 134 leaving the HMG DNA binding domain intact. Both sox1a -/mutants show no overt developmental abnormalities and are viable as adults. However, further analyses showed some variable expressivity of the phenotypes described below in the sox1a u5039 mutant allele and consequently the sox1a ups8 allele was used for all experiments. 140 Using Tg(foxD3:GFP) and Tg(flh:eGFP) transgenes to label parapineal cells, we observed that the parapineal migrated with normal timing and trajectory in sox1a -/mutants ( Figure 2A-B). This indicates that Sox1a is neither required for parapineal specification nor for migration. Furthermore, parapineal-specific expression of the transcription factor encoding genes otx5 (Gamse et al., 2002) and gfi1ab (Dufourcq et al., 2004) was not 145 affected in the sox1a -/mutants ( Figure 2C-D).
Although parapineal neurons form in sox1a -/mutants, efferent projections to the left habenula show reduced outgrowth and no branching ( Figure 2E). This phenotype does not necessarily reflect a cell autonomous deficit in the parapineal neurons as the changes in the left dHb of sox1a -/mutants (see below) are likely to impact its innervation by parapineal 150 axons.

sox1a -/mutants and morphants have a double-right dHb similar to parapineal-ablated larvae
Ablation studies have shown that the presence of a parapineal is required for the left dHb to elaborate left-sided patterns of gene expression and connectivity (Aizawa et al., 2005;155 Bianco et al., 2008;Concha et al., 2003;Gamse et al., 2005;Gamse et al., 2003).
Consequently, we assessed both normally asymmetric habenular gene expression and efferent connectivity of habenular neurons in sox1a -/mutants.
The overtly symmetric double-right habenular phenotype in sox1a -/mutants and morphants is comparable to the double-right habenulae development upon parapineal 170 ablation at early stages (before parapineal migration at 30 hpf) ( Figure 3A"-D") (also previously shown in Concha et al., 2003;Gamse et al., 2003). This indicates that the forming parapineal in sox1a -/mutants is not functional in terms of regulating left dHb development.
Note that the residual asymmetry in kctd12.1 mRNA expression in the dorsomedial domain of left dHb apparent in mutants, morphants and parapineal-ablated larvae alike (asterisk in 175 Figure 3A'-A" and Figure 3 -figure supplement 1D'), is similar to what has previously been described for residual asymmetries in habenular neuropil upon early parapineal ablation (Bianco et al., 2008;Concha et al., 2003). These asymmetries might be the result of Nodaldependent neurogenesis in the left dHb, that is potentially independent from parapinealregulated habenular asymmetries (Roussigne et al., 2009). 180 The symmetric double-right habenular phenotype of sox1a -/mutants was also evident in the efferent habenular projections to the IPN, as shown by anterograde axon tracing via lipophilic dye labelling ( Figure 3E-E'). In sox1a -/mutants, dorsal IPN innervation which normally arises from dHbL neurons (more prominent on the left) was almost completely lost and both dHb projected predominantly to the ventral IPN (n=15), the target 185 of dHbM neurons (Aizawa et al., 2005;Bianco et al., 2008;Gamse et al., 2005). This indicates that in sox1a -/mutants, most left dHb neurons have adopted dHbM character similar to the right dHb. Comparable efferent dHb projections predominantly targeting the ventral IPN were observed in early parapineal ablated embryos ( Figure 3E"), as previously described (Aizawa et al., 2005;Bianco et al., 2008;Gamse et al., 2005), confirming that the 190 parapineal fails to signal to the left habenula in absence of Sox1a function.
In summary, loss of function of the transcription factor Sox1a leads to double-right dHb phenotype similar to parapineal-ablated larvae, despite normal parapineal formation in the mutants.

