DEVELOPMENTAL DYNAMICS. Introduction RESEARCH ARTICLE

Transcription

1 a 243: , 2014 DOI: /DVDY RESEARCH ARTICLE Dynamic mirna Expression Patterns During Retinal Regeneration in Zebrafish: Reduced Dicer or mirna Expression Suppresses Proliferation of M uller Glia-Derived Neuronal Progenitor Cells Kamya Rajaram, 1 Rachel L. Harding, 2 Travis Bailey, 2 James G. Patton, 1 * and David R. Hyde 2 * 1 Department of Biological Sciences, Vanderbilt University, Nashville, Tennessee 2 Department of Biological Sciences and the Center for Zebrafish Research, University of Notre Dame, Notre Dame, Indiana Background: Adult zebrafish spontaneously regenerate their retinas after damage. Although a number of genes and signaling pathways involved in regeneration have been identified, the exact mechanisms regulating various aspects of regeneration are unclear. micrornas (mirnas) were examined for their potential roles in regulating zebrafish retinal regeneration. Results: To investigate the requirement of mirnas during zebrafish retinal regeneration, we knocked down the expression of Dicer in retinas prior to light-induced damage. Reduced Dicer expression significantly decreased the number of proliferating M uller gliaderived neuronal progenitor cells during regeneration. To identify individual mirnas with roles in neuronal progenitor cell proliferation, we collected retinas at different stages of light damage and performed small RNA high-throughput sequencing. We identified subsets of mirnas that were differentially expressed during active regeneration but returned to basal levels once regeneration was completed. We then knocked down five different mirnas that increased in expression and assessed the effects on retinal regeneration. Reduction of mir-142b and mir-146a expression significantly reduced INL proliferation at 51h of light treatment, while knockdown of mir-7a, mir-27c, and mir-31 expression significantly reduced INL proliferation at 72h of constant light. Conclusions: mirnas exhibit dynamic expression profiles during retinal regeneration and are necessary for neuronal progenitor cell proliferation. Developmental Dynamics 243: , VC 2014 The Authors. Developmental Dynamics published by Wiley Periodicals, Inc. on behalf of American Association of Anatomists Key words: mirnas; retinal regeneration; Dicer; zebrafish Submitted 8 April 2014; First Decision 6 August 2014; Accepted 26 August 2014; Published online 13 September 2014 Introduction A variety of different damage models demonstrated that zebrafish possess the ability to spontaneously regenerate any retinal cell type that is lost/damaged and restore a functional visual system (Cameron, 2000; Vihtelic and Hyde 2000; Fausett and Goldman, 2006; Raymond et al., 2006; Bernardos et al., 2007; Fimbel et al., 2007; Sherpa et al., 2008; Ariga et al., 2010; Montgomery et al., This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made. Additional Supporting Information may be found in the online version of this article. Grant sponsor: NIH; Grant number: R21 EY and R01 EY Kamya Rajaram and Rachel L. Harding contributed equally to this work. *Correspondence to: David Hyde, 027 Galvin Life Science Building, Department of Biological Sciences, University of Notre Dame, Notre Dame, IN dhyde@nd.edu; James G. Patton, 2325 Stevenson Center, Box 1820 Station B, Vanderbilt University, Nashville, TN James.G.Patton@Vanderbilt.edu 2010). All these damage models induce the M uller glia to dedifferentiate and divide asymmetrically to produce neuronal progenitor cells (Nagashima et al., 2013). These progenitor cells continue to proliferate and migrate to the damaged retinal layer and specifically differentiate into the neuronal class that was lost. While M uller glia are found in all vertebrate retinas (Lamba et al., 2008), the mammalian retina is largely incapable of regeneration; damage normally triggers reactive gliosis that involves hypertrophy, cell proliferation, and expression of a specific subset of genes that leads to glial scarring (Bringmann et al., 2009). Using different damage models, several groups performed microarray experiments and identified numerous candidate genes that change significantly in expression during retinal regeneration (Cameron et al., 2005; Kassen et al., 2007; Craig et al., 2008; Qin et al., 2009; Morris et al., 2011). Subsequently, a number of proteins and signaling pathways have been shown to be necessary Article is online at: /abstract VC 2014 The Authors. Developmental Dynamics published by Wiley Periodicals, Inc. on behalf of American Association of Anatomists 1591

2 1592 RAJARAM ET AL. for zebrafish retinal regeneration (Goldman, 2014; Gorsuch and Hyde, 2014; Lenkowski and Raymond, 2014). To precisely control the complex patterns of gene expression that occur during regeneration response in the zebrafish retina, multiple modes of regulation are employed. micrornas (mirnas) have emerged as an important class of post-transcriptional regulators of various biological processes including development, differentiation, cell specification, proliferation, and regeneration (Lee et al., 1993; Reinhart et al., 2000; Slack et al., 2000; Ambros, 2003; Bartel and Chen, 2004; Morris and McManus, 2005;, O Donnell et al., 2005; Slack and Weidhaas, 2006; Plasterk, 2006; Thatcher and Patton, 2010). mirnas are evolutionarily conserved, small non-coding RNAs that arise from RNA polymerase II-derived endogenous hairpin precursors (Cai et al., 2004; Lee et al., 2004). These transcripts are sequentially processed by the RNaseIII-like enzymes Drosha and Dicer into small, 22 nt, mature mirnas (Lee et al., 2003; Chendrimada et al., 2005; Gregory et al., 2004). mirnas function by imperfect base pairing with the 3 0 untranslated region (UTR) of target mrnas, causing mrna destabilization and decay and/or repression of protein production (Kloosterman and Plasterk, 2006). Complex tissue or organ regeneration in a number of species is regulated by mirnas (Thatcher and Patton, 2010, Thatcher et al., 2008, Yin et al., 2012, Yin et al., 2008). In zebrafish, mir-203 regulates fin regeneration (Thatcher et al., 2008), while mir-133 family members