Foxj1a is expressed in ependymal precursors, controls central canal position and is activated in new ependymal cells during regeneration in zebrafish

rsob.royalsocietypublishing.org Research Cite this article: Ribeiro A, Monteiro JF, Certal AC, Cristova ˜o AM, Sau ´de L. 2017 Foxj1a is expressed in ependymal precursors, controls central canal position and is activated in new ependymal cells during regeneration in zebrafish. Open Biol. 7: 170139. http://dx.doi.org/10.1098/rsob.170139 Received: 7 June 2017 Accepted: 27 October 2017 Subject Area: developmental biology/cellular biology/ neuroscience Keywords: Foxj1a, Shh, ependymal radial glia, CSF-contacting neurons, spinal cord injury, zebrafish Authors for correspondence: Ana Ribeiro e-mail: a[email protected] Leonor Sau ´de e-mail: [email protected].pt Electronic supplementary material is available online at https://dx.doi.org/10.6084/m9. figshare.c.3925786. Foxj1a is expressed in ependymal precursors, controls central canal position and is activated in new ependymal cells during regeneration in zebrafish Ana Ribeiro1, Joana F. Monteiro3, Ana C. Certal3, Ana M. Cristova ˜o1 and Leonor Sau ´de2 1 Instituto de Medicina Molecular, and 2 Instituto de Medicina Molecular e Instituto de Histologia e Biologia do Desenvolvimento, Faculdade de Medicina da Universidade de Lisboa, 1649-028 Lisboa, Portugal 3 Champalimaud Research, Champalimaud Centre for the Unknown, Lisboa, Portugal AR, 0000-0003-0132-9740; JFM, 0000-0002-0057-7307; ACC, 0000-0002-5091-0083; LS, 0000-0001-5933-8872 Zebrafish are able to regenerate the spinal cord and recover motor and sensory functions upon severe injury, through the activation of cells located at the ependymal canal. Here, we show that cells surrounding the ependymal canal in the adult zebrafish spinal cord express Foxj1a. We demonstrate that ependymal cells express Foxj1a from their birth in the embryonic neural tube and that Foxj1a activity is required for the final positioning of the ependymal canal. We also show that in response to spinal cord injury, Foxj1a ependymal cells actively proliferate and contribute to the restoration of the spinal cord structure. Finally, this study reveals that Foxj1a expression in the injured spinal cord is regulated by regulatory elements activated during regeneration. These data establish Foxj1a as a pan-ependymal marker in development, homeostasis and regeneration and may help identify the signals that enable this progenitor population to replace lost cells after spinal cord injury. 1. Introduction The spinal cord develops from the embryonic neural tube, an initially homogeneous neuroepithelium. During development, progenitor cells in the neural tube acquire different identities depending on their position along the dorsoventral axis. The roof plate on the dorsal side (via BMPs and Wnts) and the floor plate on the ventral side (via Shh) provide positional information that assigns cells to distinct progenitor domains [1,2]. Each progenitor domain can give rise to several cell fates that emerge sequentially. The initial step in neuroepithelial patterning specifies different neuronal subtypes. Soon after, the progenitor region switches from a neuronal to a glial fate and begins generating oligodendrocytes and astrocytes that support neurons [3]. After the differentiation and migration of neurons and glial cells from the neuroepithelium, the remaining cells now lining the central canal give rise to an additional cell fate—ependymal cells. Ependymal cells (i.e. all the cells that contact the central canal) descend predominantly from the ventral progenitor domains p2 and pMN [4,5] and the floor plate [6], and require the ventral signal Shh for their correct specification [7]. As the remnant of the embryonic neuroepithelium, the ependymal region in the adult spinal cord retains not only the pseudo-stratified epithelial organization, but also a stem/progenitor character. These cells have neurosphere-forming capacity and are able to give rise to neurons, astrocytes and oligodendrocytes in vitro, both in mouse [8,9] and human [10,11]. By contrast, in vivo ependymal cells display a more restricted lineage potential, in the context of spinal cord injury. Upon damage to the spinal cord ependymal cells show a strong &2017 The Authors. Published by the Royal Society under the terms of the Creative Commons Attribution License http://creativecommons.org/licenses/by/4.0/, which permits unrestricted use, provided the original author and source are credited. Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 proliferative response, but fail to generate neurons, forming mostly astrocytes that incorporate the glial scar and a small number of oligodendrocytes [9,12]. The discrepancy between the in vitro and in vivo data suggests that neural stem cells in the ependymal region have the potential to replace all lost cells, including neurons, but the neuronal fate is inhibited by the microenvironment in the injured spinal cord. Therefore, if these inhibitory signals were removed it could be possible to direct the endogenous stem cells towards a neuronal lineage. In contrast to mammals, adult zebrafish are able to efficiently regenerate the spinal cord due to its ability to regrow damaged axons and form new neurons, while avoiding the formation of a glial scar [13]. Injury-induced neurons arise from the ependymal region [14], suggesting that stem/progenitor cells in the zebrafish spinal cord have a wider regenerative potential than in mammals and could be used to identify the signals that help promote the neuronal fate. The study of the behaviour of zebrafish ependymal cells during regeneration would be facilitated by the use of ependymal-specific molecular markers, but these are limited in zebrafish. A good candidate is Foxj1, which is specifically expressed by ependymal cells in the mouse spinal cord [9] and is also detected in the ependymal zone in the human spinal cord [15]. Moreover, Foxj1-expressing cells were shown to enter into a proliferative state after spinal cord injury, contributing mainly to astrocytes in vivo but with the ability to differentiate into other cell types, including neurons, when cultured in vitro [9]. In zebrafish, the Foxj1 homologue— Foxj1a—is expressed in the floor plate of the developing spinal cord [16] and was shown to be elevated after injury in embryos [17]. Yet, the cellular details of the Foxj1a distribution and response to injury in the adult spinal cord were not explored. In this study we determined if Foxj1a can be used to identify ependymal cells and whether Foxj1a-expressing cells participate in the repair of the spinal cord. We report that Foxj1a expression in the ependymal region is conserved in zebrafish and accompanies ependymal cells from their genesis until adulthood. We also show that Foxj1 activity is important for the formation of the central canal, through the modulation of the Shh signalling pathway. Moreover, we confirm that Foxj1a-positive cells expand in response to injury through a Shh-dependent mechanism and contribute to the restoration of the spinal cord structure in zebrafish. 2. Results 2.1. Adult zebrafish ERGs express Foxj1a To determine if Foxj1a is expressed in the adult zebrafish spinal cord, we used the reporter transgenic line Tg(0.6foxj1a:GFP),in which a small enhancer sequence drives the expression of the fluorescent protein GFP [18]. This transgenic line was shown to reproduce Foxj1a endogenous expression in several tissues, including the developing neural tube. In transverse sections of the adult spinal cord, the Foxj1a reporter labelled most cells surrounding the central canal—ependymal cells (figure 1a,a0). The majority of GFP þ ependymal cells had ependymo-radial glia (ERG) morphology [19], with projections that extended to the pial surface (figure 1a–b0). We also confirmed by fluorescent in situ RNA hybridization (FISH) on transgenic sections that the distribution of the foxj1a:GFP reporter is similar to the endogenous foxj1a gene, which is also detected in the cells surrounding the central canal (figure 1c–c000). A general feature of ependymal cells is the presence of cilia, and Foxj1(a) is associated with the formation of motile cilia [16,20–22]. To determine if Foxj1a þ ependymal cells in zebrafish are ciliated, we labelled cilia with acetylated a-tubulin, which revealed long cilia projecting into the central canal (figure 1d,d0). At a closer view, we detected one or two cilia per ependymal cell (electronic supplementary material, figure S1). Transmission electron microscopy (TEM) revealed that most cilia at the apical edge of cells around the central canal had a central pair of microtubules (n¼21/25) (figure 1e, arrow), as well as outer dynein arms (figure 1e, arrowhead). These results confirm that cells lining the central canal have long and likely motile cilia, features that are consistent with the expression of Foxj1a. The adult zebrafish ependymal region was previously shown to retain an embryonic-like pattern of expression of several progenitor markers [23]. To assess the dorsoventral identity of Foxj1a þ cells, we tested the expression of ventral marker Nkx6.1. This analysis showed the expression of Foxj1a in both the Nkx6.1 þ ventral half of the ERGs and in the Nkx6.1-negative dorsal region that corresponded to a Pax6 þ domain (figure 1f,f0). Foxj1a expression was also more widespread in the ependymal region than that of other molecular markers associated with radial glial cells (GFAP and vimentin; electronic supplementary material, figure S2). Foxj1a þ ERGs were also in close proximity to blood vessels (BVs) (figure 1g; electronic supplementary material, figure S2), which provide support to neural stem cell niches [24,25]. These data settle Foxj1a as a general ependymal marker that labels a heterogeneous group of immature/progenitor cells. Although the majority of cells surrounding the central canal had a radial glial morphology, a subset of foxj1a:GFPexpressing cells displayed a more rounded morphology with a bulbous apical attachment (figure 1h00). This morphology is reminiscent of cerebrospinal-fluid-contacting neurons (CSF-cNs) [26]. To test