Daam1a mediates asymmetric habenular morphogenesis by regulating dendritic and axonal outgrowth

Neuroanatomical and functional differences between the left and right sides of the nervous system are conserved and crucial features that offer selective advantages for neuronal computation and behavioural performance (Concha et al., 2012). One central question that has engaged the attention of investigators for years is how asymmetric circuits develop in the embryonic nervous system. In vertebrates, this question has begun to be systematically addressed using zebrafish, a model organism that besides offering general advantages for genetic manipulation and direct visualisation of developmental processes, also reveals conspicuous asymmetries in the diencephalic epithalamus (Concha, 2004; Concha et al., 2009; Roussigne et al., 2011). The epithalamic region is formed by two main neuronal groups: the pineal complex (composed of pineal and parapineal nuclei) and the habenula (Hb) (Concha and Wilson, 2001). Among these nuclei, the Hb functions as a relay station that links limbic and striatal forebrain nuclei to monoaminergic systems of the ventral midbrain (Bianco and Wilson, 2009; Aizawa et al., 2011). In zebrafish, the Hb comprises dorsal (d-Hb) and ventral (v-Hb) domains that are homologous to the medial and lateral Hb of mammals, respectively (Aizawa et al., 2005; Aizawa et al., 2011). In contrast to mammals, the zebrafish d-Hb develops pronounced cellular and molecular asymmetries from about 36 hours post-fertilisation (hpf) that appear to rely on lateralised differences in the timing of neurogenesis (Aizawa et al., 2007). Such genetically encoded asymmetries are proposed to depend, at least in part, on early interactions established between the Hb and the left-sided-positioned parapineal nucleus (Concha et al., 2003; Gamse et al., 2003; Bianco et al., 2008), and results in the asymmetric organisation of molecular markers that define lateral (d-Hb_L) and medial (d-Hb_M) cellular subdomains of the d-Hb (Gamse et al., 2003; Aizawa et al., 2005; Agetsuma et al., 2010). One important additional feature exhibited by unipolar neurons of the d-Hb_L and d-Hb_M is their distinct axonal terminal morphology and target selectivity for dorsal and ventral domains of the interpeduncular nucleus (IPN) (Gamse et al., 2003; Aizawa et al., 2005; Bianco et al., 2008). As a consequence of such selectivity and the asymmetric organisation of d-Hb_L and d-Hb_M domains, neuronal projections from the left (primarily d-Hb_L) and right (primarily d-Hb_M) are largely segregated within dorsal and ventral domains of the IPN, respectively (Aizawa et al., 2005; Gamse et al., 2005), and this distinctive neuronal substrate appears to have important roles in the modulation of behavioural responses to fear, anxiety and stress-inducing stimuli (Facchin et al., 2009; Agetsuma et al., 2010; Lee et al., 2010; Jesuthasan, 2011).

Besides the observed cellular and molecular asymmetries, the zebrafish d-Hb also develops structural asymmetries in the size and organisation of neuropil domains (Concha et al., 2000) within which dendritic extensions from habenular neurons mix and synapse with axonal afferents arriving from other regions of the forebrain through the stria medullaris (Hendricks and Jesuthasan, 2007a; Bianco et al., 2008; Miyasaka et al., 2009). Habenular neuropil formation begins at ∼48 hpf, subsequent to the specification of d-Hb_L and d-Hb_M cellular domains, and becomes progressively enlarged on the left side reaching a conspicuous asymmetric array by 4 days of development (Concha et al., 2000; Concha et al., 2003). Importantly, organisation of neuropil asymmetry resembles the asymmetric distribution of kctd12.1 transcripts (Gamse et al., 2003), suggesting a causative link between neuropil organisation and the asymmetric distribution of cell subtypes in the d-Hb. Consistent with this idea, loss or gain of Kctd12.1 function affects habenular neuropil formation (Taylor et al., 2011), and the asymmetry of habenular neuropil becomes disrupted in conditions that affect the asymmetric expression of kctd12.1, such as after ablation of the parapineal nucleus (Concha et al., 2003; Bianco et al., 2008) and in mutants with defective Fibroblast growth factor 8 (Fgf8) (acerebellar^-/-, ace^-/-) (Regan et al., 2009) or Wnt/β-catenin (masterblind^-/-, mbl^-/-) (Carl et al., 2007) signalling. In these cases, the d-Hb develops into an almost bilaterally symmetric structure.