regulate fin, spinal cord and heart regeneration (Yin et al., 2008; Yin et al., 2012; Yu et al., 2011). The let-7a mirna was demonstrated to repress the expression of several proteins that are necessary for M uller glia dedifferentiation and proliferation in the puncture-damaged zebrafish retina, which suggests that let-7a expression must be repressed for the M uller glia to initiate a regeneration response (Ramachandran et al., 2010). Additionally, we showed that an intact mirna biogenesis pathway is necessary for zebrafish caudal fin regeneration (Thatcher et al., 2008). However, it remains unexplored if there is a similar global requirement of mirnas for successful retinal regeneration. Here, we demonstrate that the Dicer-dependent mirna biogenesis pathway is essential for normal retinal regeneration in zebrafish. Using small RNA high-throughput sequencing, we show that distinct subsets of mirnas are differentially expressed during retinal regeneration, but return to basal expression levels after completion of regeneration. Using loss-of-function studies in the regenerating retina, we illustrate that differentially expressed mirnas function to regulate the number of proliferating M uller glia-derived neuronal progenitor cells during adult zebrafish retinal regeneration. Results Loss of Dicer Inhibits Retinal Regeneration Fig. 1. Morpholino-mediated knockdown of Dicer protein expression. Lissamine-tagged Standard Control morpholino (Std Ctl MO) or dicer morpholino were intravitreally injected and electroporated into darkadapted albino retinas. After 35h constant light, dosal retinas were isolated and the expression of Dicer and Actin were analyzed by immunoblots (A). After incubation with the secondary antibody, the membrane was separated to allow a longer exposure of the Dicer region relative to the Actin region. After adjusting for the difference in the amount of the Actin protein in the two samples, the dicer morphant retinas possessed 77% less Dicer protein relative to the Standard Control morphant retina (B, *P<0.05, n¼5). Pre-miRNAs are processed in the cytoplasm by an RNAse III-like enzyme, Dicer. To determine if mirnas are important for zebrafish retinal regeneration, we knocked down Dicer protein expression in adult albino zebrafish retina prior to constant intense light damage using morpholinos (MO). We intravitreally injected and electroporated either the dicer morpholino (Wienholds et al., 2003) or the Standard Control morpholino (which is not complementary to any known sequence in the zebrafish genome, Gene- Tools) into dark-adapted albino retinas. After exposing the fish to constant intense light for 35h, the dorsal retinas (where the morpholinos were most efficiently electroporated) were isolated and protein homogenates were analyzed by immunoblots (Fig. 1A). The dicer morphant retinas possessed 77% less Dicer protein relative to the Standard Control morphant retina (Fig. 1B, P<0.05, n¼5). Thus, the dicer morpholino significantly reduced the amount of Dicer protein in the light-damaged retina. To examine the effect of Dicer knockdown on M uller glia and neuronal progenitor cell proliferation in the light-damaged retina, dark-adapted zebrafish retinas were intravitreally injected and electroporated with either a 5-base mismatch control morpholino or a dicer morpholino. Fish were then exposed to constant intense light for 0, 16, 35, 51, 68, and 96h and retinal sections were immunolabeled for PCNA (Proliferating Cell Nuclear Antigen) as a marker for proliferating cells (Fig. 2). At the start of light damage (0h of light), PCNA expression was nearly absent in the INL of 5-mis control ( ) and dicer morphant ( ) retinas (Fig. 2I), with very few PCNApositive nuclei detected in the ONL of 5-mis control and dicer morphant retinas ( and , respectively; Fig. 2J). After 16h of light damage, when photoreceptor cell death is nearing its maximal level (Nelson et al., 2013), there was no difference in the number of PCNA-positive nuclei between the 5-mis and dicer morphants, in either the ONL (1061 and 862, respectively) or the INL (160.2 and 160.4, respectively). Furthermore, there was no significant difference in the number of TUNEL-

3 FUNCTIONS OF MIRNAS DURING RETINAL REGENERATION 1593 Fig. 2. Dicer knockdown decreased INL proliferation in the light-damaged retina. Lissamine-tagged dicer 5-base mismatch (5-mis) control morpholino or dicer morpholino were intravitreally injected and electroporated into dark-adapted adult albino zebrafish. Retinas were collected at 0, 35, 51, 68, and 96h of light damage and immunostained with anti-pcna (green) antibodies and TOPRO3 nuclear stain (blue). A, E: At 35h, single PCNA-positive M uller glia were observed in the INL of dicer 5-mis control and dicer morphant retinas (arrows). B, F: Both dicer 5-mis control and dicer morphant retinas contain doublet nuclei at 51h of light damage (arrows). C: Clusters of proliferating progenitor cells are present in the INL of dicer 5-mis control retinas (arrow) at 68h of light damage. G: Single (arrow) or doublet (arrowhead) nuclei predominated in dicer morphant retinas. D: Columns of proliferating progenitors were observed in the INL of dicer 5-mis control morphant retinas at 96h of light damage. H: At 96h of light damage, doublet nuclei (arrowhead) predominated the INL of dicer morphants. I: Significantly fewer PCNA-positive INL cells were present in dicer morphant retinas compared to dicer 5-mis control morphant retinas beginning at 68h of light damage. J: dicer morphant retinas contained significantly fewer PCNA-positive ONL cells at 96h of light damage. dicer 5-mis MO, dicer 5-base mismatch control morphant; dicer MO, dicer morphant; INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar in A ¼ 50 mm and applies to B H; data represent mean 6 s.e.m; *P< 0.05 using two-way ANOVA with a Tukey s post-hoc test, n¼11. positive ONL photoreceptors between the 5-mis and dicer morphants (2565 and 2064, respectively). After 35h of constant light, when M uller glia reenter the cell cycle (Kassen et al. 2007), nearly equivalent numbers of PCNApositive cells were present in the INL of both 