Sulf2a controls Shh-dependent neural fate specification in the developing spinal cord

Sulf2a belongs to the Sulf family of extracellular sulfatases which selectively remove 6-O-sulfate groups from heparan sulfates, a critical regulation level for their role in modulating the activity of signalling molecules. Data presented here define Sulf2a as a novel player in the control of Sonic Hedgehog (Shh)-mediated cell type specification during spinal cord development. We show that Sulf2a depletion in zebrafish results in overproduction of V3 interneurons at the expense of motor neurons and also impedes generation of oligodendrocyte precursor cells (OPCs), three cell types that depend on Shh for their generation. We provide evidence that Sulf2a, expressed in a spatially restricted progenitor domain, acts by maintaining the correct patterning and specification of ventral progenitors. More specifically, Sulf2a prevents Olig2 progenitors to activate high-threshold Shh response and, thereby, to adopt a V3 interneuron fate, thus ensuring proper production of motor neurons and OPCs. We propose a model in which Sulf2a reduces Shh signalling levels in responding cells by decreasing their sensitivity to the morphogen factor. More generally, our work, revealing that, in contrast to its paralog Sulf1, Sulf2a regulates neural fate specification in Shh target cells, provides direct evidence of non-redundant functions of Sulfs in the developing spinal cord.

www.nature.com/scientificreports/ fate change depends on the formation of a novel source of Shh, named the lateral floor plate, that forms within the Nkx2.2-expressing progenitors of the p3 domain 12 . Again, Sulf1 must be up-regulated in these Shh-secreting cells so that these cells provide higher Shh signal to neighbouring pMN cells which, in turn, activate expression of Nkx2.2 that no longer represses Olig2 at this stage [17][18][19][20] . In this way, Sulf1 contributes to induce formation of a new progenitor domain, named the p* domain, populated by cells co-expressing Olig2 and Nkx2.2 fated to generate OPCs. Thus, in the developing spinal cord, the key role of Sulf1 is to change the inductive properties of Shh source cells to trigger high-threshold response to Shh.
Much less is known about the role of Sulf2 in regulating morphogen signalling. Sulf1 and Sulf2 have been proposed to display compensatory functions in many developmental processes 3,4,[21][22][23] . However, whether Sulf1 and Sulf2 exert specific roles in the developing spinal cord remains to be addressed. In this tissue, in contrast to Sulf1 whose expression is restricted to medial and lateral floor plate cells, Sulf2 is expressed broadly in both floor plate cells and neural progenitors [7][8][9][10][11][24][25][26] . Comparison of Sulf1 and Sulf2 mouse knockout phenotypes led to the conclusion that the two enzymes act in a similar way at the level of Shh source cells to regulate Shhdependent patterning at the time of OPC specification 11 . However, supporting the view that Sulf2 has a unique and additional function, conditional depletion of Sulf2 specifically in neural progenitors expressing Olig2 but not in Shh source cells is sufficient to impair gliogenesis in mouse 26 . Whether this function is linked to regulation of the Shh signal remains an open question.
In zebrafish, three Sulf members have been identified: Sulf1 and two Sulf2 paralogs, named Sulf2a and Sulf2b 27 . Here, we report that Sulf2a is the unique Sulf member whose expression is restricted to neural progenitors, then offering a simple context to study Sulf function specifically in Shh-responding cells of the developing spinal cord. We show that Sulf2a is required for proper generation of ventral neuronal and oligodendroglial cell subtypes and provide evidence that Sulf2a plays a key role in controlling neural progenitor identity. Overall, our data support a model whereby Sulf2a, by reducing the sensitivity of target cells to Shh, promotes low-threshold response to the morphogen over spinal cord development.

Results
Sulf2a expression is restricted to neural progenitors of the developing spinal cord. As a first step to investigate the function of Sulf2 proteins in zebrafish, we analysed expression patterns of sulf2a and sulf2b. We first focused on two development stages: 24 hpf, while ventral neural pattern is established and neuronal production is ongoing and 36 hpf, corresponding to the time of p* domain formation and onset of OPC production in the ventral spinal cord. As previously reported 27 , we found that, at neurogenic stage (24 hpf), sulf2b is expressed in medial floor plate cells but also in adjacent neural progenitors of the p3 domain (Fig. 1a,  b). At the onset of gliogenesis (36 hpf), sulf2b signal, although weaker, was still detected in the medial floor plate and in adjacent cells which form to the lateral floor plate at that stage (Fig. 1c, d). Thus, as previously reported for Sulf1 7 , sulf2b is expressed in Shh-producing cells of the developing spinal cord. Interestingly, we found a very different pattern of expression for sulf2a whose mRNA was detected in a broad domain of the ventral spinal cord at both neurogenic and gliogenic stages ( Fig. 1e-h). Noticeably, sulf2a mRNA was undetectable either in medial floor plate or in lateral floor plate cells but was found expressed in neural progenitors located at distance from Shh-producing cells (Fig. 1f, h). Double detection of sulf2a and shh mRNAs further confirmed that sulf2a and shh domains of expression do not overlap neither at neurogenic (24 hpf) nor gliogenic (36 hpf) stages (Fig. 1k, l). These data therefore revealed a unique property of sulf2a being exclusively expressed in neural progenitors.