A wild-type parapineal induces left habenula characteristics in sox1a -/mutants 195
The results described above are consistent with Sox1a in parapineal cells regulating the ability of these cells to impart left-sided character to the left dHb. However, as sox1a is expressed elsewhere in and around the nervous system, it is also possible that the habenular phenotype is a consequence of a role for Sox1a outside of the parapineal. To directly test whether Sox1a function is required within the parapineal to elicit habenular 200 phenotypes, we transplanted wild-type parapineal cells or control pineal cells into sox1a -/-Tg(foxD3:GFP);(flh:eGFP) embryos at 32 hpf, either to the left or right side of the endogenous pineal complex and assessed dHb character at 4 dpf ( Figure 4). Subsequent to transplantation, 3-4 transplanted parapineal cells with projections to the adjacent habenula could be detected by live imaging at 50 hpf ( Figure 4A-B and 4D-E), whereas transplanted 205 control pineal cells usually re-integrated into the pineal ( Figure 4C and 4F).  Figure 4B-B' and 4E-E').
As expected, transplanted parapineal cells were also able to induce left dHb characteristics in wild-type right habenula, with as few as two parapineal cells being sufficient to change the laterality of the adjacent right dHb (n=2; Figure 4  To conclude, sox1a -/habenulae are competent to respond to wild-type parapineal 230 signals and adopt left-type character, confirming that the sox1a -/double-right habenular phenotype results from impaired signalling between the parapineal and the left dHb.
Furthermore, both left and right habenula are competent to acquire left dHb character in response to parapineal signals, demonstrating that it is the left-sided migration of the parapineal that underlies asymmetric development of the zebrafish epithalamus. 235 The parapineal regulates habenular asymmetry at several developmental stages The temporal progression in the elaboration of habenular asymmetry spans from early asymmetric neurogenesis starting on the left side as early as 28 hpf ( In line with previous studies (Bianco et al., 2008;Concha et al., 2003;Gamse et al., 2003), early ablations at 30 hpf led to overtly double-right habenula development, as was evident for all studied markers (kctd12.1, n=25; nrp1a, n=6; VAChTb, n=11) ( Figure 5B, G, L) 255 as well as for efferent projections (n=3) ( Figure 5Q). In contrast, parapineal ablations at 35 hpf (n=16) and 50 hpf (n=15) did not obviously affect kctd12.1 expression at 4 dpf ( Figure   5C, D). Volumetric analysis did reveal a mild reduction in the average volume of the left dHb These results indicate that habenular asymmetries are regulated at several developmental stages by the parapineal, firstly at the time of left dHb neurogenesis and thereafter at the level of differentiation (axonal outgrowth and neurotransmitter domains). 280

Asymmetric habenular neurogenesis is regulated by the parapineal
The parapineal ablation experiments described above are consistent with the, as yet untested, possibility that the parapineal promotes early, asymmetric neurogenesis in the left dHb. This could contribute to promotion of dHbL character (more prominent on the left), as dHbL neurons tend to be born earlier than dHbM neurons (more prominent on the 285 right) (Aizawa et al., 2007). To assess if the parapineal does influence dHb neurogenesis, we carried out BrdU birth-date analysis for dHb neurons in wild-type and parapineal-ablated embryos. Control and 30 hpf parapineal-ablated Tg(foxD3:GFP);(flh:eGFP) embryos were exposed to a 20-minute BrdU pulse at 32 hpf, followed by a chase period until 4 dpf to allow differentiation of the habenular neurons. The number of neurons born around 32 hpf were 290 visualised with BrdU immunofluorescence and habenular asymmetries were assessed by kctd12.1 in situ hybridisation ( Figure 6A-B').
Birthdating analysis demonstrated that the parapineal does influence neurogenesis in the left dHb. In control wild-type embryos (n=22), significantly more cells were born in the left dHb compared to the right at 32 hpf as expected (p<6x10 -5 , Wilcoxon signed rank 295 test) ( Figure 6C). This asymmetry between the left and right dHb was markedly reduced in parapineal-ablated embryos (n=18, p=0.002, Wilcoxon signed rank test), due to decreased neurogenesis in the left dHb compared to controls (p=0.004, Wilcoxon-Mann-Whitney test), while the right habenula was unaffected (p=0.478, Wilcoxon-Mann-Whitney test) ( Figure   6C). Concomitantly, kctd12.1 expression revealed the expected double-right dHb phenotype 300 in the parapineal-ablated embryos ( Figure 6B).
These results demonstrate that the parapineal is required for the early wave of neurogenesis that is more prominent in the left dHb. The residual asymmetry in the number of BrdU positive cells between the left and right dHb in parapineal-ablated embryos most likely results from a Nodal-dependent (and parapineal-independent) influence upon 305 neurogenesis (Roussigne et al., 2009).

