Cachd1 is a novel Frizzled- and LRP6-interacting protein required for neurons to acquire left-right asymmetric character

Neurons on left and right sides of the nervous system frequently show asymmetric properties but how these differences arise is poorly understood. Through a forward genetic screen in zebrafish, we find that loss of function of the transmembrane protein Cachd1 results in right-sided habenula neurons adopting left-sided character. Cachd1 is expressed in habenula neuron progenitors, functions symmetrically downstream of asymmetric environmental signals that determine laterality and influences timing of the normally left-right asymmetric patterns of neurogenesis. Unbiased screening for Cachd1 partners identified the Wnt co-receptor Frizzled7 and further biochemical and structural analysis revealed Cachd1 can bind simultaneously to Fzd proteins and Lrp6, bridging between these two Wnt co-receptors. Consistent with these structural studies, lrp6 mutant zebrafish show symmetric habenulae with left-sided character and epistasis experiments with other Wnt pathway genes support an in vivo role for Cachd1 in modulating Wnt pathway activity in the brain. Together, these studies identify Cachd1 as a conserved novel Wnt-receptor interacting protein with roles in regulating neurogenesis and neuronal identity.


Introduction
The nervous systems of bilaterian animals are frequently, perhaps universally, left-right (LR) asymmetric with respect to neuroanatomy, processing of information and control of behaviour (1,2). Within vertebrates, the epithalamus shows evolutionarily conserved LR asymmetries (3,4). In zebrafish, the epithalamic dorsal habenulae (dHb) comprise medial (dHb M ) and lateral (dHb L ) domains each containing distinct subtypes of projection neuron; the dHb L is larger on the left whereas the dHb M is larger on the right (5)(6)(7). Functional asymmetry mirrors neuroanatomy; for instance, in young fish, light activates predominantly left-sided dHb L neurons, whereas a higher proportion of right-sided dHb M neurons respond to odour (8). Afferent innervation is also asymmetric with a subset of mitral cells innervating the right dHb and the photosensitive parapineal nucleus innervating the left dHb e. g. (9)(10)(11).
The development of epithalamic asymmetry is dependent upon sequential interactions between cell groups that coordinate lateralisation of circuit components (12)(13)(14). Although the developmental mechanisms by which asymmetry arises remain poorly understood, genetic analyses in zebrafish have revealed multiple roles for Wnt signalling during the establishment of lateralised circuitry. For instance, the habenulae of fish with compromised function of Axin1, a scaffolding protein in the ß-catenin degradation complex, have symmetric habenulae with rightsided character whereas the habenulae are symmetric with left-sided character in fish lacking function of the Tcf7l2 transcriptional effector of Wnt signalling (15,16). Wnt signalling also impacts habenula size, as fish with compromised function of Wls, a protein involved in Wnt secretion, have small, but asymmetric habenulae (17). These and other studies (18) indicate that Wnt signalling regulates several distinct aspects of epithalamic development suggesting complex regulation of Wnt pathway activity during this process.
The Wnt signalling pathway is involved in a wide array of biological processes during embryonic development, throughout life and in many disease states, including cancer (19). Wnt ligands can activate several signalling cascades and pathway activation is tightly regulated at many different steps from ligand production to ligand/receptor interactions, cytoplasmic signal transduction and transcriptional activation/repression (20)(21)(22). Through studying the role of Wnt signalling in the establishment of brain asymmetry, here we identify Cachd1 as a novel transmembrane component of this highly conserved and multi-functional signalling pathway.