if Foxj1a is expressed in CSF-cNs, we examined the expression of the early neuronal marker HuC/D and a CSF-cN-specific marker, polycystic kidney disease 2-like 1 (PKD2L1) [26]. This analysis showed that foxj1a:GFP þ cells with CSF-cN morphology express HuC/D (figure 1h–h000)andpkd2l1 (figure 1i–i00), indicating that Foxj1a is expressed not only in ERGs, but also in CSF-cNs. This initial characterization of Foxj1a expression in the adult zebrafish spinal cord demonstrated that, as in mammals, Foxj1a is expressed in the cells that line the central canal, which we broadly term as ependymal cells. The Foxj1a þ population is composed of various cell types, including different subtypes of radial glial progenitors and more differentiated CSF-cNs (figure 1j). However, despite the common expression of Foxj1, zebrafish and mammalian ependymal cells show some differences, namely in the distribution of markers such as GFAP (absent in mouse) and vimentin (more widespread in mouse) and in the proportion of cells with radial morphology (lower numbers in mouse) [9]. 2.2. Foxj1a is expressed in the ERG precursors in the developing neural tube We next asked if Foxj1a plays a role in the assembly of the ependymal region during neural tube development. Foxj1(a) is expressed in the embryonic floor plate [16,27] and roof plate [28]. In zebrafish, floor plate and roof plate cells participate in rsob.royalsocietypublishing.org Open Biol. 7: 170139 2 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 GFP Ac. a-tubulin TEM CSF-c neurons Foxj1a+ Hu+ pkd2l1+ dorsal cells Foxj1a+ ventrallateral cells Foxj1a+ Nkx6.1+ foxj1a:GFP DAPIfoxj1a:GFP foxj1a DAPI foxj1a:GFP Ac. a-tubulin DAPI foxj1a:GFP Nkx6.1 DAPI foxj1a:GFP flk1:mcherry DAPI foxj1a:GFP pkd2l1 DAPI foxj1a:GFP Hu DAPI GFP GFP GFPHu foxj1a GFP blood vessels (a) (a¢) (b¢) (c)(d) (b) (e) (f) (g)(h)(i)(j) (c¢¢) (c¢) (c¢¢¢) (d¢) (h¢) (i¢) (i¢¢) (h¢¢) (h¢¢¢) (f¢) Figure 1. Foxj1a is expressed in ependymal cells in the zebrafish adult spinal cord. (a–b0) Confocal stack projection of a spinal cord transverse section (a,a0) and sagittal section (b,b0) in adult Tg(0.6foxj1a:GFP) transgenic zebrafish. (c–c000)FISHoffoxj1a (magenta) in a transverse section of a spinal cord expressing the foxj1a:GFP reporter (green), showing a similar pattern of expression. (d,d0) Immunostaining with acetylated a-tubulin (magenta) to label cilia present on the apical surface of Foxj1a-expressing cells (green). (e) TEM image of a cilium at the apical ependymal region, with a central pair of microtubules (arrow) and outer dynein arms (arrowheads). ( f,f0) Ventral foxj1a:GFP þ cells express the progenitor marker Nkx6.1 (magenta). (g) BVs (mCherry þ endothelial cells in magenta) are present in the proximity of GFP þ ERGs (arrowheads). (h) Pan-neuronal marker HuC/D (Hu, magenta) immunostaining and foxj1a:GFP expression in a spinal cord transverse section. (h0–h000) Magnification of the central canal highlighting Hu/Foxj1a double positive cells (arrowheads). (i–i00)FISHofpkd2l1 (magenta) showing co-expression with foxj1a:GFP (arrowheads) in the ependymal region. ( j) Scheme of the spinal cord ependymal region showing subtypes of Foxj1a-expressing cells, based on molecular expression and cell morphology. The term ependymal cells is not consensual, as other authors suggest that adult zebrafish spinal cords only contain ERGs. DAPI-labelled nuclei are shown in grey. Scale bars: 20 mmina,b,f–h;10mminc000;5mmind; 100 nm in e; 10 mmini. rsob.royalsocietypublishing.org Open Biol. 7: 170139 3 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 contraction of the primitive lumen to form the central canal [29], which occurs between 48 and 72 hours post-fertilization(hpf) as neural precursors exit the ventricular zone [30]. To determine if Foxj1a is expressed in the cells that help form the central canal, we examined the behaviour of foxj1a:GFP þ cells during this embryonic period. The tightjunction protein ZO-1 was used to reveal the outline of the primitive lumen. Expression of the reporter transgene was detectable in the floor plate at 24 hpf (figure 2a,a0) and continued to be expressed in the ventral midline at the subsequent time-points analysed (figure 2b–f0). At 48 hpf foxj1a:GFP started to be expressed in roof plate cells (figure 2b,b0), indicating that the expression pattern observed in chick is conserved in zebrafish [28]. Endogenous foxj1a expression in the roof plate was confirmed by FISH in sections of 54 hpf embryos (figure 2h). However, the analysis of the distribution of foxj1a transcripts also uncovered a domain of foxj1a expression that was not reproduced by the reporter transgene—a region of strong expression in the middle of the neural tube (figure 2h; electronic supplementary material, figure S3). Nevertheless, the reporter could still be used to follow the behaviour of the roof plate and floor plate during lumen contraction. At 52 hpf, the roof plate started to extend as the primitive lumen started to contract, and the apical projection of roof plate cells continued to stretch until the conversion of the