Transcriptional regulatory regions of gap43 needed in developing and regenerating retinal ganglion cells



Mammals and fish differ in their ability to express axon growth-associated genes in response to CNS injury, which contributes to the differences in their ability for CNS regeneration. Previously we demonstrated that for the axon growth-associated gene, gap43, regions of the rat promoter that are sufficient to promote reporter gene expression in the developing zebrafish nervous system are not sufficient to promote expression in regenerating retinal ganglion cells in zebrafish. Recently, we identified a 3.6-kb gap43 promoter fragment from the pufferfish, Takifugu rubripes (fugu), that can promote reporter gene expression during both development and regeneration. Using promoter deletion analysis, we have found regions of the 3.6-kb fugu gap43 promoter that are necessary for expression in regenerating, but not developing, retinal ganglion cells. Within the 3.6-kb promoter, we have identified elements that are highly conserved among fish, as well as elements conserved among fish, mammals, and birds. Developmental Dynamics 239:482–495, 2010. © 2009 Wiley-Liss, Inc.


The GAP-43 protein is abundantly expressed in axonal growth cones where it plays a role in axon growth and guidance during nervous system development and regeneration (Jacobson et al., 1986). In its phosphorylated form, GAP-43 acts to potentiate filopodial formation in growth cones by stabilizing f-actin (He et al., 1997; Dent and Meiri, 1998; Nguyen et al., 2009a). In its unphosphorylated form, GAP-43 acts as a barbed-end actin capper, preventing actin polymerization, but promoting microtubule-based neurite outgrowth (He et al., 1997; Nguyen et al., 2009a). GAP-43 over-expression in transgenic mice leads to extensive activity-induced sprouting at the adult neuromuscular junction (Caroni et al., 1997). In contrast, GAP-43 knockout mice die early in the postnatal period and exhibit pronounced guidance defects at commissural junctures and in cortical topography (Strittmatter et al., 1995; Kruger et al., 1998; Maier et al., 1999; Shen et al., 2002). Together these studies demonstrate the importance of GAP-43-mediated regulation of the growth cone cytoskeleton during axon growth and guidance.

Ectopic gap43 gene expression in the CNS of adult mice can stimulate axon re-growth in neurons that do not normally regenerate (Bomze et al., 2001; Zhang et al., 2005). For example, simultaneous ectopic expression of two neuronal growth-associated proteins (nGAPs), GAP-43 and CAP-23, in adult neurons of transgenic mice results in regenerative growth of ascending spinal axons in response to dorsal hemisection of the spinal cord (Bomze et al., 2001). In cerebellar granule cells, ectopic expression of GAP-43 protein with the cell adhesion protein L1 also stimulates regenerative axon growth in adult transgenic mice (Zhang et al., 2005). These results demonstrate the importance of GAP-43 in regenerative axon growth and underscore the value of understanding how the gap43 gene is regulated, and how its expression can be stimulated and sustained in damaged neurons to induce regenerative axon growth.

The gap43 gene is transcriptionally up-regulated in both differentiating and regenerating neurons, and is down-regulated in most mature neurons (Skene, 1989). Several transcriptional regulatory elements within the gap43 gene have been characterized with respect to their roles in neuron-specific expression (Nedivi et al., 1992; Reinhard et al., 1994; Starr et al., 1994; Chiaramello et al., 1996; McCormick et al., 1996; Weber and Skene, 1997, 1998). However, there is little information about the regulatory elements responsible for the temporal regulation of gap43 within neurons. Although some of the same elements may be involved in both spatial and temporal regulation of gap43, we have shown that a 1-kb promoter element from rat gap43 was sufficient for expression in developing neurons, but not sufficient for expression in regenerating CNS neurons in zebrafish (Udvadia et al., 2001). Thus, although gap43 is expressed during axon development and regeneration, it is apparently regulated by different mechanisms. Similarly, a difference in developmental and regenerative growth-associated expression is observed with another nGAP, tuba1 (Goldman and Ding, 2000; Senut et al., 2004). These findings indicate the existence of overlapping yet distinct pathways for axon growth during development and CNS regeneration.

Differences in gene regulation in response to CNS injury between mammals and fish are well established (see review by Becker and Becker, 2007). However, differences between the cis-regulatory elements that regulate genes associated with axon growth in mammals and fish have not been explored. Previously, Goldman and colleagues discovered an element in the tuba1 gene promoter that was conserved between fish and rats, and was essential for expression during optic nerve regeneration (Senut et al., 2004). Here we describe a region sufficient for developmental expression of gap43 that contains elements conserved between fish and mammals, as well as regions necessary for regenerative gap43 expression containing elements highly conserved among fish. These results support the hypothesis that developmental and regenerative gap43 gene expression are differentially regulated and suggest that pathways regulating regenerative growth have diverged more significantly between amniotes and anamniotes than those regulating developmental growth.


We recently described a 3.6-kb promoter/enhancer sequence of fugu gap43 sufficient to promote expression in the developing zebrafish nervous system and in regenerating retinal ganglion cells of the adult fish (Udvadia, 2008). Using transient injection assays to study the effects of gap43 promoter deletion mutations, we further defined the regions necessary for nervous system–specific expression. We performed a 5′-end deletion analysis of the 3.6-kb region to determine the smallest fragment that could promote nervous system–specific expression of a GFP reporter gene. A series of five GFP reporter plasmids containing sequential 5′-end deletions of the 3.6-kb promoter were constructed (Fig. 1A). Transgenes were injected into zebrafish embryos at the one-cell stage and GFP expression was assessed at 30 hpf. Each embryo was scored for the presence of GFP-expressing cells in nervous system, muscle, notochord, skin, and the enveloping layer (EVL). All plasmids caused preferential transgene expression in neurons with varying low levels of extra-neuronal expression (Fig. 1B). Thus, the removal of the distal 2.9-kb sequence fragment (GfG43-708) resulted in a promoter that was still sufficient to promote selective expression in developing zebrafish neurons.

