Characterization of Proliferating Neural Progenitors after Spinal Cord Injury in Adult Zebrafish

Zebrafish can repair their injured brain and spinal cord after injury unlike adult mammalian central nervous system. Any injury to zebrafish spinal cord would lead to increased proliferation and neurogenesis. There are presences of proliferating progenitors from which both neuronal and glial loss can be reversed by appropriately generating new neurons and glia. We have demonstrated the presence of multiple progenitors, which are different types of proliferating populations like Sox2+ neural progenitor, A2B5+ astrocyte/ glial progenitor, NG2+ oligodendrocyte progenitor, radial glia and Schwann cell like progenitor. We analyzed the expression levels of two common markers of dedifferentiation like msx-b and vimentin during regeneration along with some of the pluripotency associated factors to explore the possible role of these two processes. Among the several key factors related to pluripotency, pou5f1 and sox2 are upregulated during regeneration and associated with activation of neural progenitor cells. Uncovering the molecular mechanism for endogenous regeneration of adult zebrafish spinal cord would give us more clues on important targets for future therapeutic approach in mammalian spinal cord repair and regeneration.


Introduction
Unlike fish and urodele amphibians which can regenerate their CNS in adult life, the adult mammalian central nervous system (CNS) shows rather limited capacity to regenerate after injury. Any spinal cord that undergoes successful regeneration would require rapid growth and proliferation leading to neurogenesis and axonogenesis. Moreover, injury induced tissue loss after CNS injury would require replenishment of lost cells both by neurogenesis and gliogenesis.
Neurogenesis in adult mammals is tightly restricted to the subependymal zone (SEZ) of the lateral wall of the ventricle and the subgranular zone (SGZ) of the hippocampus but
For TEM analysis spinal cord tissues were also fixed, sectioned and observed for TEM as described in [6]. Application of Fiji software was used to measure the nuclear and cytoplasmic lengths for identifying progenitor cells from TEM images.

Microscopy and cell quantification
Immunostained tissue sections were photographed by using an Olympus fluorescent microscope (model; BX 51) with ImagePro Express software. In all colocalization study, the quantification of cell was done from all optical images obtained by using a Zeiss LSM 510 Meta (inverted) confocal microscope. Six representative sections from 1 mm uninjured or injured from each spinal cord were averaged to compute percentage of particular cells and colocalized cells for each animal. In case of injured cord three representative sections from 500 μm both rostral and caudal side of injury epicenter were averaged for quantification. Cell counting was carried out manually using a cell counter application of ImageJ software. During the quantification process, cells were considered positive if cells were found stained with specific markers and also colocalized with DAPI. The colocalized cells and proliferating cells were considered positive only when both markers showed clear overlapping and cell nucleus overlapping with BrdU respectively. For quantifying NG2 + cells only bipolar single nucleated cells were considered positive as possible oligodendrocyte progenitors. Long, tubular, multi-nucleated cells were excluded in this study.

Immunoblotting
Zebrafish spinal cord was collected 7 and 10 days after injury along with the control, uninjured cords. Tissues were prepared in extraction buffer (37.5 mM Tris, 75 mM Nacl, 0.5% Triton X-100, protease inhibitor cocktail) and then the resulting tissue lysates were subjected to either 7.5% or 10% sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE). After the electrophoresis, proteins were transferred onto nitrocellulose membranes and subjected to western blotting. The western blots were developed with antibody anti-CNPase (1:1000, Millipore, USA), anti-MAG (1:100, Santa Cruz Biotechnology, USA), anti-OCT4 (1:200, Millipore, USA) and anti-NG2 (1:500, Millipore, USA) followed by anti-mouse or anti-rabbit alkaline phosphatase coupled secondary antibody (1:1,000 Jackson laboratory, USA). GAPDH was used as internal loading control protein. Protein bands were visualized using NBT/BCIP as substrate. PageRuler Broad Range Protein Ladder (Thermo Scientific, USA) was used as a standard molecular weight marker.