5-mis control and dicer morphant retinas (2966 and 1864, respectively), as well as in the ONL (1261 and 1162, respectively; Fig. 2A, E, I, and J). At 35h, there remained no significant difference in the number of TUNEL-positive ONL photoreceptors between the 5-mis and dicer morphants (1164 and 762, respectively). After 51h of light damage, when M uller glia begin producing neuronal progenitor cells (Kassen et al., 2007), we observed two closely associated (doublet) PCNA-positive nuclei in the INL of both 5-mis control and dicer morphant retinas (arrows, Fig. 2B and 2F, respectively), with the 5-mis control and dicer morphants containing approximately the same number of PCNA-positive INL cells (6666 and 4167, respectively; Fig. 2I). Similarly, nearly equivalent numbers of PCNA-positive ONL cells were present in the 5-mis control and dicer morphant retinas (2063 and 1762, respectively; Fig. 2J). At 68h of constant light, when clusters of proliferating PCNA-positive neuronal progenitor cells were observed in the INL of 5-mis control morphants (Fig. 2C, arrow), only individual or doublet PCNA-positive INL nuclei persisted in dicer morphant retinas (Fig. 2G, arrow and arrowhead, respectively). This suggested that either the M uller glia, or initial neuronal progenitor cells, failed to continue proliferating. Additionally, the dicer morphants contained significantly fewer PCNA-positive INL cells relative to the 5-mis morphants (6368 and 9067, P¼0.042, n¼11; Fig. 2I). In contrast, there was an equivalent number of PCNApositive ONL cells in the dicer and 5-mis control retinas (2664 and 3364, respectively; Fig. 2J). Because the dicer morphants did not possess fewer TUNEL-positive ONL neurons at either 16 or 35h relative to 5-mis control morphants, the reduction in the number of PCNA-positive neuronal progenitor cells is likely a direct effect on neuronal progenitor cell proliferation rather than reducing the number of dying retinal neurons. By 96h of light treatment, single, doublet (Fig. 2H, arrowhead), and small clusters of PCNA-positive INL cells persisted in dicer morphant retinas relative to the primarily large clusters of PCNApositive cells migrating from the INL to ONL in the 5-mis control morphant retinas (Fig. 2D, arrow). However, the dicer morphant contained approximately the same number of PCNA-positive INL cells compared to the 5-mis control morphant retinas (9667 and

4 1594 RAJARAM ET AL , respectively; Fig. 2I), but significantly fewer PCNApositive ONL cells (4364 and 6768, respectively, P¼0.01, n¼11; Fig. 2J). This could result from the significant difference in the number of PCNA-positive INL cells that were observed at 68h that then migrated to the ONL by 96h. Overall, these observations indicated that loss of Dicer expression reduced, without completely blocking, M uller glia-derived neuronal progenitor cell proliferation in the light-damaged retina. We confirmed that the reduced proliferation observed in the dicer morphant retina was not due to the loss of M uller glia by knocking down Dicer expression in Tg(gfap:egfp)nt11 transgenic zebrafish that express EGFP in all M uller glia from the gfap (glial fibrillary acidic protein) promoter (Kassen et al., 2007). We immunostained for EGFP (Fig. 3A D, I L) and PCNA (Fig. 3E -L) in dicer morphants after 35 (Fig. 3B, F, and J) and 51h of constant light (Fig. 3D, H, and L) and the Standard Control morphant (Fig. 3A, C, E, G, I, and K), or the 5-mis control morphant (data not shown). As expected, there was no statistical difference in the number of EGFP-positive M uller glia in either the INL or the ONL between the dicer and control morphant retinas at either timepoint (Fig. 3O and P, respectively), demonstrating that the reduction in PCNA-positive cells in the dicer morphant was not the result of fewer M uller glia. The EGFP-positive ONL cells likely represent M uller glia that migrated from the INL to the ONL through interkinetic nuclear migration (Nagashima et al., 2013; Lahne and Hyde, data not shown). As observed in the albino retinas (Fig. 2), the dicer morphant did not exhibit any statistically significant difference in the number of PCNA-positive cells in either the INL or ONL relative to standard control morphant retinas after 35h of constant light (Fig. 2M and N, respectively). However, the dicer morphant did exhibit significantly fewer PCNA-positive cells in the INL relative to the control after 51h of constant light (Fig. 2M, 5369 and 7966, respectively, P¼0.011), when the M uller glia-derived neuronal progenitors are beginning to amplify in number. Thus, both the albino and Tg(gfap:egfp)nt11 transgenic retinas reveal that knockdown of Dicer expression results in significantly fewer proliferating neuronal progenitor cells in the regenerating light-damaged retina. mirnas Are Differentially Expressed During Retinal Regeneration The Dicer loss-of-function experiments suggest that global mirna biogenesis is necessary for maximal production of proliferating M uller glia-derived neuronal progenitors. To identify specific mirnas that are critical for neuronal progenitor cell proliferation during regeneration of the light-damaged retina, we performed small RNA high-throughput sequencing. RNA was isolated and sequencing libraries were prepared from albino zebrafish retinas before damage (0h of light), at two different time points during active retinal regeneration (35h and 72h of light) and after completion of regeneration (28 days post recovery). These time points were chosen because M uller glia dedifferentiate and reenter the cell cycle by 35h of light exposure and M uller glia-derived neuronal progenitors are maximally proliferating and forming clusters that begin migrating to the ONL by 72h of light exposure (Kassen et al., 2007). To first confirm the reliability of library preparation and sequencing, we prepared and sequenced two independent small RNA libraries from undamaged retinas (0h of light). Global mirna expression was nearly identical between the biological replicate libraries (Pearson s correlation r¼0.98; R 2 ¼0.97). Additionally, we sequenced the second library on two different Illumina machines (GAII and HiSeq 2000) and found that the datasets were highly comparable between the technical replicates (Pearson s correlation r¼0.99; R 2 ¼0.99), ensuring consistent and reliable library production. Sequencing identified a total of 199 unique mirnas that are expressed in the adult zebrafish retina (see Supp. Tables S1 and S2, which are available online). Global mirna expression changes during the four time points tested are illustrated in the heat map in Figure 4. Most mirna expression levels remained unchanged or underwent only small changes during regeneration relative to the undamaged retina (Fig. 5A C). However, small subsets of mirnas either increased or decreased in expression at the different time points (Pearson s correlation r¼0.88 for 0h vs. 35h and 0.94 for 0h vs. 72h; R 2 ¼0.78 for 0h vs. 35h and 0.88 for 0h vs. 72h). Significantly, nearly all of these mirnas returned to their baseline expression (0h) levels following completion of regeneration by the 28d time point (Fig. 5C) (Pearson s correlation r¼0.96; R 2 ¼0.97 for 0h vs. 28d). Differences in read numbers were then normalized to total read numbers in each library and fold changes were determined. Of the mirnas examined, 36 had over a 2-fold change in expression at the 35h and/or 72h timepoints relative to 0h (Fig. 5D F), including 13 upregulated mir- NAs (Fig. 5D), 22 downregulated mirnas (Fig. 5E), and one mirna that was both up and down-regulated at different stages of regeneration (denoted by ** in Fig. 5D E). To validate the sequencing results, we examined expression of three random mirnas (mir-21, mir-130b, mir-7a) using Taqman qpcr. To normalize the expression of these mirnas across different time points, we compared expression levels to mir-9, a mirna with mid-range expression levels in the high-throughput sequencing dataset and whose expression did not change across the time points used. The relative mirna expression changes detected by high-throughput sequencing (Fig. 6A) agreed with those revealed by qpcr (Fig. 6B C). Upregulated mirnas Are Necessary for Maximal Neuronal Progenitor Cell Proliferation During Retinal Regeneration Based on the sequencing results, we selected 6 upregulated mirnas for further analysis that displayed different expression patterns and different read abundance levels, suggesting distinct roles for these mirnas at various stages of early regeneration (Fig. 7, Table 1). We selected mirnas that exhibited increased expression at the 35 and/ or 72h timepoints so that we could use morpholinos to block their processing and determine how reducing their expression affected M uller glia and neuronal progenitor cell proliferation. mirs-7a, 2142b, and 2146a were upregulated to the same level at both 35 and 72h (Fig, 7A), which suggests that they might play roles in M uller glia dedifferentiation and proliferation, as well as neuronal progenitor cell proliferation and migration (Kassen et al., 2007). mirs-27c and 231 were upregulated by 35h light damage and peaked at 72h, implying important roles in neuronal progenitor cell proliferation and migration (Fig. 7B; Kassen et al., 2007). mir-2190 expression peaked at 35h and decreased by 72h of light, which suggested a possible role in M uller glia dedifferentiation and proliferation (Fig. 7C; Kassen et al., 2007). To test the potential roles of the mirnas on neuronal progenitor cell proliferation, dark-adapted albino zebrafish retinas were intravitreally injected and electroporated with morpholinos prior to

5 FUNCTIONS OF MIRNAS DURING RETINAL REGENERATION 1595 Fig. 3. Dicer knockdown does not affect M uller glia proliferation in light-damaged retinas. Adult albino Tg(gfap:egfp)nt11 zebrafish were intravitreally injected and electroporated with standard control or dicer morpholino prior to the start of light damage. A L: EGFP-positive M uller glia (arrows) were observed in comparable numbers between dicer and standard control morphants at all time points (A D, I L, O, P). E,F: PCNA-positive INL cells were observed as mainly single nuclei in dicer and standard control morphant retinas at 35h of light damage (arrows). IJ: Overlay of these images revealed that many M uller glia co-labeled with PCNA as they re-entered the cell cycle. G: Standard control morphant retinas at 51h of light damage exhibited doublets or early columns of PCNA-positive INL cells. H: Single PCNA-positive INL cells rather than doublets predominated in dicer morphant retinas (arrows). M N: Differences in proliferation in dicer morphants resulted in significantly fewer PCNA-positive INL cells at 51h of light, but no difference in PCNA-positive ONL cells. Std Ctl MO, Standard Control morphant; dicer MO, dicermorphant; INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar in A ¼ 50 mm and applies to B L; *P<0.05 using two-way ANOVA with a Tukey s post-hoc test, n¼7. starting intense light exposure. We did not examine the effect of the morpholinos at 35h, when the M uller glia dedifferentiate and reenter the cell cycle (Kassen et al., 2007) because knockdown of Dicer expression did not reveal any effect on M uller glia proliferation (Figs. 2 and 3). We assessed the effects of reduced mirna expression at two different time points (51h and 72h of light) during regeneration (Figs. 8 11). At 51h of light exposure, the M uller glia-derived neuronal progenitors are beginning to proliferate,

