Injury-induced ctgfa directs glial bridging and spinal cord regeneration in zebrafish

See allHide authors and affiliations

Science  04 Nov 2016:
Vol. 354, Issue 6312, pp. 630-634
DOI: 10.1126/science.aaf2679

Spinal cord regeneration in zebrafish

Unlike humans, zebrafish can regenerate their spinal cord. Mokalled et al. identified a growth factor in zebrafish that helps this process (see the Perspective by Williams and He). The protein encoded by ctgfa (connective tissue growth factor a) is secreted after injury and encourages glial cells to form a bridge across the spinal lesion. Addition of this protein improved spinal cord repair in injured zebrafish.

Science, this issue p. 630; see also p. 544


Unlike mammals, zebrafish efficiently regenerate functional nervous system tissue after major spinal cord injury. Whereas glial scarring presents a roadblock for mammalian spinal cord repair, glial cells in zebrafish form a bridge across severed spinal cord tissue and facilitate regeneration. We performed a genome-wide profiling screen for secreted factors that are up-regulated during zebrafish spinal cord regeneration. We found that connective tissue growth factor a (ctgfa) is induced in and around glial cells that participate in initial bridging events. Mutations in ctgfa disrupted spinal cord repair, and transgenic ctgfa overexpression or local delivery of human CTGF recombinant protein accelerated bridging and functional regeneration. Our study reveals that CTGF is necessary and sufficient to stimulate glial bridging and natural spinal cord regeneration.

The spinal cord of adult zebrafish recovers spontaneously after injury (Fig. 1A and fig. S1, A and B). Efficient axon growth, adult neurogenesis, and absence of glial scarring distinguish this injury response from that in mammals (1, 2). After initial inflammation, ependymal cells proliferate and glia form a bridge that is thought to provide a scaffold for axonal growth (3). The severed cord reconnects, and new neuronal connections lead to functional recovery (Fig. 1A) (4). Here, we analyzed extracellular factors up-regulated in the regenerating zebrafish spinal cord.

Fig. 1 Identification of ctgfa from a screen for regulators of spinal cord regeneration.

(A) Schematic of the multistep process of spinal cord regeneration in zebrafish. (B) A screen for secreted factors expressed during spinal cord regeneration (fpkm, fragments per kilobase of transcript per million; GO, Gene Ontology). (C) In situ hybridization on spinal cord cross sections at 1 and 2 wpi and in uninjured control tissue. Sections proximal to the lesion from the rostral side are shown; dashed lines delineate the central canals. The canal dilates after injury. (D) ctgfa in situ hybridization on longitudinal spinal cord sections at 1 and 2 wpi and in uninjured control tissue. (E) ctgfa:EGFP reporter expression and GFAP immunohistochemistry during early bridging events at 5 dpi (top) and after bridge formation at 2 wpi (bottom). The middle panel shows a high-magnification view of the boxed area in the top panel. In (D) and (E), dashed lines delineate spinal cord edges, arrows point to sites of bridging, and arrowheads point to ventral ependymal cells. Scale bars, 50 μm.

To identify potential secreted, pro-regenerative signaling molecules, we screened zebrafish transcriptomes for genes induced after spinal cord injury (Fig. 1B and table S1). Our screen identified seven genes encoding secreted extracellular proteins, including fibronectin 1 a (fn1a), previously implicated in axon growth promotion (5, 6). Transcripts for fn1a, fn1b, connective tissue growth factor a (ctgfa), myostatin b (mstnb), and stanniocalcin 1 like (stc1l) increased in ependymal cells at 1 and 2 weeks post-injury (wpi) (Fig. 1C and fig. S1C). Whereas fn1a was expressed around the entirety of the central canal at 2 wpi, mstnb and ctgfa were predominantly expressed in the dorsal and ventral ependyma, respectively (Fig. 1C).

CTGF is a matricellular, multifunctional protein that can influence the activity of multiple major signaling pathways, affecting cell adhesion, migration, proliferation, and differentiation. CTGF expression increases after central nervous system (CNS) trauma in rodents, but its function after spinal cord injury has not been elucidated (79). Because ctgfa is induced upon spinal cord injury in zebrafish, we hypothesized that it may have pro-regenerative roles. We analyzed ctgfa expression along the rostrocaudal spinal cord axis (Fig. 1D). At 1 wpi, ctgfa transcription was induced in multiple cell types across the lesion site and in ependymal cells at the central canal near the lesion. At this time point, we detected strongest ctgfa expression in the ventral ependyma. By 2 wpi, ctgfa expression localized to ventral ependymal cells and marked the cellular bridge that had formed at the lesion site. Expression declined beyond 3 wpi (fig. S2A). Thus, ctgfa expression correlates with formation of the glial bridge.

