Prdm8 regulates pMN progenitor specification for motor neuron and oligodendrocyte fates by modulating the Shh signaling response

Oligodendrocytes, one of the major glial cell types of the central nervous system (CNS) of vertebrate animals, increase the speed of axon electrical impulses and support neuron health by wrapping axons with a myelin membrane (Simons and Nave, 2016). In the spinal cord, most oligodendrocytes originate from ventral pMN progenitors (Noll and Miller, 1993; Warf et al., 1991), which express the basic helix loop helix (bHLH) transcription factor Olig2 (Lu et al., 2000; Novitch et al., 2001; Zhou and Anderson, 2002; Zhou et al., 2000). pMN progenitors first produce motor neurons followed by oligodendrocyte precursor cells (OPCs) (Richardson et al., 2000; Rowitch, 2004). After specification, some OPCs rapidly differentiate as myelinating oligodendrocytes, whereas other OPCs persist into adulthood (Dawson et al., 2003; Pringle et al., 1992; Wolswijk and Noble, 1989). The switch from motor neuron to OPC production and the subsequent regulation of oligodendrocyte differentiation require tight control of gene expression through a complex network of interacting transcription factors and extracellular cues. Although many factors that promote oligodendrocyte differentiation and myelination have been identified (Elbaz and Popko, 2019; Emery, 2010; He and Lu, 2013; Sock and Wegner, 2019; Zuchero and Barres, 2013), the mechanisms that regulate the onset of OPC specification and maintain some of them in a non-myelinating state are not well understood.

During early neural tube patterning, pMN progenitors are specified by the morphogen Sonic hedgehog (Shh). The Shh ligand, secreted by the notochord, a mesodermal structure below the ventral spinal cord, and by the floor plate, the ventralmost cells of the neural tube, signals neural progenitors to acquire ventral identities (Echelard et al., 1993; Martí et al., 1995; Roelink et al., 1994). The Shh ligand binds to its transmembrane receptor Patched (Ptch), which allows intercellular Shh signaling to be transduced by Smoothened (Smo). Upon Shh binding, Smo is internalized to promote GliA transcriptional activity by inhibiting its cleavage to GliR (Briscoe and Thérond, 2013; Danesin and Soula, 2017; Ribes and Briscoe, 2009). Graded Shh activity induces expression of genes that encode bHLH and homeodomain proteins at distinct positions on the dorsoventral axis (Briscoe and Thérond, 2013; Briscoe et al., 2000; Poh et al., 2002). The duration of Shh signaling also influences cell fate and gene expression. Initially, high ventral Shh signaling activates expression of Olig2, then sustained Shh activity promotes expression of Nkx2.2 adjacent to the floor plate, ventral to Olig2, thus forming two distinct ventral progenitor domains (Dessaud et al., 2007, 2010). The sequential induction of cross-repressive transcription factors by graded morphogen activity along the dorsoventral axis creates spatially restricted progenitor domains that sequentially give rise to specific neurons and glia (Briscoe et al., 2000; Lek et al., 2010; Nishi et al., 2015).

Shh activity is necessary for the establishment of the pMN domain and motor neuron formation, and subsequently, a transient increase in Shh activity coincides with, and is required for, timely OPC specification (Danesin and Soula, 2017). Pharmacological and genetic reduction of Shh signaling in chick, mouse and zebrafish spinal cords results in prolonged motor neuron formation and impaired OPC formation (Al Oustah et al., 2014; Danesin et al., 2006; Hashimoto et al., 2017; Jiang et al., 2017; Ravanelli and Appel, 2015; Touahri et al., 2012). Furthermore, in chick neural tube explants treated with exogenous Shh, premature formation of OPCs occurs at the expense of motor neurons (Danesin et al., 2006; Orentas et al., 1999). Thus, transient elevation of Shh activity is required to induce the transition from motor neuron to OPC production. The temporal change in Shh activity is in part due to the upregulation of Sulfatase 1/2 by p3 cells, which are ventral to pMN progenitors, before OPC specification (Al Oustah et al., 2014; Danesin et al., 2006; Jiang et al., 2017). Sulfatase expression increases the local Shh ligand concentration available to pMN progenitors (Al Oustah et al., 2014; Danesin et al., 2006), and loss of Sulfatase 1/2 functions delay the motor neuron-OPC switch (Jiang et al., 2017). Whether additional mechanisms contribute to the modulation of Shh signaling strength to regulate fate specification over time, is not well understood.

In addition to expressing distinct combinations of bHLH and homeodomain transcription factors, subsets of spinal cord progenitors express specific PRDI-BF1 and RIZ homology domain containing (Prdm) proteins (Zannino and Sagerström, 2015). This family of proteins contains an N-terminal SET domain followed by a varied number of C-terminal zinc-finger repeats and they act as transcriptional regulators or co-factors implicated in nervous system patterning, neural stem cell maintenance and differentiation (Baizabal et al., 2018; Chittka et al., 2012; Hanotel et al., 2014; Hernandez-Lagunas et al., 2011; Kinameri et al., 2008; Ross et al., 2012; Thélie et al., 2015; Yildiz et al., 2019). In the ventral mouse spinal cord, neural progenitors express Prdm8 from embryonic (E) day 9.5 to E13.5 (Kinameri et al., 2008; Komai et al., 2009), corresponding to the period of motor neuron and OPC formation. The function of Prdm8 in spinal cord development is not yet known, but in the mouse telencephalon Prdm8 forms a repressive complex with Bhlhb5, a bHLH transcription factor closely related to Olig2, to regulate axonal migration (Ross et al., 2012). Moreover, in the retina, Prdm8 promotes the formation of a subset of rod bipolar cells and regulates amacrine cell type identity (Jung et al., 2015). These data raise the possiblility that Prdm8 regulates pMN cell development.

