miRNAs are a recently discovered class of genes that regulate gene expression at the posttranscriptional level and are required for development, stem cell maintenance, and renewal (5–18). Recently, Yin et al. (19) showed that fibroblast growth factor (Fgf) signaling alters the expression of multiple miRNAs during regeneration. One of the miRNA targets of Fgf signaling, miR-133, targets mps1, which encodes a kinase that regulates blastemal proliferation. Interestingly, these authors also found that various other markers of regeneration were indirectly activated on the reduction of miR-133 levels, suggesting that overall regulation of regeneration by miRNAs might be quite complex. Here we show that an intact miRNA pathway indeed is essential for regeneration. Furthermore, we show that in addition to regulation of Fgf signaling during regeneration, Wnt signaling also subject is to miRNA regulation through miR-203 control of Lef1.

To examine global miRNA expression patterns in regenerating fins, we first conducted microarrays. Caudal fins were amputated from adult fish, and RNA was isolated from three regenerative states: adult fins, fins undergoing active regeneration, and fins that appeared to be completely regenerated. Small RNAs from each stage were isolated, fluorescently labeled, and directly hybridized to microarrays to determine the expression patterns of 346 vertebrate miRNAs (20). To obtain sufficient RNA for three independent arrays, fins were amputated from 120 adult fish, which were then returned to the aquarium temperature of 27 °C, after which regeneration was allowed to proceed for 2 or 5 weeks at 27 °C before reamputation and another round of RNA isolation. At this temperature, and based on the position of amputation, regeneration was ≈30% complete by 2 weeks [supporting information (SI) Fig. S1] and was nearly complete by 5 weeks. Heat maps illustrating global changes in miRNA expression are given in Fig. S2; expression changes for individual miRNAs, along with corresponding fold changes and P values, are given in Fig. S3 and Table S1. Some miRNAs that change during regeneration did not appear to return to their previous expression levels after 5 weeks at 27 °C, possibly due to incomplete regeneration. We hypothesize that miRNAs exhibiting decreased expression during active regeneration enable expression of genes required for regeneration (19), whereas miRNAs that are up-regulated during regeneration repress genes that normally prevent proliferation and/or maintain terminal cell differentiation.

To validate our approach, we chose to first focus on those miRNAs whose expression was altered most dramatically (either up or down) and that are predicted to target genes implicated in regeneration and/or genes whose expression changes during active regeneration (Table 1). For example, the arrays showed that expression of miR-200b increased during regeneration and that one of its predicted targets is bmp3, which has been shown to be correspondingly down-regulated (21). Similarly, miR-203 is down-regulated during regeneration and is predicted to target lef1. Regeneration requires up-regulation of lef1 in newly formed regenerative epithelia and can be used as a marker for the initiation of regeneration (21–23). Thus, significant down-regulation of miR-203 during regeneration is consistent with the up-regulation of lef1 that occurs during active regeneration. Furthermore, blastema formation and maintenance of blastema cells (24) requires an active form of the heat-shock protein 60 (hsp60) and two of the down-regulated miRNAs identified in our array (miR-2 and miR-338) are predicted to target hsp60. Finally, msxb has been postulated to regulate the rate of proliferation of blastema cells during regenerative outgrowth (25, 26), and expression of miR-301, which is predicted to target msxb, is down-regulated during regeneration. These results suggest that our array strategy was able to identify differentially expressed miRNAs that target genes involved in regeneration.

We next sought to test whether miRNAs directly regulate regeneration. First, we decided to block overall miRNA production in regenerating fins by introducing two different antisense morpholino oligonucleotides against Dicer, an enzyme required for cytoplasmic processing of miRNA precursors (27). The loss of Dicer is embryonically lethal in both fish and mice (15, 28, 29), and stem cell maintenance requires Dicer (16, 17, 30). To silence Dicer, we used a previously described antisense morpholino oligonucleotide complementary to the 5′ UTR of dicer mRNA (28) (Fig. 1), as well as a second dicer morpholino against the translational start site. For these experiments, fish were placed in tanks at 33 °C after amputation, injection, and electroporation, with regeneration then monitored over time. Regeneration proceeds normally but at about twice the rate at 33 °C compared with that at 27 °C (2) (Fig. S1). The increased rate of regeneration at this temperature allows disruption of gene function by morpholinos, which can persist for up to 72 h (31). Because the array data were obtained from fish maintained at 27 °C, we first verified, by northern blot analysis, that the patterns of expression for differentially expressed miRNAs, as determined using arrays, were similarly altered when regeneration proceeded at 33 °C (Fig. S4 a and b and data not shown). Northern blot analyses of miR-203 levels showed a dramatic decrease in expression when regeneration proceeded at 33 °C, identical to that found using microarrays performed on fins isolated from fish maintained at 27 °C.