Figure 1.

5′-end deletions of 3.6-kb fugu gap43 regulatory sequence promote reporter gene expression preferentially in neurons. A: Full-length promoter, GfG43S/A, and 5′-end deletions (GfG43B/A, GfG43-972, GfG43-895, GfG43-843, and GfG43-708) used in transient GFP reporter assays in zebrafish. B: Comparison of tissue distribution of GFP cells in 30-hpf embryos that were injected at the 1-cell stage with the various deletion constructs. Each injected embryo was scored as positive or negative for GFP-expressing cells in the nervous system, muscle, notochord, skin, or the enveloping layer (EVL = extra-embryonic peridermal tissue). Each bar represents the average value from 2–4 separate experiments. Standard errors are indicated by error bars. n = total number of fish examined for each construct.

gap43 Promoter Regions Needed for Expression in Regenerating RGCs

We created stable transgenic reporter lines in which the proximal 708-bp promoter-fragment was used to drive GFP expression to determine if the region that is sufficient to drive expression in the developing nervous system is also sufficient to promote expression in the regenerating CNS. The 708-bp promoter was chosen as it was the shortest region that still functioned to promote selective expression in neurons in our transient assays. We observed transgene expression throughout the developing nervous system in three independently segregating stable transgenic lines, although GFP expression levels were noticeably lower than in the full-length reporter lines. We observed some interline variation in weak extraneuronal expression in muscle or notochord, but not in the nervous system (data not shown). One line also displayed stronger expression in the caudal hindbrain region when compared to the other lines and to the full-length promoter lines. Given that two out of the three lines did not display strong caudal hindbrain expression, it is likely that this expression is influenced by the transgene insertion site rather than the 708-bp sequence itself.

Expression in the developing retinal ganglion cells (RGCs) in all three lines was similar to that observed in the full-length reporter lines. At 48 hpf, expression occurred throughout the ganglion cell layer (GCL; Fig. 2A, E), and by 96 hpf expression was limited to the fiber layer and the inner plexiform layer (IPL; Fig. 2B, F). Thus, the spatial and temporal pattern of expression promoted by the 708-bp and 3.6-kb promoter fragments was similar in the developing retina. For this reason, we compared the expression in the regenerating retina of adult fish.

Figure 2.

The 708-bp promoter is sufficient to promote expression in developing but not regenerating retina. Native GFP expression (green) and DAPI staining (blue) in transverse sections of embryonic, larval, and adult retinas. The GfG43SA and GfG43-708 constructs both express GFP in the retinal ganglion cell layer (gcl) and their axons (*) observed in the fiber layer at 48 hpf (A,E) and in the inner plexiform layer (ipl) within the retina at 96 hpf (B, F). In control and regenerating adult retina (7 days post-crush) of the GfG43SA line, low levels of GFP are observed in control retina (C) and high levels in the GCL and fiber layer of regenerating retina (D). In the GfG43-708 line, there is no GFP expression in the GCL of either control (G) or regenerating (H) retina. Developing retina scale bar (F) = 25 μm. Adult retina scale bar (H) = 50 μm.

In order to determine if the 708-bp fragment was also sufficient to promote expression in the regenerating CNS, we crushed the optic nerve of transgenic fish and tested for transgene expression in regenerating retinal ganglion cells. Endogenous gap43 mRNA increases 1 day post-crush, maintains peak levels between 4–8 days post-crush, and then slowly returns to basal levels after regeneration is completed (Bormann et al., 1998; Kaneda et al., 2008). We monitored transgene expression at both 5 and 7 days post-crush. Unlike the 3.6-kb reporter line, transgene expression was not detected in the adult retina either constitutively in the IPL or in response to optic nerve injury in the GCL in the 708-bp reporter lines (compare Fig. 2C, D with Fig. 2G, H). These results suggest that the distal 2.9-kb fragment of the fugu gap43 promoter contains sequences that are necessary for promoting expression in the adult retina within the IPL and the regenerating GCL. However, the distal 2.9 fragment is not necessary for expression in the developing retina.

Regulation of Developmental and Regenerative gap43 Expression

In order to identify sequences necessary for regeneration-associated expression of gap43, we generated gap43/GFP reporter plasmids containing various 5′ end and internal deletions of the 2.9-kb fugu gap43 distal promoter sequence, which is necessary for regenerative expression. This 2.9-kb sequence was divided into three regions, A (664 bp), B (1,057 bp), and C (1,195 bp), based on clustering of conserved sequences that will be described in detail in the next section. The remaining 708-bp proximal promoter region, which is sufficient for developmental expression, was designated as region D and will be referred to as the “core promoter.” Six different deletion constructs were created containing all possible combinations of A, B, and C in conjunction with region D (Fig. 3; ΔA, ΔB, ΔC, ΔAB, ΔAC, and ΔBC). We established 2–3 independently segregating stable transgenic lines with each of the promoter deletion constructs to test if any of the constructs were sufficient for promoting expression in adult regenerating retinal ganglion cells (results summarized in Table 1).

Figure 3.

Deletion mutations for stable reporter lines. The Fugu 3.6-kb gap43 promoter is represented by the black ruler at the top. Numbers on the ruler refer to distance (bp) from the translation start site (ATG), and tss refers to the transcription start site. The fugu 3.6-kb promoter was divided into 4 regions based on clusters of conserved sequence. All constructs contain the D region, which is equivalent to the 708-bp promoter, that is sufficient for promoting expression in developing but not regenerating neurons. All possible combinations of regions A, B, and C were generated. Dotted lines represent deleted sequences.