lumen into the central canal was completed by 72 hpf (figure 2c–e0). By contrast, the floor plate showed little change during this process. The quantification of the size of the lumen in different embryos showed that the reduction of the lumen reproducibly occurred from 48 to 72 hpf, with a dorsal-to-ventral zipping mechanism (figure 2g). After lumen contraction at 72 hpf, the size of the ventral and dorsal regions remained constant until the last time-point analysed, 120 hpf (figure 2f,f0). The initiation of Foxj1a expression in the neural tube also coincided with the appearance of long cilia. When Foxj1a started to be expressed in the floor plate (24 hpf) and the roof plate (48 hpf), GFP þ cells exhibited long cilia (figure 2i–j0). The correlation between Foxj1a expression and appearance of long cilia is consistent with the role of Foxj1a in the regulation of motile ciliogenesis. 24 hpf 48 hpf 52 hpf 54 hpf ** 56 hpf 60 40 20 distance from lumen centre (mm) 0 –20 –4020 30 40 50 60 70 hours post-fertilization 80 90 100 110 120 dorsal region lumen ventral region foxj1a:GFP ZO-1 DAPI foxj1a DAPI foxj1a:GFP Ac. a-tubulin DAPI 72 hpf 120 hpf 24 hpf 52 hpf (a) (d) (g) (e)(i) (j) (j¢) (j¢) (f) (b) (c) (h) (a¢) (d¢) (e¢) (f¢) (b¢) (c¢) Figure 2. Foxj1a þ cells participate in the formation of the spinal cord central canal. (a–f0) Representative images of the neural tube region in transverse sections of Tg(0.6foxj1a:GFP) transgenic zebrafish embryos/larvae ranging from 24 to 120 hpf. The apical edge of the cells surrounding the lumen is identified by ZO-1 immunostaining (magenta) and the GFP reporter labels Foxj1a-expressing cells (green). (g) Quantification of lumen closure from 24 to 120 hpf. The positions of the floor plate, ventral and dorsal points of the lumen and roof plate are normalized to the middle point of the lumen. Sample number: 24 hpf (n¼12); 48 hpf (n¼6); 52 hpf (n¼9); 56 hpf (n¼9); 72 hpf (n¼11); 120 hpf (n¼8). (h) Confocal image of a FISH of foxj1a (magenta) in a 54 hpf embryo. foxj1a is expressed in the floor plate, ventro-lateral cells, roof plate (arrowhead) and pronephric (asterisks). (i,j,j0) Long cilia (arrowhead), labelled by acetylated a-tubulin (magenta), are present in Foxj1a-positive floor plate cells at 24 hpf (i) and in Foxj1a-expressing roof plate cells in the roof plate at 52 hpf ( j,j0). DAPI-labelled nuclei are shown in grey. Scale bars: 10 mmina–j;5mminj0. rsob.royalsocietypublishing.org Open Biol. 7: 170139 4 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 These results show that Foxj1a is expressed in ERG precursors in the floor plate, roof plate and in cells close to the lumen in the middle region, which continue to surround the central canal after lumen contraction. These data also revealed that the foxj1a:GFP reporter transgene is expressed only in a subset of the endogenous foxj1a domains (floor plate and roof plate) in the neural tube, indicating that the enhancer in the construct lacks some of the foxj1a regulatory elements. 2.3. Loss of Foxj1a activity affects the timing of lumen contraction To test if Foxj1a is required for the transition from primitive lumen to central canal, we examined the morphology of the lumen (visualized with ZO-1) in larvae injected with antisense morpholino oligonucleotides (MO) designed to block Foxj1a protein translation by targeting the start codon [16,31,32] (electronic supplementary material, figure S3). We first analysed larvae at 3 days post-fertilization (dpf), which showed a round and narrow lumen when injected with a standard control MO (figure 3a,a0). By contrast, foxj1a MO-injected larvae exhibited an elongated lumen, indicating that lumen contraction was delayed when compared to the controls (figure 3d,d0,g,g0). The degree of the delay varied between morphant larvae, but the size of the lumen was significantly larger and the dorsal region significantly smaller when compared to controls (figure 3j,k). These data suggest that the delay in lumen contraction observed in morphants results from a lagging dorsal zipping. To determine if foxj1a morphants are able to complete the zipping of the lumen, we analysed MO-injected larvae at 5 dpf. At this stage control larvae displayed a small and ventrally positioned central canal (figure 3b,b0). In morphant larvae, the size of the lumen was equivalent to controls, but the position of the canal was altered (figure 3j,k). In foxj1a morphants with a milder phenotype, the position of the central canal was only modestly dorsalized (figure 3e,e0), but in more severe phenotypes the canal was displaced to the middle of the spinal cord (figure 3h,h0). The measurement of the size of the ventral region showed that floor plate cells in the morphant had stretched significantly (figure 3j,k). These results indicate that the lumen size phenotype observed in 3 dpf morphants is caused by a delay that ultimately affects the final position of the canal. The co-injection of the foxj1a MO with a MO-resistant form of the foxj1a mRNA resulted in the rescue of the curved body phenotype and the position of the central