We next investigated sulf2a expression at earlier and later stages of development. Ruling out the possibility that sulf2a plays a role in establishment of progenitor domains, we found that sulf2a is not yet expressed at 16 hpf (Fig. 1i), while neural tube patterning has yet been established and neuronal production has already been initiated 7,28 . Much later, at the end of embryogenesis (72 hpf), sulf2a expression was maintained in neural progenitors and was still excluded from Shh-producing cells (Fig. 1j), thus indicating that sulf2a remains expressed in neural progenitors after the onset of neuronal production and all along gliogenesis.
Together, these results, showing that sulf2a is the only sulf gene activated specifically in neural progenitors open the possibility that this Sulf member plays a specific role in regulating signal reception at the surface of spinal progenitors as they generate neurons and glial cells.

Sulf2a controls generation of both V3 interneurons and OPCs.
Our data showing that sulf2a is activated after initiation of neuronal production and persists along gliogenesis, prompted us to examine its function in regulating generation of V3 interneurons and OPCs produced by ventral neural progenitors 29,30 . To address the role of Sulf2a, we first used a morpholino (MO) knockdown approach, using a translation (sul-f2aMO ATG ) and a splice (sulf2aMO splice ) blocking morpholino. We performed RT-PCR which highlighted the presence of unspliced forms of the sulf2a transcript in sulf2aMO splice -injected embryos, showing knockdown efficiency in these experiments ( Supplementary Fig. 1a, b). As similar results were obtained using either MO in all experiments, they are referred to as sulf2aMO in the text. We first assessed V3 interneuron generation by detecting sim1a expression as a specific marker of these neurons originating from the p3 domain 29,31 . At 48 hpf, as neuronal production has been completed, we found that the number of sim1a + cells was significantly higher in sulf2aMO-injected embryos compared to the control MO (ctrlMO)-injected ones, with more than a 20% increase in the number of V3 interneurons in sulf2a morphants (Fig. 2a-c, f). To confirm these observations, we generated a sulf2a mutant zebrafish line using the CRISPR/Cas9 system. We targeted the hydrophilic domain (HD) region of Sulf2a, known to be required for enzymatic activity and substrate recognition 3,32-34 . We obtained an indel mutation in the coding region which leads to a premature translation termination and a predicted protein missing functional HD domain ( Supplementary Fig. 2a). Validating that this mutation is indeed a loss-of-function allele, sulf2a mRNA was found nearly undetectable in embryos carrying homozygous  Fig. 2b). Adult sulf2a-/-individuals showed no overt phenotype, are viable, fertile, and produce offsprings that show no obvious morphological defects. Confirming the requirement of Sulf2a to restrict V3 interneuron production, we found that the number of sim1a + neurons increased by 25% in sulf2a-/-embryos compared to wild-type siblings ( Fig. 2d-f). We conclude from these data that Sulf2a plays a role in limiting production of V3 interneurons. Sulf2a expression being maintained at 36 hpf, the time of OPC specification 30 , we next asked whether Sulf2a might also play a role in the control of OPC generation. To identify OPCs, we used either Tg(olig2:GFP) or Tg(olig2:DsRed2) transgenic lines in which fluorescently-labelled OPCs can be identified by their morphology and position 35,36 . Due to expression of olig2 in progenitors of the pMN domain as well as perdurance of GFP or DsRed2 in newly generated MNs [35][36][37] , these reporter lines however do not allow distinguishing OPCs from other cell types in the ventral spinal cord. We bypassed this problem by performing immunodetection of Sox10, the earliest specific marker of oligodendroglial cells 37 . At 48 hpf, when OPC generation has just began, we found olig2:GFP + or olig2:DsRed2 + cells co-expressing Sox10 in the ventral spinal cord of both ctrlMO-injected and wild-type embryos (Fig. 3a, d). At identical stage, the number of OPCs was significantly reduced either in   www.nature.com/scientificreports/ sulf2aMO-injected or in sulf2a−/− embryos, with a decrease of more than 45% and 33% compared to controls (Fig. 3b, c, e, f). In order to assess the specific involvement of sulf2a in OPC development, we performed rescue experiments and found that co-injection of sulf2aMO splice together with sulf2a mRNA, which thus does not carry the MO targeted nucleotide sequence, restored the OPC number to control level (Supplementary Fig. 1c-f). Thus, our data evidenced that Sulf2a activity is also required for proper generation of OPCs but, instead of limiting the number of cells as observed for V3 interneurons, the enzyme contributes to promote OPC production. Together, these results, showing that Sulf2a limits production of V3 interneurons during neurogenesis but promotes generation of OPCs at initiation of gliogenesis, reveal specific and differential functions of Sulf2a in controlling generation of ventral neural cell subtypes.