Despite our growing understanding of the signalling events that specify neuronal asymmetries in the d-Hb, we still know very little about how structural asymmetries in neuropil distribution arise. In this study, we identify a novel genetic component of the pathways that control asymmetric neuropil formation in the d-Hb of zebrafish. dishevelled associated activator of morphogenesis 1a (daam1a) was isolated from a reverse genetic screen searching for genes differentially expressed in left and right sides of the zebrafish brain. daam1a encodes a Diaphanous-related formin (Drf) protein that belongs to the phylogenetically conserved Formin family of actin assembly factors (Wallar and Alberts, 2003). The extent of Daam1a expression matches the asymmetric growth of habenular neuropil during embryonic and larval stages of zebrafish. Local loss and gain of Daam1a function in the left Hb prior to the onset of neuropil formation results in decreased or increased left habenular neuropil, respectively, without affecting neurogenesis or cell subtype specification. At the level of single habenular neurons, knockdown of Daam1a results in impaired growth of both dendritic and axonal extensions. Our results indicate that Daam1a is a key modulator of asymmetric habenular morphogenesis, mediating the outgrowth of dendritic and axonal processes in dorsal habenular neurons.

Embryos of zebrafish (Danio rerio) were obtained by natural spawning, raised at 28°C in embryo medium and staged according to morphology (Kimmel et al., 1995) and age (hours and days post-fertilisation; hpf and dpf). Zebrafish lines used were: wild-type Tübingen, Tg(foxD3:GFP) (Gilmour et al., 2002), Tg(pou4f1-hsp70l:GFP) (Aizawa et al., 2005), Tg(ET16:GFP) (Parinov et al., 2004), acerebellar^ti282 (Reifers et al., 1998) and masterblind^tm213 (Heisenberg et al., 2001). All animal protocols were approved by the Bioethics Committee of the Faculty of Medicine, University of Chile.

Right (R) and left (L) halves of juvenile (1-month-old) zebrafish brains were microdissected, mRNA isolated with Oligotex Direct mRNA Kit (QIAGEN) and cDNA synthesised by the Gubler-Hoffman method (Gubler and Hoffman, 1983). Suppression subtractive hybridisation (SSH) was performed using the PCR-Select cDNA Subtraction Kit (BD Biosciences-Clontech). For direct SSH, R cDNA was used as tester with L cDNA as driver (R-L) and the opposite was used for reverse SSH (L-R). Amplified R and L cDNAs were labelled with [α³²P] dCTP by random priming (Prime-a-Gene Labelling System, Promega) and used as probes to hybridise zebrafish cDNA commercial libraries: 611 Zebrafish Brain cDNA pt.2 330.1.545, 611 Zebrafish Brain cDNA pt.1 330.1.521 and 609 Zebrafish EST pt.1.357.1.512 (RZPD). Differential clones were sequenced and analysed in silico (www.ncbi.nlm.nih.gov, www.ensembl.org).

Total RNA was isolated from three independent pools of ten L and ten R halves of zebrafish brains, and cDNA generated using SuperScript II Reverse Transcriptase (Invitrogen). Primers used were: 5′-GGAGGTCATGGCGCGTCC-3′ (sense) and 5′-CCTCCC GAAGACGGTAGGTG-3′ (antisense) for daam1a; 5′-AGTTCTTTCAGCTGCGGGACCTTA-3′ (sense) and 5′-GCGACGGACAGTGTGCGAGAG-3′ (antisense) for kctd12.1; 5′-AGCTGTTTTAGATGGGGTGTTGTC-3′ (sense) and 5′-AATGTCCTGGTTCTGCCTTAC-3′ (antisense) for cpd2 (cadps2); and 5′-GGCACCGGTTCTGGCTTCAC-3′ (sense) and 5′-CGGAGATCACTGGGGCATAGGTA-3′ (antisense) for α-tubulin (control). Quantitative PCR was carried out on a 7300 real-time PCR system (Applied Biosystems, Carlsbad, CA, USA) with Platinum SYBR Green qPCR Supermix-UDG (Invitrogen). Data were compiled and collected using MxPro QPCR (Agilent Technologies), and presented as the fold change in gene expression normalised to α-tubulin and expressed as L relative to R or R relative to L. Relative fold changes were calculated by a comparative C (T) method (Schmittgen and Livak, 2008).