6 1596 RAJARAM ET AL. Fig. 4. Global mirna expression patterns during retinal regeneration. Dark-adapted zebrafish (0h) were placed in constant intense light for 35h or 72h, or standard light conditions for 28 days after 96h of constant light. RNA was isolated from the retinas at all four timepoints and analyzed by high-throughput small RNA sequencing. Log 2 values of normalized mirna reads from each sequencing time point (0h, 35h, 72h, 28d) are represented in the heat map. Yellow indicates high expression, blue indicates low expression, and grey indicates no expression. mirnas tested by loss-of-function morpholino studies are highlighted with a red asterisk (*). while by 72h of light exposure, the M uller glia-derived neuronal progenitor cells are rapidly proliferating and generating clusters of PCNA-positive neuronal progenitor cells that are beginning to migrate to the ONL (Kassen et al., 2007). We sought to determine what effect loss of each of these mirnas would play early and late in neuronal progenitor cell amplification. mir-142b Of the six mirnas tested, loss of mir-142b had the most dramatic effect relative to the standard control morphant retina (Fig. 8A D), with a significant reduction in the number of PCNApositive INL cells at both 51h (4269 and 8665, respectively, P¼0.016, Fig. 8E) and 72h (77610 and 11769, respectively, P¼0.026, Fig. 8E) of constant light treatment. This suggests that mir-142b regulates early proliferation of the neuronal progenitor cells. The mir-142b morphant also possessed significantly fewer PCNA-positive ONL cells at 72h relative to the control (3366 and 5865, respectively, P¼0.017, Fig. 8F), although there was not a significant difference at 51h (2263 and 3063, respectively, Fig. 8F). This difference in the number of PCNA-positive ONL cells at 72h, but not at 51h, could be due to either a reduced number of M uller glia-derived neuronal progenitor cells that migrate to the ONL or a role for mir-142b in the continued proliferation of progenitor cells in the ONL. mir-146a Reduction of mir-146a expression resulted in significantly fewer PCNA-positive neuronal progenitor cells at 51h relative to the standard control morpholino (3066 and 8665, respectively, P¼0.006; Fig. 9A B, E). However, there was not a significant

7 FUNCTIONS OF MIRNAS DURING RETINAL REGENERATION 1597 Fig. 5. mirnas are differentially expressed during intense light damage induced retinal regeneration. A C: Global mirna expression changes during and after retinal regeneration. mirna expression at 35h (A) and 72h (B) of intense light exposure, and after 28 days of recovery in normal light conditions (C) was compared to the control expression levels (0h in intense light). Each data point represents one mirna. Pearson s correlation (r) and R 2 values are indicated. D F: mirnas with >2-fold change in expression during regeneration. D: Upregulated mirnas. mirnas tested by loss-of-function morpholino studies are highlighted in red. E: Downregulated mirnas. F: Two mirnas with expression changes out of scale in D and E (indicated with *). One mirna that is both upregulated and downregulated at different stages of regeneration is denoted with **. difference in the number of proliferating neuronal progenitor cells in the mir-146a relative to the control at 72h ( and 11769; Fig. 9C E). This suggests that expression of mir-146a plays an important role in M uller glia-derived neuronal progenitor cell proliferation. Additionally, there was no significant difference in the number of PCNA-positive ONL cells at either 51h or 72h between the mir-146a morphant and standard control morphant (Fig. 9F). The ability of the mir-146a morphant to reach the same number of proliferating INL neuronal progenitor cells as the standard control morphant at 72h could be due to (1) compensation of the early mir-146a loss by additional mirnas, most likely by the similar family member, mir-146b, that is also upregulated during regeneration (Fig. 5D), (2) cell cycle re-entry of additional M uller glia to compensate for the lack of progenitors, or (3) the short half life of mir-146a MO that could lead to de-repression of mir-146a (Selbach et al., 2008; Miska et al., 2007). mir-7a, mir-27c, and mir-31 Knockdown of either mir-7a, mir-27c, or mir-31 resulted in statistically equivalent numbers of PCNA-positive INL and ONL cells at 51h relative to the standard control morphant (Fig. 10A D, I-N). However, the mir-7a, mir-27c, and mir-31 morphant retinas contained fewer PCNA-positive cells than the standard control retina in both the INL and ONL at 72h (Fig. 10E H). At 72h, the mir-7a morphant contained statistically fewer PCNA-positive relative to the control in the INL (53613 and 11769, respectively, P¼0.001, Fig. 10I) and the ONL (1866 and 5865, respectively, P<0.001, Fig. 10L). Similarly, the mir-27c morphant contained statistically fewer PCNA-positive relative to the control in the INL (41614 and 11769, respectively, P¼0.001, Fig. 10J) and the ONL (2365 and 5865, respectively, P<0.001, Fig. 10M). Additionally, the mir-31 morphant contained statistically fewer PCNA-positive relative to the standard control morphant in the INL (7663 and 11769, respectively, P<0.05, Fig. 10K) and the ONL (3164 and 5865, respectively, P<0.05, Fig. 10N). The reduced number of PCNA-positive INL cells at 72h relative to 51h suggests that these mirnas may play important roles in the continued proliferation of M uller glia-derived neuronal progenitor cells, similar to Pax6a being necessary for later neuronal progenitor cell proliferation (Thummel et al., 2010). mir-2190 Knockdown of mir-2190 had no effect on the number of INL or ONL PCNA-positive cells during regeneration relative to the

8 1598 RAJARAM ET AL. Fig. 6. Validation of sequencing results. Three mirnas were randomly selected for validation of the sequencing results using Taqman qpcr. The expression of these mirnas across different time points was normalized to mir-9 expression, which exhibits mid-range expression levels in the high-throughput sequencing dataset and whose expression did not change across the time points used. A: mirna expression patterns during regeneration by high-throughput sequencing. B,C: qpcr verification of mirna expression levels relative to mir-9 is shown. Data represent mean 6 s.e.m. standard control morphant at either 51h (Fig. 11A, B, E, F) or 72h (Fig. 11C, D, E, F). This was a little surprising considering the large increase in mir-2190 expression at 35h and 51h relative to the undamaged retina (Fig. 7C). While this work was in progress, however, Wei et al (2012) discovered that mir-2190 is likely not a true mirna, but is a product of rrna degradation, which could account for the mir-2190 knockdown not affecting neuronal progenitor cell proliferation in light-damaged retinas. However, the mir-2190 morphant serves as a good control for Fig. 7. Expression patterns of mirnas chosen for MO-mediated knockdowns. Six