rorschach u761 mutants show symmetric habenulae due to a lesion in the cachd1
gene To identify novel genes involved in the establishment of brain asymmetry, we screened zebrafish embryos carrying ENU-induced mutations that alter asymmetric habenular expression of the kctd12.1 gene (16) and identified the rorschach u761 mutant (rch). In 4 dpf larvae homozygous for the u761 mutation, kctd12.1 expression (5) in the right habenula was increased to the level within the left habenula suggesting that both habenulae exhibit left-sided character (Fig. 1A). Other than this fully penetrant habenular phenotype, homozygous rch mutants were morphologically indistinguishable from wild types with normal asymmetry and laterality of the viscera.
Mapping using bulked segregant analysis followed by high resolution simple sequence length and single nucleotide polymorphism analyses placed the rch mutation in a 0. 28  To confirm that the u761 mutation in cachd1 is causative of the symmetric habenular phenotype, we analysed phenotypes after abrogation of cachd1 in other contexts. Embryos homozygous for a likely null mutation in cachd1 (sa17010), that makes no detectable Cachd1 protein (by immunohistochemistry; Fig. S1C, Table S1), showed the same habenular double leftphenotype, as did transheterozygote cachd1 u761 /cachd1 sa17010 mutants (n = 10/10, Fig. 1D; n = 6/6,
(B) Schematic of Cachd1 protein: two dCache domains (cyan and dark blue), a VWA domain (purple stripes), a Fzd binding domain (FZI, grey stripes), a transmembrane domain (white) and an unstructured cytoplasmic tail. Residues affected in sa17010 and u761 alleles are marked in red at approximate positions in primary sequence.

Cachd1 is expressed in neuroepithelial cells along the dorsal midline of the brain
To determine where and when cachd1 is expressed within the brain, we performed fluorescent in situ hybridisation and immunohistochemistry using an antibody raised and purified against the extracellular domain of zebrafish Cachd1 (Fig. 2, Fig. S1). Prior to neuronal differentiation, cachd1 is present broadly within the dorsal diencephalon ( Fig. 2A), and co-localises with dbx1b, an early marker of habenula neuron precursors ( Fig. 2A, B) (23). During the period of habenular neurogenesis (24), cachd1/Cachd1 expression becomes restricted to a proliferative neuroepithelial domain close to the midline adjacent to mature habenula neurons ( Fig. 2C-E′, Fig. S5). Although the cachd1 mutant only shows an overt mutant phenotype on the right side of the brain, we detected no obvious asymmetry in cachd1/Cachd1 expression until much later in development, long after habenula asymmetry has been established (Fig. S6). Early Nodal signalling-dependent brain (25,26) and visceral (27) asymmetries were unperturbed in cachd1 mutant embryos (Fig.   S7). Taken together, these results suggest that cachd1 functions locally within the progenitor domain that gives rise to habenula neurons.