canal in a subset of larvae (electronic supplementary material, figures S4 and S5). Moreover, the injection of Cas9/ gRNA complexes targeting the foxj1a locus resulted in a small number of larvae with a curved body and a central canal shifted dorsally (electronic supplementary material, figure S6). Together, these data argue that the morphant phenotype is specific to Foxj1a downregulation. To determine if the dorsalization of the central canal in foxj1a morphants affects the identity of the cells encircling the canal, we analysed the expression of the ventral marker Nkx6.1. At 5 dpf, Nkx6.1 was expressed in cells lining the ventral side of the central canal in both control and morphant larvae (figure 3c,c0,f,f0,i,i0), suggesting that the dorsoventral identity of ependymal cells was not changed by the loss of Foxj1a function. Together, these results indicate that Foxj1a activity influences the position of the central canal, but not the specification of the ependymal precursor cells surrounding the canal. 2.4. Foxj1a attenuates Shh activity and prolongs the proliferative phase The lumen closure phenotype observed in foxj1a morphants was reminiscent of larvae lacking the microRNA miR-219 [30]. In miR-219 morphants, the extended lumen was associated with enhanced Shh activity and Shh-induced proliferation [33]. To test if Foxj1a also affects the activity of the Shh pathway, we assayed the expression of the Shh target gene patched2 [34] by FISH in 54 hpf embryos, when the lumen is closing. Standard MO-injected larvae presented a ventral-to-dorsal accumulation of patched2 transcripts near the lumen (figure 4a,c), which was increased in foxj1a morphants (figure 4b,d). This increase in the expression of a Shh target gene in the absence of Foxj1a is consistent with the hypothesis that Foxj1 dampens Shh signal transduction [27]. The increase in patched2 expression is not a result of increased shh levels in the floor plate of foxj1a morphants, as assessed by FISH (electronic supplementary material, figure S7). The Shh-signalling pathway has both patterning and mitogenic roles in the neural tube [35] and the attenuation of Shh activity is required for progenitor cells to exit the cell cycle and differentiate [33]. To test ifthe loss ofFoxj1a affectsthe transition from proliferation to differentiation, we quantified the number of mitotic cells (labelled by phospho-histone H3(pHH3)) in control and morphant 54 hpf embryos. At this stage, very few mitotic cells were detected in the neural tube of Standard MO embryos (0.1744+0.0437 cells/section) (figure 4e,g), but in foxj1a morphants the number of pHH3 þ cells was significantly increased (1.25+0.1326 cells/section) (figure 4f,g). This result suggests that a decrease in Foxj1a activity leads to an extended period of neural progenitor proliferation, likely due to the ectopic activation of the Shh pathway. This delay in the transition to differentiation may explain the delay in lumen closure observed in Foxj1a morphants. Together, these data propose a role for Foxj1a in the triggering of progenitor cell cycle exit and induction of lumen contraction, through the modulation of the Shh signal transduction. 2.5. Foxj1a-expressing cells participate in the regeneration of the spinal cord To determine if adult zebrafish Foxj1a þ ependymal cells contribute to the repair of the spinal cord after an injury, we examined the proliferative response of Fox1a þ cells after a compression injury protocol. In this protocol, the vertebral column was exposed and the spinal cord was compressed along the dorsoventral axis using forceps [36]. The site of the injury and the adjacent rostral and caudal regions (350 mm from the injury centre) were analysed in transverse sections. Proliferation was assayed using proliferating cell nuclear antigen (PCNA) antibody to label cells in G1 and S phases of the cell cycle [14]. In adult fish that had undergone only sham-injury, the number of PCNA þ cells was very low (figure 5a). The quantification of number and position of PCNA and GFPexpressing cells, plotted in a composite map of all samples, showed foxj1a:GFP þ cells around the central canal and very few PCNA þ cells (figure 5a0,h). At 3 and 7 days post-injury (dpi), the sections at the injury site were too disrupted to analyse; therefore, only the adjacent sections were quantified. At 3 dpi the number of proliferating cells had significantly increased (figure 5b,h) and dividing foxj1a:GFP þ ERGs rsob.royalsocietypublishing.org Open Biol. 7: 170139 5 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 around the central canal comprised almost 50% of all proliferating cells (47.05% +4.2%) (figure 5b0). At day 7 the number of proliferating cells had further increased, but the proportion of PCNA þ cells that expressed foxj1a:GFP was lower than at 3 dpi (23.5%+7%) (figure 5c,c0,h). The expansion of the ependymal area argued that new ependymal cells dorsal region Foxj1a_MO Standard_MO Standard_MO foxj1a:GFP ZO-1 DAPI foxj1a:GFP Nkx6.1 DAPI Foxj1a_MO ventral region 06040 size (mm) 20 3 dpf 3 dpf n=28 n=22 n=28 