Given that each reporter construct contained the core promoter in the D region, we predicted that all six reporter genes would be expressed in developing neurons. Surprisingly, we found that the addition of region A and/or B to the core promoter in the absence of C (ΔAC, ΔBC, ΔC,) resulted in the complete lack of expression in developing embryos (data not shown). We identified 2–3 independent transgenic founders for these lines using PCR, but we did not detect GFP fluorescence in the founders or any of the F1 or F2 progeny from these founders. These results suggest that sequences within regions A and B repress the developmental expression of gap43, while sequences within region C can mitigate the repressive activity.

The remaining reporter genes (ΔA, ΔB, ΔAB) were expressed throughout the developing nervous system, although overall levels of GFP fluorescence were noticeably less than that displayed in lines carrying the full-length 3.6-kb promoter (Fig. 4A, E, I, M, Q). The overall decrease in transgene expression observed in the deletion mutations suggests that regions A, B, and C also contain enhancer elements that contribute to the higher levels of expression observed with the full-length 3.6-kb promoter. In the ΔAB lines, the addition of C to the core promoter resulted in expression throughout the developing nervous system (Fig. 4M, Q). The deletion of the B region (ΔB) led to subtle differences in expression pattern in the developing hindbrain (Fig. 4J, N). We also detected extraneuronal expression in the heart of all three ΔB lines beginning at 3 days post-fertilization (data not shown), suggesting the B region suppresses expression in the heart. The deletion of the A region alone, or in the context of the ΔABC mutant, led to decreased expression in the optic tectum (Fig. 4G, S). These results suggest that the B region may also suppress expression in the tectum, while regions A and/or C promote expression in the tectum. Despite these regional differences in expression among mutant reporter lines, there was a similar pattern of GFP expression in the forebrain, olfactory system, and retinal ganglion cells in all lines (Fig. 4 D, H, L, P, T), suggesting that the retina of the adult was a good model to test mutant reporter construct expression during regeneration.

Figure 4.

Changes in developmental expression pattern as a result of promoter deletions. Transgene expression in stable GFP reporter fish carrying the Fugu 3.6-kb gap43 promoter (A–D), or promoter deletions: ΔA (E–H), ΔB (I–L), ΔAB (M–P), and ΔABC (Q–T). Images of 48-hpf embryos captured on the fluorescent stereoscope (A, A inset, E, I, M, Q) show that promoter deletions result in overall lower levels of GFP expression (scale bar in Q = 500 μm). The image in (A) was taken at 1/3 the exposure time, while the overexposed image in the inset was taken at the same exposure time as E, I, M, Q. Higher magnification views of the embryos (B–D, F–H, J–L, N–P, R–T; scale bar in T = 100 μm). Dorsal view of hindbrain at 24 hpf (B, F, J, N, R). Lateral view of brain at 48 hpf (C, G, K, O, S). Ventral view of head at 48 hpf (D, H, L, P, T). cb, cerebellum; hb, hindbrain; ob, olfactory bulb; oc, optic chiasm; oe, olfactory epithelium; ot, optic tectum; rgc, retinal ganglion cells.

In order to determine if regions A, B, and/or C contributed to expression in regenerating CNS neurons, we crushed the optic nerve in adult transgenic lines. The promoter deletions, ΔA, ΔB, and ΔAB, showed patterns of expression in the developing retina (Fig. 5A, B, E, F, I, J) similar to the full-length promoter and the core promoter (Fig. 2A, B, E, F). As in developing embryos, there was no expression in regenerating retinal ganglion cells of transgenic fish that lacked the C region (not shown). However, the addition of region C to the core promoter could restore regenerative expression (Fig. 5L), although it was not sufficient to restore the constitutive expression in the IPL (Fig. 5K). The addition of A or B with C could restore constitutive expression in the IPL (Fig. 5C, G). This suggests that constitutive expression in the IPL of the adult retina requires region A or B, although these regions are not required for IPL expression in the developing retina (Fig. 5B, F, J). Addition of A and C also resulted in slightly higher expression in regenerating GCL (Fig. 5 H). Interestingly, expression in control and regenerating retina of the ΔAB lines was higher than in the ΔA lines (Fig. 5L, H), suggesting B may contain elements responsible for suppressing expression. Together, these results demonstrate that both A and C contain elements that activate expression in response to optic nerve crush; however, the elements within C are necessary for this response.

Figure 5.

Transgene expression in regenerating retina requires region C. Native GFP expression (green) and DAPI staining (blue) in transverse sections of embryonic (48 hpf; A, E, I), larval (4 dpf; B, F, J), and adult retinas (control; C, G, K; and regenerating; D, H, L) from promoter deletion GFP reporter lines: ΔA (A–D); ΔB (E–H); ΔAB (I–L). All lines displayed the same spatial and temporal pattern of expression in the developing retina (A, B, E, F, I, J) as previously observed for the full-length promoter (see Fig. 2A, B). Unlike ΔABC reporter lines, which did not express GFP in regenerating adult retina (see Fig. 2G, H), addition of the C region in the ΔAB lines was sufficient to restore regenerative expression (L). (*) Axons of RGCs in fiber layer. dpc, days post-crush; gcl, ganglion cell layer; ipl, inner plexiform layer; inl, inner nuclear layer. Developing retina scale bar (J) = 25 μm. Adult retina scale bar (L) = 50 μm.