Whole-mount in situ hybridisation was performed as described (Thisse and Thisse, 2008) using antisense probes for kctd12.1, pou4f1 (Aizawa et al., 2005), lefty1 (Bisgrove et al., 1999), southpaw (Long et al., 2003) and pitx2 (Essner et al., 2000). Antisense daam1a was synthesised from the partial coding sequence contained in a commercial clone (UCDMp611A02132Q14, RZPD). The anti-Daam1 antibody (Abnova, 1:50) (Liu et al., 2008) recognised 111 amino acids of the N-terminal region of the human DAAM1, which shares 84% identity with the zebrafish Daam1a protein. For anti-Daam1 and anti-GFP (Invitrogen, 1:1000), fixed embryos were treated with 1 mg/ml collagenase (Sigma) (Burgess et al., 2009). Immunostaining for anti-acetylated α-tubulin (Sigma-Aldrich, 1:1000) was performed as described (Concha et al., 2003). For anti-α-tubulin (Sigma, 1:50) and Alexa Fluor 568 phalloidin (Invitrogen, 1:200), embryos were fixed in 3.7% formaldehyde (Merck) in 0.8% MeOH/PBS, or in 4% paraformaldehyde (PFA) (Polysciences). Alexa Fluor 488-, 568- and 647-conjugated secondary antibodies were used (Invitrogen, 1:200). For nuclear counterstaining, embryos were incubated at 4°C with NeuroTrace 488 green Nissl (Molecular Probes, 1:250) (Signore et al., 2009) and TO-PRO-3 iodide stain (642/661) (Molecular Probes, 1:1000) (Regan et al., 2009).

Antisense morpholino oligonucleotides (MOs) against southpaw (spaw-MO, 10 ng/embryo) (Long et al., 2003), no tail (ntl-MO, 1.8 ng/embryo) (Feldman and Stemple, 2001), ulk2 (ulk2^spl-MO, 2 ng/embryo) (Taylor et al., 2011) and daam1a (daam1a-MO; 8 ng/embryo; 5′-AGCTATGACCCCCTCTCAAAATGGC-3′, tagged with lissamine at the 3′) (Kida et al., 2007), and control morpholino (Co-MO, 8 ng/embryo; 5′-AGCTATGACCCCCTCTCAAAATGGC-3′, tagged with lissamine at the 3′) were injected in one- to four-cell zebrafish embryos using standard procedures. Efficiency of gene knockdown was assessed by phenotype, epithalamic lefty1 expression (absent in spaw-MO and bilateral in ntl-MO), or Daam1a immunostaining (reduced in daam1a-MO). pCS2-GFP-hDaam1 and unc-51-like kinase (ulk2) [gifts from Drs R. Habas (College of Science and Technology, Temple University, Philadelphia, PA, USA) and J. Gamse (Medical Center, Vanderbilt University, Nashville, TN, USA), respectively] were used for standard mRNA in vitro synthesis. For rescue experiments, embryos were co-injected with one of the following combinations: 200 pg GFP-hDAAM1 mRNA + 8 ng daam1a-MO; 200 pg GFP-hDAAM1 mRNA + 8 ng Co-MO; 200 pg GFP-hDAAM1 mRNA + 2 ng ulk2^spl-MO; 150 pg ulk2 mRNA + 2 ng ulk2^spl-MO; or 150 pg ulk2 mRNA + 8 ng daam1a-MO.

Focal electroporation was performed as described (Tawk et al., 2009) using pCS2-GAP43-mCherry or pCS2-GAP43-GFP plasmids. Local electroporation was adapted from a previous report (Hendricks and Jesuthasan, 2007b). In brief, glass injection needles were filled with daam1a-MO (1 mM), Co-MO (1 mM), pcDNA-C-DAAM1-HA (1.5-2 μg/μl) or pCS2-GAP43-GFP (1.5-2 μg/μl). MO or DNA was injected into the third ventricle using a Picospritzer III (Parker Hannifin). Five 25-volt pulses, each lasting 1 msecond, were manually applied at 20 Hz using a Grass DS9 stimulator.

Larvae were fixed in 4% PFA in PBS and the skin covering the dorsal diencephalon and eyes was removed. Crystals of the lipophilic dyes DiD and DiO (Molecular Probes) were applied to L and R Hb, respectively, using tungsten needles connected to a micromanipulator (Aizawa et al., 2005; Signore et al., 2009). Labelled larvae were incubated in 0.5% PFA in PBS at 48°C for 2 days in the darkness, to allow lipophilic dyes to reach the IPN.

Embryos were anaesthetised (0.003% tricaine), mounted in 1% low melting agarose and imaged as described (Concha et al., 2003; Signore et al., 2009) using UltraView RS spinning disc microscope (PerkinElmer) with C-Apochromat 40×/1.2 W or 63×/1.2 W, or Fluoview1000 Spectral confocal microscope (Olympus) with Plan-Apochromat 40×/1.0 W. Images were deconvolved to reduce blurring and noise (Huygens Professional, SVI) and optimised for brightness/contrast keeping the dynamic range of raw images. 3D image projections were obtained using Volocity (PerkinElmer), ImageJ (http://rsbweb.nih.gov/ij/) and SCIAN-Soft (http://www.scian.cl). Final figure panels were selected according to both consensus criteria and representative phenotype.