n=21 3 dpf 5 dpf 5 dpf 5 dpf 5 dpf n.s. n.s. **** **** ******** 5 dpf 3 dpf Standard_MO mild phenotypestrong phenotype Foxj1a_MO 5 dpf 5 dpf 60 –40 –30 –20 –10 0 10 20 30 distance from lumen centre (mm) 40 50 3 dpf 3 dpf lumen (k)(j) (a)(b)(c)(a¢)(b¢)(c¢) (g)(h)(i)(g¢)(h¢)(i¢) (d)(e)(f)(d¢)(e¢)(f¢) Figure 3. Neural tube lumen closure is perturbed in the absence of Foxj1a. (a–i0) Representative images of transverse sections of the neural tube of Tg(0.6foxj1a:GFP) transgenic larvae injected with Standard Morpholino (MO) (a–c0)orfoxj1a MO (d–i0) at one-cell stage and analysed at 3 and 5 dpf. foxj1a MO-injected larvae showed lumen closure phenotypes with variable degrees of severity, between mild (d–f0) and strong (g–i0), as highlighted by the apical ZO-1 immunostaining (magenta) (a,b0,d,e0,g,h0). (c,c0,f,f0,i,i0) Nkx6.1 þ cells (magenta) are still present lining the ventral half of the central canal in foxj1a morphants. DAPI-labelled nuclei are shown in grey. Scale bars: 10 mm. ( j) Quantification of lumen closure in Standard MO and foxj1a MO larvae at 3 and 5 dpf. The positions are normalized to the middle point of the lumen. Sample number is shown in the graph and includes data from three independent experiments. (k) Quantification of the size of the lumen and the ventral and dorsal regions, derived from the data shown in ( j). Each point represents one individual and the mean and s.d. bars are also shown. pvalues calculated using two-tailed unpaired t-test (****p,0.0001; n.s., not significant). rsob.royalsocietypublishing.org Open Biol. 7: 170139 6 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 were being formed, but not all new cells expressed foxj1a:GFP. The transgene was also downregulated in new neurons as they left the ependymal zone and remained expressed only in CSFcNs (electronic supplementary material, figure S8). At 14 dpi a large number of cells, mostly foxj1a:GFP-low or negative, was still proliferating at the injury epicentre (figure 5e,e0,h), but the number of dividing cells had decreased in adjacent regions (figure 5d,d0,h). By 30 dpi, the ependymal zone remained expanded, but the number of proliferating cells had considerably decreased, both far and close to the injury (figure 5f–g0,h). The dynamic behaviour of ERGs contrasted with the population of CSF-cNs in the ependymal region, which remained almost unchanged in response to injury (electronic supplementary material, figure S9). These data reveal that the Foxj1a þ ERG population is sensitive to injury and initiates a proliferative programme that contributes to the expansion of the ependymal area. However, the presence of GFP-low or negative cells in the ependymal region raised the possibility that Foxj1a expression was not induced in new ERGs or that the reporter transgene was not responsive in these cells because it lacks the necessary regulatory elements, as observed in the embryonic neural tube. 2.6. Shh signalling activity induces Foxj1a expression during regeneration To determine if foxj1a:GFP expression reproduced endogenous foxj1a in proliferating ERGs during regeneration, we compared the expression of the GFP reporter and the distribution of foxj1a transcripts in 7 dpi spinal cords. In sham-injury spinal cords, foxj1a:GFP labelled all cells around the central canal (figure 6a,a0). In adjoining sections, FISH for foxj1a revealed low levels of transcripts distributed around the central canal (figure 6b,b0). In injured spinal cords, the foxj1a:GFP reporter displayed high levels in the ventral and dorsal regions and low or absent expression in the middle region, where most PCNA þ cells were located (figure 6c,c0). The reverse pattern was seen for foxj1a transcripts: high transcript levels in the PCNA þ lateral cells and low signal in the dorsal and ventral regions (figure 6d,d0). These data show that Foxj1a is highly expressed by proliferating cells after an injury, confirming that Foxj1a þ ERGs actively participate in the repair of the lesioned spinal cord. In addition, these results reveal that the upregulation of foxj1a expression in response to injury is not replicated by the reporter transgene, suggesting that the injury-dependent expression of Foxj1a requires regulatory regions that are not present in the promoter sequence used in this transgenic line. The activation of Foxj1a expression in new ependymal cells during regeneration raises the question of whether the injuryassociated expression of Foxj1a is modulated by the signals used during development. Since in the developing neural tube Foxj1a expression in the floor plate is regulated by the Shh signalling pathway [16], we examined Foxj1a expression in lesioned spinal cords exposed to an inhibitor of the Shh pathway—cyclopamine. This experiment was carried out in larvae, which are more tractable than adults in terms of compound delivery and tissue penetrance [37–39]. We performed spinal cord transection in 5 dpf Tg(0.6foxj1a:GFP) larvae (figure 6e) and confirmed that foxj1a expression is also induced after spinal cord injury in 3 dpi larvae (figure 6g,i). Moreover, the distribution