Conservation of gap43 Regulatory Regions Among Fish

In order to identify conserved regions that might be responsible for directing developmental and regenerative gap43 expression, we compared the gap43 upstream regulatory region to that from three other fish species whose gap43 genomic regions are available. It was not possible to make comparisons of these sequences to the zebrafish or amphibian gap43 promoters because sequences upstream of and including gap43 exon 1 in these species are not yet posted in public genomic sequence databases (Zv7 and JGI 4.1, respectively). The zebrafish promoter sequences have been resistant to our cloning and sequencing efforts (unpublished data), which is likely due to its telomeric location on chromosome 15. However, the fugu promoter can accurately direct transcription in zebrafish in a temporal and spatial pattern that mimics endogenous gap43, suggesting functional conservation of these sequences in species that diverged over 300 million years ago (Yamanoue et al., 2006).

Our first comparison was between the fugu gap43 promoter and that of a closely related pufferfish, Tetraodon nigroviridis. In general, intergenic and intronic sequences are less likely to be conserved across species than protein-coding sequences since changes in coding sequence can lead to changes in protein function. Therefore, non-coding sequences that are highly conserved across species may have undergone selective pressure to retain their functional significance (Venkatesh and Yap, 2005). We have previously shown that the fugu 3.6-kb gap43 genomic fragment can regulate the spatial and temporal pattern of gene expression in differentiating and regenerating neurons in a manner similar to the endogenous zebrafish promoter (Udvadia, 2008). Within the fugu sequence, we identified conserved elements with potential regulatory significance by conducting two-way sequence alignments (BLAST) between fugu and Tetraodon, which diverged approximately 85 million years ago (Yamanoue et al., 2006). We found 10 distinct sequences within the 3.6-kb fragment that are highly conserved (>80% sequence identity) between the two species (Fig. 6, red bars). They range in length from 37 to 468 bp and can be divided into three groups: distal, intermediate, and proximal to the translational start site. In addition to sequence similarity, these regions are organized in the same spatial order and most are similar distances from the translational start site in both species.

Figure 6.

Teleost conserved sequence elements within the gap43 promoter. Comparison of gap43 promoter regions from four different fish species (Fugu, Tetraodon, Medaka, and Stickleback) using 2-way BLAST shows that there are several regions of highly conserved sequence. The conserved regions, represented by the red, orange, and yellow boxes below the ruler [Tetraodon (red), medaka (orange), and stickleback (yellow)] share >80% sequence identity with similar regions in the Fugu 3.6 kb gap43 promoter. The ruler only applies to the fugu sequence. However, the relative position of all conserved sequences, in relation to the translational start site is also conserved. EvoPrinter (Odenwald et al., 2005) was also used to identify teleost conserved sequence elements (TCSEs, black rectangles) from Fugu, Tetraodon, and Medaka, which were manually aligned with sequences from Stickleback.

To determine the divergence of gap43 cis-regulatory regions over a longer evolutionary period, we compared the fugu 3.6-kb enhancer/promoter sequence to genomic sequences from more distantly related fish. The medaka and stickleback diverged from pufferfishes approximately 191 and 183 million years ago, respectively (Yamanoue et al., 2006). There was less similarity between the fugu gap43 promoter sequence and that of either medaka or stickleback than with Tetraodon. However, eight regions ranging from 15–60 bp in length were highly conserved between medaka and fugu (Fig. 6, orange bars) and five regions ranging from 52–71 bp in length were highly conserved between stickleback and fugu (Fig. 6, yellow bars). All of these sequences were contained within the regions conserved in the 2 pufferfish species (Fig. 6). Alignment of sequences from fugu, Tetraodon, medaka, and stickleback shows they overlap and are highly conserved in all four species (Fig. 6). In addition to 2-way BLAST alignments, we also used EvoPrinter (Odenwald et al., 2005) combined with manual alignment to find sequences conserved among fish (Fig. 6, black bars). EvoPrinter uses the same principles for finding sequences of potential functional importance, but it is capable of finding shorter sequences (6 bp) that fall below the statistical significance cut off for BLAST based on sequence length (Odenwald et al., 2005; Brody et al., 2008). Sequences that were conserved between 3 or all 4 of the fish were grouped into Teleost Conserved Elements or TCSEs. This analysis identified specific nucleotides and putative transcription factor binding sites that are conserved among all four fish species versus those that are conserved between three or two (Table 2). Together, these comparisons have identified specific sequences that are highly conserved among a wide range of teleost species in both the proximal 708-bp core promoter region, which is sufficient to promote developmental expression, and the distal 2.9-kb promoter fragment, which is necessary to promote expression in regenerating CNS neurons.1

Table 1. GFP Expression Patterns During Development and Regeneration Using Different Reporter Constructsa
 Developing EmbryoAdult retina
TransgeneSpinal cordForebrainMidbrainHindbrainRetinaControl GCLRegenerating GCLControl fiber layerRegenerating fiber layer
  • a

    Relative intensity of expression is indicated: (−) indicating no GFP expression, (+) indicating GFP expression with (++) and (+++) representing increasing intensity. Expression in the GCL refers to expression in retinal ganglion cell bodies, while expression in the fiber layer refers to expression in retinal ganglion cell axons.

Table 2. Teleost Conserved Sequence Elements (TCSEs)a
NameRegionSequencePutative TFBS
  • a

    Under the “Sequence” heading, sequences conserved between at least two species are capitalized, sequences conserved between at least 3 species are underlined, and sequences shared by all 4 fish species are bolded. Putative transcription factor–binding sites (TFBS) within the TCSEs using TESS or MatFinder are listed.