Sulf2a controls the identity of ventral neural progenitors. We next focused on the question of how the same enzyme can fulfil differential regulatory functions to ensure production of a correct number of neural derivatives. Sulf2a being expressed in neural progenitors, we postulated that the enzyme might be involved in regulating progenitor patterning and subsequent cell diversification. We started analysing progenitor domain organisation at 24 hpf while neurons are being produced. We focused on the p3 and pMN domains characterised by the expression of Nkx2.2a and Olig2, respectively. Nkx2.2a expression was assessed detecting mEGFP in the Tg(nkx2.2a:mEGFP) line 36 , while Olig2 expression was monitored by detecting the olig2 mRNA using fluorescent RNA in situ hybridisation. Immunodetection of Sox2, a generic marker of neural progenitors 38,39 , was also performed to distinguish these cells from newly generated neurons. Analysis of ctrlMO-injected embryos revealed nkx2.2a:mEGFP + progenitors organised in a row of one to two cells in the ventral progenitor zone and comparison with olig2 mRNA staining showed that olig2 expression defined the dorsally adjacent nonoverlapping progenitor domain ( Fig. 4a-a"), thus reflecting ventral patterning establishment. Similar analysis performed in sulf2aMO-injected embryos showed that the nkx2.2a:mEGFP + domain extended dorsally, covering rows of two to three cells, reflecting a more than 35% broadening of the p3 domain compared to controls ( Fig. 4b, b", c, c", d). Thus, in agreement with the excess of V3 interneurons observed in Sulf2a-depleted embryos, p3 progenitor number is increased in these embryos. By contrast, olig2 expression was found markedly reduced in sulf2aMO-injected embryos ( Fig. 4b' c"). Counting of olig2 + progenitors indicated a more than 59% decrease in sulf2a morphants compared to control embryos (Fig. 4e). Accordingly, the number of MNs, assessed by immunostaining of Islet1/2 in Tg(olig2:DsRed2) embryos, was reduced by 20-30% in 48hpf sulf2a-depleted embryos ( Supplementary Fig. 3). Together, these data indicated that Sulf2a is part of the regulatory process that controls dorsal boundary positioning of the Nkx2.2a/p3 domain and thus contributes to maintain the Olig2/ pMN domain over the period of neuronal production.
We next examined neural patterning at the onset of gliogenesis (36 hpf). As expected, we found that, in control embryos, the nkx2.2a:mEGFP + progenitor domain was broader than that observed at the neurogenic period (compare Figs. 4 and 5) and partially overlapped the olig2-expressing domain (Fig. 5a", a"'), reflecting formation of the p* domain at this stage. As observed at neurogenic period, sulf2a depletion caused dorsal expansion of the nkx2.2a:mEGFP + domain, with a more than 25% increase in the number of nkx2.2a:mEGFP + progenitors compared to controls (Fig. 5b, b", b"' , c", c"' , d), thus indicating that the dorsal boundary position of the Nkx2.2aexpressing domain is also affected by Sulf2a depletion at the time of patterning rearrangement. Counting of olig2 + progenitors in sulf2a morphants showed that their number decreased by 40-50%, a reduction still significant but less pronounced than that observed at neurogenic stage ( Fig. 5b' , c' , e). Noticeably and in agreement with the loss of repressive activity of Nkx2.2a on olig2 expression at this stage, no significant difference was found in the number of progenitors co-expressing nkx2.2a:mEGFP and olig2 between control and sulf2a morphant embryos ( Fig. 5a"-c"' , f). The logical corollary of this latter observation is that, upon sulf2a depletion, the observed reduction in the number of Olig2 progenitors should be due to a specific decrease in the number of progenitors expressing Olig2 but not Nkx2.2a. Accordingly, quantification of these cells showed a decrease by more than 60% in sulf2a-depleted embryos ( Fig. 5a"-c"' , g). Thus, at gliogenic stage, Sulf2a depletion causes dorsal expansion of the Nkx2.2a-expressing domain, does not affect generation of the p* domain but leads to a drastic loss of the dorsally located progenitors that express Olig2 but not Nkx2.2a. From these data, we conclude that, at stage of patterning rearrangement, Sulf2a function is again required to position dorsal boundary expression of the Nkx2.2a domain, but also to ensure that a population of Olig2-expressing progenitors that do not activate Nkx2.2a are generated.