Discussion
Using two alternative approaches -cell ablations/transplants and genetic manipulation - In light of these observations, it is tempting to hypothesise that during teleost evolution, the left-sided parapineal has become a dominant signalling centre in the regulation of habenular asymmetries, whereas eft-sided Nodal pathway activation is primarily required for determination of laterality (left-sided migration of the parapineal 340 (Concha et al., 2000;Concha et al., 2003;Gamse et al., 2003) and parapineal size (Garric et al., 2014). In this scenario, the parapineal has taken over an ancestral role of Nodal in the regulation of habenular neurogenesis, possibly due to restrictions in developmental timing and duration of Nodal cascade activity (Signore and Concha, 2017;Signore et al., 2016).

sox1a -/mutants have symmetric habenulae with largely double-right character
Zebrafish have long been an excellent model to study genetic regulation of brain asymmetry from development to function (Concha et al., 2012;Concha et al., 2009;Duboue, 2017;Roussigne et al., 2012). Here we have shown that the double-right habenula phenotype of 365 sox1a -/mutants is identical to that in parapineal-ablated larvae, revealing a genetic factor behind the development of epithalamic asymmetries in zebrafish. Furthermore, the normal formation and migration of the parapineal in sox1a -/mutants despite loss of Sox1a in the parapineal indicates that Sox1a specifically functions in the regulation of signalling between the parapineal and left dHb rather than in parapineal specification. 370 Despite broad expression of sox1a in the embryonic zebrafish brain, sox1a -/mutants do not seem to have severe developmental defects other than loss of dHb asymmetry.
However, the other teleost sox1 paralogue -sox1b -has a nearly identical expression pattern with sox1a at early stages with the exception of the parapineal (Gerber et al., 2019;Okuda et al., 2006), and 80% sequence similarity with sox1a in the ORF, suggesting that 375 these two sox1 genes are likely to have redundant functions in the developing CNS.
Likewise, redundant functions of B1 sox genes have been described for the Sox1 knock-out mouse, in which formation of the lens (where only Sox1 is expressed) is severely disrupted whereas the CNS shows only mild developmental abnormalities (due to overlapping expression of Sox1 with Sox2 and Sox3) (Ekonomou et al., 2005;Malas et al., 2003;380 Nishiguchi et al., 1998). The functional redundancy between B1 sox genes has also been suggested in early embryogenesis of zebrafish by combinations of sox2/3/19a/19b knockdowns (Okuda et al., 2010).
In sox1a -/mutants, the parapineal forms normally, although parapineal cells do not send fully developed projections to the left habenula. Rather than a cell-autonomous 385 phenotype, this is most likely a secondary effect due to left habenula character not being specified. Indeed, previous studies have shown that lateralised afferent innervation of the dHb depends on the lateralised character of the left and right dHb (deCarvalho et al., 2013;Dreosti et al., 2014). It is also unlikely that any of the double-right dHb characteristics described here for sox1a -/mutant larvae are caused by the abnormal extension of 390 parapineal axons as parapineal axotomies have no apparent effect on asymmetric habenular efferent connectivity.
To conclude, the parapineal-specific expression of sox1a and the overlapping expression of different B1 sox genes in the rest of the zebrafish CNS renders the sox1a -/mutant a valuable model for studying genetic regulation of brain asymmetry development 395 in a context without overt defects in other aspects of brain development.

