Prdm14 acts upstream of islet2 transcription to regulate axon growth of primary motoneurons in zebrafish

The formation of interconnected neural circuit requires the proper positioning of different populations of neurons, the correct projection of axons and dendrites, and the remodeling of complicated neural connections. Extrinsic cues, such as neurotrophic factors, have long been thought to play major roles in setting up neural circuits, especially in promoting axon outgrowth (Markus et al., 2002). In addition to extrinsic cues, accumulating evidence also indicates that cell-intrinsic programs are required for axon growth and target projection. For example, the upregulation of transcription factors such as SnoN, Id2, NFAT and CREB has been reported to induce axon elongation (Kanning et al., 2010; Lee and Pfaff, 2001; Mizuguchi et al., 2001; Pattyn et al., 2003). Interestingly, activation of some of the transcription factors is dependent on extrinsic cues, suggesting that, at least in some cases, extrinsic cues and intrinsic cues are interconnected (Lee and Pfaff, 2001; Mizuguchi et al., 2001; Pattyn et al., 2003). Thus, combinational functions of intrinsic and extrinsic cues might be essential for properly setting up neural circuits.

Among neural networks, motor circuits are responsible for the execution and coordination of animal movement. Within motor circuits, motoneurons (MNs) provide the action connections between the nervous system and motor output. A number of transcription factors have been identified to contribute to the initial MN diversification and in maintaining the identities of MN subtypes (Kanning et al., 2010; Lee and Pfaff, 2001; Lewis and Eisen, 2003). MNs come from the ventrolateral MN domain region, in which the transcription factors Nkx6.1 and Pax6 are expressed. Combinational activities of these two factors allow expression of Olig2, a bHLH transcription factor, in the domain (Kanning et al., 2010; Pattyn et al., 2003). Olig2, combined with another bHLH factor, Neurogenin 2, activates MN-specific transcription factors such as Mnr2, Hb9, Lim3 and Islet1/2 to specify MN progenitors (Mizuguchi et al., 2001). Two classes of transcription factors encoded by the LIM-HD and Hox genes, in combination of extrinsic factors, also control MN axon outgrowth (Bonanomi and Pfaff, 2010).

In zebrafish, the primary MNs (PMNs) are derived from the ventral spinal cord regions on both sides of the floor plate. PMNs are classified into caudal primary (CaP), middle primary (MiP) and rostral primary (RoP) MNs. Separately, variable primary (VaP) MNs initially develop equivalently to CaP but later degenerate (Eisen et al., 1986; Eisen et al., 1990; Myers et al., 1986; Westerfield et al., 1986). The three major PMNs can be recognized by their stereotypical cell body positions and axon projection patterns (Westerfield et al., 1986). The projection patterns are due to their initial developmental axon selections, rather than being the result of the elimination of wrongly formed projections (Eisen et al., 1986). CaP, MiP and RoP axons leave the spinal cord sequentially at the same ventral root and extend ventrally until they reach the horizontal myoseptum (HMS). After a brief pause at the HMS, the three axon subtypes diverge in their growth pathways within the myotomes. CaP axons further extend ventrally, whereas MiP axons branch and extend dorsally, and RoP axons extend laterally in the HMS (Eisen et al., 1986; Myers et al., 1986; Westerfield et al., 1986).

It is conceivable that different PMNs require unique molecular programs for their specification and morphogenesis (Kanning et al., 2010; Lewis and Eisen, 2003). Among key factors controlling zebrafish PMN development, two LIM-HD transcription factors, Islet1 and Islet2, are essential for PMN development. islet1 expression is initially detected in all postmitotic PMNs, but its expression is only maintained in MiP and RoP at later stages [15 hours postfertilization (hpf)], whereas islet2 expression enriches specifically in CaP and VaP (Appel et al., 1995; Tokumoto et al., 1995). In the absence of Islet1, zebrafish PMN precursors switch fates and form interneurons (INs) (Hutchinson and Eisen, 2006). By contrast, inhibition of zebrafish Islet2 function by a dominant-negative mutant LIM^isl2 results in failure of CaP neurons to project their axons beyond the HMS (Segawa et al., 2001), indicating that Islet2 is essential for CaP axon outgrowth. Despite the highly regulated expression patterns of Islet1 and Islet2, their functions may be redundant for PMN specification in terms of axon projection trajectories (Hutchinson and Eisen, 2006). Thus, PMN subtype establishment and axon outgrowth differences might be controlled by different upstream transcription factors that differentially regulate islet1 and islet2. For example, late-phase islet1 expression in MiP is activated by Nkx6. Disruption of Nkx6 causes MiP to develop a more IN-like morphology (Hutchinson et al., 2007). However, the immediate upstream regulator of islet2 remains unknown.