The number of HuC (Elavl3)- and HuD (Elavl4)-positive (+) nuclei in the Hb labelled by anti-HuC/D immunostaining (Invitrogen, 1:500) was determined from confocal image z-stacks as described (Roussigné et al., 2009). Volume was measured in 3D reconstructions using custom-made image-processing routines written in SCIAN-Soft, which involved thresholding, manual and automated segmentation, and generation of binary masks of the habenular neuropil and IPN.

Statistical analyses were performed using Origin Pro8 software. Comparison of individual datasets from different conditions was performed using the Kolmogorov-Smirnov (KS) test. A t-test was used to compare differences in IPN volume. The Mann-Whitney non-parametric test was used to compare differences in neuropil volume and values of L-R gene expression obtained by qPCR. Significance was set at P<0.05.

To identify genes with asymmetric expression in the zebrafish brain, cDNAs obtained through reverse transcription of mRNA samples isolated from L and R halves of juvenile brains were used for direct (R-L) and reverse (L-R) SSH (Fig. 1A). Differential cDNA products were used as probes to screen zebrafish cDNA libraries. Library clones that were differentially hybridised by probes from either direct or reverse SSH but not both, and those for which strength of hybridisation significantly differed between the two SSH conditions (Fig. 1B), were chosen for further in silico analysis and assessment of mRNA expression through whole-mount in situ hybridisation. We identified daam1a as a gene enriched on the left side of the zebrafish brain. qPCR analysis corroborated the presence of higher levels of daam1a mRNA in the L compared with the R side of juvenile brains (Mann-Whitney test, P<0.025) (Fig. 1C).

daam1a mRNA was detected in a distinct domain of the d-Hb that is larger on the L than on the R. Asymmetric expression was detected from ∼48 hpf (Fig. 1D,E), became conspicuous at 72 hpf (Fig. 1F), and continued throughout larval (Fig. 1G) and juvenile (Fig. 1H) stages into adulthood (data not shown). In the larvae, daam1a mRNA localised to the asymmetric habenular neuropil (Fig. 1G). Whole-mount immunofluorescence (using an antibody directed against human DAAM1 that also detects zebrafish Daam1a) (Liu et al., 2008) revealed a modest asymmetric expression of Daam1 at the outset of habenular neuropil formation (Fig. 1I) that progressively increased following the changes in neuropil growth (Fig. 1J-L). At the subcellular level, Daam1a was detected in a punctate pattern that was highly enriched in the habenular neuropil (Fig. 1L) and could be detected in the soma (Fig. 1M, arrow) and dendritic processes (Fig. 1N, arrowheads) of single habenular neurons. Consistent with Daam1a levels mirroring the growth of habenular neuropil, expression of daam1a was symmetric in the Hb of ace^-/- and mbl^-/- mutants, which develop symmetric habenular neuropil (Carl et al., 2007; Regan et al., 2009) (supplementary material Fig. S1; Table 1). Conditions in which Nodal signalling in the brain is either absent (e.g. after southpaw morpholino injection, spaw-MO) (Long et al., 2003) or bilateral (e.g. after no tail morpholino injection, ntl-MO) (Feldman and Stemple, 2001), and which result in randomisation of habenular neuropil asymmetry (Concha et al., 2000; Gamse et al., 2005), also resulted in randomised directionality of daam1a asymmetric expression in the Hb (supplementary material Fig. S1; Table 1). Altogether, these results indicate that expression levels of Daam1a mirror the asymmetric growth of habenular neuropil in the Hb, suggesting that Daam1a is a readout of asymmetric habenular morphogenesis.

To investigate a possible role of Daam1a in asymmetric habenular morphogenesis, we initially performed global Daam1a knockdown by injection at the one- to four-cell stage of a morpholino antisense oligonucleotide designed to block daam1a translation (daam1a-MO) (Kida et al., 2007). Embryos injected with control morpholino (Co-MO) developed a characteristic left-sided enlarged habenular neuropil as assessed by anti-acetylated α-tubulin immunostaining (Fig. 2A, brackets) (Concha et al., 2000; Concha et al., 2003). In addition, axons from habenular neurons labelled with the pou4f1-hsp70l:GFP transgene (Aizawa et al., 2005) projected through the fasciculi retroflexus to the IPN, where they organised in a characteristic ring-shape pattern (Fig. 2K). Global loss of Daam1a function induced two main classes of phenotypes in the Hb. First, some daam1a-MO embryos (32%; n=324) displayed reversed directionality of habenular asymmetry with neuropil in the R Hb showing L-side characteristics and L Hb displaying neuropil properties of the R Hb (Table 2). We found that such complete reversal of neuropil asymmetry was due to an early role of Daam1a in the control of embryo asymmetry, which resulted in disruption of Nodal signalling in both the lateral plate mesoderm and the epithalamus, and subsequently in randomised directionality of epithalamic asymmetry (data not shown).