Cachd1 functions in both habenulae to promote right-sided and/or suppress leftsided character
The right dorsal habenula (dHb) differs from the left dorsal habenula with respect to gene expression, organisation of synaptic neuropil and targeting of projection neuron connections (5-7, 11, 28); asymmetries of all these features in cachd1 mutants were reduced or absent with the right habenula closely resembling the left habenula ( Fig. 3A-B′ and Fig. S8). The dHb contain two broad sub-types of projection neuron present in different frequencies on right and left (6,7,24,28). In the left dHb, neurons that project to the dorsal interpeduncular nucleus (dIPN; termed dHb L neurons)
(D-E′) Dorsal views of habenulae after immunohistochemistry with anti-Cachd1 antibody (cyan) co-stained with either anti-HuC/D antibody to mark differentiated neurons (red) at 2 dpf (D) or anti-GFP antibody (red) to mark habenula neurons expressing GFP in Tg(110316_GFP)u775 (u775Tg) embryos at 3 dpf (E). All images are maximum projections of confocal stacks, except for a single confocal slice in (D). and DiD (green) labelling of left-and right-sided habenula neuron axon terminals predominantly innervating the dIPN and vIPN respectively, and raphe (r), in 5 dpf wildtype (A, n = 3) or cachd1 u761 mutant (B, n = 8) larvae.
(F-H) Segmentation of confocal stacks from Et(gata2a:eGFP)pku588 wildtype or cachd1 u761 mutant larvae incubated at 24 (F), 32 (G) and 48 hpf (H) with a pulse of BrdU then processed at 5 dpf as in (E). Double positive cells are represented in magenta; BrdU-positive only cells are represented in green. Time of pulse indicated in top right corner.
(I) Quantification of the proportion of BrdU-positive neurons that also expressed Et(gata2a:eGFP)pku588 (magenta) in 5 dpf wildtype or cachd1 u761 larvae incubated with a pulse of BrdU at 24 hpf (all timepoints presented in Fig. S9). Error bars represent 95% confidence intervals. Total number of cells and larvae for each genotype indicated in axis label in brackets. Qʹ test of equality of proportions (all timepoints, degrees of freedom = 15, χ 2 = 747.49, p = 1.39 × 10 -149 ), post hoc pairwise comparisons using a modified Marascuilo procedure with Benjamini-Hochberg correction for multiple testing, **** p < 0.005.
Scale bars = 50 µm (A-H) are predominant whereas on the right, the majority of neurons project to the ventral IPN (vIPN; termed dHb M neurons). Unlike in wildtypes, in cachd1 u761 mutants, DiI/DiD labelling revealed that the right dHb extensively innervated the dIPN, consistent with a higher proportion of right-sided dHb neurons adopting dHb L character (n = 3 wildtype siblings, 8 u761 mutants; Fig. 3A-B′).
Together these results show that on the right side of the brain, Cachd1 function promotes dHb M and/or suppresses dHb L character. However, they do not reveal whether Cachd1 has any function in determining the molecular character of the left side of the brain.
The parapineal nucleus, a small group of cells present on the left side of the brain, is critical for the elaboration of most aspects of left-sided habenula character (5,7,11,29). Consequently, if the parapineal is ablated ( Similarly, in cachd1 u761 , sox1a ups8 double mutants (n = 3/3; Fig. 3D) the cachd1 mutant phenotype was epistatic to the sox1a mutant phenotype. Taken together, these results imply that Cachd1 functions on both sides of the brain to suppress left-sided character and/or promote right-sided character. As a corollary to this, it also implies that the role of the parapineal is to abrogate the function of Cachd1 within the left habenula.
Both the timing of neurogenesis and the environment into which habenula neurons are born influence their sub-type identity (16,24). dHb L neurons tend to be generated earlier than dHb M neurons and neurogenesis is initiated earlier in the left habenula than in the right habenula (24).
Furthermore, early born neurons differentiating on the left have a higher probability of adopting dHb L character than neurons born on the right (16). To elucidate how Cachd1 impacts these asymmetries in neurogenesis, we performed birth dating experiments to assess both the extent of habenular neurogenesis at different stages and the timing of birth of dHb L neurons, marked with the transgene Et(gata2a:eGFP)pku588 (pku588Et, Fig. 3E-I). These analyses showed that neurogenesis began early in cachd1 u761 mutants compared to wildtype, was symmetric on the left and right sides of the brain ( Fig. 3F-G, Fig. S9) and diminished over time (Fig. 3H, Fig. S9). In addition, early born neurons in the right habenula of cachd1 mutants had a higher likelihood of taking on dHb L -character than in wildtype embryos, as assessed by subsequent expression of the pku588Et transgene ( Fig. 3I; Fig. S9) and the dHb L marker kctd12.1 (Fig. S10).

CACHD1 binds to Wnt pathway receptors
Given that the phenotype of embryos homozygous for the u761 allele is a consequence of the absence of Cachd1 on the cell surface, we reasoned that the extracellular domain of Cachd1 is likely essential for its function and performed experiments to identify cell surface proteins that could physically interact with this receptor. In cultured cells, CACHD1, which has homology to the α2δ family of auxiliary subunits of voltage-gated Ca 2+ channels (VGCCs), can alter VGCC activity and compete with other α2δ proteins (30,31). However, to date there is no evidence of the necessity of this interaction during development in vivo and so we took an unbiased screening approach to find other interacting partners.
We initially identified FZD7 as a potential binding partner in a Retrogenix Cell Microarray Technology screen using a human CACHD1 ectodomain (ECD) multimer as a prey protein (see
Taken together, these results demonstrate conserved interactions between CACHD1 and the Wnt co-receptors LRP6 and Frizzled family proteins.