In addition to the transcription factors mentioned above, a number of Prdm family (PRDI-BF1 and RIZ homologous region) proteins play diverse roles at different stages of neural development. For example, mouse Prdm16 is expressed in neural stem cells and has been shown to promote cell identity by modulating oxidative stress (Chuikov et al., 2010). In Drosophila, Hamlet, the homolog of Prdm16, controls odorant gene choice and axon targeting specificity through modifying cellular responses to Notch signals in a context-dependent manner (Endo et al., 2011). In zebrafish, Prdm1a is expressed at the edge of the neural plate and is required for the specification of neural crest and Rohon-Beard (RB) sensory neurons (Hernandez-Lagunas et al., 2005; Olesnicky et al., 2010; Rossi et al., 2009). Prdm6, 8, 12, 13 and 16 were shown to be expressed in the developing mouse CNS (Kinameri et al., 2008), but their exact functions remain to be defined. Another important member of the family is Prdm14, a role for which has not been reported in neural development. Instead, it is indispensable for the establishment of primordial germ cells (PGCs), with roles in the reacquisition of pluripotency and epigenetic reprogramming (Ohinata et al., 2009; Yamaji et al., 2008). Recent studies have revealed that PRDM14 is also a key component of the core transcriptional regulatory network for maintaining embryonic stem cell identity by directly regulating the pluripotency gene POU5F1, and may prevent embryonic stem cells from differentiating into extra-embryonic endoderm (Chia et al., 2010; Ma et al., 2011; Tsuneyoshi et al., 2008).

Here, we report a zebrafish gene-trap mutant, slg, in which prdm14 gene expression is disrupted, resulting in axon growth defects in CaP. We show that Prdm14 is required for islet2 activation and that Islet2 mediates major functions of Prdm14 in CaP. These results strongly argue that, in addition to extrinsic factors, intrinsic factors such as Prdm14 specifically control CaP axon outgrowth in zebrafish.

Zebrafish husbandry and embryo manipulations were performed as described (Westerfield, 2000). The T2ASAd gene-trap vector was modified from T2AL200R150 (Urasaki et al., 2006) by replacing the EF1a promoter and second intron of rabbit β-globin with the splicing acceptor of the first intron of zebrafish bcl2 and adding transcription termination signals (see supplementary material Table S1 for sequences). T2ASAd plasmid (25 ng/μl) and Tol2 transposase mRNA (50 ng/μl) were injected (1 nl each) into 1-cell embryos. Over 300 injected embryos were raised and outcrossed with wild-type fish. GFP expression was examined at different developmental stages until 72 hpf. Ten gene-trap fish lines with obvious EGFP expression were identified, four of which showed neuronal EGFP expression patterns, including the slg mutant.

To identify the gene trapped in the slg mutant, we used Tail-PCR (Liu and Chen, 2007). prdm14 sequence (GenBank NM_001163831.1) was used to design primers prdm14-5 (5′-AAAGGATCCATGGCTATGTCGGTTTCTCTCTCCAG-3′) and prdm14-3′ (5′-AAACTCGAGTTAGTTCCAGGGTCTGTACTC-3′). The resulting cDNA was cloned into the pCS2+ plasmid (from Dr David Turner, University of Michigan, USA) to generate CS2-Prdm14.

Digoxigenin-labeled or fluorescein-labeled probes were synthesized using an in vitro transcription system (Roche) and probes longer than 2 kb were hydrolyzed using limited alkaline hydrolysis buffer. Whole-mount in situ hybridization and double-labeling in situ hybridization were performed as described (Jowett, 2001). Whole-mount immunohistochemistry was performed as described (Macdonald, 1999). Embryos were photographed with a Leica MZ16F or Nikon Elips50i microscope using a Nikon DS-Ri1 digital camera or imaged with an Olympus FV1000 confocal microscope. Region of interest (ROI) scores of Islet signals were calculated and analyzed using FV10-ASW 2.1 software (Olympus).