TCSE 7CACAAATACGGAGTGTGGGTGTTTTGCATGGACATGCACCTGCCAHNF-3, Elf-1, Runx1, Oct-1, Oct-2, Oct-3, MyoD, ASH-1, Tal-1, NF-X3, E-box, Snail, EGR-2, FoxK2, p53

Conservation of gap43 Regulatory Regions Between Fish and Mammals

Comparison of the fugu gap43 regulatory regions to those in animals that are not capable of CNS regeneration showed conservation of sequences proximal but not distal to the start of translation. We used MultiPipMaker (Schwartz et al., 2000) to search for sequences similar to the fugu 3.6-kb gap43 promoter/enhancer in sequences spanning the entire length of the 5′ intergenic region and intron 1 from Tetraodon, medaka, stickleback, chicken, mouse, rat, and human (Fig. 7). This program aligns genomic sequences using algorithms that detect short stretches of similarity between the reference DNA that can be found anywhere along the length of the test DNA. This strategy is useful for detecting conserved sequences from species that may have diverged significantly over time. Similar strategies have been used successfully to identify conserved sequence elements with regulatory significance in zebrafish and/or mouse (Elnitski et al., 2002; Santini et al., 2003). Our analysis identified upstream regions of similarity within approximately 700 bp of the gap43 translational start site (Fig. 7). The similarities between fugu and chicken, mouse, rat, or human are not as strong as those between fugu and Tetraodon or medaka. Nonetheless, the gap43 regulatory sequences conserved between all six genomes are contained within the region, which is sufficient for developmental expression within the nervous system. MultiPipMaker analysis of sequences spanning more than 300,000 bp upstream of the GAP-43 translational start site as well as intron 1 sequences in chicken, mouse, rat, and human genomes, did not identify significant similarities with the distal conserved elements in the 3.6-kb fugu promoter.

Figure 7.

Distal gap43 regulatory regions contain elements that are highly conserved among fish. Potential gap43 regulatory sequences from Tetraodon, medaka, stickleback, chicken, mouse, rat, and human compared to the fugu gap43 3.6-kb 5′ flanking sequence using MultiPipMaker. Lengths of 5′ flanking sequences that were compared to the 3.6-kb fugu sequence are shown in parentheses. Intron 1 sequences were also included in the comparison for chicken, mouse, rat, and human (length listed in parentheses after “+”). A–C: Correspond to the regions involved in the deletion series. D: Corresponds to the 708-bp promoter that is sufficient for expression in developing, but not regenerating retina. Black stripes correspond to the TCSEs in Figure 6. Top bar represents fugu 3.6-kb gap43 sequence. Red, orange, and yellow stripes correspond to the Tetraodon, medaka, and stickleback conserved elements as determined by 2-way BLAST and in Figure 6. Dark blue stripe corresponds to 30-bp coding region of exon 1. Light grey, grey, teal, and mauve stripes correspond to repeat sequences detected by Repeat Masker. Green shading refers to sequences within the fugu gap43 promoter that share 60–100% identity to sequences found somewhere in the sampled region of gap43 from the indicated species; pink shading refers to sequences that share 100 bp or more, without gaps and with ≥70% identity.

In the process of identifying gap43-coding sequences within fish, chicken, and mammalian genomes, we noticed that synteny was conserved with regard to the 3′ flanking gene, but not with regard to the 5′ flanking gene. In fish, chicken, and mammals, gap43 is flanked on its 3′-side by the lsamp gene (limbic system-associated membrane protein; Fig. 8). However, the 5′-flanking gene in fish is different from that in chicken and mammals. In fish, it is the eva1 gene (epithelial V-like antigen 1), while in chicken and mammals, it is zbtb20 gene (zinc finger and BTB domain containing 20 gene; Fig. 8). In chicken and mammals, the eva1 gene is located on a completely different chromosome, but in a location with conserved synteny, supporting the idea that this difference in genomic arrangement is common to the avian and mammalian lineages. These results suggest that the chromosomal rearrangement affected regulatory sequences distal to the 5′ end of gap43 more dramatically than the proximal 5′ promoter sequences, resulting in conservation of sequences necessary for developmental but not regenerative retinal expression.

Figure 8.

Differences in synteny around the gap43 gene locus between fish and other species. The lsamp gene (orange) lies 3′ to the gap43 gene (yellow) in fish, chicken, and mammals. However, the gene 5′ of gap43 in fish (eva1; blue) is different from that in birds and mammals (zbtb20; green). In addition, the orientation of eva1 in fish is the same as gap43, while zbtb20 in birds and mammals is in the opposite orientation. The distance from gap43 to its neighboring genes is indicated below the dotted lines to the left and right of the arrow representing gap43. The yellow, orange, green, and blue boxes are not to scale and represent only the coding sequences of the genes. Note that the medaka and stickleback genomes contain gaps in these regions so distances reflect the gap size predicted by Ensembl, and thus may not be entirely accurate.


A number of factors, both intrinsic and extrinsic to neurons, are known to contribute to the failure of most adult neurons in the mammalian CNS to regenerate after axonal damage. Here, we have focused on the differential regulation of an intrinsic factor, the axon growth-associated gene, gap43, whose expression is tightly correlated with axon extension in developing and regenerating nerves. Transcription of gap43 is strongly induced and sustained in response to CNS injury in fish, which have a robust capacity for functional CNS regeneration (reviewed in Becker and Becker, 2007). A recent study examining GAP-43 expression and phosphorylation during optic nerve regeneration in the zebrafish confirmed that both the mRNA and phosphoprotein levels were upregulated 8–12-fold in the retina in response to injury and then remained elevated (approximately 4-fold) throughout the prolonged period of synaptic refinement (80–100 days) until full functional regeneration was achieved (Kaneda et al., 2008). In mammals, injury to the optic nerve normally results in transient expression of GAP-43 in a limited number of retinal ganglion cells, which is correlated with localized sprouting, but not regeneration (Doster et al., 1991; Meyer et al., 1994; Schaden et al., 1994). We are interested in understanding how gap43 is regulated in response to a CNS injury that stimulates a sustained response and full functional regeneration.