mirnas that exhibited increased expression during the light-treatment timecourse were selected for morpholino-mediated knockdown. Log 2 values of mirna fold changes relative to 0h control values are shown. A: Three mirnas that are upregulated at both 51h and 72h (mir-7a, mir-142b, mir-146a). B: Two mirnas that peak at 72h (mir-27c, mir-31). C: One putative mirna, which is probably a product of rrna degradation, peaks at 35h (mir-2190). non-specific effects of the morpholinos on M uller glia and neuronal progenitor cell proliferation. Discussion We demonstrate that mirna processing is necessary for proliferation of the M uller glia-derived neuronal progenitor cells. Additionally this study represents the first thorough examination of the dynamic changes in mirna expression during regeneration of the adult zebrafish retina. Finally, we demonstrate that loss of individual mirnas can affect neuronal progenitor cell proliferation both early and late during regeneration. We first examined the relative importance of mirnas in retinal regeneration by knocking down the expression of Dicer. It was previously shown that individual mirnas play important roles in the regeneration of several different zebrafish tissues. let-7a is

9 FUNCTIONS OF MIRNAS DURING RETINAL REGENERATION 1599 TABLE 1. Upregulated mirnas Tested in Loss-of-Function Experiments a mirna Peak upregulation time point Log 2 (Peak fold change) mir-7a 35h, 72h 3 mir-27c 72h 3 mir-31 72h 4 mir-142b 35h, 72h 1 mir-146a 35h, 72h 5 mir h 9 a Based on the sequencing results, we selected 6 upregulated mirnas for further analysis that displayed different expression patterns and different read abundance levels, suggesting distinct roles for these mirnas. necessary for M uller glia dedifferentiation and proliferation by repressing the expression of several genes required for retinal regeneration (Ramachandran et al., 2010). Increased expression of mir-133b is necessary for spinal cord regeneration in zebrafish (Yu et al., 2011). mir-133 expression regulates zebrafish heart regeneration (Yin et al., 2012), while mir-133, mir-196, and mir-203 have all been shown to play important roles in zebrafish fin regeneration (Yin et al., 2008; Thatcher et al., 2008; Sehm et al., 2009). However, it was not known if mirnas played a critical role later in retinal regeneration, such as regulating neuronal progenitor cell proliferation. Morpholino-mediated reduction of Dicer expression revealed a significant reduction in neuronal progenitor cell proliferation, but not a significant change in the number of PCNA-positive M uller glia. This is consistent with the observed reduction of let-7a expression in the damaged zebrafish retina being necessary for M uller glia dedifferentiation and proliferation (Ramachandran et al., 2010). While it is possible that expression of some Dicer-dependent mirnas are required for M uller glia dedifferentiation and proliferation, these data suggest that neuronal progenitor cell proliferation is more dependent on Dicermediated mirna expression. This is consistent with only a relatively small subset of mirnas being expressed in stem/progenitor cells; increasing numbers are expressed as cells begin to differentiate (Thatcher et al., 2007; Wienholds et al., 2003). Further, conditional Dicer loss-of-function studies in developing mice retinas demonstrated mirnas promote shifts of retinal progenitors from early to late competence states (La Torre et al., 2013; Georgi and Reh, 2010). Loss of mirnas resulted in progenitors being stuck in early competence states, leading to excess production of early-born RGC and loss of later-born cell types. These results suggest that M uller glia dedifferentiation may require expression of only a small subset of mirnas in the regenerating retina with the resulting neuronal progenitor cells requiring expression of larger numbers of mirnas. This could explain the mild effect of Dicer loss-of-function on M uller glia dedifferentiation relative to the robust effect on progenitor cell proliferation. It will be interesting to follow the fate of neuronal progenitor cells in the dicer morphant retinas to assess if they retain the ability to differentiate into functional retinal neurons. Additionally, the dicer morphant cannot reveal the function of mirnas that normally decrease in expression during regeneration, such as the role of let-7a in M uller glia dedifferentiation and proliferation. To gain a deeper appreciation of how the expression of individual mirnas change during retinal regeneration, we performed high-throughput sequencing of zebrafish retinas before, during, and after regeneration. This analysis identified subsets of mirnas that are differentially expressed during retinal regeneration. Importantly, almost all the differentially expressed mirnas returned to their basal expression levels (0h of constant light treatment) once regeneration was completed (24 days after 4 days of constant light treatment). This dynamic mirna expression profile was reminiscent of our similar finding during zebrafish caudal fin regeneration (Thatcher et al., 2008). Our dataset also confirmed that let-7a decreased in expression from 0h to 35h of constant light, consistent with the puncture-damaged retina (Ramachandran et al., 2010). However, let-7a expression then increased from 35h to 72h in the light-damaged retina, suggesting that it may play another role later in neuronal progenitor cell proliferation, migration, or commitment. Dynamic mirna expression patterns emphasizes their importance for regenerating diverse tissues. To test if differential expression of individual mirnas play important roles in neuronal progenitor cell proliferation during retinal regeneration, we performed morpholino-mediated knockdowns of five upregulated mirnas. Reduced mir-142b expression decreased the number of proliferating neuronal progenitors at both 51h and 72h of light relative to the control, while knockdown of mir-146a reduced neuronal progenitor cell proliferation at only 51h relative to the control. In contrast, the mir-7a, mir- 27c, and mir-31 morphant retinas possessed fewer PCNApositive neuronal