Together, these data highlight the key role of Sulf2a in assigning and maintaining neural progenitor identity at neurogenic and gliogenic stages.

Sulf2a regulates allocation of oligodendrocyte lineage cell fates. In zebrafish, newly specified
OPCs, all expressing Olig2, represent a heterogeneous cell population in regard to Nkx2.2a expression and fate. While some OPCs express Nkx2.2a and differentiate rapidly as oligodendrocytes, others, that do not express Nkx2.2a, remain as non-myelinating OPCs 40,41 . Hereafter we will refer to these two OPC subsets as myelinating and non-myelinating OPCs, respectively. Whether these two OPC subsets emerge from Olig2 neural progenitors already distinguishable by differential Nkx2.2a expression remained an open question. Our data, highlighting the requirement of Sulf2a specifically for maintaining progenitors expressing only Olig2, led us to ask whether Sulf2a contributes specifically to generate the non-myelinating OPC subset. We thus examined the consequences of Sulf2a depletion on the development of the two distinct OPC populations by performing immunodetection of the pan-OPC marker Sox10 in Tg(nkx2.2a:mEGFP; olig2:DsRed2) embryos in which the two OPC subtypes can be distinguished by differential reporter expression 36 . As expected, at early gliogenic period (48 hpf), we found that the global population of OPCs, i.e. the Sox10 + /olig2:DsRed2 + cells, was reduced in number either in sulf2aMO-injected or in sulf2a−/− embryos, decreasing by 40% compared to controls (Fig. 6a- www.nature.com/scientificreports/ considered separately the myelinating (olig2:DsRed2 + /nkx2.2a:mEGFP +) and the non-myelinating (olig2:DsR ed2 + /nkx2.2a:mEGFP−) OPC subtypes. Quantification of these cells pointed to a decrease in number of both OPC populations but to different extent. In sulf2a morphants, non-myelinating OPCs were found decreased in number by 64% (Fig. 6a, b", g) while the myelinating population was reduced only by 31% compared to controls (Fig. 6a, b", f). A similar tendency was obtained analysing the two OPC subsets in sulf2a−/− embryos (Fig. 6c, d", f, g). Thus, the impaired development of OPCs observed in Sulf2a-depleted embryos is mostly attributable to a decreased number of OPCs that do not express Nkx2.2a.
These results, showing that Sulf2a plays a prominent role in promoting the development of non-myelinating OPCs, support the view that Sulf2a contributes to the early segregation of the two OPC lineages by preserving a pool of progenitors that only express Olig2.   . 7a-d). These observations suggested that, at 24 hpf, sulf2a expression overlaps the entire Olig2 + domain while it is not the case at 36 hpf, since, at that stage, the dorsal-most Nkx2.2a + cells are known to co-express Olig2. To clarify this, we analysed expression of sulf2a mRNA at 36 hpf in Tg(olig2:GFP) embryos and found expression of sulf2a in a subset of Olig2 progenitors (Fig. 7e, f). Sulf2a expression being excluded from the Nkx2.2a-expressing domain, we conclude that these cells correspond to Olig2 progenitors that are not included in the p* domain. Overall, these data, showing that sulf2a is excluded from progenitors that respond to high Shh signalling activity (Nkx2.2a + cells) but expressed in progenitors that do not activate high-threshold response to Shh (Olig2 + cells), support the view that Sulf2a might act at the surface of Olig2 + progenitors to reduce their sensitivity to Shh signalling. They moreover highlight differential expression of sulf2a in Olig2 progenitors at gliogenic stage. An attractive hypothesis was thus that Sulf2 restricts the range of high-threshold response to Shh along the ventral to dorsal axis of the progenitor zone. To test this possibility, we used Tg(GBS-ptch2:EGFP) reporter line in which cells responding to Shh can be visualized by EGFP expression 44 . At 24 hpf, in control embryos, GBS-ptch2:EGFP expression was detected in the ventral neural tube in medial floor plate cells as well as adjacent progenitors that both express Sox2 (Fig. 8a,b). We interpreted the EGFP signal in medial floor plate cells as perdurance of the reporter protein since, while responsible for medial floor plate induction at earlier stages, Shh signalling is known to be inactive in these cells at 24 hpf 45,46 . Moreover, EGFP was not detected at distance from the Nkx2.2a/p3 domain, indicating that the EGFP signal only allowed detecting the highest levels of Shh signalling. Having this in mind, we next examined EGFP expression in sulf2aMO-injected embryos. We found that the EGFP expression domain was enlarged and cell quantification confirmed a slight but significant increase in the number of EGFP + /Sox2 + cells compared to controls (Fig. 8c-e), thus validating the role of Sulf2a in limiting the dorsal extent of high-threshold Shh signalling activity.