Conclusions
Here, we have shown that the parapineal is essential for the development of habenular asymmetries in the larval zebrafish at several stages. sox1a mutant fish exhibit an almost complete loss of left habenula characteristics despite the formation of a parapineal nucleus 400 providing an excellent genetic tool to study the signalling events responsible for establishing habenular asymmetries. In addition, precise, time-controlled parapineal ablation and transplant experiments demonstrate the step-wise manner of habenula asymmetry regulation by the parapineal and the incredible potency of the parapineal cells to induce left habenula characteristics in both left and right habenulae. 405

Fish lines and maintenance
Zebrafish (Danio rerio) were maintained in the University College London Fish Facility at 28°C and standard light conditions (14 h light/10 h dark). Embryos were obtained from natural spawning, raised at 28.5°C and staged as hours or days post fertilisation (hpf, dpf) 410 according to (Kimmel et al., 1995). 0.003% 1-phenyl-2-thiourea (PTU) was added to the water at 24-26 hpf to prevent pigmentation. For live-imaging, 0.04 mg/ml (0.02%) Tricaine (ethyl 3-aminobenzoate methanesulfonate) (Sigma) was added to the water for anesthesia.

Fixation and dissection
Embryos and larvae were fixed in 4% (w/v) paraformaldehyde (PFA) in phosphate buffered saline (PBS) by over-night immersion at 4°C. For BrdU immunohistochemistry and lipophilic dye labelling, the brain of 4 dpf larvae was dissected out by manual dissection with larvae 420 pinned in sylgard (Turner et al., 2014).

Generation of sox1a -/mutants by CRISPR/cas9
sox1a mutant lines were generated by CRISPR/Cas9 targeted genome editing relying on non-homologous end joining repair mechanism, as described in detailed protocols provided

Whole-mount in situ hybridisation (ISH, FISH)
Digoxygenin (Roche) labelled RNA probes were made using standard protocols and spanned a minimum of 800 bp. To enhance permeabilisation, fixed embryos or larvae were dehydrated in methanol for a minimum of one hour at -20°C, rehydrated in PBST (PBS with 0.5% Tween-20, Sigma) and treated with 0.02 mg/ml proteinase K (PK, Sigma) for 10-40 440 minutes depending on the age of the fish. Probe hybridisation was carried out at 70°C in standard hybridisation solution containing 50% formamide over-night, with 2 ng/l of RNA probe. Embryos were washed at 70°C through a graded series of hybridisation solution and 2x saline sodium citrate (SSC), followed by further washes with 0.2x SSC and PBST at room temperature. Blocking was carried out in maleic acid buffer (150 mM maleic acid, 100 mM 445 NaCl, 2% sheep serum, 2 mg/ml BSA) for 2-3 hours. Probes were detected by over-night incubation with anti-Digoxigenin-AP Fab fragments (1:5000) (Roche) and stained with standard Nitro Blue Tetrazolium (NBT) and 5-Bromo-4-chloro-3-indolyl phosphate (BCIP) (Roche) ISH protocol. Fluorescent in situ hybridisation (FISH) was carried out using either Fast Red tablets according to manufacturer's instructions (Roche, discontinued from 450 manufacturing) or Fast Blue BB Salt (Sigma) and NAMP (Sigma) staining as previously described (Lauter et al., 2011).