Enzyme linked immunosorbant assay (ELISA)
Analysis of Msx-1 protein by using ELISA assay was done according to [6]. Briefly, tissue extract of spinal cord tissue samples (65ug protein /well) were added to the wells of ELISA plate and incubated overnight at 4°C. Subsequently blocking buffer (1% BSA in PBS) containing primary antibody, anti-Msx1 (1:100, Santa Cruz Biotech, USA) were added and incubated for 1 hr at 37°C. After incubation with primary antibody, wells were washed thrice in washing buffer (0.5% BSA, 0.5% NE-40 in PBS) followed by incubation with secondary antibody (HRP conjugated anti-rabbit antibody (1:1000 dilutions) for 1 hr at 37°C. Several washes were given before adding the substrate TMB (tetramethyl benzidine). Color reaction was developed in dark and stopped by addition of 1M H 2 SO 4 . The optical density was measured at 450nm by using an ELISA reader (Bio RAD, USA). Quantitative real-Time RT-PCR (qRT-PCR): RNA was prepared from 3 biological replicates. For each biological replicate, spinal cord from about 15-20 fishes were pooled in both the control and in the regenerating samples. Control fishes and fishes possessing injury in the spinal cord were anesthetized deeply for 5 minutes in 0.1% tricaine (MS222; Sigma, USA) and approximately 1mm length of spinal cord from injury epicenter were dissected out (both rostral and caudal) from 15-20 fishes in each batch and pooled for RNA extraction. The extraction of RNA and qRT-PCR was done following the method mentioned in Hui et al., 2014. The specific gene primers like zebrafish sox2 (F 5' AGAGTTT CTCACTGAAGGTACATAG 3' and R 5' TGCCTCTGTTCGTTCTCTCA 3'), pou5f1 (F 5' AACTCCGAGAACCCTCAGGA 3' and R 5' TCGTTTTCTTTTTCGCGTGTCG 3'), msx1-b/msxB (F 5' GAACAGAGCACTTGGTCAAACTC 3' and R 5' TCCTGTTCGTCTTGT GCTTGC 3', msx-c (F 5' AGTGACAAGGGACAGTCCGGCT 3' and R 5' TAAACGGGGT CCGCGGTTTTCG 3'), msx1-a/msxe (F 5' TTCTGCAGCTGCCGGTGAAG 3' and R 5' TTTT CTCAGAGGGCACGCGG 3') the internal control gene β-actin1 (F 5' GCTCACCATGGAT GATGATATCGC 3' and R 5' GGAGGAGCAATGATCTTGATCTTC 3') were amplified in separate reaction tubes. No primer -dimers were obtained for either the target genes or beta-actin as assessed by melt curve analysis. The specificity of the products was also confirmed by melt curve analysis. The PCR cycles in all cases were started with Taq activation at 94°C for 5 mins and followed by final extension of 72ºC for 7 mins.

In situ hybridization tissue section
For in situ hybridization, pCRIITOPO vector (Invitrogen) containing pou5fl1 and Msx-b cDNA clones were obtained from Genome Institute of Singapore, zebrafish genome resources. Both sense and antisense RNA probes were generated using a digoxigenin (DIG) RNA labeling kit (Roche Diagnostics, Laval, Que´bec, Canada) following the manufacturer's instructions. Hybridization of sense and antisense probe were carried out on tissue section as mentioned by [18]. Photographs were taken under light microscope using an Olympus microscope (model; BX 51) and a Leica Microsystem microscope with build camera (DFC 290).

Statistical analysis
The statistical analyses were performed with Microsoft Excel (Office 2000) and all experimental data have been expressed as Mean ± s.e.m. and numbers of spinal cord samples used are mentioned in the corresponding legends of each figure. The data obtained were compared using unpaired two-tailed Student's t-test when comparing two groups and one-way ANOVA followed by Bonferroni post-hoc test when comparing more than two groups. The P values were determined using either GraphPad Prism or Statistica software. Error bars represent s.e. m. and statistical significance are represented as ÃÃÃ P 0.001; ÃÃ P 0.01; Ã P 0.05. P>0.05 was not considered as significant.