To investigate prdm8 expression and function in pMN progenitors, we used the developing zebrafish spinal cord as a model. Our expression analysis showed that pMN progenitors express prdm8 before and during the switch from motor neuron to OPC production, and that, subsequently, differentiating oligodendrocytes downregulate prdm8 expression. Because Prdm8 can control cell fate, we therefore hypothesized that Prdm8 regulates motor neuron and OPC specification. To test this hypothesis, we performed a series of experiments to assess changes in pMN cell fates resulting from loss of prdm8 function. Our data reveal that prdm8 mutant embryos have a deficit of late-born motor neurons, excess differentiating oligodendrocytes and a deficit of OPCs. Birth-dating studies showed that the motor neuron deficit results from a premature switch from motor neuron to OPC production. prdm8 mutant embryos have abnormally high levels of Shh signaling and pharmacological suppression of Shh signaling rescued the motor neuron deficit but not the formation of excess oligodendrocytes. Taken together, our data suggest that Prdm8 dampens Shh signaling activity to modulate the timing of the motor neuron-OPC switch, and secondarily regulates the myelinating fate of oligodendrocyte lineage cells.

To begin our investigation of prdm8 function, we first assessed the temporal and spatial features of prdm8 expression in the zebrafish spinal cord during development. To do so, we performed in situ RNA hybridization (RNA ISH) using transverse sections obtained from Tg(olig2:EGFP) embryos and larvae. pMN cells in these fish express EGFP driven by olig2 regulatory DNA (Shin et al., 2003). pMN cells and cells dorsal to the pMN domain expressed prdm8 at 24, 36 and 48 h postfertilization (hpf) (Fig. 1A). This is consistent with previous data revealing that cells of the pMN, p2 and p1 domains in the developing mouse spinal cord express Prdm8 (Kinameri et al., 2008; Komai et al., 2009). Next we evaluated pMN cell expression of prdm8 through development using a single cell RNA-seq (scRNA-seq) dataset obtained from olig2:EGFP⁺ cells isolated from 24, 36 and 48 hpf embryos, a period spanning the formation of most motor neurons and OPCs. An aligned Harmony clustering analysis and uniform manifold approximation and projection (UMAP) of the data, revealed that gene expression profiles formed several distinct clusters (Fig. 1B). Plotting individual gene expression profiles revealed that many olig2⁺ sox19a⁺ cells, which likely represent pMN progenitors, also expressed prdm8 (Fig. 1C-E).

Following motor neuron formation, pMN progenitors begin to form OPCs. This is initiated by the reorganization of ventral progenitor domains, such that pMN cells that enter the oligodendrocyte lineage begin to co-express Nkx2.2 and Olig2 (Agius et al., 2004; Fu et al., 2002; Kessaris et al., 2001; Soula et al., 2001; Zhou et al., 2001). Our scRNA-seq data show that at 48 hpf, cells that expressed nkx2.2a and olig2 also expressed prdm8, signifying that nascent OPCs express prdm8 (Fig. 1F). We validated these observations using fluorescent RNA ISH. Consistent with our scRNA-seq data, prdm8 mRNA puncta were present in the pMN domain marked by olig2 expression at 24, 36 and 48 hpf (Fig. 1G-I). At 24 and 36 hpf, cells that expressed prdm8 and olig2 were adjacent to more ventral nkx2.2a⁺ p3 domain cells (Fig. 1G,H), but at 48 hpf some cells at the p3/pMN border expressed all three transcripts (Fig. 1I). By 72 hpf, the expression of olig2 mRNA was mostly depleted from the pMN domain but evident at a high level in oligodendrocyte lineage cells. At this stage, some olig2⁺ cells expressed both prdm8 and nkx2.2a, some expressed only nkx2.2a and others expressed only prdm8 (Fig. 1J). These data indicate that following pMN progenitor cell expression, prdm8 expression is differentially maintained by oligodendrocyte lineage cells.

To determine the identity of prdm8⁺ cells, we compared prdm8 expression with the expression of genes characteristic of the oligodendrocyte lineage in the scRNA-seq dataset. At 48 hpf, a subset of cells that expressed prdm8 also expressed sox10, which encodes a transcription factor expressed by all oligodendrocyte lineage cells (Britsch et al., 2001; Kuhlbrodt et al., 1998; Park et al., 2002) (Fig. 2A,B). Some prdm8⁺ sox10⁺ cells also expressed myrf, which encodes Myelin regulatory factor, a transcription factor required for oligodendrocyte differentiation (Emery et al., 2009) (Fig. 2C). This dataset included only a few mbpa⁺ cells, and these appeared as a small subset of sox10⁺ myrf⁺ cells (Fig. 2D). Therefore, these cells represent pre-myelinating oligodendrocytes. A heatmap representation of these cells (Fig. 2E,F) showed that most sox10⁺ nkx2.2a⁺ cells expressed prdm8 at high levels. However, cells that also expressed myrf and mbpa at higher levels had little prdm8 expression. These data suggest that pre-myelinating oligodendrocytes downregulate prdm8 expression as they differentiate.