Structural characterisation of Cachd1 complex with FZD5 and LRP6
Guided by our in vitro measurements, we attempted co-crystallisation of CACHD1 ECD with FZD5 CRD and LRP6 P3E3P4E4 . Crystals appeared in a PEG 4000 condition (see Methods) and diffracted to 4.7Å resolution. The structure was determined by molecular replacement using crystal structures of the CACHD1 ECD :FZD5 CRD complex, previously determined in our laboratory (data not shown), and LRP6 P3E3P4E4 (33) (PDB: 4A0P). The arrangement of the ternary complex in the crystal conforms to a C2 1 space group with three complexes in an asymmetric unit (ASU). Refinement yielded complete structures of equivalent quality for all three copies (Table S2) helices and DKK-1C (Fig. 5C). This suggests that Cachd1 may also compete with Wnt3a for binding to the LRP6 P3 propeller. Taken together our biophysical and structural analyses showed that CACHD1 is a novel binder to both members of the FZD family of Wnt receptors and the LRP6

Fig. 5. CACHD1 forms a ternary complex with FZD5 and LRP6
(A) Cartoon representation of mouse CACHD1 ECD , (rainbow-colored from N-(blue) to C-(red) terminus) in complex with mouse FZD5 CRD (cartoon and surface in teal) and human LRP6 P3E3P4E4 (cartoon and surface in salmon). The position of the four cache (C-1,2,3,4), VWA and FZD interaction (FZI) domains of CACHD1 ECD are indicated.
(B) Superimposed structures of the FZD8:Wnt3 complex (PDB: 6AHY) with the FZD5 CRD :CACHD1 ECD complex. Wnt3 is shown as a violet cartoon tube with palmitoleic acid (PAM) as spheres.

cachd1 genetically interacts with Wnt pathway genes
If Cachd1 functions together with Fzd and Lrp proteins during habenular development, then we might expect that abrogation of Fzd and/or Lrp6 function would also result in habenular asymmetry phenotypes. The Fzd family is large and we expect much redundancy between family members (20) and so we focussed on analysis of Lrp6 function in habenular development. We generated several predicted lrp6 null alleles using CRISPR/Cas9 and found that homozygous mutant larvae showed a fully penetrant, symmetric double-left habenular phenotype, with visceral asymmetry unperturbed (lrp6 mutants, n = 80/84, Fig. 6A, Fig. S14, Table S3). Consequently, loss of cachd1 and lrp6 function both result in symmetric habenulae with left-sided character, consistent with these two genes functioning in the same signalling pathway in the brain.
We tested for a genetic interaction between cachd1 and lrp6 by injecting a validated cachd1 morpholino into embryos from lrp6 u349/+ heterozygous outcrosses at a dose that results in a low frequency of symmetric habenular phenotypes in injected wildtype embryos. If the two genes interact genetically in the development of habenula asymmetry, then one might expect lrp6 heterozygous embryos to be more sensitive to reduced Cachd1 function than their wildtype siblings. Indeed, we observed that heterozygous lrp6 u349/+ larvae were approximately three times more likely to be bilaterally symmetric for habenular kctd12.1 expression than wildtype siblings when injected with a low dose (1 ng) of cachd1 splice-blocking morpholino (MO1, Fig. 6B; bilateral phenotype proportion ± 95% confidence interval, n: WT: 0.14 ± 0.09, 64 larvae; lrp6 u349/+ : 0.47 ± 0.13, 55 larvae; Marascuilo procedure for comparing multiple proportions and Benjamini-Hochberg correction, degrees of freedom = 2, p = 0.0002). To confirm this difference was not due to morpholino efficacy, we injected a standard effective dose (2 ng) in parallel: a high degree of  Genetic interactions between cachd1 and two other Wnt pathway genes implicated in habenular development (tcf7l2 and axin1) were also examined (15,16). Tcf7l2 is a transcriptional effector of Wnt signalling and loss of tcf7l2 function results in symmetric habenulae with double-left character (16). tcf7l2 exI/+ heterozygotes show a wildtype habenular phenotype, however when Cachd1 levels were reduced in tcf7l2 exI/+ heterozygotes through injection of low dose cachd1 morpholino, many larvae showed a symmetric, double-left phenotype ( Fig. 6C;  Axin1 is a scaffolding protein in the β-catenin degradation complex and compromised axin1 function results in symmetric habenulae with double-right character (15), a phenotype opposite to that of cachd1 mutants. When we generated axin1 tm213 /cachd1 u761 double mutants, they exhibited the axin1 mutant phenotype (as assessed by expression of kctd12.1, Fig. 6D, kctd8, scl18b and vachtb, data not shown; n = 15/15); consequently, compromised Axin1 function is epistatic to loss of Cachd1 function. This is consistent with Axin1 functioning downstream of Cachd1 and the Fzd/Lrp6 receptor complex.
Wnt signalling often regulates expression of Wnt-pathway genes through positive and negative feedback mechanisms (20). Indeed, the spatially localised expression of cachd1 along the roofplate and in the dorsal epithalamus is similar to that of other Wnt pathway genes such as the ligand encoding wnt1, wnt3a and wnt10 genes and known Wnt pathway targets axin2 and lef1 (Fig. S15). We directly tested whether CACHD1 is itself a target of Wnt signalling, using qPCR to assess CACHD1 expression in cultured HEK293 cells treated with Wnt3a, or Wnt3a+RSpondin1 conditioned media, or a stable HEK293 cell line with a mutation in APC (named APC4) (42).
CACHD1 showed a similar level of transcriptional response to enhanced Wnt pathway activity as other Wnt target genes (Fig. 6E) suggesting that Cachd1 may be involved in Wnt signalling in other epithelial cell types. Similarly, Cachd1 expression was upregulated in murine Apc mutant organoids (APC5, Fig. 6F). By contrast, cells derived from colorectal cancers (APC mutants: DLD1, SW480; β-catenin mutants: HCT116, Ls174T) showed downregulated CACHD1 expression.
Together, these results provide compelling evidence to suggest that the structural interactions we have demonstrated are pertinent to Cachd1 function in the developing brain and that Cachd1 is a novel modulator of the Wnt signalling pathway, potentially functioning in many contexts.