Differences in Axon Growth-Related Gene Expression During Development and CNS Regeneration

We recently reported that a 3.6-kb promoter from the fugu gap43 was able to mimic developmental and CNS regeneration-associated gene expression of the endogenous zebrafish gap43 (Udvadia, 2008). In this study, we have shown that the proximal 708 bp of the fugu gap43 promoter is sufficient for promoting developmental expression and contains sequences that are broadly conserved in gap43 promoter proximal sequences in other fish species, chicken, rodents, and human. In contrast, the distal 2.9-kb region contains sequence elements that are highly conserved among fish, but not mammals or birds. These sequences appear to be necessary for expression in the regenerating retina, but not essential for expression in the developing retina. The results are consistent with previous findings that demonstrated a difference between sequences regulating developmental and CNS regenerative expression of growth-associated genes (Goldman and Ding, 2000; Udvadia et al., 2001).

Cis-regulatory elements can promote growth-associated expression in regenerating neurons, but not developing neurons (Senut et al., 2004; Bacon et al., 2007). A 64-bp sequence in the promoter of alpha-1-tubulin, another axon growth-associated gene, is necessary for regenerative but not developmental gene expression (Senut et al., 2004). Deletion of this small element from a large regulatory region that includes 1.6 kb of 5′ flanking regions as well as intronic regions has no effect on developmental expression, but eliminates regeneration-associated expression after optic nerve crush. This 64-bp fragment is conserved in the rat alpha-1-tubulin gene and contains an E-box adjacent to a homeodomain-binding site. Although we have not found a similar primary sequence in the fugu gap43 promoter, comparison of the teleost conserved elements with known transcription factor–binding sites using TESS ( and MatInspector (Cartharius et al., 2005) have localized an E-box-binding site and nearby homeodomain-binding sites to the distal-most segment (region A) of the fugu gap43 3.6-kb promoter that is conserved among fish and is involved in promoting regeneration-associated gene expression.

The promoter region of the gene encoding galanin, a neuropeptide that also serves as a trophic factor for dorsal root ganglion (DRG) sensory neurons, has two regions of 18 and 23 bp that are necessary for expression in regenerating but not developing DRGs (Bacon et al., 2007). These regions contain overlapping binding sites for three families of transcription factors: Ets, Stat, and Smad. Using bioinformatics techniques, the authors found similar arrangements of these sites in three other axotomy-induced genes in mouse including the gap43 gene. In the mouse, two such clusters occur 8–10 kb upstream of the transcriptional start site (Bacon et al., 2007). In the 3.6-kb fugu gap43 promoter, MatInspector identified overlapping sites for ets family member, elf-2, and STAT3 within the C-region, although not within regions that were conserved in all four fish (see Supp. Table S1, which is available online).

Sequences in both the A region and C region of the promoter contain a number of additional putative transcription factor–binding sites that are associated with axon growth and regeneration, which could be responsible for promoting expression in the regenerating CNS. For example, both the A and C regions contain highly conserved putative binding sites for p53, which drives axon outgrowth and gap43 expression after facial motor neuron axotomy (Tedeschi et al., 2008). Interestingly, there is also a putative NFAT binding site adjacent to the p53 site in region C (Supp. Table S1), which can act as a negative regulator of GAP-43 during neuronal maturation (Nguyen et al., 2009b). We expect that a number of overlapping, yet distinct signalling pathways converge to promote axon growth in response to CNS injury as with injury in the PNS (Cai et al., 2002; Qiu et al., 2002; Cavalli et al., 2005; Perlson et al., 2005; Cao et al., 2006; Spencer et al., 2008).

Conserved cis-Regulatory Elements

We have discovered regulatory sequences in fugu gap43 that are sufficient for developmental expression and those that are necessary for expression in regenerating CNS. Although fugu and zebrafish diverged more than 300 million years ago (Yamanoue et al., 2006), gap43 regulatory regions responsible for promoting expression in newly differentiated and regenerating neurons appear to be functionally conserved. While we do not yet have the upstream regulatory sequences of zebrafish, we have identified discrete sequence elements that are highly conserved between 4 different species of fish including two pufferfishes, medaka, and stickleback. Recent analysis of the medaka gap43 promoter using transient reporter assays shows that proximal promoter sequences are sufficient to promote neuron-specific expression in developing medaka (Fujimori et al., 2008). The conserved gap43 promoter sequences that we have identified not only share a high degree of sequence similarity among fish, but some elements also share similarity with gap43 promoter sequences from birds and mammals (Fig. 7). Several of these sequences are contained within the proximal promoter region required for gene expression in the developing nervous system. In addition, the 1-kb rat promoter used in previous experiments included these proximal regulatory regions and promoted the same developmental gene expression as endogenous gap43 (Udvadia et al., 2001). However, the 1-kb rat promoter did not promote reporter gene expression during regeneration in zebrafish (Udvadia et al., 2001). Interestingly, within the 2.9-kb region necessary for expression in regenerating RGCs, we find sequences that are highly conserved among diverse species of fish. However, we have not been able to detect similar sequences in the putative upstream or intron 1 regulatory regions of chicken, rodents, or human gap43. These results suggest that the elements correlated with regenerative expression in the CNS of fish are either missing or are significantly altered in composition and/or location in amniotes.