To identify the cell populations that express ctgfa during spinal cord repair, we generated transgenic reporter zebrafish with a 5.5-kb genomic sequence upstream of the ctgfa translational start site fused to an enhanced green fluorescent protein (EGFP) reporter cassette. ctgfa:EGFP fluorescence resembled endogenous ctgfa mRNA expression in spinal cord tissue at 2 wpi (fig. S2, B and C). As early as 5 days post-injury (dpi), domains of ctgfa:EGFP and glial fibrillary acidic protein (GFAP), a marker of multiple glial cells in the CNS, overlapped within a subpopulation of glial cells at the injury site (Fig. 1E). We interpret these to be bridging cells. Similarly, ctgfa:EGFP colocalized with the GFAP+ bridge at 2 wpi (Fig. 1E). Comparison of ctgfa:EGFP and GFAP expression on serial cross sections revealed that ~97% of GFAP+ bridging glia at the lesion core were also EGFP+ at 1 wpi (fig. S3). Further away from the lesion, ctgfa:EGFP was mainly present in ventral ependymal cells (fig. S3). We also detected ctgfa:EGFP in skeletal muscle, bone cells, and reactive fibroblast-like cells around the lesion site. ctgfa-driven expression suggested delineation of “pioneer” bridging glia.

To determine whether ctgfa is required for spinal cord regeneration, we generated a ctgfa mutant allele (ctgfabns50; referred to as ctgfa) that harbors a frameshift-causing 7-nucleotide deletion within the third exon of the ctgfa locus (fig. S4, A to C). ctgfa−/− animals are adult viable and appear to have unaffected motor function capacity (10) (fig. S4D). However, ctgfa−/− animals showed diminished swim capacity after spinal cord injury, with no significant functional recovery between 2 and 6 wpi (Fig. 2A). Heterozygous (ctgfa+/−) animals showed partial recovery of swim capacity by 6 wpi (Fig. 2A). At 4 wpi, anterograde axon tracing indicated that axon regeneration across the lesion site was reduced by ~25% in ctgfa+/− and ~60% in ctgfa−/− spinal cords proximal to the lesion, and by ~40% in ctgfa+/− and ~80% in ctgfa−/− cords distally (Fig. 2B). Thus, ctgfa is required for spinal cord regeneration.

Fig. 2 ctgfa is necessary for glial bridging and spinal cord regeneration.

(A) Swim assays assessed animals’ capacity to swim against increasing water current inside an enclosed swim tunnel. Seven wild-type (ctgfa+/+), 10 ctgfa heterozygous (ctgfa+/−), and 10 mutant (ctgfa−/−) clutchmates were assayed at 2, 4, and 6 wpi. Statistical analyses of swim times are shown for ctgfa−/− (red) and ctgfa+/− (orange) relative to wild type. Recovery of ctgfa−/− animals was not significant between 2 and 6 wpi. (B) Anterograde axon tracing in ctgfa mutant animals at 4 wpi. For quantification of axon growth at areas proximal (shown in images) and distal to the lesion core, 16 wild-type, 17 ctgfa+/−, and 20 ctgfa−/− zebrafish from two independent experiments were used. (C) GFAP immunohistochemistry in ctgfa mutant spinal cords at 4 wpi. Percent bridging was quantified for 10 wild-type, 9 ctgfa+/−, and 10 ctgfa−/− clutchmates from three independent experiments. Dashed lines delineate glial GFAP staining; arrows point to sites of bridging. (D) Glial cell proliferation in wild-type, ctgfa+/−, and ctgfa−/− spinal cords at 1 wpi. Arrowheads indicate EdU-positive gfap:GFP-positive cells. For quantification of glial proliferation indices (left) and number of EdU-positive gfap:GFP-negative cells (right), 10 wild-type, 12 ctgfa+/−, and 15 ctgfa−/− animals from two independent experiments were used. *P < 0.05, **P < 0.01, ***P < 0.001; ns, not significant. Scale bars, 100 μm.