Whole-mount immunohistochemistry
Fixed larvae were stained and imaged as whole-mounts following standard procedures (Shanmugalingam et al., 2000;Turner et al., 2014). In short, samples were dehydrated in 455 methanol for a minimum of one hour at -20°C, rehydrated in PBST and treated with 0.02 mg/ml proteinase K (PK, Sigma) for 10-40 minutes depending on the age of the fish. 10% Heat-inactivated Normal Goat Serum (NGS) (Sigma) was used for block and for over-night primary antibody incubation at 4°C with the following antibodies: rabbit anti-GFP (dilution 1:1000, Torrey Pines Biolabs, Cat# TP401), mouse anti-acetylated tubulin (dilution 1:250, 460 IgG2b, α-tubulin, Sigma Cat# T7451) and mouse monoclonal anti-BrdU antibody (1:450, Roche, Cat# 11170376001). Secondary antibody incubation was carried out over night at 4°C using Alexa Fluor 488-conjugated, 568-conjugated and 647-conjugated secondary antibodies (1:200, Molecular Probes, Cat# A32731, A21144, A21126). For immunohistochemistry after in situ hybridisation, probe hybridisation was carried out at a 465 lower temperature (65-68°C) to ensure high-quality immunolabelling. After Fast Red or Fast Blue in situ hybridisation, samples were washed 6x20 minutes in PBST followed by primary antibody incubation in PBST without NGS and immunohistochemistry as usual.

Neural tract tracing
Tracing of habenula efferent projections was carried out by labelling with membrane-bound 470 lipophilic dyes DiI (DiIC18(3), Molecular Probes, Cat# D3911) and DiD (DiIC18(5), Molecular Probes, Cat# D7757) in 4 dpf embryos. To that end, immobilised embryos (pinned down from the body with needles) were dissected to expose the brain. For a dorsal view, embryos were then placed between two needles and under a stereomicroscope, crystals of DiI (left dHb) and DiD (right dHb) were manually applied to dorsal habenulae with electrolytically 475 sharpened tungsten needles. Brains were incubated in PBS overnight at 4°C, mounted in 1.5% low melting point agarose (Sigma) in PBS and imaged by confocal laser scanning microscopy. The success rate of bilateral labelling was approximately 60%.

Parapineal laser-ablations and axotomies
Two-photon laser-ablations of the parapineal cells and for parapineal axotomies were 495 carried out with in Tg(foxD3:GFP);(flh:eGFP) embryos with double-transgenic GFP expression in the pineal complex (Concha et al., 2003), using Leica 25x/0.95 NA PL FLUOROTAR water-dipping objective on a Leica TCS SP8 Confocal microscope coupled with a multiphoton system (Chameleon Compact OPO-Vis, Coherent) and an environmental chamber at 28.5°C. Embryos were manually dechorionated and immobilised by mounting on 500 a glass slide in a drop of 1.5% low melting point agarose (Sigma) in fish water with 0.04 mg/ml Tricaine for anesthesia. Ablations were carried out at 2-3 separate z-planes, using 30-60% of the maximum output laser power (80 mW) at the wavelength of 910 nm. Each scan took on average 5-10 seconds per z-plane. For 30 hpf parapineal ablations, 1/3 of the pineal complex anlage was removed from its anterior end, the position of parapineal precursors 505 (Concha et al., 2003). Embryos were removed from agarose directly after the ablations.
Ablation success was confirmed by live confocal imaging the next day. The success rate for parapineal two-photon ablations (all cells ablated) is approximately 80% at 35 and 50 hpf but lower (60%) for 30 hpf ablations (due to regeneration of the parapineal) and 3 dpf ablations (possibly due to the compact structure of the parapineal and the blood vessels 510 covering it). Each experiment was carried out in 2-3 separate replicates with the exception of previously published results (indicated where appropriate), which were confirmed once.
The N-numbers for all ablation experiments are given in the Results section. Ablated embryos were analysed with a comparable number of control embryos (embryos mounted in agarose but not ablated). 515 Axotomies were performed at 30-40% laser power by 2-3 pulses at 910 nm using the Bleach Point function on Leica Application Suite X (LAS X) software. The axon bundle was severed approximately 10 µm from the cell body at three time-points (due to regeneration) -at 50, 60 and 72 hpf. The embryos were removed from agarose in between these time points to ensure normal development. Axotomy success was confirmed by live confocal 520 imaging of the pineal complex at 4 dpf before fixation and dissection for lipophilic labelling.
Embryos were incubated in Claw 1xE3 embryo medium (5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl2, 0.33 mM MgSO4) with 15% DMSO (Sigma, Cat# 276855) for 5 minutes on ice, followed by 20-minute 10 mM BrdU (Sigma, Cat# B5002) incubation in Claw 1xE3 embryo medium with 15% DMSO on ice. At 50 hpf, parapineal-ablated embryos and non-ablated 530 controls were mounted in agarose and ablation success was confirmed by live confocal imaging. Fail-ablated embryos (with at least one parapineal cell left unablated and/or with a damaged pineal) were excluded from further analysis. At 4 dpf, larvae were fixed and dissected to expose the brain. Brain dissection also leads to loss of the superficially positioned pineal complex in most cases. Fast Red (Roche, discontinued from 535 manufacturing) fluorescent in situ hybridisation for kctd12.1 followed by BrdU immunohistochemistry was carried out as described, with an added step of 45-minute 2N HCl treatment to expose the BrdU epitope prior to antibody staining. BrdU-positive cells from 3D reconstructions were counted using semi-automated detection in Imaris 7.7.1 (Bitplane) software. The experiment was repeated three times and the results were 540 analysed using GraphPad Prism 8.0.2 software. The N-numbers were limited by high technical difficulty of combining parapineal ablations and BrdU staining in a sort window of time (30-32 hpf), allowing recovery between the two experiments. The data did not show clear normal distribution and therefore non-parametric tests were used for statistical analysis. Wilcoxon-Mann-Whitney test was carried out for unpaired comparisons of BrdU 545 cell counts between control and parapineal-ablated embryos. Wilxocon signed rank test was used for paired analysis of BrdU cell counts in the left and right habenula.