progenitors relative to the control at only 72h. Finally, mir-2190, which is unlikely to be a true mirna (Wei et al., 2012), showed no difference in the number of proliferating INL or ONL cells relative to the control at either timepoint. Although the targets and the molecular pathways through which these mirnas function during retinal regeneration is unknown, it is worth noting that some of these mirnas have similar functions in other tissues. Previously, mir-142 was suggested to play a role in liver regeneration (Lu et al., 2013) and control the balance between mesenchymal progenitor cell proliferation and differentiation through regulating Wnt signaling (Carraro et al., 2014). This is of interest because Wnt signaling was previously shown to play an important role in M uller glia dedifferentiation and proliferation in the regenerating zebrafish retina (Ramachandran et al., 2011; Wan et al., 2012). While we did not directly test if knock down of mir-142b reduced M uller glia proliferation, the reduced number of PCNA-positive INL cells at 51h in the mir-142b morphant could be due to fewer proliferating M uller glia. Thus, it will be interesting to determine if loss of mir-142b suppresses M uller glia proliferation, thereby indirectly contributing to reduced neuronal progenitor cell proliferation at later stages, or if loss of mir-142b directly affects neuronal progenitor cell proliferation through the Wnt signaling pathway. Loss of mir-27c and mir-7a decreased the number of proliferating neuronal progenitor cells during retinal regeneration. Interestingly, this effect is reminiscent of the role of mir-27a and -b in muscle satellite cell activation, which promotes muscle progenitor cell migration, proliferation, and delays differentiation (Crist et al., 2009; Lozano-Velasco et al., 2011). mir-7a has also been implicated in neuronal stem cell (NSC) maintenance and cell

10 1600 RAJARAM ET AL. Fig. 8. Knockdown of mir-142b reduces proliferation. Adult albino zebrafish were intravitreally injected and electroporated with either standard control or mir-142b morpholino prior to the start of light damage. A,B: After 51h of constant light, doublet PCNA-positive INL cells (arrows) were present in both the standard control and mir-142b morphant retinas. C,D: After 72h, PCNA-positive INL cell clusters are present in standard control morphants (C, arrowhead), but only doublet or fewer PCNA-positive INL cells (D, arrow) remained in mir-142b morphant retinas. E,F: mir-142b knockdown significantly reduced the number of PCNA-positive INL cells at both 51 and 72h and ONL cells at 72h. INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar in A ¼ 50 mm and applies to B D; data represent mean 6 s.e.m. *P< 0.05 using two-way ANOVA with Tukey s post-hoc test, n¼6. Fig. 9. Knockdown of mir-146a reduces proliferation at 51h. Adult albino zebrafish were intravitreally injected and electroporated with either standard control or mir-146a morpholino prior to the start of light damage. A,B: After 51h of constant light, doublet PCNA-positive INL cells are present in standard control morphants (arrow), but mainly single PCNA-positive INL cells are present in mir-146a morphants (arrow). C,D: By 72h, both standard control and mir-146a morphants contained columns of INL proliferating nuclei (arrowheads). E,F: mir- 146a knockdown significantly reduced the number of proliferating nuclei at 51h, but not at 72h. There was not a significant difference in the number of PCNA-positive INL cells at either 51h or 72h between the mir-146a and standard control morphant retinas. INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar in ¼ 50 mm and applies to B D. Data represent mean 6 s.e.m. **P< 0.01 using two-way ANOVA with Tukey s post-hoc test, n¼6. fate determination in the mouse brain, where it restricts Pax6 protein to NSCs (De Chevigny et al., 2012). In the regenerating zebrafish retina, increased Pax6a and Pax6b expression in neuronal progenitor cells is required for their proliferation (Thummel et al., 2010). Further, knockdown of mir-7a expression phenocopies the inhibition of neuronal progenitor cell proliferation observed in pax6b morphant retinas (Thummel et al., 2010). It will be interesting to determine if mir-7a regulates pax6a and/or pax6b expression or if mir-7a and pax6b act in the same complex regulatory network in the regenerating zebrafish retina. Retinal cell specific mirna expression patterns also need to be determined. They could be expressed in retinal neurons or the M uller glia, in which case, they likely would regulate the expression of genes that serve as signals to the neuronal progenitor cells. Alternatively, they could be expressed in the neuronal progenitor cells to regulate the expression of proteins that act cell autonomously in progenitor cell proliferation. Identifying the site of their expression would aid in determining what role they play in neuronal progenitor cell proliferation. Conversely, if the mirna targets that affect neuronal progenitor cell proliferation were elucidated, then the spatial expression of the mirnas could be deduced. Regardless, this work lays the foundation for many potential important studies in elucidating the roles of several

11 FUNCTIONS OF MIRNAS DURING RETINAL REGENERATION 1601 Fig. 10. Knockdown of mir-7a, mir-27c and mir-31 reduced proliferation at 72h. Adult albino zebrafish were intravitreally injected and electroporated with either standard control mir-7a, mir-27c or mir-31 morpholino prior to the start of light damage. A D: After 51h of light, standard control, mir-7a, mir-27c, and mir-31 morphant retinas contained PCNA-positive nuclei in the INL (arrows). E: At 72h, standard control morphant retinas contain columns of INL proliferating nuclei (arrowhead). F H: At 72h, mir-7a, mir-27c and mir-31 morphant retinas contained mainly doublet INL and single nuclei (arrows). I K: There were no significant differences in the number of PCNA-positive INL cells at 51h between the standard control morphant retina and the mir-7a (I), mir-27c (J), and mir-31 (K) morphant retinas. In contrast, significantly fewer PCNA-positive INL cells were