Deficits in glial bridging might underlie the defects in spinal cord regeneration displayed by ctgfa mutants. Using GFAP and acetylated α-tubulin immunohistochemistry, we observed robust glial bridges in wild-type and ctgfa+/− animals at 4 wpi (Fig. 2C and fig. S5A). However, ctgfa−/− animals displayed ~71% less bridging than did wild-type clutchmates (Fig. 2C and fig. S5A). At 2 wpi, glial cells within ctgfa−/− injury sites often failed to extend projections into the lesion (fig. S5B). 5-ethynyl-2′-deoxyuridine (EdU) incorporation assays showed that ctgfa−/− zebrafish displayed a ~48% reduction in glial cell proliferation at 1 and 2 wpi (Fig. 2D and fig. S6). EdU+GFAPcell numbers were comparable in wild-type, ctgfa+/−, and ctgfa−/− tissues at 1 wpi (Fig. 2D), which suggests that the effects of ctgfa mutations were preferential to glial cells. Thus, ctgfa is required for the changes in proliferation and morphology of glial cells during spinal cord regeneration.

To examine the effects of excess ctgfa on spinal cord regeneration, we generated and injured transgenic fish that express full-length ctgfa under control of a heat-inducible promoter (hsp70:ctgfa-FL) (fig. S7). Recovery of swim capacity was improved in ctgfa-overexpressing animals given daily heat shocks and assessed at 2, 4, and 6 wpi (Fig. 3A). Swim capacity was comparable between sham-injured hsp70:ctgfa-FL and wild-type clutchmates at 2 and 6 weeks after heat shock (Fig. 3A), indicating contextual effects of ctgfa overexpression. Histology indicated increased bridging and axon regeneration in ctgfa-overexpressing fish relative to controls at 2 and 4 wpi, respectively (Fig. 3, B and C, and fig. S8A). Thus, whole-animal ctgfa overexpression promotes regeneration after spinal cord injury.

Fig. 3 ctgfa promotes glial bridging and spinal cord regeneration.

(A) Swim assays determined motor function recovery of 10 hsp70:ctgfa-FL (green) and 10 wild-type (gray) clutchmates at 2, 4, and 6 wpi. For sham controls, 8 ctgfa-FL–overexpressing (dashed green) and 7 wild-type (dashed gray) zebrafish were analyzed. Statistical analyses of swim times are shown for injured ctgfa-FL (green) relative to wild type. (B) GFAP immunohistochemistry was used to quantify glial bridging at 2 wpi in 18 ctgfa-FL–overexpressing and 16 wild-type zebrafish from three independent experiments. (C) Anterograde axon tracing at 4 wpi after ctgfa-FL overexpression. Quantification at areas proximal (shown in images) and distal to the lesion core represents 12 ctgfa-FL–overexpressing and 10 wild-type zebrafish from two independent experiments. (D) Swim assays for 8 ctgfa-CT–overexpressing (blue), 10 ctgfa-NT–overexpressing (violet), and 9 wild-type clutchmate animals (wild-type controls for CT in dashed blue and for NT in dashed violet). Statistical analyses of swim times are shown for ctgfa-CT (blue) relative to wild type. (E) Glial bridging at 2 wpi in 19 ctgfa-CT–overexpressing and 20 wild-type animals from two independent experiments. (F) Anterograde axon tracing at 4 wpi after ctgfa-CT overexpression. Quantification represents 16 ctgfa-CT–overexpressing and 16 wild-type animals from two independent experiments. (G) Swim capacity was assessed for 9 vehicle-treated (gray), 8 HR-CTGF-FL–treated (green), and 9 HR-CTGF-CT–treated (blue) animals. Statistical analyses are shown for HR-CTGF-FL (green) and HR-CTGF-CT (blue) treatments relative to vehicle controls. (H) Glial bridging at 2 wpi in 18 HR-CTGF-CT–treated and 15 vehicle-treated animals from three independent experiments. (I) Anterograde axon tracing at 4 wpi after HR-CTGF-CT treatment. Quantification represents 18 vehicle-treated, 16 HR-CTGF-FL–treated, and 14 HR-CTGF-CT–treated animals from two independent experiments. For histology in (B), (E), and (H), dashed lines delineate glial GFAP staining and arrows point to sites of bridging. *P < 0.05, **P < 0.01, ***P < 0.001. Scale bars, 100 μm.