Morpholino (MO) (Gene Tools) sequences (5′-3′) are: prdm14 splicing MO (prdm14 SplMO), TGATTCTTGGACCAACCTGTGCTGG; islet2 ATG MO (islet2 MO), GAATATCCACCATACAGGAGGGTTA (Segawa et al., 2001); prdm1a splicing MO (prdm1a MO), TGGTGTCATA - CCTCTTTGGAGTCTG (Roy and Ng, 2004). MOs were injected into embryos at the 1- to 2-cell stage. Approximately 1 nl of each MO was injected at 1.5 μg/μl for prdm14 SplMO, 6 μg/μl for islet2 MO and 0.5 μg/μl for prdm1a MO. For more information on MOs, see supplementary material Table S2, Fig. S1.

For spontaneous coiling, embryos at 22-24 hpf were kept without stimulus and recorded for 1 minute using a Nikon DS-Qi1MC camera. For touch response, embryos at 28-30 hpf were dechorionated and kept in the resting state and then stimulated with a pin. The stimulus was repeated at ∼1- to 2-second intervals and recorded using a Nikon DS-Qi1MC camera at 50 frames per second for 1 minute. Movies were analyzed with NIS-Element software (Nikon) or QuickTime (Apple). Typical twist movement frames in each group were extracted and assembled with Adobe Photoshop.

A promoter including three copies of the 125 bp mnx1 enhancer (mnx1) was used to drive transgene expression in PMNs (Zelenchuk and Brusés, 2011). mnx1-islet2 or mnx1-prdm14 and mnx1-mRFP were injected together with transposase mRNA into 1-cell slg mutant embryos; 1 nl of each plasmid (25 ng/μl) and transposase mRNA (50 ng/μl) was injected.

ChIP-PCR was performed essentially as described previously (Lindeman et al., 2009) with minor modifications. Briefly, nuclei pellets from embryos were incubated in crosslink buffer NIM (0.25 mM sucrose, 25 mM KCl, 5 mM MgCl₂, 10 mM Tris-HCl, pH 7.4) plus 1% formaldehyde for 8 minutes at room temperature. The reaction was quenched by adding glycine. The nuclei were then washed with NIM buffer and resuspended in lysis buffer. The chromatin lysate was sonicated and precleared with protein A beads (Millipore). The precleared supernatant was incubated with 2 μl rabbit polyclonal Prdm14 antibodies (see supplementary material Fig. S2). Protein A beads were then added and incubated at 4°C. The cross-linking of the bound complexes was reversed by adding proteinase K and NaCl. The recovered DNA was used as template for PCR. For ChIP-PCR in cell culture, human HCT116 cells were cotransfected with equal amounts of CS2-Prdm14 and GL3-Pislet2 [∼3 kb islet2 promoter inserted into GL3-Basic (Promega)] and ChIP-PCR performed similarly to above. Primers are described in supplementary material Table S2.

In a Tol2 transposon-mediated gene-trap screen intended to study neural development, we identified a trapped fish line showing restricted GFP expression patterns in a subset of neurons (Fig. 1A). Because CaP neurons of the homozygous mutant embryos show shortened axons, we named the mutant short lightning (slg). In the trunk region, GFP marks RB, CaP, VaP and their axons (Fig. 1B,C). However, RoP and MiP are notably GFP negative. Additionally, another subpopulation of spinal neurons is GFP positive (Fig. 1D). These cells might be INs, but further characterization is required to define their identity.

To better identify neuronal types among the GFP-positive cells, we stained the neurons with a znp1 antibody (a PMN axon marker). Colocalization of GFP with znp1 staining in ventral projected axons (Fig. 1E-E″) indicates that the GFP-positive cells are CaP/VaP neurons. GFP also labels statoacoustic ganglion (SAG) neurons, trigeminal neurons (TGNs), olfactory sensory neurons (OSNs), midbrain and forebrain neurons (MBNs and FBNs), Mauthner cells (MCs) and reticulospinal neurons (RSNs), but not other types of cells (supplementary material Fig. S3). Together, these results indicate that GFP expression in this fish line is mostly in neurons.