The morpholinos against dicer were injected into the lepidotrichia (bony rays) on the dorsal half of amputated fins, followed by electroporation to facilitate cellular uptake (25, 32) (Fig. 1A). Because newly amputated fins cannot be injected immediately, we waited 2 days after amputation, to allow wound healing, before injectiing morpholinos or RNAs. The entire fin was subjected to electroporation, but only the dorsal half was injected with morpholinos or RNAs, such that the ventral half served as an internal control. After injection of the two different morpholinos against dicer and a mismatched control morpholino, we measured the microns of regrowth on the dorsal halves versus the ventral halves of electroporated fins. As shown in Fig. 1 B and C, regeneration was significantly inhibited in the dorsal halves of fins injected with the dicer morpholinos compared with those injected with either the mismatch morpholino or uninjected controls (UICs). The effect was pronounced at 6 days postamputation (dpa), after which the fins began to regenerate over time (data not shown), which is consistent with the half-life of the morpholinos but also indicates that the injections did not irreversibly block regeneration. These results demonstrate that an intact miRNA pathway is essential for regeneration and are consistent with the hypothesis that induction of specific miRNAs is needed for proper regeneration (19).

To investigate the function of miRNAs that are down-regulated during regeneration and to test the role of individual miRNAs, we focused on miR-203. miR-203 was found to be significantly down-regulated during regeneration and is predicted to target lef1 (Table 1), a Wnt signaling-regulated transcription factor, the transcription of which is induced during regeneration and serves as a marker of the basal epidermal layer during blastema formation (22). The 3′ UTR of lef1 contains two potential miRNA recognition elements (MREs) for miR-203 (Fig. 2A). As a first test of the hypothesis that miR-203 targets lef1, we created a construct that fused the 3′ UTR of lef1 to the coding region of GFP (Fig. 2A). We then injected in vitro transcribed GFP fusion transcripts into single-cell zebrafish embryos in the presence or absence of exogenous miR-203 and examined fluorescence levels in living embryos at 1 day postfertilization (dpf) (Fig. 2B). We also co-injected antisense morpholinos complementary to miR-203 (i.e., miR-203^MO) that function to block the activity of endogenous miR-203 and thus should lead to increased fluorescence (31, 33). As a control, GFP lacking the lef1 3′ UTR also was injected into single-cell zebrafish embryos with and without exogenous miR-203 and miR-203^MO (Fig. 2B). As shown in Fig. 2B, co-injection of miR-203 and GFP+lef1 clearly reduced GFP levels in embryos, whereas co-injection of miR-203^MO and GFP+lef1 resulted in increased fluorescence. In parallel assays, western blot analyses of pooled embryo lysates were performed using antibodies against GFP and α-tubulin (Fig. 2 C and D). Consistent with the fluorescence experiments, GFP levels were decreased significantly on co-injection of miR-203, and knockdown of endogenous miR-203 by injection of miR-203^MO led to increased GFP levels (Fig. 2 D and E). GFP levels were significantly lower in the GFP+lef1 injected embryos compared with the GFP-UTR control. This lower fluorescence can be attributed to basal miR-203 expression in 1 dpf embryos (Fig. S4a). Together, the GFP reporter assays suggest that miR-203 specifically targets lef1.