The second class of habenular phenotype involved a reduction in both the extent of habenular neuropil (Fig. 2B-D, brackets) and the number of habenular axons reaching the IPN (Fig. 2L-N). Most daam1a-MO embryos displayed a moderate phenotype in both the Hb (54%; Fig. 2C,P) and the IPN (62%; Fig. 2M,Q). In these cases, which comprised left-sided and reversed phenotypes with mild or moderate neuropil reduction, the pattern of neuropil still remained asymmetric (Fig. 2C); this is in contrast to the most severe phenotypes (26%) in which both sides of the Hb developed a minimal amount of neuropil (Fig. 2D). Co-injection of daam1a-MO and human DAAM1 (hDAAM1) mRNA resulted in habenular neuropil and IPN connectivity similar to controls (Fig. 2E,O-Q), indicating that abrogation of Daam1a function was directly responsible for the observed phenotypes. Indeed, functional loss of Daam1a did not disrupt habenular neurogenesis as the number of HuC (+) post-mitotic habenular neurons was similar in control and daam1a-MO embryos at 48 hpf (daam1a-MO=230±22; Co-MO=234±17; mean±s.d.; P>0.05; n=9). Specification of habenular cell subtypes was also unaffected as revealed by the normal asymmetric organisation of GFP (+) domains after Daam1a knockdown in two transgenic lines labelling distinct asymmetrically distributed neuronal subtypes of the d-Hb, Tg(pou4f1-hsp70l:GFP) (Aizawa et al., 2005) and Tg(ET16::GFP) (Parinov et al., 2004) (Fig. 2G-I; Table 2). Furthermore, the parapineal nucleus was lateralised towards the side of enlarged habenular neuropil, and in the most severe neuropil phenotypes (Fig. 2D) the parapineal still adopted a lateral position (data not shown), suggesting that Daam1a knockdown affected events downstream of the pathways that control the lateral positioning of the parapineal nucleus. The only asymmetric habenular marker that was reduced after Daam1a knockdown, other than daam1a, was kctd12.1 (supplementary material Fig. S2; Table 2), which also localises to the habenular neuropil (Gamse et al., 2003). In conclusion, these results indicate that Daam1a is required for habenular asymmetric morphogenesis, in particular for the development of asymmetric neuropil and the establishment of connectivity from the Hb to the IPN.

To verify that Daam1a plays a direct role in neuronal morphogenesis of the Hb, we performed spatio-temporally restricted knockdown of Daam1a through local electroporation of daam1a-MO directly into the habenular region at a developmental stage that followed cell subtype specification and preceded the onset of habenular morphogenesis (44 hpf) (Fig. 3A). We focused on targeting the L Hb, as this was the primary site of endogenous daam1a expression, although we obtained similar results after bilateral electroporation of the Hb (Fig. 3; supplementary material Fig. S5). Local electroporation of daam1a-MO efficiently led to both morpholino incorporation (supplementary material Fig. S3) and reduced levels of Daam1a protein expression (supplementary material Fig. S4). Importantly, the extent of neuropil was markedly reduced in the L Hb of daam1a-MO-electroporated embryos (Fig. 3C, brackets) compared with controls (Fig. 3B, brackets), similar to the severe phenotypes observed after global knockdown of Daam1a (compare Fig. 3C with Fig. 2D). Indeed, quantification of habenular neuropil volume revealed a significant decrease in the L Hb of electroporated embryos compared with controls (L-Hb-daam1a-MO=6752±442 μm³; L-Hb-Co-MO=10,042±747 μm³; P<0.05; n=5) (Fig. 3N,O,Q) whereas neuropil of the R Hb remained unaffected (Fig. 3Q). In addition to changes in habenular neuropil, left-sided local electroporation of daam1a-MO also induced a decrease in the extent of pou4f1-hsp70l:GFP (+) axonal labelling in the IPN (Fig. 3I; supplementary material Fig. S5) and in the volume of L Hb projections reaching the dorsal IPN (dIPN) compared with controls (L-Hb-dIPN-daam1-MO=13,001±502 μm³; L-Hb-dIPN-Co-MO=29,820±1595.4 μm³; P<0.05; n=4), whereas projections to the ventral IPN (vIPN) from either R or L Hb appeared unaltered (data not shown, P>0.05; n=4) (Fig. 3K,L). As we did not observe axons projecting to ectopic brain regions (data not shown) and the number and organisation of habenular cell subtypes appeared unaffected as revealed by the normal expression of the pou4f1-hsp70l:GFP transgene (Fig. 3C,F), we interpreted the loss of IPN connectivity as a primary defect in habenular axonal extension. In summary, local knockdown experiments indicate that Daam1a is required after 44 hpf in the habenular region for the growth of asymmetric neuropil and habenular efferent connectivity to the IPN.