We used Tail-PCR to identify the gene trapped in the slg mutant (Liu and Chen, 2007). Sequencing indicated that all DNA fragments flanking the inserted transposon were derived from the prdm14 locus (Fig. 2A). The conceptually translated zebrafish Prdm14 protein contains a PR/SET (PR) domain and a six zinc-finger (ZF) domain, which are highly conserved from fish to human (Fig. 2B,C; supplementary material Fig. S4). prdm14 regulatory elements are likely to drive the observed GFP expression, as heterozygous embryos consistently showed weaker GFP fluorescence than homozygotes (Fig. 2D).

prdm14 is initially expressed at the dorsal region at 30% epiboly (Fig. 2E) and later in the anterior neural plate region and a part of the posterior neural plate region around the 75% epiboly and bud stages (Fig. 2F,G). Consistent with the GFP expression patterns, prdm14 RNA in the anterior neural plate region is expressed in MBNs and the preplacodal ectoderm region, then in OSNs, SAG neurons and TGNs (Fig. 2H-J). In the posterior neural plate region, prdm14 is restricted to spinal cord neuron precursors, including ventral neuron precursors and dorsal neuron precursors at the 1-somite stage. Around the 22-somite stage, the RNA can be detected in spinal cord neurons (Fig. 2K-M). prdm14 RNA also partially colocalizes with ngn1 (neurog1; a pan-neuronal marker) in the anterior region of the trunk and in brain at the 3-somite stage (Fig. 2N-O″). At the 26-somite stage, prdm14 shares expression patterns with islet2 (an RB and CaP marker) (Fig. 2P-P″), which is consistent with the GFP expression patterns in these neurons. In summary, GFP expression profiles of heterozygotes are highly similar to those of prdm14 RNA, indicating that the trapped gene in the slg mutant is prdm14.

To study functions of prdm14 in neuron development, we investigated whether prdm14 expression is disrupted in slg mutants. Sequencing data from Tail-PCR indicate that the transposon is inserted in the first intron of prdm14 and is likely to disrupt prdm14 expression (Fig. 3A). Indeed, prdm14 expression was greatly reduced in the homozygotes (Fig. 3B,C). Injection of a prdm14 morpholino (SplMO), which spans the boundary of exon 2 and intron 2, disrupted prdm14 RNA splicing (Fig. 3A,D). prdm14 expression in slg mutants was reduced to ∼20% of that in wild type (Fig. 3E). The Prdm14 protein level was also greatly reduced in slg mutants and prdm14 SplMO (splmo) morphants (Fig. 3F).

A previous study showed that the Tol2 transposon could be mobilized efficiently by injection of Tol2 transposase mRNA (Urasaki et al., 2008). To further establish that the transposon insertion in the prdm14 gene is responsible for its reduced expression, we excised the Tol2 transposon by injecting transposase mRNA into 1- to 2-cell homozygous embryos obtained from F0 slg mutants. The injected F1 embryos were bred and outcrossed with wild-type fish to obtain F2 embryos. If the insertion was excised in the germ lineage of F1, some F2 progeny would carry a wild-type allele and a Tol2 excised allele at the prdm14 locus and should be GFP negative. Indeed, up to 22.69% of F2 embryos were GFP negative in several independent experiments, suggesting that Tol2 excision occurred in these F1 germ cells. The F1 fish were designated as mosaic F1 revertant and the prdm14 allele excised in the mosaic F1 revertant germ line was defined as the prdm14 revertant allele (r). We crossed the mosaic F1 revertant fish with prdm14^–/– and obtained prdm14^r/– progeny and their prdm14^–/– siblings. The prdm14^r/– and prdm14^–/– embryos can be distinguished by their different GFP intensity. We found that prdm14 expression is restored in prdm14^r/– embryos (Fig. 4E,F), suggesting that the insertion is likely to be the sole cause of the reduced expression. In summary, prdm14 expression is substantially disrupted in slg mutants and splmo morphants.

Expression of prdm14 in neural tissues in early zebrafish embryos prompted us to examine whether prdm14 is involved in early neural development. We monitored the expression of olig2 (an MN progenitor marker), foxd3 (a neural crest progenitor marker), cxcr4b (an OSN progenitor marker), ngn1 and huC (elavl3 – Zebrafish Information Network) (pan-neural markers). All these markers showed similar expression patterns and expression levels in slg mutant and wild-type embryos (supplementary material Fig. S5A). In addition, cell body positions and cell numbers of examined neurons were unaffected in the mutant embryos at 2.5 days postfertilization (dpf) (supplementary material Fig. S5B,C). These results suggest that reduction of prdm14 expression does not affect early neuronal specification.