To corroborate these observations, we examined prdm8, myrf and olig2 mRNA expression using fluorescent RNA ISH in the trunk spinal cord of 72 hpf larvae. Consistent with our scRNA-seq findings, this revealed that a majority of olig2⁺ cells expressed either prdm8 or myrf, and that few olig2⁺ cells expressed both genes (Fig. 2G). To determine whether some oligodendrocyte lineage cells maintain prdm8 expression, we assessed RNA-seq data collected from cspg4⁺ OPCs and mbpa⁺ oligodendrocytes isolated from 7 days postfertilization (dpf) larvae (Ravanelli et al., 2018). We found that OPCs expressed prdm8 at 75-fold higher levels than oligodendrocytes (Fig. 2H). We validated these data using fluorescent RNA ISH to label prdm8 mRNA at 7 dpf in combination with myrf to mark oligodendrocytes (Fig. 2I), or cspg4 to mark OPCs (Fig. 2J). This revealed that OPCs, but not oligodendrocytes, expressed prdm8, thus confirming our RNA-seq data. We therefore conclude that pMN progenitors and OPCs express prdm8, and that prdm8 expression declines in oligodendrocytes undergoing differentiation.

Zebrafish prdm8 encodes a 502 amino acid protein containing an N-terminal PR/SET domain and three zinc-finger domains, similar to its human and mouse orthologs (Fig. 3A). To investigate Prdm8 function, we used CRISPR/Cas9 to create gene-disrupting mutations within the first exon (Fig. 3B). We verified the efficiency of single guide RNA (sgRNA) targeting using diagnostic fluorescent PCR and subsequently raised injected embryos to adulthood. We identified F0 adults that transmitted mutations through the germ line and selected two, prdm8^co49 and prdm8^co51, for further analysis. DNA sequencing revealed that the co49 allele contained a 5 bp insertion, whereas the co51 allele had a 4 bp deletion. Both alleles were predicted to result in translation frameshifting leading to premature translation termination and proteins truncated within the PR/SET domain (Fig. 3B). Because the C-terminal zinc finger domains of mouse Prdm8 are necessary for nuclear localization (Eom et al., 2009), we predicted that the truncated proteins produced by the co49 and co51 alleles would be non-functional. Genotyping assays revealed that F1 heterozygous adults transmitted mutant alleles to progeny with Mendelian frequencies (Fig. 3C). Homozygous mutant embryos had no discernible morphological phenotype at 24 hpf or at early larval stages (Fig. 3D). To assess the effect of the co49 allele on prdm8 transcript levels, we performed RT-PCR. This revealed a partial reduction in the levels of mutant prdm8 mRNA relative to wild type (Fig. 3E).

To determine whether prdm8 regulates the formation of oligodendrocytes, we performed RNA ISH to detect myrf expressed by wild-type and prdm8 mutant larvae. At 72 hpf, larvae homozygous for the co49 allele had almost twice as many spinal cord myrf⁺ oligodendrocytes compared with wild-type siblings (Fig. 4A,B). Heterozygous siblings were not different from wild type (Fig. 4A,B). Larvae homozygous for the co51 allele similarly had excess myrf⁺ oligodendrocytes relative to wild-type siblings (Fig. 4C,D). Larvae trans-heterozygous for the co49 and co51 alleles also had a greater number of myrf⁺ oligodendrocytes (Fig. 4E,F), indicating that this phenotype results specifically from loss of prdm8 function and not as a consequence of an off-target mutation produced by CRISPR/Cas9. We additionally examined the expression of mbpa. Consistent with our myrf data, co49 homozygous mutant larvae had ∼twofold more dorsal mbpa⁺ oligodendrocytes than wild-type siblings (Fig. 4G,H). These data indicate that Prdm8 limits oligodendrocyte formation.

To identify the source of excess oligodendrocytes, we first counted the number of oligodendrocyte lineage cells using an antibody to detect expression of Sox10 in spinal cord sections of wild-type and mutant larvae carrying the Tg(olig2:EGFP) reporter. At 72 hpf, homozygous prdm8 mutant larvae had the same number of Sox10⁺ olig2:EGFP⁺ cells as control larvae (Fig. 5A,B). To determine the proportion of Sox10⁺ cells that differentiated as oligodendrocytes, we then performed immunohistochemistry to detect Sox10 on sections obtained from 5 dpf larvae carrying a Tg(mbpa:tagRFPt) transgenic reporter. This experiment showed that homozygous prdm8 mutant larvae had a significant increase in the number of Sox10⁺ mbpa:tagRFPt⁺ oligodendrocytes without a change in the total number of Sox10⁺ cells relative to sibling controls (Fig. 5C-E). To assess the OPC population, we labeled sections from 5 dpf larvae carrying a Tg(cspg4:mCherry) transgenic reporter (Ravanelli et al., 2018) with Sox10 antibody. Homozygous prdm8 mutant larvae had fewer OPCs than wild-type siblings but total oligodendrocyte lineage cells were unchanged (Fig. 5F-H). These data indicate that Prdm8 regulates the proportion of OPCs that differentiate as oligodendrocytes.

We next investigated whether prdm8 regulates motor neuron formation from pMN progenitors, which precedes OPC specification. To do so, we used an antibody to detect Isl1/2 (Isl), which marks post-mitotic motor neurons (Ericson et al., 1992). At 24 hpf, homozygous prdm8 mutant embryos had the same number of Isl⁺ olig2:EGFP⁺ motor neurons as controls (Fig. 6A,B), signifying that prdm8 mutant embryos initiate motor neuron formation normally. By contrast, at 36 and 48 hpf, prdm8 mutant embryos had fewer motor neurons than control embryos (Fig. 6C-F), suggesting that Prdm8 is required to maintain motor neuron production from pMN progenitors.