CTGF harbors four protein interaction domains and a protease domain that self-cleaves CTGF into profibrotic N-terminal and proliferative C-terminal peptides (11, 12). To determine the active portion of zebrafish Ctgfa during spinal cord regeneration, we created transgenic fish with heat-inducible expression of either N-terminal or C-terminal ctgfa fragments (hsp70:ctgfa-NT and hsp70:ctgfa-CT) (fig. S7A). Only ctgfa-CT overexpression recapitulated the pro-regenerative effects of ctgfa-FL (Fig. 3, D to F). By 6 wpi, swim capacity was markedly increased in ctgfa-CT–overexpressing animals relative to ctgfa-NT–overexpressing or wild-type clutchmates (Fig. 3D). Anatomically, ctgfa-CT overexpression resulted in a factor of >3 increase in glial bridging at 2 wpi and a factor of >2 increase in axon growth at 4 wpi (Fig. 3, E and F, and fig. S8B). These experiments indicate that the pro-regenerative activity of Ctgfa maps to its C-terminal domains.

To examine whether the effects of Ctgfa augmentation could be reproduced by localized delivery into the spinal cord lesion site, we injured wild-type animals and applied human recombinant CTGF (HR-CTGF-FL and HR-CTGF-CT) peptides adjacent to the lesion site at 5 and 10 dpi using a gelfoam sponge. We then assessed regeneration at 2 and 4 wpi, corresponding to 9 and 23 days after initial treatment. Human CTGF and zebrafish Ctgfa are 81% identical and 87% similar at the amino acid level within the four protein interaction domains (fig. S9). Treatment with either HR-CTGF-FL or HR-CTGF-CT enhanced zebrafish spinal cord regeneration (Fig. 3, G to I, and fig. S8C). Swim capacity was improved in HR-CTGF-FL–treated or HR-CTGF-CT–treated animals by 2 wpi (Fig. 3G), near that of uninjured animals. As early as 1 wpi, we observed increased GFAP expression and a factor of >10 enhancement of glial bridging in HR-CTGF-CT–treated animals relative to vehicle-treated controls (fig. S8D). Bridging remained greater at 2 wpi by a factor of ~3 (Fig. 3H). At 4 wpi, axon regeneration was increased by a factor of ~1.5 to 3 in HR-CTGF-FL–treated or HR-CTGF-CT–treated animals relative to vehicle controls (Fig. 3I). Application of exogenous CTGF protein at the lesion site rescued functional and anatomical spinal cord regeneration in ctgfa mutants, indicating specificity of both the phenotype and treatment (fig. S10). Thus, human CTGF protein, ostensibly via its C-terminal domains, enhances spinal cord regeneration in zebrafish.

Glial cell responses are thought to dictate the outcomes of spinal cord injury across species. We found that Ctgfa promotes zebrafish spinal cord regeneration, at least in part by facilitating the proliferation and bridging activity of pioneer glial cells that also express ctgfa. Other cell types also induce ctgfa in the injured zebrafish spinal cord and might contribute additional functions during regeneration. In mammals, reactive gliosis causes scarring and inhibits regeneration (13, 14), although evidence indicates that the lesion contains a heterogeneous pool of glial cells that release both axon growth–inhibiting and axon growth–permissive factors (6, 1519). We suggest that identifying the mammalian glial subtype that can produce CTGF, or that is competent to respond to it, could reveal a pro-regenerative mammalian counterpart to the zebrafish bridging glia.

Supplementary Materials

Materials and Methods

Figs. S1 to S10

Table S1

References (2026)

References and Notes

Acknowledgments: We thank A. Johnson, C. Eroglu, and M. Bagnat for discussions; N. Lee and K. Jones for technical and bioinformatics help; and the Duke University School of Medicine Zebrafish Shared Resource for animal care. Supported by NIH training grant T32HL007101 (M.H.M.), NIH grant R01 HL081674 (K.D.P.), the Max Planck Society (C.P. and D.Y.R.S.), and Duke University School of Medicine (K.D.P.). RNA sequencing data are archived at GEO (accession number GSE77025). The supplementary materials contain additional data. M.H.M. and K.D.P. are inventors on patent applications (62/159,413 and 62/398,781) submitted by Duke University that cover the use of Ctgf for spinal cord regeneration.

Stay Connected to Science

Navigate This Article