of foxj1a transcripts coincided with the accumulation of PCNA þ cells induced by the lesion (figure 6f,h), resembling the adult spinal cord. We next examined the role of the Shh pathway in the upregulation of foxj1a in the injured tissue. Injured larvae were treated with 200 mM of cyclopamine or an equivalent volume of DMSO from day 1 to 3 post-injury (figure 6j). The concentration of cyclopamine used was sufficient to inhibit injury-induced proliferation in larvae (electronic supplementary material, figure S10). Foxj1a expression was detected by FISH in spinal 54 hpf Standard_MO Foxj1a_MO 14/17 10/12 4**** 0 1 2 3 no. pHH3+ cells/section Standard_MO Standard_MO Foxj1a_MO Fox j 1a_MO 54 hpf foxj1a:GFP pHH3 DAPI patched2 patched2 expression pattern (a)(b) (c)(d)(g) (e)(f) Figure 4. Shh signalling and proliferation are enhanced in the absence of Foxj1a. (a,b) Expression of patched2 by ISH in 54 hpf embryos injected with Standard Morpholino (MO) (a)orfoxj1a MO (b). Two types of expression patterns were observed, which are schematically represented in (c,d). The majority of foxj1a morphants display increased patched2 expression in ventral cells surrounding the lumen. (e,f) Confocal images of transverse sections of Tg(0.6foxj1a:GFP) 54 hpf embryos injected with Standard MO (e)orfoxj1a MO (f) and immunostained against phospho-histone H3 (pHH3) to detect mitotic cells (magenta). The foxj1a:GFP reporter is shown in green and DAPI-labelled nuclei are shown in grey. Scale bars, 10 mm. (g) Quantification of the number of pHH3 þ cells per section (Standard MO, n¼ 43; foxj1a MO, n¼46). Data obtained from three independent experiments. Each point represents the average of two non-consecutive sections per embryo and the mean and s.d. bars are also shown. pvalue calculated using the two-tailed unpaired t-test (****p,0.0001). rsob.royalsocietypublishing.org Open Biol. 7: 170139 7 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 cord sections of injured larvae and the fluorescence intensity of foxj1a transcripts was quantified in cells along the dorsoventral axis (figure 6k). In DMSO-treated larvae foxj1a transcripts were present at high levels in cells lining the lumen (figure 6l), with peak intensity at middle/dorsal positions (figure 6n,o). By contrast, the fluorescence intensity levels of foxj1a:GFP PCNA DAPI sham 3 dpi 350 mm rostral 350 mm rostral 350 mm rostral injury centre 7 dpi 14 dpi DAPI GFPSP PCNASP PCNA&GFP (a)(b)(c)(d)(e) 150 DV position (mm) 100 50 0 –50 –100 –100 –50 0 50 100 LR position (mm) 150 DV position (mm) 100 50 0 –50 –100 150 70 350 mm from the injury injury centre 0 sham 3 dpi 7 dpi 14 dpi 30 dpi 14 dpi 30 dpi sham 3 d p i7d p i14d p i30d p i14d p i30d p i 40 30 no. cells/section no. cells/section PCNA and GFP PCNASP 0 10 20 10 20 30 40 50 60 * *** ANOVA: F= 14.86, P< 0.0001 (****) ANOVA: F= 6.968, P= 0.0009 (***) ** ** * *** *** n.s. n.s. n.s. n.s. n.s. 100 50 0 –50 –100 –100 –50 0 50 100 LR position (mm) –100 –50 0 50 100 LR position (mm) –100 –50 0 50 100 LR position (mm) injury centre 30 dpi 350 mm rostral –100 –50 0 50 100 LR position (mm) –100 –50 0 50 100 LR position (mm) –100 –50 0 50 100 LR position (mm) 150 100 50 0 –50 –100 150 100 50 0 –50 –100 150 100 50 0 –50 –100 150 100 50 0 –50 –100 (a’)(b’)(c’)(d’)(e’) foxj1a:GFP PCNA DAPI (f) (f¢) (g) (g¢) (h) Figure 5. Spinal cord injury in adult zebrafish triggers proliferation of Foxj1a þ ERGs. (a–g) Representative confocal images of transverse sections of spinal cords of Tg(0.6foxj1a:GFP) transgenic zebrafish from 3 to 30 dpi. The control spinal cord in (a) was collected 7 days after a sham injury. The sections were located at 350 mm rostral to the injury centre (b–d,f) or at the injury site (e,g). Proliferative cells (in magenta) were labelled using an antibody against PCNA and cells expressing the GFP reporter are shown in green. DAPI-labelled nuclei are shown in grey. Scale bars, 20 mm. (a0–g0) Composite maps of cell positions quantified in transverse section images of different spinal cords for each time point (sham, 7 dpi, 14 dpi: n¼6; 3 dpi, 30 dpi: n¼5). DV, dorsal/ventral; LR, left/right; SP, single positive. (h) Quantification of the number of PCNA SP (top plot) and PCNA þ /GFP þ (bottom plot) cells per section. Each circle represents one section at 350 mm rostral or caudal from the injury centre (left side of plot) or at the injury site (right side of plot). The mean and s.d. bars are also shown and a one-way ANOVA F-test was performed between groups, followed by a Dunnett test to compare every mean to the control mean (*p,0.05; **p,0.01; ***p,0.001; n.s., non-significant). rsob.royalsocietypublishing.org Open Biol. 7: 170139 8 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022 foxj1a transcripts were dampened in larvae exposed to cyclopamine (figure 6m), with significantly lower levels in the middle/ dorsal region when compared to the controls (figure 6n,o). This result suggests that Shh signalling activity is required for the injury-dependent expression of Foxj1a in ependymal cells. Together, these data reveal that