Results
Presence of proliferating progenitor: Identification of NSC like cells Replacement of tissue due to high proliferative response after SCI in zebrafish was observed by others [13] and also by us [6]. Presence of proliferating progenitor(s) in spinal cord can be demonstrated by transmission electron microscopy (TEM) analysis and immunohistochemical localization of proliferating cell markers. To characterize proliferating progenitors, we have analysed BrdU incorporation in injured cord along with Sox2, a marker for neural progenitor (Fig 1, S1 Fig) as suggested by many authors in different organism, such as zebrafish, axolotl and Xenopus [14,48]. We found that Sox2 is expressed only in few cells of the grey matter in adult uninjured spinal cord (Fig 1A and 1C). But the expression of Sox2 is upregulated in injured spinal cord as shown by time course analysis of sox2 mRNA and the highest level of expression is observed in 3 dpi cord (Fig 1D). Immunohistological analysis also confirms increased expression of Sox2 at 3 dpi cord and which sustains until 7 dpi, where almost all ependymal cells around the central canal are expressing Sox2 as shown in representative sections taken from injury epicenter of different injured cords (Fig 1F, 1J, 1N and 1R, S1A Fig). Increased proliferation of BrdU incorporation was first observed in 3 dpi cord followed by highest proliferation in 7 dpi and subsequent decrease in proliferation in 10 and 15 dpi cords [6]. A significant increase in Sox2 + /BrdU + cells were observed in both 3 dpi (p<0.01) and 7 dpi (p<0.001) cords compared to uninjured cord ( Fig 1S, S1 Table). Proliferating Sox2 + cells are mostly present in the grey matter, particularly in the ventricular zone around the central canal and in subependyma (Fig 1I, 1R and 1R1, S1B Fig). A few cells near pial membrane are also showing colocalization of Sox2 and BrdU ( Fig 1I). Increased number of proliferating Sox2 + cells in 3 dpi (8.34%, S1 Table) and 7 dpi (11.98%, S1 Table) cords suggest that these cells are proliferating progenitors. Although the number of Sox2 positive cells are more (data not shown) but the colocalized Sox2 + /BrdU + cells are less in 3 dpi cord compared to 7 dpi cord because the proliferation rate in 3 dpi cord is lower than that of 7 dpi cord. Sox2 + cells in grey matter are also colocalized with newly formed neurons as they are expressing HuC/D protein, approximately 6% and 9% Sox2 + cells are HuC/D + in 3 day and 7 day post injured cord respectively compared to uninjured cord where less than 0.5% Sox2 + cells are HuC/D + (Fig 1M and 1S S1 Table). Colocalization of Sox2 with ependymo-radial glial marker GFAP indicates that 5.69% and 5.81% of the Sox2 positive cells are GFAP + radial glia in 3 day and 7 day injured cord respectively ( Fig 1Q and 1S, S1 Table) as confirmed by comparing their morphological characters (Fig 1P and 1Q).
Ultrastructural analysis of injured cord showed existence of progenitor cells both in early stage like 3, 7 dpi cord and in late stage like 10 dpi cord (Fig 1T-1V). Progenitors were characterized by high nuclear-cytoplasmic index (also known as the nucleus-cytoplasm ratio, N:C ratio, or N/C index), a measurement by which a ratio of the size (i.e., volume) of the nucleus of a cell to the size of the cytoplasm of that cell is analysed. The N/C index indicates the maturity of a cell, because as a cell matures the size of its nucleus generally decreases [49,50]. For determining N/C index, the nuclear and cytoplasmic lengths were measured using the application of Fiji software for identifying progenitor cells from TEM images. We observed that at early stage, progenitors have very little cytoplasm but a large nucleus thus showing very high N/C index in 3 and 7 dpi cord (Fig 1T and 1U). The obvious presence of progenitors in late stage at 10 dpi cord ( Fig 1V) also showed high N/C index and can be recognized as these cells have dense chromatin pattern in nucleus. Other features are absence of cytoplasmic filaments but appearance of a few cytoplasmic organelles such as rough endoplasmic reticulum and Golgi apparatus [51][52][53]. Characterization of glial progenitors: Proliferating astrocyte and oligodendrocyte progenitors after SCI A2B5 gangliosides are one of the earliest markers for glial restricted precursors (GRP) during mammalian CNS development. These A2B5 expressing cells identify lineage restricted glial precursors, that exist in developing mouse neural tube, among embryonic stem (ES) cells and are similar to rat GRP [54]. These restricted precursors are known to generate oligodendrocyte and astrocytes but not neurons [55,56]. Furthermore, transplantation of GRPs, isolated from fetal spinal cord into injured cord, is capable of differentiating into astrocytic and oligodendrocytic lineages but not to a neuronal lineage [57].
NG2, a sulfated proteoglycan, is a marker of oligodendrocyte progenitor cells (OPC) [58,59] during CNS development. To identify the nature of proliferating glial progenitors in CNS injury, we have used both A2B5 and NG2 as markers along with BrdU incorporation (Fig 2, S2  Fig and S3 Fig). A2B5 is expressed in both uninjured and injured cord of a zebrafish, indicating the presence of restricted glial progenitor even in an adult cord. The A2B5 expressions are mostly located in white matter but very few cells are also in grey matter. Interestingly, few cells near the pial membrane are also expressing A2B5 (Fig 2A and 2D). The number of A2B5 immunoreactive cells decrease immediately after injury in 3 dpi cord, followed by a significant increase (approximately 8%) in 7 dpi cord compared to uninjured cord ( Fig 2B, S2 Table). In 10 and 15 dpi cords, the number of A2B5 immunoreactive cells gradually decreases. When we compare A2B5 immunoreactivity after giving BrdU incorporation, the percentage of dividing A2B5 + cells (A2B5 + /BrdU + ) are very low both in uninjured (~1%) and 3 dpi cord (~5%), but a significantly higher numbers (~16%, p<0.001) are present in 7 dpi cord ( Fig 2C). Similarly, we found very few A2B5 immunoreactive cells are colocalized with GFAP in uninjured cord, and many of the same population also exist primarily in the white matter of 7 dpi cord (S2A, S2B, S2C and S2D Fig). The NG2 immunoreactivity is found in elongated cells, many of which display characteristic bipolar morphology in both adult uninjured and 7 dpi cord ( Table). Similar to A2B5, expression of NG2 is also downregulated in 3 dpi cord ( Fig 2D) and significantly upregulated (~10%, p<0.01) in 7 dpi cord ( Fig 2Q) when compared to uninjured cord ( Fig 2D, S2 Table) and are mostly distributed in white matter ( Fig  2Q). Percentage of proliferating NG2 + cells (NG2 + /BrdU + ) cells in uninjured is minimal (~0.9%), whereas in 3 dpi cord nearly 3% cells are expressing both the markers. A relatively high number of proliferating NG2 + cells (~8%) are present in 7 dpi cord ( Fig 2D, S2 Table). Very few cells express both NG2 and A2B5 are also present in -injured cord (Fig 2E, 2F and 2F1, S2E and S2F Fig). Furthermore, TEM analysis revealed that there is presence of proliferating astrocyte progenitors in 7 dpi cord, with typical heterochromatinised body in the nucleus and electron lucent cytoplasm containing very few organelles (Fig 2Z; [60]). These data further supports our immunohistochemical analysis. We also observed the presence of mature oligodendrocyte in uninjured cord with smooth endoplasmic reticulum, microtubules and quantification of proliferating SOX2 + cells, SOX2 + /HuC/D + colocalized cells and SOX2 + /GFAP + colocalized cells in uninjured, 3 dpi and 7 dpi cord after crush injury. Values represent as mean ± s.e.m. (n = 5). Statistical significance represented as p value (Student's t-test; *p<0.01, **p<0.001). T) A small sized progenitor (red arrow) with compact nucleus in the injury epicenter of a 3 dpi cord section. U) Several newly formed cells (yellow arrowheads) in the injury site of a 7 dpi cord with progenitor phenotype. V) A newly formed progenitor cell with few cytoplasmic organelles (green arrows) in the injury site of a 10 dpi cord. 'NU' and 'CY' denote nucleus and cytoplasm respectively, white dotted line demarcate the boundaries of the nucleus and the cell. The values in bar represents Mean ± s.e.m. (n = 5), statistical significance represented as p value (Student's t-test; **p<0.01, ***p<0.001). 'cc' denotes central canal of the cord. Scale bar = 50 μm (A-C, F-Q), 20 μm (R), 10 μm (R1), 2 μm (T), 1 μm (U), 500 nm (V).  showing a proliferating astrocyte progenitor with a prominent dumb-bell shaped compact dividing nucleus (blue arrows) with characteristic heterochromatin body (HC) in the injury site. Z1) A newly formed oligodendrocyte with small cytoplasmic area where microtubules (MT, red arrows) are obvious and a few cytoplasmic organelles (green arrows) like primary lysosome (pLY) is present at the injury site of 10 dpi cord section. Note that axolemma (AXL) is in close axolemma in close vicinity of cytoplasm ( Fig 2Y) and newly formed oligodendrocyte(s) in regenerating cord (Fig 1Z) as revealed by their appropriate characters like electron lucent cytoplasm, few organelles like smooth endoplasmic reticulum, primary lysosome and presence of microtubules but absence of intermediate filaments unlike astrocytes [61].