One possible explanation for these data is that in the absence of prdm8 function, pMN progenitors prematurely switch from motor neuron to OPC production, resulting in a deficit of motor neurons. To test this prediction, we exposed 24, 30 and 36 hpf embryos to a pulse of bromodeoxyuridine (BrdU) to label cells in S-phase and used immunohistochemistry at 48 hpf to identify the progeny of the labeled progenitors (Fig. 7A). Compared with stage-matched wild-type control embryos, homozygous prdm8 mutant embryos exposed to BrdU at each time point had a deficit of BrdU⁺ Isl⁺ motor neurons (Fig. 7B-H). By contrast, mutant embryos pulsed with BrdU at 24 and 30 hpf had more BrdU⁺ Sox10⁺ cells than controls (Fig. 7B,C,E,F,I), whereas those pulsed at 36 hpf had fewer (Fig. 7D,G,I). These data indicate that prdm8 mutant embryos prematurely terminate motor neuron formation and concomitantly produce OPCs earlier than normal. However, mutant embryos also prematurely terminate OPC production. Altogether, these data raise the possibility that Prdm8 prevents premature OPC specification, thus preserving pMN progenitors for motor neuron fate.

OPC specification coincides with and requires a dorsal expansion of Nkx2.2 expression from the p3 domain, resulting in co-expression of Nkx2.2 and Olig2 (Kessaris et al., 2001; Kucenas et al., 2008; Soula et al., 2001; Xu et al., 2000). Thus, one possible mechanism by which prdm8 prevents premature OPC specification is regulation of the time at which pMN progenitors express nkx2.2a. To examine this possibility, we first used RNA ISH to evaluate the area of nkx2.2a expression within the olig2:EGFP⁺ pMN domain. At 24 and 36 hpf, nkx2.2a expression appeared to be expanded dorsolaterally into the pMN domain of mutant embryos (Fig. 8A,B). Measurement of the area of overlap of nkx2.2a and olig2:EGFP expression confirmed this observation, revealing a significant increase in the area of coincident olig2:EGFP and nkx2.2a expression in mutant embryos (Fig. 8D). By 48 hpf, there was no difference in nkx2.2a expression in the pMN domain between wild-type and mutant embryos (Fig. 8C,F). However, the number of dorsal olig2:EGFP⁺ OPCs that expressed nkx2.2a at 48 hpf was increased almost threefold in prdm8 mutant embryos compared with controls (Fig. 8C,E). We also used fluorescent RNA ISH to label nkx2.2a and olig2 mRNA, and quantified nkx2.2a puncta in the pMN domain at 28 hpf, before OPC specification. prdm8 mutant embryos had more nkx2.2a mRNA puncta localized to pMN cells than controls (Fig. 8F,G). Altogether, these data suggest that prdm8 controls the timing of OPC specification by controlling the time at which pMN cells initiate nkx2.2a expression.

At the end of neurogenesis, ventral spinal cord cells transiently elevate Shh signaling activity, which is necessary for OPC specification (Orentas et al., 1999; Soula et al., 2001; Touahri et al., 2012). Experimentally increasing Shh levels caused premature termination of motor neuron formation and precocious OPC formation (Danesin et al., 2006), similar to the loss of prdm8 function and thereby raising the possibility that Prdm8 suppresses Shh activity in the ventral spinal cord. To test this possibility, we probed for expression of ptch2, a transcriptional target of the Shh signaling pathway. At 24 hpf, prdm8 mutant embryos appeared to express more ptch2 than wild-type embryos (Fig. 9A). By 48 hpf, there was no visible difference in ptch2 expression between genotypes (Fig. 9A). Next, we used fluorescent RNA ISH to quantify ptch2 expression. At 24 and 36 hpf, prdm8 mutant embryos expressed more ptch2 mRNA relative to total spinal cord area compared with controls (Fig. 9B-F). Wild-type and prdm8 mutant embryos similarly expressed shha, which encodes a Shh ligand, suggesting that the elevated level of ptch2 expression results from increased Shh signaling activity and is independent of ligand expression (Fig. 9G). To further validate that loss of prdm8 leads to elevated Shh activity, we probed for expression of boc1, which encodes a Shh co-receptor that is negatively regulated by Shh signaling (Tenzen et al., 2006). At 24 hpf, prdm8 mutant embryos expressed less boc1 than controls, which is in line with the idea that loss of prdm8 function leads to increased Shh activity (Fig. 9H,I). These results are consistent with the possibility that Prdm8 suppresses Shh response in pMN cells to regulate the transition between motor neuron and OPC formation.

Because the premature transition between motor neuron and OPC production resulting from lack of prdm8 function resembled the effect of abnormally elevated Shh signaling, we predicted that the number of motor neurons in prdm8 mutant embryos could be rescued by inhibiting Shh activity. To test this prediction, we treated prdm8 mutant and wild-type embryos with a low concentration of cyclopamine to partially block Shh signal transduction from 18 to 30 hpf and assessed motor neuron number at 48 hpf. The number of motor neurons was similar in prdm8 mutant embryos treated with cyclopamine and wild-type embryos treated with vehicle control (Fig. 10A,B), consistent with our prediction. Furthermore, both wild-type and prdm8 mutant embryos treated with cyclopamine had more motor neurons than their genotype-matched controls (Fig. 10B), raising the possibility that suppression of Shh signaling delays the motor neuron to OPC switch, resulting in the formation of excess motor neurons. By contrast, treatment with cyclopamine from 30 to 42 hpf, after most spinal cord neurogenesis is normally completed, had no effect on motor neuron number in either wild-type or prdm8 mutant embryos (Fig. 10C,D). Consistent with our previous assessments, vehicle control-treated mutant embryos had fewer motor neurons than wild-type siblings (Fig. 10A-D). These data therefore support the possibility that Prdm8 suppresses Shh signaling within pMN cells to regulate the termination of motor neuron production and the timing of the neuron-glia switch.