To determine whether increased levels of Daam1a activity could promote habenular neuropil formation and axon extension, we performed local electroporation of a plasmid coding for a truncated form of DAAM1 that lacks the N-terminal regulatory domains (C-Daam1) and works as a dominant-positive (constitutively active) form (Fig. 3C) (Habas et al., 2001; Matusek et al., 2006; Ju et al., 2010). Local electroporation of pcDNA-C-DAAM1-HA in the habenular region at 44 hpf resulted in an ipsilateral increase of neuropil when directed to the L (Fig. 3D; n=8) but not to the R (supplementary material Fig. S6; n=9), compared with embryos electroporated with a control plasmid (data not shown; n=6), a finding that was corroborated by quantification of neuropil volume (L-Hb-C-DAAM1=14,473±119 μm³; L-Hb-Co-plasmid=10,239±824 μm³; P<0.05; n=5) (Fig. 3P,Q). By contrast to the differential effect in habenular neuropil growth, local electroporation of pcDNA-DAAM1-HA led to increase of pou4f1-hsp70l:GFP (+) habenular axons in the IPN when directed to either L (Fig. 3J,M) or R (supplementary material Fig. S6). In left-side-electroporated embryos, the volume of L Hb projections in the dIPN increased significantly (L-Hb-dIPN-C-DAAM1=36,147±818 μm³; L-Hb-dIPN-Co-plasmid=27,603±1600 μm³; P<0.05; n=3) whereas projections to the vIPN from either L or R Hb appeared unaffected (data not shown, P>0.05; n=3) (Fig. 3K,M). Habenular expression of pou4f1-hsp70l:GFP remained unchanged irrespective of the side (L, R or bilateral) of pcDNA-C-DAAM1-HA electroporation (Fig. 3G; supplementary material Fig. S6), indicating that habenular cell subtype organisation was unaffected. We thus conclude that over-activation of Daam1 in the habenular region, prior to and/or during the onset of neuropil and axonal morphogenesis, is sufficient to promote the growth of both habenular neuropil (when directed to the L) and axonal efferents to the IPN (when directed to either L or R).

To assess directly whether dendritic and axonal outgrowth was sensitive to the modulation of Daam1a, we analysed the pattern of dendritic arborisation and shape of axonal terminals of individual habenular neurons labelled through focal electroporation of pCS2:GAP43-GFP. In control embryos, we recognised the stereotypical unipolar morphology of habenular neurons (Bianco et al., 2008), with a dense dendritic arbour (Fig. 4A, arrowheads) and a single unbranched axon extending from the soma (Fig. 4A, arrow; n=6). In addition, we observed the characteristic complex pattern of axon terminal morphology in the IPN, with bilateral and spiral branches (Fig. 4E) (Bianco et al., 2008). By contrast, habenular neurons from daam1a-MO embryos showed a greatly pruned dendritic arbour (Fig. 4B, arrowheads) and an incomplete and less complex pattern of axonal terminals in the IPN (Fig. 4F; n=5), compared with controls. Electroporated neurons from embryos co-injected with daam1a-MO and hDAAM1 mRNA at the one- to two-cell stage developed a pattern of dendritic arborisation similar to controls (Fig. 4C, arrowheads) and showed a partial rescue of axonal terminal morphology in the IPN (Fig. 4G). Finally, habenular neurons from embryos locally electroporated with pcDNA-C-DAAM1-HA at 44 hpf showed increased dendritic arborisation (Fig. 4D, arrowheads) and axonal extension (Fig. 4H) in the IPN. Altogether, these findings corroborate that Daam1a regulates dendritic and axonal outgrowth in dorsal habenular neurons.