To bring additional support to the view that Sulf2a acts as a negative regulator of Shh signalling, we analysed expression of patched2, a as a marker for Shh response 47 . Fluorescent in situ hybridisation of patched2 combined with Sox2 immunodetection was performed at 24 hpf and the relative fluorescent in situ signal was quantified along the ventral to dorsal axis of the neural tube ( Fig. 8f-h). In control embryos, as previously reported 48 , we found graded expression of patched2, decreasing dorsally in neural progenitors (Fig. 8f,h). Noticeably, dorsally to the highest ventral signal, the intensity curve declines with a bending and then flattens dorsally, indicating that the response to Shh does not decrease linearly from the ventral region of the neural tube. Similarly, in sulf2a-MO-injected embryos, patched2 gradient progressively decreased dorsally (Fig. 8g,h). However, difference was observed in the shape of the curve which does bend but appeared decreasing dorsally in a linear manner (Fig. 8h). Comparison of patched2 relative signal intensities further confirmed significant higher levels of patched2 expression at the level of the intermediate neural tube in sulf2a-depleted embryos compared to controls (Fig. 8h), thus reflecting higher levels of Shh signalling activity in neural progenitors that should have expressed sulf2a. These results thus reinforce the view that sulf2a acts to limit response to Shh. Importantly, we found that shh expression was not modified in sulf2a-depleted embryos ( Supplementary  Fig. 4), indicating that the influence of Sulf2a on Shh signalling activity does not depend on transcriptional regulation of the morphogen factor.
These data, showing that Sulf2a activity is required to downregulate Shh activity and in particular to limit the dorsal extent of high-threshold Shh response, highlight a functional relationship between Sulf2a and Shh, the enzyme acting by reducing the sensitivity of target cells to the Shh signal.

Discussion
In the ventral spinal cord, generation of distinct neural cell types is a spatio-temporally regulated process that mainly depends on the morphogenetic activity of Shh. During development, neural progenitors are exposed to varying concentrations of Shh which drive them to adopt specific transcriptional identities and eventually generate a particular neuronal or glial cell fate. We show here that Sulf2a contributes to assign specific neural identities to Shh-responding progenitors. We provide evidence that expression of sulf2a in Olig2 progenitors is essential to prevent highest levels of Shh signalling within these cells, thus allowing them to orient toward the production of, first MNs and, later on, of a particular subtype of OPC. Our data support a model in which Sulf2a spatially restricts expression of high-threshold Shh responsive genes to cells closest to the source of Shh (Fig. 9).
The first main finding of our work is that, in contrast to other vertebrates where the different Sulfs display overlapping expression domains in the developing spinal cord 10,11,[24][25][26] , in zebrafish, a subset of spinal neural progenitors only expresses Sulf2a throughout neurogenesis and gliogenesis. Starting out from this observation, studying the role of Sulf2a in zebrafish represented a unique opportunity to investigate the still debated question of whether Sulf proteins fulfill redundant functions or contribute differentially to spinal cord development 9,11,26 . During the neurogenic phase, which takes place between 10 and 30 hpf in the ventral zebrafish spinal cord 28  www.nature.com/scientificreports/ Quantifications in (f,g) correspond to Sox2 + progenitors that either co-express olig2 and nkx2.2a:mEGFP (f) or express olig2 mRNA but not nkx2.2a:mEGFP (g). Except in e where datasets were compared with Student's t-test (unpaired two-tailed), datasets were compared with Mann Whitney's test (two-tailed). Results are presented as a mean number of cells along the dorso-ventral axis per embryo ± s.d (*p < 0,05, ** p < 0,005, ***p < 0.001,****p < 0.0001).