In this report, we focus on Prdm14 functions in neuronal development in the trunk, specifically CaP development. One CaP neuron is centrally positioned in each myotome and extends its axon to the distal part of the ventral muscle, which can be easily recognized in 1-day-old embryos (Fig. 4A). CaP cell numbers and cell body positions are normal in slg mutant and splmo morphant embryos (supplementary material Fig. S6A). Expression of GABA, an IN transmitter that is often misexpressed in fate-changed PMNs, is not detected in CaP in slg mutants (supplementary material Fig. S6B), suggesting that the CaP neurons might not switch to an IN-like fate. These results indicate that the early specification of CaP neurons in slg mutants is largely normal.

In wild-type embryos, CaP growth cones leave the spinal cord at ∼17 hpf and extend ventrally until they arrive at the first intermediate target – the muscle pioneers located at the HMS. The growth cones pause at the HMS for 1-2 hours and then continue to extend along the ventral myotome next to the notochord and finally reach the ventral-most region of ventral myotome. Four regions along the axon pathway can be defined: HMS; ventral edge of the notochord (VNC); proximal portion of the ventral muscle (PVM); and distal portion of the ventral muscle (DVM) (Hilario et al., 2009; Rodino-Klapac and Beattie, 2004) (Fig. 4A). One of the most prominent cellular phenotypes in slg mutant embryos is their significantly shorter CaP axons, as compared with the heterozygotes at 26 hpf (Fig. 4B-C′). Similarly, CaP axons in splmo morphants are also shortened (Fig. 4D,D′). Moreover, abnormal axon branching was evident in some CaP neurons in slg mutant and splmo morphant embryos (Fig. 4C′,D′). Staining by znp1 and SV2 antibodies (PMN axon markers) confirmed the axon outgrowth defects in the mutant and morphant embryos (Fig. 4E; supplementary material Fig. S6C). Statistical analysis indicated that the majority of the CaP axons in slg mutant embryos only reached the PVM, with some growth cones even stalled in the HMS and VNC (Fig. 4F). In the prdm14^r/– revertant embryos, CaP axon morphology was restored to that observed in the heterozygous embryos (Fig. 4E,F). Together, these results indicate that Prdm14 is required for CaP axon outgrowth to the ventral-most regions in the trunk at 26 hpf.

The CaP axon defects in slg mutants clearly suggest that their early embryonic movements might be compromised. In zebrafish, there are two major types of embryonic movement at early stages (17-30 hpf): spontaneous tail coiling and the touch response. The embryos begin to exhibit spontaneous tail coiling shortly after PMN axons first extend out of the spinal cord (17 hpf). At later stages (21 hpf), they can respond to touch with an over-the-head fast coiling of the trunk, known as the touch response. Both movements require a functional early spinal circuit, including PMNs (Brustein et al., 2003).

We compared spontaneous tail coiling in wild-type, slg heterozygous, homozygous and splmo morphant embryos at 22-24 hpf. The coiling frequency of homozygous or morphant embryos was reduced to less than once per minute compared with two to three times per minute in wild-type and heterozygous embryos (Fig. 5A). Moreover, coiling amplitude was also greatly decreased in the mutants and morphants (data not shown). We also checked the touch response of embryos at 28-30 hpf, at which time embryos are touch responsive but no longer coil spontaneously (Tallafuss and Eisen, 2008). Wild-type embryos responded to tail touch with a stereotyped bend of the trunk. In slg homozygous and splmo morphant embryos, the trunk bending amplitude was much smaller and the movement was uncoordinated (Fig. 5B,C; supplementary material Movies 1-4). As expected, the prdm14^r/– revertant embryos showed normal spontaneous tail coiling and touch response, as in wild-type or heterozygous embryos (Fig. 5B,C; supplementary material Movie 5). Thus, the disruption of the prdm14 locus is likely to be the sole cause of the movement defects.