As members of the Drf subfamily of Formin proteins act as actin assembly factors (Wallar and Alberts, 2003) and also play a less common role in regulating the stability of microtubules (Ju et al., 2010), we investigated whether the global loss of Daam1a function affected the organisation of F-actin and α-tubulin in the Hb. Embryos injected with Co-MO and co-stained with fluorescent phalloidin (to label F-actin) and Nissl (to delineate the cellular context of the Hb) showed a distinct smooth fibrillar organisation of F-actin in the Hb, with increased density of labelling in the asymmetric neuropil domains (Fig. 5A,D,G; n=12). By contrast, F-actin staining in the Hb of daam1a-MO embryos was irregular in shape and showed thicker and shorter bundles compared with controls (Fig. 5B,E,H; n=10). In these embryos, α-tubulin was also decreased and exhibited a less elaborate distribution than in controls (Fig. 5K,N,Q; n=10). Co-injection of daam1a-MO and hDAAM1 mRNA resulted in partial rescue of the normal architecture of F-actin organisation in the Hb (Fig. 5C,F,I; n=10) and the strength of α-tubulin staining was increased (Fig. 5L,O,R; n=10). Altogether, these results indicate that Daam1a is required for the normal assembly and organisation of actin filaments in the developing zebrafish Hb, and also suggest a possible additional role for Daam1a in regulating microtubules.

As the zebrafish Unc-51-like 2 (Ulk2) kinase has been recently implicated in the control of habenular neuropil morphogenesis downstream of Kctd proteins (Taylor et al., 2011), and the functional loss of Ulk2 induces a reduced habenular neuropil phenotype that resembles Daam1a knockdown (compare Fig. 6B and 6C), we investigated whether Daam1a and Ulk2 interact in the promotion of neuropil growth. Co-injection of morpholino directed against daam1a and ulk2 (ulk2^spl-MO) at concentrations that in the context of single knockdowns induce only a modest reduction of habenular neuropil (Fig. 6D,E), led to severe reduction of habenular neuropil (Fig. 6F). Furthermore, hDAAM1 mRNA rescued the habenular neuropil phenotype induced by loss of Ulk2 (Fig. 6H) whereas ulk2 mRNA retained the habenular neuropil phenotype caused by knockdown of Daam1a (Fig. 6I). Altogether, these findings indicate that Daam1a and Ulk2 play a synergistic role in the promotion of neuropil growth, and that Daam1a is likely to work downstream of Ulk2 in the process.

In this study, we used a reverse genetic approach to identify daam1a as a new genetic component of the pathways that control asymmetric habenular morphogenesis. Expression of daam1a in the developing Hb follows the specification of cell subtypes and mirrors the asymmetric changes in neuropil growth at embryonic and larval stages. Consistent with a role of Daam1a downstream of the main pathways that control the development of brain asymmetry (e.g. Fgf and Wnt) and its directionality (Nodal) in the brain, habenular daam1a expression becomes symmetric in ace^-/- and mbl^-/- mutants, and randomised after loss of Spaw or Ntl function. Importantly, loss and gain of Daam1a function decreases and increases, respectively, the formation of habenular neuropil and efferent axonal connectivity to the IPN without affecting habenular neurogenesis or cell subtype specification. These findings, and the fact that Daam1a promotes neuropil growth downstream of Ulk2, indicate that Daam1a plays a key role in asymmetric habenular morphogenesis, mediating the growth of habenular neuropil and efferent axonal connectivity to the IPN.