◂ showed that Sulf2a prevents excess production of V3 interneurons and fosters MN generation. Later on, from 36 hpf, Sulf2a, while not essential for OPC production, has proven to play a key role in triggering generation of non-myelinating OPCs. This markedly contrasts with the function of Sulf1 which stimulates V3 interneuron production, has no impact on MN generation and proves to affect more OPC generation when depleted 7 . Thus, our work makes clear that Sulf2a and Sulf1 exert differential functions in the ventral spinal cord. We next provide evidence that Sulf2a is part of the regulatory process that provides positional identity to ventral neural progenitors. Specifically, Sulf2a is required, either at neurogenic or gliogenic stage, to maintain a pool of progenitors that express Olig2 but not Nkx2.2a. At neurogenic stages, sulf2a expression defines the ventral boundary of the Olig2/pMN domain and acts to prevent dorsal expansion of Nkx2.2a expression within this domain. In agreement with the well-known repressive activity of Nkx2.2a on olig2 expression at this stage 17,20,31,49 , dorsal misexpression of Nkx2.2a observed in Sulf2a-depleted embryos is accompanied by a marked reduction in the number of Olig2 progenitors. Thus, the progenitor fate change caused by Sulf2a depletion is sufficient to explain the overproduction of V3 interneurons at the expense of MN development observed in these embryos. However, also likely contributing to V3 interneuron overproduction, is the presence of Nkx2.2a/p3 progenitors that persist in Sulf2a-depleted embryos at gliogenic stage. Indeed, in contrast to the wild-type situation, in these embryos, we observed that the Nkx2.2a/p3 domain extends dorsally to the Shh-expressing domain (lateral floor plate).
A somewhat similar but more complex situation was found regarding progenitor identity regulation by Sulf2a at the onset of gliogenesis where sulf2a expression, still excluded from the Nkx2.2a-expressing domain, thus from the p* domain, is detected in a subset of Olig2 progenitors. Consistently, the enzyme appears dispensable for formation of the p* domain but required for proper development of progenitors that only express Olig2. How Sulf2a might control development of this particular progenitor population however remains elusive. Since Nkx2.2a no more represses Olig2 at initiation of gliogenesis, dorsal misexpression of Nkx2.2a cannot provide a simple explanation to the loss of Olig2 progenitors caused by Sulf2a depletion as it is the case at neurogenic stage. An alternative explanation might be that Sulf2a plays a role in progenitor recruitment, a process that occurs at initiation of gliogenesis and during which progenitors positioned dorsally to the pMN domain descend to this domain and only then initiate Olig2 expression, a process known to depend on Shh signalling 41,50 .
Quite apart from the question of how Sulf2a regulates neural patterning, our data shed also new light on the still open question of the origin and relationship of progenitors that give rise to the two OPC subtypes in zebrafish. Two distinct types of spinal OPCs, ones expressing Nkx2.2a and differentiating rapidly as oligodendrocytes and others, not expressing Nkx2.2a and remaining as non-myelinating OPCs have recently being shown to arise from distinct Olig2 progenitor pools 40,41 . However, how these OPC populations, heterogeneous with respect to Nkx2.2a expression and fate, are generated remained elusive. Our data, showing that Sulf2a depletion preferentially affects generation of both non-myelinating OPCs and Olig2 progenitors that do not express Nkx2.2a, strongly support the view of a lineage relationship between these cells. In apparent contradiction with this proposal, Nkx2.2a-expressing OPCs are also affected, although more moderately, by Sulf2a depletion despite the presence of a correct number of Olig2/Nkx2.2a-co-expressing progenitors. Since neuronal signals and activity can alter OPC development (see 51 for review), this result can be potentially explained by the aberrant neuronal populations generated upon Sulf2a depletion.