Discussion
Our results identify CACHD1 as a novel Wnt pathway component that bridges FZD and LRP6 Wnt co-receptors and functions in the developing brain and potentially other Wnt pathway contexts (43).
We have demonstrated evolutionary conserved interactions between CACHD1 and multiple FZD receptors through a previously unidentified FZI domain. Our crystal structure suggests that this domain could potentially compete with Wnts binding to FZDs through their PAM moiety.
Similarly, CACHD1 binding to LRP6 P3E3P4E4 through the first dCache domain suggests potential competition with either Wnt ligands or the Wnt inhibitor, DKK.
The simultaneous binding of Cachd1 to both Fzd and Lrp6 co-receptors suggests it may activate downstream signalling by clustering of the cytoplasmic signalling apparatus, as observed with artificial ligands (44). A role for Cachd1 in activating signalling would be consistent with the observed similarity of habenular phenotype in cachd1, lrp6 and tcf7l2 mutants, where the latter is thought to be due to a loss of pathway activation (16), and contrast to the phenotype of axin1 mutants, in which the pathway is overactivated (15). However, the complexity of Wnt signalling frequently confounds simple interpretations and further studies are required to determine the in vivo signalling consequences of Cachd1/Fzd/Lrp6 interactions. For instance, the binding of Cachd1 to Fzd/Lrp receptors may simultaneously activate signalling and outcompete binding of Wnts, indirectly impacting activity of the pathway that might otherwise be set by the secreted ligand.
Indeed, the relative differences in binding affinity to different Fzd proteins may render the consequences of Cachd1/Fzd/Lrp6 interactions highly context dependent.
Our study suggests that asymmetric Cachd1-dependent modulation of Wnt signalling leads to lateralisation of habenula neurons by altering both timing of neurogenesis and the probabilistic selection between alternate neuronal fates. We show that Cachd1 is present and can function on both sides of the brain but its activity on the left is antagonised by an unknown signal(s) from the parapineal. During habenular development, as in many other contexts, Wnt signalling appears to function at multiple stages and in multiple processes, from proliferation, through timing of neurogenesis, to acquisition and maintenance of neuronal identity (this study; (15)(16)(17)(18)). It is largely unclear how this complexity of pathway activity and outcome is effected and an attractive possibility is that context-dependent activity of Cachd1 may contribute to this poorly understood aspect of Wnt signalling.