Image analysis
Confocal imaging was carried out using a Leica TCS SP8 system with a 25x/0.95 NA PL FLUOROTAR water-dipping objective for live-imaging or 25x/0.95 NA PL IRAPO water-550 immersion objective with coverslip correction for fixed samples. Image analysis was performed using Fiji (ImageJ) and Imaris 7.7.1 (Bitplane) software. Images and figures were assembled using Adobe Photoshop and Adobe Illustrator.

Author contributions
IL and VD designed the research with input from AF, PB and SWW. VD generated the sox1a ups8 mutant line and performed initial phenotypic analyses; IL confirmed and 560 extended the phenotypic analyses in both sox1a mutant lines and performed ablation and transplantation experiments; SN carried out morpholino experiments together with IL. All authors contributed to interpretation of data. IL and SW wrote the article with input from all authors.

565
The authors declare no competing interests.

Funding
This study was supported by a Wellcome Trust Investigator Award to SWW      An 11 bp (sox1a ups8 ) and a 10 bp (sox1a u5039 ) deletion (orange) was introduced into the single 1011 bp exon of sox1a, leading to a premature stop codon at positions 197 and 412, respectively. In sox1a ups8 , the deletion leads to disruption of the sequence coding for the HMG DNA binding domain (blue), whereas in sox1a u5039 , the HMG domain sequence is intact.   (C-D,F) sox1a ups8 allele exhibits complete penetrance with all homozygous mutants having a doubleright dHb, as confirmed by nrp1a mRNA in situ hybridisation at 4 dpf in Tg(foxD3:GFP);(flh:eGFP) 880 background with GFP expressed in the pineal complex (C and Figure 3). sox1a +/ups8 heterozygotes (n = 45) (D) have a phenotype identical to wild-type siblings (F).

(E)
The two mutant sox1a alleles do not complement each other -all trans-heterozygote larvae are double-right for nrp1a mRNA expression (n = 45/45). Similar to sox1a ups8 , trans-heterozygotes of the two mutant alleles exhibit stalled outgrowth of parapineal axons (white arrows in C and E).

935
Note that innervation of dorsal interpeduncular nucleus (dIPN) from the left habenula is lost only upon ablation of all parapineal cells. The asterisk indicates a small tuft of neurites in the dorsal anterior IPN that are present upon complete parapineal ablation (Bianco et al., 2008) and can also be detected in sox1a mutants (Fig. 3E'). All scale bars 25 µm