To test whether miR-203 regulates lef1 during actual regeneration, we amputated caudal fins and injected excess miR-203, then performed electroporation. Evaluation of fin regeneration under these conditions revealed a definite loss of regenerative outgrowth in the dorsal halves of fins injected with miR-203 compared with the control ventral halves (UICs) (Fig. 3 A and B) or with injection of miR-15b, a miRNA whose expression does not change during regeneration (Fig. S5a). To directly examine Lef1 levels under these conditions, we immunostained whole regenerating fins with antibodies against Lef1. Lef1 levels were dramatically decreased in the outermost epithelial layer on injection of miR-203 in the dorsal halves, but not in the ventral halves, of regenerating fins (Fig. 3A). Similar Lef1 levels were detected in both the dorsal and ventral halves of regenerating UIC fins (Fig. 3B). As an additional control, we injected morpholinos against a region including the translation start site of lef1 mRNA to determine whether a similar phenotype to miR-203 overexpression could be found. Consistent with the regulation of lef1 by miR-203, knockdown of Lef1 inhibited regeneration in the dorsal halves of injected fins (Fig. S5b). These results are entirely consistent with the hypothesis that miR-203 regulates Lef1 levels during regeneration.

We next injected miR-203^MO into regenerating fins to determine the effect of loss of miR-203 on regeneration. Two different morpholinos were used, the same miR-203^MO as before and a second morpholino against the pre-miR-203 loop region. In both cases, we observed a dramatic and remarkable increase in fin growth (Fig. 3C and Fig. S5 c and d). Some of the morpholino experiments resulted in increased fin length, whereas others demonstrated widening of the regenerating dorsal side, such that we quantitated the increase in regeneration for these experiments by measuring the increase in total surface area between the dorsal and ventral halves (Fig. 3E). Consistent with the increase in size being due to knockdown of miR-203, immunostaining and western blot analyses of Lef1 protein levels in the dorsal halves of injected fins demonstrated significantly increased expression of Lef1 (Fig. 3 C and F). Thus, regulation of Lef1 by miR-203 appears to be a critical limiting step in regeneration. The data also suggest that reestablishment of miR-203 levels is necessary for proper termination of regeneration.

To further test the hypothesis that miR-203 regulates lef1, we performed in situ hybridization experiments to localize miR-203 and lef1 RNAs on sections from a single bony ray flanked by adjacent mesenchymal tissue. We found a perfect correlation between the presence of miR-203 and the loss of lef1 (Fig. S4c). Normally, miRNAs repress translation, but continued repression can lead to mRNA decay (34, 35). We also sectioned fins and conducted immunostaining with antibodies against Lef1 or, as a control, β-catenin. As before, Lef1 levels increased during normal regeneration, decreased on co-injection of miR-203, and were greatly elevated by co-injection of miR-203^MO, whereas the β-catenin levels did not change under the different conditions (data not shown).

To ensure that the effect of miR-203 is direct, we evaluated whether the expression of Lef1 from mRNAs lacking a normal 3′ UTR would be able to rescue the repression observed in the presence of miR-203. Co-injection of miR-203 with lef1 mRNAs lacking MREs led to a substantial rescue of both fin regeneration and Lef1 protein levels (Fig. 3 D and E). The overall rescue was dose-dependent (Fig. S6). Finally, we injected morpholinos against lef1 in the presence and absence of co-injected miR-203^MO (Fig. S5b, e, and f). The loss of lef1 led to inhibition of regeneration, and the overgrowth normally observed on injection of miR-203^MO was blocked by co-injection of lef1^MO, consistent with the hypothesis that miR-203 directly regulates lef1 expression.

Our gain-of-function and loss-of-function regeneration experiments validate our array data and also suggest that a possible mechanism for termination of regeneration may be the increased expression of specific miRNAs that control the genes essential for regeneration. Consistent with this, the microarray and northern blot analysis results (Fig. S4b) demonstrate that miR-203 is readily detectable in adult fins and postregeneration fins but mostly undetectable in regenerating fins, suggesting transcriptional regulation of miR-203. Our results point to a key limiting role for lef1 in regeneration, but the possibility remains that miR-203 regulates other genes involved in regeneration as well. Considering the Dicer experiments shown in Fig. 1, along with data indicating a role for miR-133 in regeneration (19), it seems clear that the overall regeneration process is subject to extensive control by miRNAs (Fig. 4).