Presence of Schwann cells in injured spinal cord
We observe the presence of a small population of NG2 immunoreactive cells, which are probably oligodendrocyte progenitors. Oligodendrocytes are known to have been involved in myelination of CNS axons. Our TEM analysis previously demonstrated presence of oligodendrocyte in uninjured cord [6]. Furthermore, there are many newly formed Schwann cells in 10 dpi cord adjacent to regenerating axons indicates involvement of Schwann cells in remyelination process as characterized by TEM in injured cords. To identify both myelinating and non-myelinating cells in injured cord, we have used several markers (Fig 3, S4A Fig), such as CNPase/MAG for myelinating cells and CNPase/GFAP for non-myelinating cells [62,63]. MAG and CNPase expressing cells shown restricted localization in the white matter of the cord and very few CNPase expressing cells are present in the uninjured cord (Fig 3A-3C). For obvious reason myelinating cell populations are greater in injured cord. This is quite evident as we observed significant upregulation (p<0.01) of both CNPase and MAG expression compared to that of uninjured cord (Fig 3D and 3E, S4A Fig). Electron microscopic observations revealed that there is presence of oligodendrocytes in the uninjured cord, but to our surprise in injured cord we see fewer oligodendrocytes without myelinated axon fibers ( Fig 3L) but many newly formed Schwann cells (Fig 3M). The basal lamina of these Schwann cells can be seen around the plasma membrane of unmyelinated axon fibers (Fig 3M) in injured cord. These Schwann cells are also shown with myelinated axon fibers, rough endoplasmic reticulum, microtubules, and small mitochondria [64] and are probably the Schwann cell progenitors involved in the remyelination process of denuded axons in regenerating cord ( Fig 3N).

Characterization of radial glia subtypes in uninjured and injured spinal cord
Analysis of cortical radial glial cells during development showed lineage heterogeneity, with subtypes of neurogenic and gliogenic radial glia [41]. Proliferating ventricular radial glia serves as precursor and can yield neurons after spinal cord injury [6,13,65,66]. To investigate the presence as well as the role of different subtypes of radial glia in injured cord, we have used different subtype markers of radial glia. Interestingly, all such markers like GLAST, BLBP and GFAP (Fig 4A, 4E, 4B and 4F) are expressed in uninjured spinal cord and the expression is confined in cells with characteristic radial glial morphology (Fig 4I, 4L and 4O). There are three different subtypes of radial glial population namely GLAST + /BLBP + , GLAST + /GFAP + and GFAP + / GLASTexists in uninjured cord (Fig 4D, 4D.1, 4H and 4H.1). The presence of radial glia as proliferating precursor was shown by using BrdU along with other radial glial markers in injured cord. The 7 dpi cord showed BrdU + radial glial cells in the ependyma expressed different markers like GLAST, BLBP and GFAP (Fig 4I-4K, 4L-4N and 4O-4Q respectively).
All these three subtypes of proliferating radial glia like GLAST + /BrdU + , BLBP + /BrdU + , GFAP + /BrdU + cells are present in uninjured cord and their numbers do not change   Progenitors in Regenerating Zebrafish Cord significantly in 3 dpi cord compared to uninjured cord. Interestingly, their numbers in 7 dpi cord are increased significantly (p<0.01 and p<0.001) and highest for all 3 phenotypes corresponding with highest proliferation rate (Fig 4S-4U). Ultrastructural analysis also confirmed presence of radial glia and many multi-ciliated ependymal cells around the central canal in the 7 dpi cord (Fig 4V and 4W). These radial glial cells with cytoplasmic processes contain intermediate filaments and glycogen granules and the ependymal cells have characteristic multiple cilia ending towards the central canal [67,68]. These cells also resemble to glial progenitors and may be implicated in migration of newly formed cells adjacent to glial process as shown in Fig 4X.