We next tested whether suppressing Shh signaling in prdm8 mutant embryos affected oligodendrocyte development. As above, there was no difference in the number of Sox10⁺ cells between wild-type and mutant larvae treated with a control solution between 18 and 30 hpf (Fig. 10E,F). Whereas wild-type embryos treated with control solution or 0.5 μM cyclopamine from 18 to 30 hpf had similar numbers of Sox10⁺ cells, prdm8 mutant embryos treated with cyclopamine had a significant deficit of Sox10⁺ cells compared with mutant embryos treated with control solution and wild-type embryos treated with cyclopamine (Fig. 10E,F). Treating embryos with cyclopamine from 30 to 42 hpf similarly caused prdm8 mutant embryos to have fewer Sox10⁺ cells than mutant embryos treated with control solution and wild-type embryos treated with cyclopamine (Fig. 10G,H). One possible explanation for these observations is that, although cyclopamine treatment delays the premature neuron-glia transition in prdm8 mutant embryos, it does not rescue the premature termination of OPC production, thereby resulting in a deficit of oligodendrocyte lineage cells.

We have shown above that prdm8 mutant larvae have excess oligodendrocytes and a deficit of cspg4⁺ OPCs. To determine whether this phenotype results from misregulated Shh signaling, we treated embryos with cyclopamine from 30 to 42 hpf and examined expression of myrf as a marker for myelinating oligodendrocytes. prdm8 mutant larvae treated with control solution and cyclopamine had similar numbers of myrf⁺ oligodendrocytes (Fig. 10I,J). Thus, suppressing Shh signaling did not rescue the excess oligodendrocyte phenotype of prdm8 mutant larvae, raising the possibility that Prdm8 regulates oligodendrocyte formation independently of its role in controlling the timing of a neuron-glia switch.

The neuron-glia switch, whereby neural progenitors produce neurons followed by glia, is a general feature of developing nervous systems (Rowitch and Kriegstein, 2010). Despite its important role in diversifying neural cell fate, the mechanisms that cause the switch and determine its timing remain poorly understood. In the ventral spinal cord, a temporally regulated rise in Shh signaling activity appears to trigger pMN progenitors to switch from motor neuron to OPC production (Danesin and Soula, 2017). Our results now indicate that Prdm8 suppresses Shh signaling activity within pMN progenitors to control the timing of the motor neuron-OPC switch.

Distinct types of neurons and glia arise from distinct subpopulations of progenitor cells aligned along the dorsoventral axis of the spinal cord. A large body of work conducted over the past 30 years has shown that the identities of these progenitor populations are determined by combinatorial expression of an extensive array of bHLHs and homeodomain transcription factors (Sagner and Briscoe, 2019). Additionally, specific subpopulations of spinal cord progenitors also express members of the Prdm protein family, although how these factors contribute to spinal cord development has received considerably less attention (Zannino and Sagerström, 2015). For example, dorsal progenitors express Prdm13, which regulates the balance between excitatory and inhibitory interneuron production by blocking the activity of bHLH transcription factors that drive expression of genes required for excitatory neuron differentiation (Chang et al., 2013; Mona et al., 2017). p1 progenitors express Prdm12, which is required for V1 interneuron formation (Thélie et al., 2015), and Prdm14 promotes Islet2 expression and axon outgrowth in motor neurons (Liu et al., 2012). Finally, pMN, p1 and p2 progenitors express Prdm8 (Kinameri et al., 2008; Komai et al., 2009). Although the spinal cord function of Prdm8 had not been previously investigated, evidence indicating that Prdm8 regulates the specification of retinal cells (Jung et al., 2015), and that Prdm8 can interact with bHLH transcription factors (Ross et al., 2012; Yildiz et al., 2019), raises the possibility that Prdm8 contributes to mechanisms that determine spinal cord progenitor fate.

The first main finding of our work is that zebrafish spinal cord cells express prdm8 similarly to mouse (Kinameri et al., 2008; Komai et al., 2009). Our data show that pMN progenitors express prdm8 throughout developmental neurogenesis and gliogenesis, but prdm8 expression subsides from the ventral spinal cord as pMN progenitors are depleted at the onset of the larval period. Furthermore, our analysis extends the mouse expression data by showing that oligodendrocyte lineage cells also express prdm8. Specifically, our data indicate that newly specified OPCs express prdm8 but then downregulate it as they initiate oligodendrocyte differentiation. By contrast, larval OPCs, marked by cspg4 expression, maintain prdm8 expression. Altogether, these observations raise the possibility that Prdm8 regulates pMN progenitor fate specification and, subsequently, OPC differentiation.

Our second main conclusion is that Prdm8 regulates the timing of the motor neuron to OPC switch by determining how strongly pMN progenitors respond to Shh signaling. Specifically, we found that prdm8 mutant embryos have a deficit of late-born motor neurons because of a premature neuron-glia switch, that mutant spinal cord cells express abnormally high levels of ptch2 and that the motor neuron deficit was rescued by treating mutant embryos with a low concentration of a Shh inhibitor. This finding supports previous evidence that a transient burst of Shh signaling activity initiates the switch from motor neuron to OPC production (Danesin and Soula, 2017). This burst is mediated, at least in part, by sulfatase function (Al Oustah et al., 2014; Danesin et al., 2006; Jiang et al., 2017). Sulfatases are secreted by ventral spinal cord cells and increase the range of Shh ligand in the extracellular matrix by regulating the sulfation state of heparan sulfate proteoglycans (Farzan et al., 2008; Yan and Lin, 2009). How cells receive and process extracellular signals also can influence signaling strength. In particular, Notch signaling increases the sensitivity of neural cells to Shh signaling (Kong et al., 2015; Ravanelli et al., 2018; Stasiulewicz et al., 2015). Currently, we do not know how Prdm8 suppresses Shh signaling activity within pMN progenitors. Because Prdm8 functions as a transcriptional inhibitor (Chen et al., 2018; Eom et al., 2009; Iwai et al., 2018; Ross et al., 2012), it might suppress expression of factors that transduce Shh signaling. For example, Prdm8 could suppress expression of the Shh co-receptors Cdon and Gas1, which enhance cell response to Shh (Allen et al., 2007, 2011). Alternatively, Prdm8 could limit expression of Notch signaling effectors that enhance Shh signaling. The identification of genes misregulated in prdm8 mutant embryos, combined with the determination of genomic loci targeted by Prdm8, should help uncover the regulatory function of Prdm8 in pMN progenitor specification.