In addition to the neuron defects, abnormal muscle development can also cause defective movement (Granato et al., 1996). To exclude this possibility, we examined the development of muscle pioneers, fast muscle and slow muscle. Immunohistochemistry results indicated that muscle development appeared normal in slg mutants (supplementary material Fig. S7A-C). Thus, it is axon shortening caused by prdm14 disruption that contributes to the embryonic movement defects in slg mutant and splmo morphant embryos.

Prdm14 belongs to the PR domain-containing protein family (Fig. 2B), some members of which show protein methyltransferase activity (Hayashi et al., 2005; Kim et al., 2003). We could not detect any methylation activity of zebrafish Prdm14, however, using various substrates (supplementary material Fig. S8; data not shown). Previously, mouse and human PRDM14 were shown to directly bind to regulatory elements of pluripotency-related genes (Chia et al., 2010; Ma et al., 2011), suggesting that they might function as transcription factors. Zebrafish Prdm14 is a nuclear protein and its localization depends on its intact ZF domain (supplementary material Fig. S9A,B); thus, it might also function as a transcription factor in MNs. To identify potential Prdm14 targets, we analyzed a number of genes involved in PMN differentiation/maturation. Among them, islet2 is specifically expressed in zebrafish CaP, but not in RoP or MiP. Consistent with its expression pattern, Islet2 plays an important role in the late-stage differentiation/maturation of CaP (Segawa et al., 2001). Further, islet2 morphants also show shortened CaP axons (Fig. 6A) (Segawa et al., 2001). Importantly, the islet2 RNA level is greatly decreased in CaP in slg mutant and splmo morphant embryos (Fig. 6B). By contrast, islet1 expression in RoP and MiP is unaffected (Fig. 6B). In addition, the expression of mnx1 (a ventral PMN marker), olig2 (a ventral neuron marker), huC (a pan-neuronal marker) and pax2a (an IN and midbrain-hindbrain boundary marker) is not altered in slg mutants (Fig. 6B,D; supplementary material Fig. S10A).

We also examined Islet2 protein expression. Because a zebrafish Islet2-specific antibody is not available, we relied on a general Islet antibody that recognizes both Islet1 and Islet2. Although it was difficult to distinguish the Islet1 signal from that of Islet2, we showed that Islet-positive cells (one or two cells in each somite segment) with GFP expression are CaP or VaP (Fig. 1C; Fig. 6C), suggesting that Islet2 and Prdm14 overlap in expression in Cap/VaP. Indeed, previous studies have shown that Islet2 is expressed in these cells (Segawa et al., 2001; Tokumoto et al., 1995). Thus, we refer to Islet staining in GFP-positive cells as indicating Islet2 and Islet staining in GFP-negative cells in ventral spinal cord as indicating Islet1. Islet2 staining in CaP (GFP positive) greatly decreased in slg mutant and splmo morphant embryos [Fig. 6C; ROI values (mean±s.e.m.): slg heterozygote, 3147.2±162.6; slg homozygote, 1567.8±147.1; splmo, 1829.3±90.8]. By contrast, the Islet1 signal in GFP-negative cells (presumably RoP and MiP) remained unchanged, consistent with the fact that prdm14 is not expressed in RoP and MiP [Fig. 6C; ROI values (mean±s.e.m.): slg heterozygote, 3161.6±79.8; slg homozygote, 3284.1±52.1; splmo, 3436.6±150.2].

Although prdm14 is also expressed in RB (Fig. 1B; Fig. 2P-P″), islet2 expression is not downregulated in RB in slg mutant embryos (Fig. 6B,D). Previous studies suggested that another PR domain protein, Prdm1a, regulates islet2 expression in RB but not in CaP (Olesnicky et al., 2010). Thus, it is likely that Prdm1a and Prdm14 regulate islet2 expression in RB and CaP, respectively. To test this hypothesis, we reduced Prdm14 and Prdm1a expression separately or in combination. Downregulation of prdm14 led to decreased islet2 expression in CaP but not in RB (Fig. 6B,Da′). Conversely, prdm1a MO reduced islet2 expression in RB but not in CaP (Fig. 6Da″). Injection of prdm1a MO into slg mutant embryos resulted in loss of islet2 expression in both RB and CaP (Fig. 6Da″′). As a control, islet2 expression at the distal end of the yolk extension (cloaca) was not affected (Fig. 6Da-a″′). Expression of mnx1 and islet1 was largely intact for all manipulations (Fig. 6Db-c″′). These data indicate that islet2 expression is mainly regulated by Prdm1a in RB and by Prdm14 in CaP.