Expression of pluripotent stem cell like markers by proliferating cells after SCI
Based on our cDNA array analysis [18], we have identified 35 pluripotency related genes expressed in regenerating zebrafish spinal cord. Among these several pluripotency related genes which are also expressed in regenerating fin, limb and tail in different species, we observe upregulation of two common genes pou5f1 and sox2, in regenerating cord (S3 Table). Expression of pou5f1 which is a zebrafish homolog of mammalian OCT4 [69] has been documented at different time points after injury (Fig 5). Quantitative analysis by qRT-PCR corroborates well with in situ hybridization and immunohistochemistry observations (Fig 5).
The in situ hybridization analysis showed pou5f1 mRNA is abundant in neuron like cells localized in the subependymal zone in grey matter of uninjured cord (Fig 5A and 5A1). Interestingly, cellular pattern of pou5f1 expression in 7 dpi cord differs from uninjured one, as it is upregulated in different cell types after injury. High levels of pou5f1 transcripts are present in the injury epicenter, in 7 dpi cord. Pou5f1 mRNA expression is also found in the ependymal cells around the regenerating central canal, in newly formed neuron like cells and in other population of glial cells in white matter (Fig 5C and 5C1). In injured samples, a unique observation is the upregulation of pou5f1 expression in white matter which is more prominent near the epicenter and down-graded at the adjacent region. Expression at far adjacent region is near normal and seen only in grey matter as in uninjured cord (Fig 5F and 5G). The qRT-PCR analysis showed 2.8 fold increase of pou5f1mRNA expression in 7 dpi cord than uninjured cord (Fig 5D). However, pou5f1 transcripts are reduced in 3 dpi cord and 10 dpi cord (Fig 5D). The level of expression in 15 dpi is more or less similar to that of uninjured cord (Fig 5D). We have later characterized protein level expression of OCT4 by using antibody along with HuC/D, a neuronal marker and a proliferation marker, BrdU (Fig 5, S6 Fig and S4B Fig). We found that there are OCT4 + /HuC/D + neurons in the grey matter of uninjured cord (Fig 5I and 5I1). The number of OCT4 expressing cells increased significantly in 7 dpi cord, both in grey and white matter (Fig 5J  and 5J1) and this data also corroborates well with the in situ hybridization analysis of pou5f1. In uninjured cord, the neurons are OCT4 + /BrdU -, although a very few colocalized cells are in grey matter (Fig 5K and 5K1). A sharp rise in the number of OCT4 + /BrdU + cells in 7 dpi cord suggests OCT4 is associated with proliferating progenitors (Fig 5L and 5L1). In 7 dpi cord, many OCT4 + /HuC/D + colocalized cells are present in grey matter (Fig 5J1), and are similar to small newborn neurons as confirmed by observing the morphology and ultrastructure in our earlier analysis [6]. There are cells near subpial membrane which are also OCT4 + /BrdU + and a few OCT4 immunoreactive cells present in white matter, co-express A2B5 (Fig 5L and 5M).

Expression of Vimentin and Msx in neural progenitor
Our cDNA array data [18] refers to expression of several genes related to dedifferentiation mechanism and epithelial-mesenchymal (E-M) transition event. We have studied expression of vimentin [44] and Msx-1 [70][71][72] as markers for dedifferentiation and E-M transition [46,73].
Expression of Msx-1 was evaluated by different methodologies such as qRT-PCR, in situ hybridization (Fig 6) and ELISA (S4C Fig). All the three zebrafish homologs of msx genes (msx-b, msx-c and msx-e) showed similar expression pattern in injured state (Fig 6A-6C). A high level of msx-b expression was observed in the 3 dpi cord but decreased in 7 dpi cord ( Fig  6A) as revealed by qRT-PCR. Similarly, in situ hybridization analysis showed no expression of msx-b transcript in uninjured cord (Fig 6E), upregulation in 3 dpi cord, predominantly in grey matter cells and it was decreased in 7 dpi cord (Fig 6F-6H). Immunohistochemical localization shows presence of vimentin in the cells in ependymal layer around the central canal of uninjured cord (Fig 6I and 6I1), whereas in 7 dpi cord upregulation of vimentin expression is observed in the cells in ependymal layer forming the ependymal bulb-a structure widely recognized in injured cord in both urodele [43] and fish [6] ( Fig 6K and 6L). There is an initial loss of expression of vimentin in 3 dpi cord in the injury epicenter (Fig 6J), followed by an increased filamentous nature of expression in 7 dpi cord compared to uninjured cord (Fig 6K-6M). Expression of vimentin is associated with radial glia like cells around the central canal and with the mesenchymal cells around the ependymal bulb ( Fig  6L). Many of these vimentin positive cells are in proliferating state and hence colocalized with PCNA in the 7 dpi cord (Fig 6M and 6M1).