Finally, we found that prdm8 mutant larvae have excess oligodendrocytes at the apparent expense of OPCs. There are at least two possible explanations for this phenotype. Because our expression data show that cells undergoing oligodendrocyte differentiation downregulate prdm8 expression, the first possibility is that Prdm8 inhibits OPC differentiation and, therefore, in its absence, OPCs that normally persist into the larval stage instead develop as myelinating oligodendrocytes. A second possibility is that Prdm8 regulates the allocation of pMN progenitors for distinct oligodendrocyte lineage cell fates. Previously, in a process we called progenitor recruitment, we showed that motor neurons, OPCs that rapidly differentiate and OPCs that persist into the larval stage arise from distinct pMN progenitors that sequentially initiate olig2 expression (Ravanelli and Appel, 2015; Ravanelli et al., 2018). We also found that slightly higher levels of Shh signaling favor the formation of oligodendrocytes over larval OPCs, which is similar to the oligodendrocyte phenotype of prdm8 mutant animals. However, inhibiting Shh with cyclopamine did not restore oligodendrocytes and OPCs to their normal numbers, raising the possibility that Prdm8 regulates oligodendrocyte lineage cell fate independently of Shh signaling. Identifying Prdm8 regulatory targets combined with detailed cell lineage analysis will help us to discriminate between these possibilities.

All animal work was approved by the Institutional Animal Care and Use Committee at the University of Colorado School of Medicine. All non-transgenic embryos were obtained from pairwise crosses of males and females from the AB strain. Embryos were raised at 28.5°C in E3 media [5 mM NaCl, 0.17 mM KCl, 0.33 mM CaCl₂ and 0.33 mM MgSO₄ (pH 7.4), with sodium bicarbonate], sorted for good health and staged according to developmental morphological features and hpf (Kimmel et al., 1995). Developmental stages are described in the Results section for individual experiments. Sex cannot be determined at embryonic and larval stages. Embryos were randomly assigned to control and experimental conditions for BrdU and pharmacological treatments. The transgenic lines used were Tg(olig2:EGFP)^vu12 (Shin et al., 2003), Tg(mbpa:tagRFPt)^co25 (Hines et al., 2015) and Tg(cspg4:mCherry)^co28 (Ravanelli et al., 2018). All transgenic embryos were obtained from pairwise crosses of males or females from the AB strain to males or females of each transgenic line used.

We designed a sgRNA for the zebrafish prdm8 gene using the CRISPOR web tool (www.crispor.tefor.net) (Table S1). The sgRNA was constructed by annealing sense and anti-sense single-stranded oligonucleotides containing 5′ BsaI overhangs, and was inserted into BsaI-linearized pDR274 using the Quick Ligation Kit (New England BioLabs) (Table S1). The plasmid was transformed into chemically competent DH5a cells and purified from individual colony liquid cultures using a Qiagen Spin Miniprep Kit. To make the sgRNA, we linearized purified pDR274, containing the guide sequence, with DraI and used a T7 RNA polymerase for in vitro transcription (New England BioLabs). The pMLM3613 plasmid encoding cas9 was used for in vitro transcription using the SP6 mMessage mMachine Kit (Ambion) according to manufacturer's instructions. The sgRNA and cas9 mRNA were co-injected into single cell AB zebrafish embryos at the following concentrations: 200 ng/μl cas9 mRNA and 150 ng/μl prdm8 mRNA.

The following day, injected embryos were assayed for sgRNA activity by DNA extraction and two rounds of PCR amplification, first to amplify the prdm8 CRISPR target with gene-specific primers containing a M13F extension to the 5′ end of the forward primer (5′-TGTAAAACGACGGCCAGT-3′) and a second to add a fluorescein tag to the 5′ end of the amplified region (Table S2). The fluorescein-tagged PCR product was analyzed using capillary gel electrophoresis to detect product length. To detect F0 founders, we set up pairwise crosses of injected adults with ABs and screened their offspring for mutagenic events by fluorescent PCR and capillary gel electrophoresis. We used Sanger sequencing to determine the sequence of mutant alleles. We identified two mutant alleles, one with a 5 bp insertion (prdm8^co49) and another with a 4 bp deletion (prdm8^co51) (Fig. 3B). We were unable to recover homozygous mutant adults; therefore, these lines were maintained as heterozygotes through pairwise crosses with ABs or transgenic lines.

To genotype embryos and adults, we designed a dCAPs assay to insert a restriction site into either the mutant or wild-type allele via PCR. Specific forward primers were designed for each allele, one that added a BsrGI restriction site into the prdm8^co49 allele and another that added a NdeI restriction site into the wild-type allele for prdm8^co51 identification (Table S2). PCR products were digested with the appropriate enzymes and samples were run on a 2.5% agarose gel; the prdm8^co49 digest creates 267 and 37 bp digested mutant fragments and a 299 bp undigested wild-type fragment; the prdm8^co51 digest creates two digested wild-type fragments of 260 and 55 bp, and a 315 bp undigested mutant fragment (Fig. 3C).