The opposite modulatory effect shown by loss and gain of Daam1a function in the ability to promote dendritic and axonal growth provides strong evidence in support of a role of Daam1a as a key modulator of neuronal morphogenesis in the Hb. The punctate distribution of Daam1a expression in the developing habenular neuropil, and more specifically in dendrites of habenular neurons, is also consistent with Daam1a participating in neuronal process outgrowth. Indeed, functional abrogation of Daam1a markedly disrupts the formation of dendritic processes examined in single habenular neurons, also leading to an incomplete and less complex pattern of axonal terminals in the IPN. Consistent with this, Drosophila Daam (dDaam) and mouse Daam1 (mDaam1) both regulate filopodial formation and neurite outgrowth in developing neurons (Matusek et al., 2008; Salomon et al., 2008; Gonçalves-Pimentel et al., 2011). dDaam shows a punctate distribution in F-actin-rich structures in the growth cones of cultured primary neurons (Matusek et al., 2008), whereas puncta of mDaam1 expression are observed in the dendritic shaft of pyramidal cells of the hippocampus, and in synaptic regions where Daam1 colocalises with pre-synaptic (Sv2) and post-synaptic (Psd95; also known as Dlg4) markers (Salomon et al., 2008). Importantly, dDaam is required for filopodial extension at axonal growth cones (Matusek et al., 2008; Gonçalves-Pimentel et al., 2011), whereas mDaam1 appears to play a role in dendritic extension (Salomon et al., 2008). In zebrafish, habenular neurons require Daam1a for both dendritic and axonal morphogenesis, although these two events appear to show differential Daam1a requirements, as they are affected to different extents after Daam1a knockdown (dendritic>axonal) and can be rescued to distinct degrees by overexpression of hDAAM1 (again, dendritic>axonal). In addition, the extension of dendritic processes appears to be subject to local growth control depending on the side of the Hb. For example, local electroporation of C-DAAM1 induces habenular neuropil growth only when directed to the L side, which suggests a mechanism that moderates the influence of Daam1 on neuropil growth in the R Hb, without affecting its ability to promote axonal extensions to the IPN. These results appear to be consistent with a recently proposed model of asymmetric local modulation of habenular neuropil growth mediated by Kctd proteins in zebrafish (Taylor et al., 2011). In this model, Kctd12.1 and Kctd12.2 are expressed primarily in the L and R Hb, respectively, and exert a stronger inhibition of neuropil growth on the R than the L Hb through interacting with Ulk2, a kinase that is expressed bilaterally in the Hb from 48 hpf (Taylor et al., 2011). Whether Kctd proteins inhibit the ability of Daam1a to promote neuropil growth in the R Hb is yet to be determined. However, the onset of asymmetric daam1a expression coincides with the onset of ulk2 expression and both are concurrent with neuropil growth in the Hb, suggesting a link between these events. Indeed, Daam1a and Ulk2 play a synergistic role in the promotion of neuropil growth. Furthermore, hDAAM1 overexpression rescues the impaired neuropil phenotype induced by functional loss of Ulk2 (whereas Ulk2 appears to be unable to rescue the phenotype induced by Daam1a knockdown) suggesting that Daam1a works downstream of Ulk2. Altogether, these findings are consistent with a collaborative role of Daam1a and Ulk2 in the promotion of neuropil growth downstream of the pathways that establish cell subtype specification in the Hb. Where and how such collaboration occurs remains to be determined. An intriguing possibility is the endocytic pathway, as Ulk1 and Ulk2 localise to vesicular structures in growth cones and regulate neuronal process extension through early endosome trafficking at nascent neurites (Tomoda et al., 2004; Zhou et al., 2007; Toda et al., 2008), whereas zebrafish Daam1a is actively transported to endocytic vesicles and its asymmetric subcellular localisation appears to be relevant for the convergent-extension movements of notochordal cells (Kida et al., 2007).

At the effector level, Daam1a regulates both F-actin and microtubule assembly during habenular morphogenesis. The former role is consistent with previous reports indicating that Daam1 works as an F-actin nucleation and elongation factor induced by Rho GTPases (Prokop et al., 2011), and that dDaam cooperates with Arp2 and Arp3 in a common mechanism of filopodia formation that depends on Enabled and is regulated through Profilin (Chickadee - FlyBase) activity (Gonçalves-Pimentel et al., 2011). Reports linking Daam1 to microtubule assembly and stabilisation are, however, scarce (Ju et al., 2010). The observed increase in intensity of α-tubulin (+) labelling after expression of hDAAM1 suggests that zebrafish Daam1a regulates microtubule dynamics in the Hb. Regulation of F-actin and microtubule dynamics might result from Daam1a interacting with other signalling pathways, such as Wnt/PCP signalling, which is known to modulate the growth of axons and dendrites (Wang and Nathans, 2007; Tissir and Goffinet, 2010). Daam1 works as an intermediary of Wnt/PCP signalling that leads to Rho activation and subsequent remodelling of the cytoskeleton during Xenopus and zebrafish gastrulation (Habas et al., 2001; Kida et al., 2007), and some ligands and core components of Wnt/PCP are expressed in the developing Hb at the time of neuropil formation (data not shown).

Here, we demonstrate that Daam1a plays a key role in asymmetric habenular morphogenesis by regulating dendritic and axonal outgrowth in dorsal habenular neurons, possibly through the modulation of cytoskeletal dynamics. Although Daam1 and Ulk2 have a synergistic role in the promotion of habenular neuropil growth, downstream of the pathways that establish asymmetric cell subtype specification, further studies are needed to address how these two factors interact and become differentially modulated on L and R sides of the Hb.

We thank Ricardo Armisen for critical reading of the manuscript; Drs Raymond Habas and Joshua Gamse for providing plasmids; members of the Laboratory of Experimental Ontogeny Fish Facility for fish care; Erika Labbe and Jorge Jara for technical help with image processing; Carmen Gloria Lemus for advice in the statistical approach; and Luis Michea for sharing the use of the qPCR machine.