prdm14 expression precedes that of islet2 (supplementary material Fig. S9C,D). We examined whether zebrafish Prdm14 can directly bind to the islet2 promoter. Using a conserved Prdm14 binding sequence identified in mammals to scan the zebrafish islet2 promoter region, we identified a putative binding site (pBS) for Prdm14 near the transcription start site (Fig. 6E). In ChIP-PCR, Prdm14 antibody specifically pulled down an islet2 PCR fragment containing the pBS (Fig. 6F). However, the antibody failed to precipitate the fragment from slg mutant embryos (Fig. 6G). In addition, Prdm14 antibody can precipitate an islet2 DNA fragment containing the pBS from the cell extract of mammalian cells expressing Prdm14 (Fig. 6H). When Prdm14 was not expressed or the pBS was mutated, the antibody failed to pull down the fragments (Fig. 6H). In addition, the Prdm14 pBS in the islet2 promoter may be required for islet2 expression in CaP (supplementary material Fig. S11). Taken together, Prdm14 can bind to the conserved pBS in the islet2 promoter and may therefore play a role in activating islet2 transcription.

Because Prdm14 activates islet2 expression, it is possible that islet2 might mediate the major output of prdm14 functions in CaP. To test this, we used an islet2 transgene to rescue the CaP axon defects in slg mutant embryos. Previously, mnx1 was shown to be strictly expressed in PMNs, and an mnx1-3×125bp enhancer sequence (mnx1) was used successfully to drive expression of a GFP reporter in PMNs (Zelenchuk and Brusés, 2011). We used the mnx1 sequence to drive islet2 or prdm14 transgene expression in PMNs and used the mnx1-mRFP transgene to label PMNs with islet2 or prdm14 transgene expression (Fig. 7A). Expression of prdm14 or islet2 transgenes in CaP rescued the axon outgrowth defects of slg mutants (Fig. 7B,C). These results suggest that islet2 is a major downstream target of prdm14 and is likely to mediate major functions of Prdm14 in CaP (Fig. 7D).

Here, we show that prdm14 is expressed in selected neurons, including CaP, in early zebrafish embryos. Prdm14 is required for CaP axon outgrowth and early embryonic movement. We further identify islet2 as a major downstream target of Prdm14 in CaP (Fig. 7D). Our results provide an example of an intrinsic factor that is crucial for the development of a specific MN axon.

We showed that prdm14 expression is mainly restricted to neural lineages in early zebrafish embryos (Fig. 1; supplementary material Fig. S3). Despite the early expression of prdm14 in the neural system, we did not observe neural specification defects in slg mutant embryos, suggesting that Prdm14 might be dispensable for this early process (supplementary material Fig. S5A). It remains possible that redundant factors mask the impact of Prdm14 loss in early neural specification. Thus, we cannot rule out the possibility that prdm14 might play a role in early neural development.

At 26 hpf, prdm14 is prominently expressed in RB, INs and PMNs. In PMNs, prdm14 is only expressed in CaP/VaP, but not RoP and MiP. Similarly, only a proportion of INs are marked by prdm14. These observations suggest that expression of prdm14 itself is highly regulated. prdm14 is also expressed in the nervous system of mouse and human (Kinameri et al., 2008) (Allen Brain Atlas, http://www.brain-map.org/). In human embryonic stem cells, PRDM14 may regulate genes involved in neurogenesis (Chia et al., 2010). Thus, it would be intriguing to investigate whether PRDM14 also plays roles in mammalian neuron development, especially PMN axon outgrowth.

CaP development is a well-characterized system used to study mechanisms that regulate branching and projection of MN axons (Lewis and Eisen, 2003). Using slg mutants and prdm14 morphants, we found that disruption of Prdm14 expression does not appear to affect CaP specification and initial axon extension, but leads to premature branching and shortened axon outgrowth (Fig. 4). Muscle development was largely normal in the mutant (supplementary material Fig. S7). Thus, the defects are likely to be due to a cell-autonomous function of Prdm14 in the neurons. Previous studies showed that Plexin:Nrp1a/Semaphorin, the extracellular matrix (ECM) component Tenascin-C, the glycosyltransferase Lh3 and secreted glycoprotein PGRN are all essential for CaP axon development (Chitramuthu et al., 2010; Feldner et al., 2007; Sato-Maeda et al., 2006; Schneider and Granato, 2006; Schweitzer et al., 2005; Tanaka et al., 2007). It remains to be investigated whether Prdm14 directly or indirectly regulates these factors. Previously, the Prdm family protein Hamlet was implicated in the control of axon targeting in Drosophila (Endo et al., 2011). Thus, some members of the Prdm family might be widely used to regulate axon morphogenesis.