We genotyped and decapitated euthanized 36 hpf embryos to isolate trunk tissue. We then added TRIzol to the trunks of each embryo and pooled trunks according to genotype. The pooled tissue was homogenized by trituration with a needle and phenol/chloroform was used to extract RNA. After precipitating RNA with isopropyl alcohol, cDNA was made using the SuperScript IV First-Strand Synthesis kit (Thermo Fisher Scientific) according to the manufacturer's instructions. To assess the expression of prdm8 mRNA, we designed primers overlaying the junctions between exons 1/2 and exons 2/3, resulting in the amplification of a 250 bp fragment. As a control, we used primers overlapping exons 1/2 and 3/4 to detect rpl13 mRNA. The same amount of cDNA was added to each PCR reaction and the products were detected on a 2.5% gel.

Embryos and larvae were dechorionated, incubated in a 20 μM BrdU solution for 40 min on ice at indicated time points (Fig. 7A) and subsequently washed four times (5 min each time) with embryo medium. Embryos and larvae were allowed to develop until 2 dpf in embryo E3 media. Samples were fixed in 4% paraformaldehyde (PFA) in 1× PBS, embedded (1.5% agar and 5% sucrose), sectioned and prepared for immunohistochemistry as described below.

Cyclopamine (Cayman Chemical, 11321) was reconstituted in ethanol to make a 10 mM stock and stored at −20°C. Dechorionated embryos were treated with 0.5 μM cyclopamine or an equal concentration of ethanol alone in E3 media at indicated time points. Following treatment, embryos were washed three times with E3 media and grown to designated time points before fixation.

In situ RNA hybridizations were performed as described previously (Hauptmann and Gerster, 2000). Probes included ptch2 (Concordet et al., 1996), nkx2.2a (Barth and Wilson, 1995), myrf, mbpa (Brösamle and Halpern, 2002) and prdm8 (Table S2). Plasmids were linearized with the appropriate restriction enzymes and cRNA preparation was carried out using Roche DIG-labeling reagents and T3, T7 or SP6 RNA polymerases (New England Biolabs). After staining, embryos were embedded in 1.5% agar/5% sucrose and frozen over dry ice. Transverse sections (20 μm) were cut using a Leica CM 1950 cryostat, collected on microscope slides and mounted with 75% glycerol.

Fluorescent in situ RNA hybridization was performed using the RNAScope Multiplex Fluorescent V2 Assay Kit (Advanced Cell Diagnostics) on 12 μm PFA-fixed and agarose-embedded cryosections according to the manufacturer's instructions with the following modification: slides were covered with Parafilm for all 40°C incubations to maintain moisture and disperse reagents across the sections. The zebrafish olig2-C1, nkx2.2a-C2, ptch2-C2, myrf-C2, cspg4-C2, boc1-C3 and prdm8-C3 transcript probes were designed and synthesized by the manufacturer, and used at a dilution of 1:50. Transcripts were fluorescently labeled with Opal520 (1:1500), Opal570 (1:500) and Opal650 (1:1500) using the Opal 7 Kit (Perkin Elmer, NEL797001KT).

Tg(olig2:EGFP) euthanized embryos (at 24, 36 and 48 hpf) were collected in 1.7 ml microcentrifuge tubes and deyolked in 100 μl of prechilled Ca-free Ringer's solution [116 mM NaCl, 2.6 mM KCl and 5 mM HEPES (pH 7.0)] on ice. Embryos were pipetted intermittently with a p200 micropipette for 15 min and left for 5 min. Protease solution (500 μl; 10 mg/ml BI protease, 125 U/ml DNase, 2.5 mM EDTA and 1× PBS) was added to microcentrifuge tubes on ice for 15 min and embryos were homogenized every 3 min with a p100 micropipette for 15 min. Stop solution (200 μl; 30% FBS, 0.8 mM CaCl₂ and 1× PBS) was then mixed into the tubes. Samples were then spun down at 400 g for 5 min at 4°C, and supernatant was removed. On ice, 1 ml of chilled suspension medium [1% fetal bovine serum (FBS), 0.8 mM CaCl₂, 50 U/ml penicillin and 0.05 mg/ml streptomycin] was added to the samples, which were then spun down again at 400 g for 5 min at 4°C. The supernatant was removed, and 400 μl of chilled suspension medium was added and the solution was filtered through a 35 μm strainer into a collection tube. Cells were FAC sorted to distinguish EGFP⁺ cells using a MoFlo XDP100 cell sorter at the CU-SOM Cancer Center Flow Cytometry Shared Resource and collected in 1.7 ml FBS-coated microcentrifuge tubes in 200 μl of 1× PBS.

The Chromium Box from 10x Genomics was used to capture cells using a Chromium Single Cell 3′ Reagent Kit (PN-1000075). Libraries were sequenced on the Illumina NovaSeq 6000 Instrument. FASTQ files were analyzed using Cell Ranger Software. We obtained 2174 (24 h), 2555 (36 h) and 3177 (48 h) cells yielding a mean of 118,014 (24 h), 65,182 (36 h) and 96,053 (48 h) reads per cell with a median of 1929 (24 h), 1229 (36 h) and 1699 (48 h) genes identified per cell.

Raw sequencing reads were demultiplexed, mapped to the zebrafish reference genome (build GRCz11/danRer11) and summarized into gene expression matrices using Cell Ranger (version 3.0.1). The resulting count matrices were further filtered in Seurat 3.1.0 (www.satijalab.org/seurat/) to remove cell barcodes with fewer than 250 detectable genes, more than 5% of unique molecular identifiers (UMIs) derived from mitochondrial genes, or more than 50,000 UMIs (to exclude putative doublets). This filtering resulted in 6489 single cells across all samples (1952 from 24 hpf, 2147 from 36 hpf and 2390 from 48 hpf). After standard Seurat normalization, principal component analysis was carried out using the 1291 most variable genes. Next, dimensionality reduction was performed using uniform manifold approximation and projection on the first 15 principal components. Differential expression and marker gene identification was performed using MAST (Finak et al., 2015).