Acknowledgements
We thank many colleagues for support and advice during the course of this project, staff at Diamond Light Source for assistance with X-ray data collection,

Author Contributions
The senior authors wish to emphasise that all four lead authors made equally important contributions to this study and are happy for individuals to list the joint authors in whichever order they wish on CVs and other documents.

Competing interests
The authors declare no competing financial interests.

Data and materials availability
Further information and requests relating to zebrafish resources and reagents, including mutants generated in this study, should be directed to Steve W. Wilson (s.wilson@ucl.ac.uk), and those relating to structural biology and biochemistry, to E. Yvonne Jones (yvonne.jones@strubi.ox.ac.uk).

Zebrafish husbandry and fish lines
Zebrafish were maintained in a designated facility according to UK Home Office and local regulations, on a 14h/10h light:dark cycle. Embryos were routinely stored in fish system water supplemented with methylene blue, or E3 embryo medium at 28°C. Zebrafish experiments and husbandry in the United States of America followed standard protocols in accordance with University of Washington Institutional Animal Care and Use Committee guidelines.

Generation of mutant and transgenic lines
The u761 mutant was generated by ENU mutagenesis. Mutations were induced in wild-type male AB/TL fish by four rounds of 3 mM ENU treatment as previously described (45).
The sa17010 allele of cachd1 was acquired from the Zebrafish Mutation Project (46).

Cloning and genotyping of u761
Having used a combination of backgrounds to generate our F2s, we mapped u761 in F3 embryos. We used bulked segregant analysis (49) followed by high resolution SSLP and SNP analyses to localize u761 to a 0.28 MB interval on LG6 between a SNP in the first coding exon of ak4 (2/5212 recombinants; ak4 e1 primers; see Table S5) and an SSLP in intron 8-9 of cachd1 (1/5212 recombinants; cachd1 i8-9 primers; see Table S5). Sequencing of cachd1 cDNA revealed a T to A transversion in the 24th exon of cachd1 that causes a valine to aspartic acid amino acid substitution in its transmembrane domain (reverse strand 6:31607781 T>A, 1122V>D, Zv11 assembly). Mutants were subsequently genotyped with DCAPs primers (Table S5, u761-AloI primers) and the restriction enzyme AloI, which cuts the mutant allele, and then more routinely by KASP assay (see below).  BioLabs; see Table S5 for primers). The phosphorylated fragments were then cloned into the StuI site of a pCS2+ vector treated with Antarctic Phosphatase (New England Biolabs) to prevent recircularization. The resulting vectors were cut with BamHI and SnaBI and the cachd1-containing fragments cloned into the BamHI and EcoRV sites of the pTol2 HSE:GFP vector to obtain the pTol2 HSE:cachd1,GFP construct for injection.

DNA extraction, KASP and HRMA genotyping
pTol2 gng8:GFP: a 3060 bp promoter region of the gng8 gene was amplified by PCR (see Table S5 for primer details) and cloned into a TOPO-TA vector. This fragment was subcloned into pEGFP-N1, upstream of the eGFP open reading frame, and the subsequent gng8:eGFP fragment cloned into pTol2.
For flow cytometry protein production, the coding sequence for the ectodomain of human and zebrafish CACHD1 (truncated before the transmembrane domain at P1095/P1108 respectively) was codon optimised for HEK cells and synthesised by GeneArt (Thermo Fisher Scientific, Waltham, MA, USA). These fragments had NotI and AscI target sequences at the 5′ and 3′ ends, respectively, for subcloning into prey protein and ectodomain bait protein production vectors.
Human and zebrafish CACHD1 prey protein expression constructs (ectodomain fused to a COMP domain, β-lactamase domain and FLAG tag) and zebrafish cachd1 ectodomain production constructs (ectodomain fused to hexahistidine and BirA ligase peptide substrate tags) were prepared by NotI/AscI restriction enzyme double digest (New England Biolabs) of pTT3-based vector backbones (50) (Addgene IDs 71471 and 36153) and shuttle vectors containing the synthesised fragments, followed by ligation with T4 ligase (New England Biolabs). The resulting constructs were screened by Sanger sequencing to confirm correct in-frame insertion.
To create human and zebrafish FZD-eGFP bait protein constructs, IMAGE consortium clones (51) (see Table S6 for details) were used as templates in PCR reactions to generate full length inserts (including the seven transmembrane domains) with NotI and AscI target sequences at the 5′ and 3′ ends, respectively, except for fzd4 and fzd9a where the insert was synthesised by GenScript (Piscataway, NJ, USA) as no complete full length clone was available. fzd1 and fzd8b both had NotI/AscI restriction sites in the respective coding sequences, so fusion PCR was used to generate full length inserts with synonymous mutations in the recognition sequences (see Table   S5 for primer sequences). The PCR products were purified using a Qiaquick PCR purification kit (Qiagen, Hilden, Germany) and then digested with NotI/AscI (New England Biolabs) and ligated to a pTT3 vector containing eGFP (see below). The resulting constructs were verified by Sanger sequencing.
The pTT3-eGFP vector was constructed by replacing the C-terminal tag encoding region of a bait protein vector (52) (Addgene ID 36150) with eGFP. The bait protein vector was digested with AscI/BamHI (New England Biolabs) to remove the tag encoding region and then ligated to an eGFP insert generated by PCR using primers with AscI and BamHI tails (see Table S5 for primer sequences). The resulting vector was verified by Sanger sequencing to ensure in-frame insertion of the eGFP coding sequence.