Finally, examination of Shh signalling activity upon Sulf2a depletion together with the well-known dependence of Olig2 and Nkx2.2a expression on differential Shh thresholds 37,42,43 are consistent with a role of Sulf2a in attenuating Shh activity at the surface of Olig2 progenitors. Consistently, development of the non-myelinating OPC subset, which is particularly affected by Sulf2a depletion, depends on lower doses of Shh as compared to myelinating OPCs 41 . Our work thus brings support to an essential role of Sulf2a-dependant HS sulfation in regulating Shh signal reception. This is in line with previous reports showing that the sulfation level of glypicans, a family of GPI-anchored HSPGs acting either as co-receptors or as competitors of Shh binding to its receptor Patched, is an important level of regulation of their function at the surface of Shh receiving cells [52][53][54][55] . Importantly, this work reveals that Sulf2a plays an opposite role than Sulf1 which stimulates Shh signalling in the developing spinal cord 7-10 . The differential expression of Sulf2a in neural progenitors and of Sulf1 in Shh-producing cells indicates that Sulfs can fulfill opposite functions on the very same signalling cascade by acting either at the source or in receiving cells. This is reminiscent to the dual function of DSulf1, the unique Sulfatase gene in Drosophila, www.nature.com/scientificreports/ which is expressed both in Hh-producing and -receiving compartments where it exerts a positive and a negative regulatory function on Hh signalling, respectively 56 . Overall, our work identifies Sulf2a as a crucial regulator of neural cell diversification within the ventral developing spinal cord through a negative action on Shh signalling. Sulf2a contributes to pattern gene expression in neural progenitor cells, sustaining low-threshold response to Shh morphogen at the two critical periods of neuronal and glial production. This work not only unravels a novel function for Sulf2a but also highlights how extracellular sulfatases contribute in different ways to a given signalling pathway depending on their spatial expression in a complex tissue. Generation of sulf2a mutant by Crispr/Cas9 technology and genotyping. SgRNA targeting the exon 11 of sulf2a gene was designed using the CRISPOR selection tool 58 . The following primers were purchased from Integrated DNA Technologies (IDT): 5′ TAG GCT TCC TCT TGA GAG GCG CTG 3′ and 5′ AAA CCA GCG CCT CTC AAG AGG AAG 3′, annealed and cloned into a pDR274 host vector (Addgene). After sequence-based selection, this vector was linearized and used as a template for producing RNAs by in vitro transcription (T7 mMessage mMachine, Ambion). The obtained sgRNA solution was mixed with Cas9 nuclease (NEB Cas9 Nuclease, S.pyogenes 20 mM, M0386M) at 45 ng/µl final concentration and used for microinjection in Tg(Olig2:GFP) one-cell stage embryos. F0 individuals carrying mutations were screened by sequencing PCR products using the following oligonucleotides: 5′ TTC CTC TGT GCA GAA GTG GC 3′ and 5′ TCA ATC TGA CAA CCT GCG CT 3′. F0 fish positive for mutation was outcrossed to wild-type fish. F1 offsprings were thus screened by sequencing PCR products using the above primers to confirm germline transmission, which led to identifying an indel mutation (16 nucleotide deletion and 9 nucleotide insertion) generating a STOP codon at position 519. F1 carrier fish was selected and crossed to wild-type fish in order to generate F2 individuals that were further used to produce the subsequent generations of embryos for phenotypic analysis. Embryos were obtained by crossing heterozygous fish and were further individually genotyped before analysis. Genotyping was performed from genomic DNA extracted from tail clipping for adults or from dissected heads from embryos, using the DirectPCR Lysis Reagent RT-PCR experiments. Total RNAs were extracted using Trizol from 20 injected embryos at 24 hpf for each condition. 500 ng of RNAs were used to perform reverse transcription following provider guidelines (Super-Script II, Invitrogen). Resulting cDNA was used as template for PCR reaction using a forward oligonucleotide www.nature.com/scientificreports/ hybridizing in exon 11 5′ GCT GCC AAC AGA GGA CAG AT 3′ and a reverse oligonucleotide hybridizing in exon 12 5′ TTG AGG TTG AGG GGA CGG TA 3′ of sulf2a gene. PCR product size was analysed on agarose gel, sulf2a spliced mRNAs give rise to a 221 bp band while unspliced forms generate a 2430 bp size PCR product (Supplementary Fig. 1a,b).
Staining procedures. Fixation and storage. Embryos were raised and staged according to standard protocols 59 . Staged embryos were released from the chorion using forceps and placed in a 3.7% formaldehyde solution overnight at 4 °C for in situ hybridisation or 1 h at room temperature for immunostainings. Then embryos were dehydrated through a methanol series and stored in 100% methanol at − 20 °C.