Larvae were fixed using 4% PFA/1× PBS overnight at 4°C. Embryos were washed with 1× PBS, rocked at room temperature, embedded in 1.5% agar/5% sucrose, frozen over dry ice and sectioned in 20 or 15 µm transverse increments using a cryostat microtome. Slides were place in Sequenza racks (Thermo Scientific), washed three times (for 5 min each time) in 0.1% Triton X-100/1× PBS (PBSTx), blocked for 1 h in 2% goat serum/2% bovine serum albumin/PBSTx and then placed in primary antibody (in block) overnight at 4°C. The primary antibodies used included: rabbit anti-Sox10 (1:500; Park et al., 2005); mouse anti-Islet (1:500; Developmental Studies Hybridoma Bank, AB2314683); rat anti-BrdU (1:100; Abcam, AB6326); or mouse JL-8 Living Colors (1:500; Clontech, 632380) to restore Tg(olig2:EGFP) fluorescence after RNA ISH. Sections were washed for 1 h at room temperature with PBSTx and then incubated for 2 h at room temperature with secondary antibodies at a 1:200 dilution in block. The secondary antibodies used included: AlexaFluor 488 anti-rabbit (Invitrogen, A11008), AlexaFluor 588 anti-rabbit (Invitrogen, A11011), AlexaFluor 647 anti-rabbit (Jackson ImmunoResearch, 111606144), AlexaFluor 488 anti-mouse (Life Technologies, A11001), AlexaFluor 568 anti-mouse (Invitrogen, A11004) and AlexaFluor 568 anti-rat (Invitrogen, A11077). Sections were washed for 1 h with PBSTx and mounted in VectaShield (Vector Laboratories).

Fixed sections of embryos and larvae were imaged on a Zeiss Cell Observer SD 25 spinning disk confocal system or a Zeiss Axiovert 200 microscope equipped with a PerkinElmer spinning disk confocal system. IHC cell counts were collected using a 20× objective (n.a. 0.8) and representative images were collected using a 40× oil immersion objective (n.a. 1.3). Wild-type 1 dpf larvae were positioned on top of a 2% agarose plate and imaged using a Leica M165FC dissection scope with a SPOT RT3 camera. RNA ISH sections were imaged using differential interference contrast optics and a Zeiss AxioObserver compound microscope. Cell counts and representative images were acquired at a 40× magnification (n.a. 0.75). Images are presented as extended z-projections or a single plan (RNA ISH) collected using Volocity (PerkinElmer) or Zen (Carl Zeiss) imaging software. Image brightness and contrast were adjusted in Photoshop (Adobe) or ImageJ (National Institutes of Health).

Evaluations of nkx2.2a and olig2:EGFP domain overlap were performed in transverse sections by collecting single plane wide-field images of each section. The area of overlap was quantified using ImageJ by outlining the area of expression of olig2:EGFP and the area of nkx2.2a expression within the total olig2:EGFP area. Quantifications of fluorescent cell numbers in transverse sections were performed by collecting confocal z stacks of the entire section. Quantifications of RNA ISH cell numbers in transverse sections were performed by viewing the entire z plane. Data for each embryo were collected from ten consecutive trunk spinal cord sections and n represents the average number of cells per section in one embryo. All cell counts on sections were performed by a researcher who was blinded to the identities of the slides except Fig 4A-D and Fig. 6E,F.

Quantification of fluorescent RNA ISH hybridization was carried out on z projections collected at identical exposures. All quantifications were performed with ImageJ Fiji using a custom script created by Karlie Fedder, University of Colorado, Department of Pediatrics, https://github.com/rebeccaorourke-cu/Prdm8-regulates-pMN-progenitor-specification. First, ten 0.5 μm z intervals were maximum z projected and background was subtracted using a 2-rolling ball. The image was then thresholded by taking two standard deviations above the mean fluorescence intensity. A region of interest was drawn around the pMN domain or spinal cord, and puncta were analyzed using the ‘analyze particles’ feature with a size of 0.01 to infinity and circularity of 0.00 to 1.00. All thresholded puncta were inspected to ensure single molecules were selected. Puncta with an area of only one pixel were removed from the dataset. Data for each embryo were collected from five consecutive trunk spinal cord sections and n represents the average number of puncta in a region of interest per section in a single embryo.

We plotted all data and performed all the statistical analyses in GraphPad Prism. All data are expressed as mean±s.e.m. Normality was assessed with a D'Agostino and Pearson omnibus test. For statistical analysis, we used an unpaired Student's two-tailed t-test for all data with normal distributions, or Mann–Whitney tests for non-normal data. Unless otherwise stated, all graphs represent data collected from one laboratory replicate, sampling fish from multiple crosses with no inclusion or exclusion criteria. P values not provided in the graphs are indicated as follows: *P<0.05, **P<0.001, ***P<0.0001.

We thank Christina Kearns for isolating cells for scRNA-seq and members of the Appel lab and the Section of Developmental Biology for discussions and advice. Cell sorting was performed by the University of Colorado Cancer Center Flow Cytometry Shared Resource, supported by a Cancer Center Support grant (P30CA046934). scRNA-seq was performed by the University of Colorado Anschutz Medical Campus Genomics Shared Resource Core Facility, supported by a Cancer Center Support grant (P30CA046934). Single cell RNA-seq and bioinformatics analysis was supported by a pilot award from the University of Colorado RNA Bioscience Initiative. The University of Colorado Anschutz Medical Campus Zebrafish Core Facility was supported by a National Institutes of Health grant (P30 NS048154).

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.191023.reviewer-comments.pdf