Discussions
Presence of proliferating neural progenitors CNS regeneration in zebrafish involves both axonal regrowth and reorganization of ependymal cells that proliferate, migrate and differentiate to give rise to the lost tissue. One of the primary responses following an injury is the proliferation that would eventually yield to both neuronal and glial cells. Adult CNS harbors a slowly dividing quiescent cell population but upon injury, these cell enters cell cycle, hence become transit amplifying and hence rapidly dividing [6,7,36]. Here we provide information on characters of different proliferating progenitors involved during neurogenesis and gliogenesis in the spinal cord of zebrafish before and after a crush injury.

Proliferating progenitors are Sox2 positive
Injury induced proliferation in fish cord occur both in grey as well as in the white matter of the cord and proliferation can be detected in 3 dpi cord, but the proliferative response peaks at 7 dpi cord [6]. Here we found that after injury almost all cells present around the ependyma are expressing Sox2 and some are expressing proliferation markers. Similarly, Sox2 expressing progenitor cell population also exists in fish optic tectum [14]. In the brain, these Sox2 + /PCNA + population of cells are capable of self-renewal and hence referred as NSC [35]. Attempt has been made, to identify and analyse NSC in fish brain to exploit their potential to become different types of neural cells [5,14,36]. Sox2 + cells are also thought to be neural progenitors in amphibian cord [30,48,74,75]. Radial glia, expressing GFAP, vimentin and aromatase are known to be involved in generation of new neurons [34,44,45,76,77] and providing vital support to migrating neurons [78,79]. Deletion of Sox2 in the axolotl cord resulted in defective proliferation of GFAP + cells. It has been suggested that Sox2 mediate expansion of neural stem cell pools are required for regenerating axolotl spinal cord [74]. Others reported Sox2 and GFAP are expressed in ependymo-radial glia in adult newt brain and they are in proliferating state [80]. Presence of both Sox2 + /GFAP + ependymo-radial glia and Sox + /NeuN + cells in subependyma have been demonstrated in axolotl spinal cord after transection injury [30]. We reported earlier the presence of Sox2 + /BrdU + cells in regenerating spinal cord, where we studied the expression of Sox2 to validate our cDNA array hybridization data [18]. In the present analysis, we have found that a significant rise of Sox2 + /BrdU + cells (8-12%) in injured cord suggesting the presence of NSC like proliferating progenitor during regeneration of adult zebrafish spinal cord. There is also a high fold increase (~5 fold) in sox2 mRNA level in injured cord (3 dpi) compared to uninjured cord. Further immunohistological analysis also supported our qRT-PCR analysis. Since we found that 6-9% of Sox2 expressing cells are also HuC/D positive, a marker expressed in newborn neurons, indicating that Sox2 positive cells are contributing to regenerative neurogenesis. [19] also reported upregulation of Sox2 in ependymal cells following transection injury in zebrafish spinal cord and suggested its role as proliferation initiator in ependymal cells as these Sox2 + cells also expressed PCNA, another proliferation marker at later time point of regeneration. Although they showed no evidence to suggest that these cells are expressing neuronal marker after injury and thus contributing to regenerative neurogenesis. While experiment involving regeneration competent Xenopus larvae demonstrated massive proliferation Sox2/3 + cells following injury, suggesting occurrence of neurogenesis as there are upregulations of several markers, such as doublecortin, α-tubulin [75].
The expression of Sox2 in our analysis identifies a particular progenitor and is associated with neurogenic state of the progenitor, which was not shown earlier in regenerating zebrafish spinal cord [19]. These proliferating progenitors indeed generate new neurons as confirmed by the neuronal marker. Identification of newly born neurons and their morphological resemblance also corroborates with our previous analysis of the cellular identity in regenerating cord [6]. Apart from HuC/D + neuronal cells, GFAP + radial glias also express Sox2. Expression Sox2 in radial glia is probably related to neurogenic potential of radial glia [41,81] as shown in regenerating axolotl cord [30,74].