Protein production and purification
Conditioned media was harvested from transfected cultures, pooled and filtered through 0.2 µm filters and stored at 4˚C until use.
Prey protein transfections were quantified by β-lactamase assay, measuring the turnover of nitrocefin substrate by changing absorbance at 485 nm over time (50), then normalised by dilution.
Biotinylated bait ectodomain transfections were dialysed against PBS using SnakeSkin dialysis tubing (molecular weight cut-off 10,000 Da; Thermo Scientific) and several buffer changes (approximately 25 -30 L in total). Biotinylated protein concentration was quantified by ELISA, using streptavidin-coated microplates and a monoclonal antibody to detect the CD4d3+4 tag (50) (Nunc Immobilizer, Thermo Fisher Scientific).
Unbiotinylated ectodomain transfections were collected and quantified by ELISA using nickelcoated microplates and pooled for purification using nickel-sepharose columns (HisTrap HP, GE Healthcare, Chicago, IL, USA) and an AKTAxpress chromatography system (GE Healthcare).

Antibody generation and purification
To characterise the expression pattern of the receptor protein, we raised and affinity purified a polyclonal antibody against the recombinant extracellular domain of zebrafish Cachd1.
Briefly, purified zebrafish Cachd1 ectodomain was prepared (see above) and sent to Cambridge Research Biochemicals (Billingham, United Kingdom) for a rabbit immunisation protocol. Activity against the Cachd1 ectodomain in rabbit blood sera was confirmed by ELISA.
The blood serum was then affinity purified against biotinylated recombinant ectodomain immobilised on a streptavidin sepharose column, using an AKTAxpress chromatography system.
Purified antibodies were eluted in fractions using a low pH buffer, then immediately neutralised.
Peak fractions were tested for anti-Cachd1 activity, then pooled and dialysed against PBS. Total protein concentration was determined by absorbance at 280 nm by Nanodrop. The affinity purified antibody was checked for purity by SDS-PAGE and then validated by western blot, immunohistochemistry and flow cytometry (see Fig. S1 and data not shown) (30).

Retrogenix Cell Microarray Technology
Cell Microarray Technology (55) was used to identify potential binding partners for multimerised human CACHD1 ectodomain (prepared as above) and was performed by Charles

Flow cytometry
To The same procedure was followed for experiments testing the ability of OMP-18R5 to block Cachd1 prey-FZD-eGFP interactions, but with an additional incubation step before the application of prey proteins: cells were resuspended in OMP-18R5 diluted in 1% BSA in PBS (1:800) or 1% BSA in PBS only (control) and incubated for 30 minutes on ice, washed three times in PBS, and then resuspended in prey protein dilutions.
Mock transfection controls were used to determine forward and side scatter voltages for samples prior to data collection, and for background gating thresholds in data analysis. "Cells only" (no prey/primary or secondary antibodies) and "secondary antibody only" controls were included in every experiment. Flow cytometry data was analysed using FlowJo V10 (FlowJo, Ashland, OR, USA). Single cell populations were isolated using forward and side scatter values, bisected into eGFP-negative (untransfected) and eGFP-positive subpopulations and then the median value for phycoethryin fluorescence (indicating prey binding) calculated for each (Fig. S11B). Binding of prey protein to eGFP-positive cells was quantified by taking the ratio of the medians: ∆ = ln( + − ⁄ ).