Comparative analysis of genomic organization between vertebrate species in the vicinity of the gap43 locus suggests that differences in the distal elements could be due to a chromosomal rearrangement that occurred during vertebrate evolution. One possibility is that the more distal elements regulating gap43 in fish were relocated to a different chromosome in birds and mammals along with eval during a reorganization event in this region of the chromosome. Other possibilities are that the distal elements were gained in fish or deleted from birds and mammals after teleosts split from ancestors of amniotes. There is also the possibility that the core binding sites within the conserved elements in the distal 2.9-kb region are present in mammals and birds, but the primary sequence has diverged beyond recognition by our methods, and is not activated after mammalian CNS injuries.

Differential Regulation of CNS and PNS Regeneration

Unlike the CNS, the PNS of all species has a robust capacity for regeneration. Sensory neurons of the dorsal root ganglion (DRG) are a widely studied model used to test the differences between the regenerative capacity of CNS and PNS neurons. The DRG cell bodies reside within the PNS and extend two axons, one to the periphery and the other centrally through the dorsal columns of the spinal cord. In mammals, injury to the dorsal spinal cord does not elicit gap43 expression or a regenerative response of DRG neurons, but regeneration can be stimulated in the spinal cord after a preconditioning lesion to the peripheral processes (Richardson and Issa, 1984; Neumann et al., 2005). The peripheral nerve lesion stimulates a transcription-dependent transition into a growth-competent state (Smith and Skene, 1997). Although administering the peripheral lesion prior to the central lesion is necessary to evoke re-growth of the central axons, it was recently demonstrated that a peripheral lesion after a central lesion promotes the induction of nGAP gene expression in DRGs even in the absence of CNS regeneration (Ylera et al., 2009). Furthermore, it was found that if the central lesion minimized glial scarring, the central processes could regenerate even if the conditioning peripheral lesion occurred after the central lesion (Ylera et al., 2009). Together these results confirm that CNS regeneration in amniotes requires both the neuron-intrinsic induction of nGAPs and the reduction of growth-inhibitory extrinsic factors, such as scarring.

Here we have demonstrated that gap43 promoter regions necessary for driving expression in regenerating retinal ganglion cells are separated from regions that are sufficient for promoting developmental expression. Furthermore, the regions regulating regenerative expression are well conserved across a broad range of teleost species, but are not shared between teleosts and mammals. It is possible that while axon growth during development is a highly conserved process among all vertebrates, mechanisms regulating regenerative axon growth have diverged. These differences suggest that specific regulatory elements of gap43 and other nGAP genes are essential to sustain a regenerative response in the CNS. Further delineation of regeneration-specific sequences and the factors that bind them, as well as investigation into epigenetic changes, will focus on the regulatory regions identified in this study.


Zebrafish Husbandry

Zebrafish husbandry and all experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC). Zebrafish colonies were maintained as previously described (Westerfield, 2000). Adult fish were housed in recirculating rack systems (Aquatic Habitats, Apopka, FL) at 28.5°C on a 14-hr light, 10-hr dark cycle, and fed twice daily with Artemia and/or Zeigler Adult Zebrafish Complete Diet (VWR, West Chester, PA). Fry were raised in an incubator at the same temperature and under identical light conditions as the adults in static tanks with daily water changes for 2–3 weeks and fed with GP Larval Diet (50–100 and 100–200 μm; Brine Shrimp Direct, Ogden, UT) and Artemia.

Fugu gap43 Reporter Plasmid

The pW1GfG43SA reporter plasmid driving a membrane-targeted GFP under the regulation of a 3.6-kb fragment of the Takifugu rubripes (fugu) gap43 promoter has been previously described (Udvadia, 2008). Five 5′-end deletions of the promoter were created for analysis in transient transgenic analysis using PCR to amplify successively shorter promoter fragments that were cloned into pW1EGFP (Udvadia, 2008) using the Bgl II and Age I restriction sites. The deletion plasmids, GfG43BA, GfG43-972, GfG43-895, GfG43-845, and GfG43-708 contained 1,904, 972, 895, 845, and 708 bp, of the promoter, respectively (Fig. 1A).

In order to facilitate stable transgenesis with the pW1GfG43-708 plasmid, we also generated a plasmid in which the 708-bp promoter and GFP sequences were flanked with I-SceI meganuclease recognition sites (GfG43ΔABC) as previously described (Thermes et al., 2002). Six additional GFP reporter plasmids with flanking I-SceI sites were constructed for creating stable transgenic lines with deletions in the distal promoter regions necessary for regeneration-associated expression. These reporter constructs contained both 5′-end and internal deletions of the 3.6-kb fugu gap43 promoter fragment: GfG43ΔA, GfG43ΔB, GfG43ΔAB, GfG43ΔC, GfG43ΔAC, and GfG43ΔBC (Fig. 3B).

Generation and Analysis of Transient and Stable Transgenic Fish

Initial analysis of transgene expression patterns was carried out using transient transgenic assays of 24–30-hr-old embryos injected at the 1-cell stage with linearized transgene DNA. GFP expression was visualized and documented in live embryos using fluorescence microscopy (TE 2000, Nikon, Melville, NY). Embryos were scored for GFP expression detected by live fluorescent imaging in different tissues including nervous system (brain, spinal cord, peripheral nerve ganglia), muscle, notochord, skin, and enveloping layer (EVL). The percentage of total injected embryos expressing GFP+ cells in these tissues was averaged over 2–4 experiments with a minimum of 20 injected embryos analyzed for each experiment.