It is not surprising that slg mutants display defective embryonic movements in spontaneous tail coiling and touch response (Fig. 5) given that CaP axons in the mutants failed to reach the DVM region (Fig. 4). However, swimming behaviors are generally unaffected in slg mutant embryos at 2.5 dpf (data not shown), suggesting that the underlying neuron circuits function normally at later stages. In agreement with this observation, the majority of CaP axons reach the ventral-most myotome and form connections with muscles at this stage (supplementary material Fig. S5C). These data suggest that Prdm14 could be required for the timely connection of the axons to their targets. Alternatively, the residual Prdm14 expression in slg mutant embryos might be enough for full axon extension, albeit at a slow rate. We favor the second possibility because prdm14 SplMO injection, which is more efficient in blocking prdm14 expression (supplementary material Fig. S6D), causes more severe axon outgrowth defects and slows the full growth of the axons to their final targets (data not shown). Generation and analysis of a prdm14 null mutant might provide further insight into the requirements of Prdm14 activity in CaP axon development.

Although islet1 and islet2 show specific expression patterns in later PMNs, either is sufficient to specify MiP and CaP subtypes (Hutchinson and Eisen, 2006). It is hypothesized that factors upstream of the Islet genes regulate the specification of different PMNs. Indeed, Nkx6 proteins expressed in MiP act upstream of islet1 and are required to maintain MiP identity (Cheesman et al., 2004; Hutchinson et al., 2007). An upstream regulator of islet2 in CaP had not been identified prior to this study. We show here that Prdm14 directly binds to the promoter region of the islet2 gene and is required for islet2 expression in CaP (Fig. 6). Islet2 plays a major role in neuronal subtype specification and MN subclass differentiation (Lee and Pfaff, 2001; Shirasaki and Pfaff, 2002). In addition, islet2 is an essential factor for Slit2 to induce axonal branching and elongation of TGNs, suggesting that downstream factors of Islet2 are crucial for mediating axonal development (Yeo et al., 2004). We show here that Islet2 is a major mediator of Prdm14 function in CaP axon outgrowth.

To further investigate axon development in CaP, it would be helpful to identify downstream targets of Islet2. Several proteins have been implicated in CaP axon morphogenesis. For example, Plexin A3 and Nrp1a expressed in CaP are required for axon outgrowth (Feldner et al., 2005; Feldner et al., 2007; Sato-Maeda et al., 2006). However, the expression patterns of plexin A3 and nrp1a are not altered in slg mutants (supplementary material Fig. S10B). Similarly, expression of agrin and c-met, which encode two further factors involved in CaP axon growth (Kim et al., 2007; Tallafuss and Eisen, 2008), is largely unaffected in slg mutants (supplementary material Fig. S10B). However, other changes, such as post-translational modifications to these molecules in CaP of slg mutants, cannot be ruled out. In addition, components of the ECM, such as Tenascin-C and Collagen XVIII, are also important in CaP axon outgrowth (Schneider and Granato, 2006; Schweitzer et al., 2005). Possible changes to signal transducers of the ECM in CaP could result in slg mutant phenotypes. It would be interesting to investigate connections between the ECM and Prdm14 or Islet2 activities. prdm14 is also expressed in RB (Fig. 1). However, islet2 expression is only downregulated in CaP but not in RB in slg mutant embryos. Our data suggest that Prdm14 and Prdm1a regulate islet2 expression in CaP and RB, respectively (Fig. 6D) (Olesnicky et al., 2010). The functions of Prdm14 in RB remain to be investigated.

We thank members of the J.Z. laboratory for discussions; Dr Shanshan Zhu for bioinformatics assistance; Drs Catherina G. Becker, Kristin Bruk Artinger, Yoshihiro Yoshihara and Juan L. Bruses for plasmids; and Dr Koichi Kawakami for the original Tol2 system.