Presence of different types of glial progenitors
We have identified glial restricted precursors (GRP) of both astrocytic lineage and oligodendrocytic lineage by using A2B5 and NG2 in fish cord. GRP are also found in developing mammalian neural tube and that can give rise to astrocytes and oligodendrocytes [55,56]. A2B5 also represents O-2A progenitors and defines an intermediate glial precursor in embryonic cord [82][83][84][85]. We have shown that the number of A2B5 cells decrease after injury probably because there is initial tissue loss immediately after injury. Although injury epicentre of 3 dpi cord shows presence of some cells because of cell migration and proliferation and it continues until 7 dpi cord, where we see plenty of accumulated cells at the injury epicentre. The number of proliferating A2B5 expressing cells increased significantly in 7 dpi zebrafish cord. These cells are mostly present in white matter and may probably generate cells of astrocyte lineage. We found a good percentage of A2B5 immunoreactive cells are also expressing GFAP, as observed by others in developing mammalian CNS [86]. The presence of astrocyte precursors in injured zebrafish cord was further confirmed by TEM, where we have shown the presence of proliferating astrocyte progenitors in 7 dpi cord.
Recently it has been reported that A2B5 + cell line generated from another teleost brain exhibited markers for both astroglia and oligodendrocytes [33], although we observed very few A2B5 + cells expressing NG2 -a marker for oligodendrocyte progenitor in injured cord. Others [87] reported that in adult human brain A2B5 + progenitors could undergo a limited number of cell divisions in vitro and can generate oligodendrocyte. Adult A2B5 + cells retain properties of progenitor cells when compared to post mitotic mature human oligodendrocytes and are more committed to the oligodendrocyte lineage than their fetal counterparts [88]. In adult teleost fish CNS, there is heterogeneity of stem cell; an overwhelming majority of adult stem cell in brain stem /spinal cord area is glia [89]. However, in regenerating zebrafish cord, only a minor population of A2B5 expressing cells colocalize with NG2. Our data also suggests that both A2B5 and NG2 expressing progenitor like cells exist in adult uninjured cord indicative of its embryonic character in the adult [90,91].
In mammalian SCI, upregulation of NG2 was observed in the oligodendrocyte progenitors and macrophages near injury epicenter [92]. In injured fish cord increase of both proliferating A2B5 and NG2 immunoreactive cells suggest injury induces proliferation of these glial progenitors. Both populations identify GRP like cells which are present predominantly in white matter. The cell type expressing NG2 in regenerating cord are few small cells and many elongated cells with long bipolar processes, unlike developing mammalian CNS, where large cells with multiple processes were observed [86]. It would be important to analyse the fates of various glial restricted precursors in fish spinal cord, since identity of these precursor may vary from mammalian counterpart, where new cells are generated from astrocytic stem cell [4,93,94]. Most importantly these glial precursors can respond to injury and participate in repair process in fish cord. Furthermore, unlike mammal, in teleost cord it has been reported recently that FGF dependent glial bridge formation facilitates axonal regeneration, suggesting that gliosis at injury site augments regeneration in fish cord [34].

Radial glia represents a heterogeneous population
We showed the presence of radial glia and its injury induced proliferation in adult cord [6], where we showed both slowly dividing and rapidly dividing cells. Among the radial glia in adult CNS, some are slowly dividing, at quiescent state, have capacity for self-renewal and while others can produce transient amplifying neuronal precursors during neurogenesis [95]. Therefore, dividing radial glia is not a homogeneous population. These cells are also associated with the remarkable ability of the CNS regeneration both in urodele and fish CNS [68,96]. Radial glial cells are neurogenic in a number of non-mammalian CNS both in adult and developing stage [79,97]. In this context it is important to mention that in mammalian CNS, radial glia can come from multipotent neural stem cells and represent the remnant of stem cells in adult CNS and can generate both astrocytes and neurons [98,99]. Adult zebrafish telencephalon harbors heterogeneous population of radial glia with respect to their rates of division and expression of different markers like GFAP, GLAST, BLBP and RC2 [5,9,10,41]. We have identified different subtypes of radial glia in zebrafish cord before and after a crush injury. In uninjured cord these cells are present surrounding the ependymal canal and express markers like GFAP, BLBP and GLAST. Following an injury, increased incorporation of BrdU in cells expressing these markers suggests that many radial glial cells had started to proliferate. These radial glial cells in the zebrafish cord have clear ependymal feature, and possess motile cilia as observed in our TEM analysis and confirmed by others [68].
At least three different types of radial glial populations namely GLAST + /BLBP + , GLAST + / GFAP + and GFAP + /GLASTare present in uninjured cord. We also confirm the heterogeneity among these glial populations even when these are in proliferating state after injury. Radial glia are neurogenic in developing mammalian CNS [95,100] and others have shown that differentiated glia can express high level of GFAP and activated glia loose GFAP expression. These cells can also dedifferentiate to generate neurons and radial glia following injury induced proliferation in zebrafish cord [13,34]. Approximately 8% of the proliferating cells in zebrafish spinal cord differentiate into newly generated motor neurons after transection injury [13]. The ependymo-radial glia or radial glia in adult zebrafish CNS display key features of stem cells, like slowly dividing quiescent population, self-renewal and generation of different cell types [6,13,101]. Furthermore, distinct domain of ependymo-radial glia in adult injured cord can give rise to different types of neurons as different domains are specified by different combination of transcription factors. For examples motor neuron progenitor (pMN) like domain would express olig2/nkx6.1/pax6 and would generate motor neurons [68,102]. All these evidence indicates involvement of radial glia in neurogenesis of regenerating zebrafish cord.
Vimentin, an intermediate filament protein, labels radial glia in developing cortex [79,81]. In regenerating urodele cord, it is considered to be a NSC marker [44] and associated with epithelial-mesenchymal transition [43]. Expression of vimentin by injury-reactive ependymal cells in urodele cord was interpreted as the creation of embryonic environment where we see dedifferentiation and return of multipotent state progenitors [44]. In zebrafish cord, we show similar expression of vimentin in radial glia situated in the ependymal layer. An increased level expression of dividing radial glia with vimentin in injured cord indicates that the most dividing progenitor radial glia express vimentin.