A stable transgenic line bearing the GfG43-708 transgene was generated by injecting purified, linear transgene DNA into the yolk-free cytoplasm of 1-cell-stage embryos (Linney and Udvadia, 2004). Additional transgenic lines were also created by injecting the GfG43ΔA, GfG43ΔB, GfG43ΔAB, GfG43ΔC, GfG43ΔAC, GfG43ΔBC, and GfG43ΔABC plasmids along with the I-SceI enzyme (New England Biolabs, Ipswich, MA) as previously described (Thermes et al., 2002). Injected embryos were raised to maturity and progeny from pair matings of the injected fish were subsequently screened for evidence of germline transmission of the transgene as previously described (Linney and Udvadia, 2004). Live embryos were screened for GFP expression using fluorescence microscopy, and GFP (+) F1 embryos were raised to establish each line. In order to identify founders for lines in which GFP was not detectable in embryos (GfG43ΔC, GfG43ΔAC, and GfG43ΔBC), genomic DNA extracts were prepared from pools of embryos (Linney and Udvadia, 2004) and screened by PCR using transgene-specific primers that could distinguish between the different deletion mutations. Subsequent clutches of embryos from identified founder fish were raised for approximately 8 weeks, at which time F1 transgene carriers were identified by PCR using DNA extracted from caudal fin clips (Amsterdam et al., 1999).

Analysis of Developmental Transgene Expression

Live imaging of transgenic fish was carried out on fish from the F2–F6 generation. Low-magnification views of each line were captured using a fluorescent stereoscope (SZX12, Olympus, Center Valley, PA) equipped with a digital color camera (Olympus DP70 with DP controller software). For higher resolution images, embryos were anesthetized in Tricaine and embedded in 0.5% agarose to immobilize for image capture using an inverted fluorescence microscope (Nikon TE2000) equipped with a cooled CCD camera (CoolSNAP ES, Photometrics, Tucson, AZ) using Metamorph (Molecular Devices, Sunnyvale, CA) for image capture and processing.

Optic Nerve Crush and Histology

Optic nerve crush lesions were performed on adult zebrafish, 6–12 months of age, as previously described (Bormann et al., 1998). Briefly, the left optic nerve of anesthetized fish was exposed and crushed for 5 sec using Dumont no. 5 forceps. The intact right optic nerve served as an unoperated control. Fish were sacrificed 7 days after optic nerve crush. Eyes were dissected and fixed overnight in 4% paraformaldehyde and prepared for frozen sectioning as previously described (Barthel and Raymond, 1990). Twenty-micron sections were collected on Superfrost plus slides, dried at room temperature, and coverslipped in Vetashield Hardset Mounting Media with DAPI (Vector Laboratories, Burlingame, CA). Images of GFP and DAPI staining in the sectioned retinas were captured with the same inverted microscope setup described above. Images were overlaid, cropped, oriented, and assembled using Adobe Photoshop (San Jose, CA).


A 3.6-kb sequence, which includes 5′ flanking regions and exon 1 from the fugu gap43 (Japanese pufferfish; FUGU assembly 4.0), was compared with sequences from 5′ flanking regions and exon 1 sequences from Tetraodon nigroviridis (Green spotted pufferfish; TETRAODON assembly 7), Oryzias laptipes (medaka; MEDAKA assembly 1), and Gasterosteus aculeatus (stickleback; assembly BROAD S1). The 3.6-kb fugu sequence was also compared to 5′ flanking regions, exon 1, and intron 1 sequences from Gallus gallus (chicken; assembly WASHUC2), Mus musculus (mouse; NCBI assembly m36), Rattus norvegicus (rat; RGSC assembly 3.4) and Homo sapiens (human; NCBI assembly 36). Intron 1 sequences of the mammalian gap43 genes were included because they have previously been implicated in promoting neuron specificity in transgenic mice (Vanselow et al., 1994). Sequences were compared using three different publicly available programs (1) BLAST 2 Sequences (; (Altschul et al., 1990), (2) EvoPrinter (; (Odenwald et al., 2005), and (3) Multi PipMaker (; (Schwartz et al., 2000; Elnitski et al., 2002). Repeat sequences in the 3.6-kb promoter fragment were detected using the RepeatMasker program (Smit, AFA, Hubley, R & Green, P. RepeatMasker Open-3.0. 1996–2004 <>). In some cases, annotation of exons within the gap43 genomic region in Ensembl genomic sequence databases did not match the known cDNA sequence. Therefore, gap43 orthologues for the different species were identified using the gap43 cDNA sequences to query species-specific genomic sequence databases at the Ensembl website using BLAST. A similar method was used to identify eva1 and zbtb20 in medaka and chicken, respectively. Regions of the gap43 promoter/enhancer that were conserved between pufferfish, medaka, and stickleback were analyzed for putative transcription factor binding sites using TESS (Transcription Element Search System,; Schug, 2003) and MatInspector (; Cartharius et al., 2005). Default TESS parameters were used with the exception of the following: (1) Maximum Allowable String Mismatch % (tmm) = 5, (2) Minimum lg likelihood ratio (ta) = 10, (3) Factor attribute = Organism classification, metazoa. These changes increased the stringency of the search so that only the most specific matches were detected. MatInspector sites with matrix similarity scores of 0.8 or higher were reported.


We acknowledge Angela Schmoldt and undergraduate assistants, Dana Dirkintis and Carolyn Umbreit, for expert management of the fish facility and maintenance of the many lines created for this work. We also thank Drs. R. D. Heathcote and J. L. Witten for constructive comments and criticism of the manuscript prior to submission. Finally, we acknowledge the The University of Wisconsin-Milwaukee startup funds.