Single in situ hybridisation. Single in situ hybridisation was adapted from Xu et al. 62  Fluorescent in situ hybridisation coupled with immunodetection. Fluorescent in situ hybridisation coupled with immunodetection was adapted from Denkers et al. 63 . After hybridisation with the olig2 digoxygenin-labelled probe, HRP-coupled anti-digoxygenin antibody (1/1000, Roche) was incubated together with chick anti-GFP and rabbit anti-Sox2 primary antibodies overnight. Subsequently, Alexa 488-conjugated goat anti-chick and Alexa 647-coupled goat anti-rabbit secondary antibodies were applied 5 h at room temperature. After PBT washes, detection of the HRP-coupled antibody was achieved by applying Cyanine3 TSA reagent (Perkin Elmer) 30 min at room temperature. Embryos were mounted in mowiol after at least 1 h of PBT washes.
Sectioning. After staining, embryos were prepared as described in Andrieu et al. 64   www.nature.com/scientificreports/ exposures. Ten 0.5 μm z intervals were maximum z-projected and patched2 fluorescent signal was measured in a rectangle of 9 µm wide centred on the spinal cord lumen. To compare signal intensity profiles between different experimental conditions, data were normalised to the maximum value in each section analysed. Data for each embryo were collected from five trunk spinal cord sections and values represent the average of fluorescence intensity per section from five embryos for each experimental condition. Statistical analyses were performed using Prism5 software (GraphPad). Normality of data sets was tested using Kolmogorov-Smirnov's, D' Agostino and Pearson's and Shapiro-Wilk's tests using Prism5 (GraphPad). We considered datasets as normal when found normal in the three tests. Datasets following a normal distribution were compared with Student's t-test (unpaired two-tailed) while datasets that did not follow a normal distribution were compared using Mann Whitney's test (two-tailed). Patched2 intensity profiles were analysed using two-way ANOVA (two-tailed) multiple comparison test. For each experiment, the statistical test used is indicated in figure legends. Statistical significance was accepted at a p < 0.05. All data are expressed as mean values per embryo ± standard deviation (s.d).
Received: 21 August 2020; Accepted: 16 December 2020 Figure 9. Model for Sulf2a function. Scheme showing Sulf2a and Shh-dependent gene expression in ventral neural progenitors during neuronal production (24 hpf, a,a') and at the onset of gliogenesis (36 hpf, b,b') in wild-type (a,b) and sulf2a depleted (a' ,b') contexts. In 24 hpf wild-type embryos (a), the ventral-most progenitors of the p3 domain (green), adjacent to the medial floor plate (MFP, brown), express the highthreshold Shh responsive gene Nkx2.2a and generate V3 interneurons. Sulf2a (turquoise blue frame), expressed in dorsally located Olig2/pMN progenitors (red), limits the dorsal extent of high-threshold Shh response to prevent activation of Nkx2.2a and thus allows maintenance of Olig2 expression (red) required to orient these cells toward the MN lineage. When sulf2a is depleted (a'), the range of high-threshold Shh response extends dorsally, leading to dorsal misexpression of Nkx2.2a which in turn represses Olig2. Subsequent changes in the sizes of the p3 and pMN domains cause overproduction of V3 interneurons at the expense of MNs. In 36hpf wild-type embryos (b), the Shh source has expanded in the former Nkx2.2a/p3 domain to form the lateral floor plate (LFP, brown). Immediately adjacent progenitors, i.e. the ventral-most cells of the Olig2/pMN domain, because they activate high-threshold Shh response, upregulate Nkx2.2a. At that stage, Nkx2.2a no more represses Olig2, leading to formation of the p* domain (yellow), populated by progenitors co-expressing Olig2 and Nkx2.2a. Sulf2a expression (turquoise blue frame) is excluded from p* cells but maintained in dorsally located Olig2 progenitors (red) where the enzyme, again, prevents high-threshold Shh response. Then, the two distinct populations of Olig2 progenitors, expressing Nkx2.2a (yellow) or not (red) generate the myelinating (yellow) and non-myelinating (red) OPC, respectively. In 36 hpf sulf2a-depleted embryos (b'), LFP formation is not affected but, because of the dorsal expansion of the Nkx2.2a/p3 domain that occurred at earlier stage, the dorsally adjacent progenitors express Nkx2.2a but not Olig2 (green), in contrast to the wild-type situation. Nonetheless, Sulf2a depletion, again, allows activation of high-threshold Shh response that causes dorsal misexpression of Nkx2.2a. Subsequently, the p* domain (yellow), while positioned dorsally, forms properly but, the reduced size of the Olig2/pMN domain leads to a deficit in progenitors that only express Olig2 and, consequently to a strong reduction in the population of non-myelinating OPCs (red). It has to be noted that the population of myelinating OPC (yellow) is also mildly reduced possibly due to a secondary effect of aberrant neuronal populations (see discussion).