Expression of markers related to cellular dedifferentiation and reprogramming in regenerating cord
Cellular dedifferentiation is a common phenomenon during the regeneration of various organs [29]. Epimorphic regeneration in urodele amphibians and teleost fish involve dedifferentiation, where terminally differentiated cells can convert to undifferentiated or less differentiated progenitors [103][104][105]. In zebrafish, dedifferentiation represents a key phenomenon during regeneration of heart, retina and fin where blastemal cells show lineage restriction [106][107][108]. Increased expression of transcription factor msx-1 is responsible for driving the terminally differentiated state towards undifferentiated state through the process of dedifferentiation [21,109]. During the process of dedifferentiation, proliferation provides the basis for tissue regeneration and creation of new cell lineages [29]. Msx also keeps the blastemal cells in several organs in proliferating state [110][111][112]. Interestingly in Xenopus, Msx promotes spinal cord regeneration even during refractory period, from stage 45-47 [113]. Both msx-b (zebrafish homologue of Msx-1) and vimentin are upregulated in proliferating cells near or in the ependymal layer in zebrafish cord and thus suggesting a possible involvement of ependymal cells in the dedifferentiation process. Another possibility is the involvement of Schwann cell as a dedifferentiating cell population [29,106]. During development, Schwann cell precursor can dedifferentiate and redifferentiate to mature cells [114]. Similar presence of dedifferentiating and proliferating Schwann cells has been observed after nerve injury [115]. However, remyelinating Schwann cells in CNS are known to migrate from peripheral nervous system through spinal and cranial roots, meningeal fibres or autonomic supplies [116]. Presence of Schwann cell progenitors in regenerating zebrafish spinal cord is a unique feature, never been reported earlier and are involved in remyelination process as shown by TEM analysis. Although we cannot ascertain the origin of Schwann cells in injured spinal cord but we show presence of newly formed Schwann cells in the injured cord that may have generated through dedifferentiation and proliferation of mature Schwann cells. In future role of Schwann cell in regenerating cord need s to investigated, since these cells may play important role in creating regeneration permissive niche.
Another promising procedure to be exploited for regenerative therapies is cellular reprogramming where somatic cells can be converted into stem cells. Reprogramming process is known to involve four core transcription factors Oct4 (zebrafish homologue of Pou5f1), Sox2, cMyc and Klf4 [25,27,117,118]. It has also been suggested that dedifferentiation is a common cellular mechanism during induction of pluripotent stem cells and during appendage regeneration in fish and urodele amphibians [29,103,105]. Whether CNS regeneration in zebrafish is due to presence of resident stem cell population or due to reprogramming of mature cells is not clear. However during appendage regeneration, it has been suggested that both mechanisms i.e., dedifferentiation and activation of stem cell population are involved [119,120]. To understand the origin and fate of the new cells generated during CNS regeneration, the developmental potential (multipotent) of the suggested cell types needs to be explored. Expression of several reprogramming genes has been reported during regeneration of lens, fin and limb [25,121]. Among the four most commonly used reprogramming genes, we have shown upregulated expression of pou5f1 and Sox2 in neural progenitors which are probably involved in specifying neural progenitors during neurogenesis after injury. Similar role in neurogenesis for Sox2 has also been described in mouse [122] and zebrafish brain [11,36].
Our qRT-PCR analysis and cDNA array data indicate that there is increased expression of pou5f1 in regenerating zebrafish cord [18]. Others reported, expression of pou5f1 in normal and regenerating fins concluding that pou5f1 is required for fin regeneration but not crucial for blastema formation [25]. Three genes pou5f1, sox2 and msx-b all are required for the regenerative outgrowth [25]. We did not obtain any blastema like structure in regenerating cord because of the nature of injury but interestingly all these three genes are upregulated after injury suggestive of a common underlying genetic mechanism regulating the regenerative processes in this species. Pou5f1 is normally expressed in subependymal regions housed by neurons in uninjured cords suggestive of the fact that even uninjured zebrafish cord is neurogenic and pou5f1 is associated with neurogenesis [38,123]. Upon injury, OCT4 is associated with different kind of cell types initially but expression later becomes concentrated in newborn neuronal cells suggesting their role in neurogenesis. It is unlikely that it confers real pluripotency, although direct reprogramming of human neural stem cells by OCT4 has been reported in vitro [28,124].
In future, the involvement of pou5f1 and other genes in reprogramming of neural cells in zebrafish should be addressed by designing more thorough genetic and functional analysis. In the present context of understanding the mechanism of in vivo regeneration, we can interpret that regenerating spinal cord harbors multiple proliferating progenitors as identified by us using known markers of different progenitors. There are high levels of injury induced proliferation, both glia and neurons are generated from endogenous progenitor cells and different types progenitors are involved during regeneration of spinal cord. The genes related to cellular dedifferentiation are also expressed in various cell types. Further genetic dissection of endogenous regenerative potential of adult spinal cord would allow us to use zebrafish as an ideal model to facilitate the development of improved therapeutic strategy for CNS injury and neurodegenerative diseases.  Table. Quantification of A2B5 + , NG2 + , A2B5 + /BrdU + and NG2 + /BrdU + cells in uninjured and injured cord sections of various time points. Values represented as mean ± s.e.m. (n = 5), Statistical significance as p value (Student's t-test; ÃÃ p<0.01, ÃÃÃ p<0.001). (DOC) S3 Table. Comparison of pluripotency related genes expressed in various regenerating tissues in different species. (DOC) and Arijita Basu for running immunohistochemical analysis with vimentin antibody. We are grateful to Dr. S. Mukhopadhyay of University of Calcutta and Dr. David Gardiner, UC Irvine for critically reading the manuscript.