activation of the Shh signalling pathway after injury triggers upregulation of Foxj1a expression and promotes proliferation in the regenerating spinal cord. 3. Discussion Ependymal cells were viewed simply as the lining of the central canal until the discovery of neural stem cells within this spinal cord population [8,10]. The identification of resident stem cells in the mammalian spinal cord led to a new therapeutic perspective in which ependymal cells could be used as an alternative to stem cell transplantation to help replace cell types lost after spinal cord injury [40]. However, further work is needed to understand how to overcome the astrocytic differentiation bias of these endogenous stem cells and to favour their differentiation into new neurons. Insight may be gained from the study of organisms that are able to spontaneously form new neurons and regain motor function after spinal cord injury, such as zebrafish [14]. Here, we identify Foxj1a as a general ependymal marker in zebrafish that can help follow the response of ependymal cells during spinal cord regeneration. Unlike other described molecular markers such as Sox2, GFAP and vimentin [41], Foxj1a is not restricted to subsets of ependymal cells. The expression of Foxj1a in the adult zebrafish spinal cord, together with the expression of other ependymal markers, reinforces the molecular similarities of stem/progenitor cells found in zebrafish and mammals. Nevertheless, when comparing our results obtained in zebrafish with the ones reported in mouse we could see differences in the ependymal region in terms of protein distribution (GFAP and vimentin) and radial glial morphology, which may be associated with differences in cellular function and could have an impact on their regenerative potential. We further show that Foxj1a is expressed by neurons that maintain attachments to the apical surface (CSF-cNs), indicating that Foxj1a labels not just ERGs but all cells contacting the central canal. This study also reveals that Foxj1a is expressed in ependymal cells not only in the adult spinal cord but from the time of formation of the central canal in the embryonic neural tube. We show that Foxj1a is induced in the floor plate and roof plate (a)(b)(a¢)(b¢)(c)(d)(c¢)(d¢) (e)(f) (g) (h) (i)(l)(m) (n) (o) (j)(k) sham uninjured (8 dpf) 3 dpi (8 dpf) SCI analysis 100 [80–100] DMSO Cyc DMSO Cyc [20–80] [0–20] 0 10 mm 60 0 n.s. 40 0 [0–20] [20–80] [80–100] relative p osition 10 20 30 100 relative position dorsalventral 908070 **** *** 605040302010 10 20 30 40 50 (n)(o) 5 dpf 1 dpi 2 dpi 3 dpi (8 dpf) 200 mM Cyc 200 mM Cyc DMSO cyclopamine 3 dpi 7 dpi foxj1a:GFP PCNA/BV DAPIfoxj1a:GFP PCNA DAPIfoxj1a DAPI foxj1a:GFP PCNA DAPI foxj1a fluorescence intensity (arb. units) foxj1a fluorescence intensity (arb. units) foxj1a DAPI foxj1a DAPI foxj1a:GFP PCNA/BV DAPI foxj1a PCNA DAPI foxj1a PCNA DAPI Figure 6. Endogenous Foxj1a expression is increased in response to injury in a Shh-dependent manner. (a–d0) Confocal images of transverse sections of Tg(0.6foxj1a:GFP) spinal cords immunostained against GFP (a,c) and FISH for foxj1a transcripts (b,d) in adjacent sections of sham-injury spinal cord (a,b) (n¼2) and 7 dpi spinal cord (c,d)(n¼8). The co-labelling with an antibody against PCNA (in magenta) shows that the majority of proliferative ERGs express foxj1a (d,d0) but not the foxj1a:GFP transgene (c,c0). BVs are also visible in (a) and (c) due to the expression of the transgene Tg(flk1:mCherry).(e) Illustration of the site of spinal cord transection in 5 dpf larvae. ( f–i) Representative confocal images of sections of 8 dpf Tg(0.6foxj1a:GFP) transgenic larvae, uninjured ( f,g)(n¼8) or 3 dpi (h,i)(n¼10). Adjoining sections were either immunostained against GFP (green) and PCNA (magenta) ( f,h) or processed with FISH for foxj1a transcripts (green) (g,i). ( j) Schematic of the cyclopamine (Cyc) treatment experiments in injured larvae. (k) Schematic of the neural tube region selected to quantify the fluorescence levels of foxj1a transcripts plotted in (n) and (o). (l,m) Representative confocal images of transverse sections of 3 dpi spinal cords of larvae treated for 2 days with DMSO (l) or 200 mM cyclopamine (m) in the medium. The sections were collected 100 mm rostral to the injury site and foxj1a transcripts were detected with FISH. (n) Fluorescence intensity (f.i.) profile of foxj1a transcripts along the dorsoventral axis (relative to the size of the neural tube) show reduced foxj1a levels in Cyc-treated larvae (blue) (DMSO: n¼14; Cyc: n¼13) (a.u., arbitrary units). The line represents the mean f.i. and shaded regions correspond to the standard deviation intervals. (o) Quantification of the average foxj1a f.i. in the regions shown in (k). The mean and s.d. bars are also shown in (n,o) and the pvalues were calculated using two-tailed unpaired t-test (n.s., not significant; ***p,0.001; ****p,0.0001). DAPI-labelled nuclei are shown in grey. Scale bars, 20 mm. rsob.royalsocietypublishing.org Open Biol. 7: 170139 9 Downloaded from https://royalsocietypublishing.org/ on 23 February 2022