In situ hybridization and Immunohistochemistry
Embryos or larvae were fixed in 4% paraformaldehyde and in situs performed following standard protocols (59). To create plasmid templates for in situ probe generation, regions of the zgc:101731, slc18a3b, aoc1 and cachd1 genes were PCR-amplified (see Table S5 for primer sequences) and TA-cloned into the pCRII vector. The kiss1 in situ probe template was generated directly by PCR (see Table S5). Previously published in situ probes used include (see Table S7 Table   S6. Antisense probes were generated with digoxigenin and fluorescein labelling kits (Roche). Antidigoxigenin-AP and anti-fluorescein-AP antibodies (Roche) coupled with 5-bromo-4-chloro-3′indolyphosphate and nitro-blue tetrazolium chloride were used to visualize colorimetric in situs.
Anti-digoxigenin-POD and anti-fluorescein-POD antibodies and Alexa Fluor-conjugated tyramides (Molecular Probes) were utilized for detection in fluorescent in situ hybridization.
In situ hybridisation chain reaction was performed according to published protocol (60) using Alexa Fluor-conjugated hairpin amplifiers and hybridisation buffers from Molecular Instruments Inc.
(Los Angeles, CA, USA). Probe sets for cachd1 and lrp6 are detailed in Table S8.
For immunohistochemistry, embryos were stained according to published protocol, with the exception of using freshly fixed embryos without storage in methanol (61). Antibodies used in this study were: anti-acetylated α-tubulin (Antibody registry ID: AB_477585, clone 6-11B-1, mouse

Heat shock, laser cell ablation, BrdU, labelling of habenular projections and transplantation experiments
For rescue experiments, embryos transgenic for Tg(HSE:cachd1,GFP)w160 were heat shocked for 30 minutes in a 40˚C water bath, then raised at standard temperature to 4 dpf and fixed in 4% paraformaldehyde.
Laser cell ablation, BrdU incorporation experiments and lipophilic dye labelling of habenular efferent projections were performed as previously described (16).

Imaging
For transmitted light pictures, larvae were mounted in glycerol and imaged using differential interference contrast optics (Leica CTR6000; 20× and 40× objectives; Leica Microsystems, Wetzlar, Germany). For confocal microscopy, heads were mounted in 1.2% low-melt agarose in glass-bottom dishes (MatTek, Ashland, MA, USA or LabTek, Grand Rapids, MI USA).
Wnt3a CM was generated from L cells.

Quantitative RT-PCR
RNA was extracted from cell culture or organoids according to the manufacturer's instructions (Qiagen RNeasy; Qiagen). cDNA was prepared using Maxima first strand cDNA synthesis kit with dsDNase (#1672, Thermo Fisher Scientific). Quantitative PCR detection was performed using PowerUp SYBR Green Master Mix (A25742, Applied Biosystems, Waltham, MA, USA). Assays for each sample were done in triplicate and were normalized to housekeeping genes ACTB (human β-ACTIN) or Hrpt1 (mouse). Primer sequences are listed in the Table S5. The Q′ test for equal proportions and modified Marascuilo procedure for multiple testing (using a Wilson variance calculation) are described in (63). Where the proportion was 0.1 <̂< 0.9

Statistics
and/or > 20, confidence intervals were calculated using a normal assumption; otherwise by the Wilson count method. Fluorescent and confocal microscopy images were adjusted globally for brightness and contrast using FIJI (v1.53n), scale bars added and then flattened into RGB images and exported as TIFFs.

Figure and manuscript preparation
Colour balance of wholemount in situ hybridisation images was adjusted in Adobe Photoshop CS6 (64 bit).