present in all three mirna morphant retinas relative to standard control morphant retinas at 72h. L N: There were no significant differences in the number of PCNA-positive ONL cells at 51h between the standard control morphant retina and the mir-7a (L), mir-27c (M), and mir- 31 (N) morphant retinas. In contrast, significantly fewer PCNA-positive INL cells were present in all three mirna morphant retinas compared to standard control morphant retinas at 72h. INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar in A ¼ 50 mm and applies to B H. Data represent mean 6 s.e.m. *P< 0.05; ** P < 0.01; *** P <0.001 using two-way ANOVA with Tukey s post-hoc test, n¼6. mirnas in potentially different aspects of retinal regeneration in zebrafish. Experimental Procedures Fish Maintenance and Adult Zebrafish Light Lesioning Zebrafish were maintained in 14h:10h light:dark cycles at 28.5 C. Constant intense light lesioning of albino zebrafish was performed as described (Vihtelic and Hyde, 2000). Briefly, either albino or albino Tg(gfap:EGFP)nt11 transgenic adult (6 9 months of age, 4 5 cm in length) zebrafish were dark-adapted for 14 days and then transferred to constant intense light (20,000 lux) with the temperature maintained at C. Eyes were collected at 0, 16, 35, 51, 68, 72, or 96 hours (h) of light damage. For the recovery time points, fish were returned to normal light conditions (14h-light, 10h-dark cycles; 28.5 C) after 96h of intense light exposure and allowed to recover for 28 days. All experimental protocols in this study were approved by the University of Notre Dame Animal Care and Use Committee and are in compliance with the Association for Research in Vision and Ophthalmology statement for the use of animals in vision research. Western Blots Western blots were performed as described (Nelson et al., 2013). Briefly, dorsal retinas were isolated from morpholinoelectroporated and light-damaged (35h) retinas, homogenized in 1x PBS/1% Triton X-100 and protease inhibitor (Roche Diagnostics, Indianapolis, IN), and incubated on ice for 1h. Lysates were briefly centrifuged to remove Triton X-100-insoluble debris and the equivalent of two dorsal retinas were combined with 2X Laemmli sample buffer and 10x reducing agent (Invitrogen, Grand Island, NY). Proteins were processed, electrophoresed, and transferred to Hybond PDVF membranes (GE Healthcare,

12 1602 RAJARAM ET AL. separately to X-ray film to allow long exposures of the Dicer signal relative to the short exposures for the actin signal. Immunoblots were quantified using ImageJ software from three biological triplicates and analyzed using the Student s t-test. Fig. 11. Knockdown of mir-2190 does not affect proliferation. Adult albino zebrafish were intravitreally injected and electroporated with either standard control or mir-2190 morpholino prior to the start of light damage. A,B: After 51h of constant light, standard control and mir-2190 morphant retinas both display PCNA-positive doublet INL proliferating nuclei (arrows). C,D: After 72h of light, standard control and mir-2190 morphant retinas display clusters of INL nuclei (arrowheads). E,F: Statistically equivalent numbers of INL and ONL PCNApositive nuclei are present in standard control and mir-2190 morphant retinas. INL, inner nuclear layer; ONL, outer nuclear layer. Scale bar in A ¼ 50 mm and applies to B D; data represent mean 6 s.e.m. P > 0.05 using two-way ANOVA with Tukey s post-hoc test, n¼6. Waukesha, WI). The membranes were blocked in PBS/5% nonfat dry milk/0.1% Tween 20 for 1h at room temperature and then incubated with either a mouse anti-dicer monoclonal antibody (#D-11, 1:50; Santa Cruz Biotech, Santa Cruz, CA) or mouse antiactin monoclonal antibody (1:5,000; Cell Signaling Technology, Danvers, MA) overnight at 4 C. The membranes were washed, incubated with horseradish peroxidase-conjugated secondary antibodies (1:5,000; GE Healthcare) at room temperature for 1h. The membranes were cut to separate the Dicer-containing region from the actin region because of the very large difference in the levels of signal between the two detected proteins. The two portions of the same blot were then simultaneously incubated with the ECL-Prime detection system (GE Healthcare) and exposed Small RNA Library Preparation and High-Throughput Sequencing Zebrafish were sacrificed in groups of 100 at each time point: (1) before (0h), (2) during intense light exposure (35h and 72h), and (3) following recovery (28d). Retinas were collected in TRIzol reagent and total RNA was extracted and size fractionated on 15% urea acrylamide gels. Small RNAs (15 30 nt) were size selected, ligated to 3 0 and 5 0 end adapters, and amplified by RT-PCR to generate small RNA libraries as described (Wei et al., 2012). The primer and adapter sequences used to construct libraries are listed in Table 2. Two biological and technical replicate libraries were prepared for the 0h time point. Libraries were sequenced on the Illumina Genome Analyzer II and Hiseq 2000 platforms at the Vanderbilt VANTAGE sequencing core. Raw data were submitted to the NCBI GEO database (GSE58702). The number of sequencing reads from each time point is listed in Suppl. Table S1. Small RNA Read Processing and mirna Expression Profile Generation Sequencing reads were processed to remove adapter sequences and mapped to the ZV9 zebrafish genome. Bowtie (Langmead et al., 2009) was used to map the resulting reads to mirna hairpin sequences from mirbase ( followed by further filtering to remove reads derived from precursor mirna loop regions or passenger strands, as described (Wei et al., 2012). Reads that mapped back to mature mirna sequences but did not contain the mirna seed sequence (nucleotides 2 6 from the 5 0 end of the mature mirna) or that mapped to the mature mirna sequence only in the 3 0 region were excluded. Mature mirna reads were normalized to the total number of mappable reads per sequencing run. Normalized reads were log 2 transformed and the resulting values were displayed using heat maps (Multi experiment viewer, Fig. 4; Saeed et al., 2003) and plotted in scattergrams (Fig. 5A C). Reads that mapped back to mirna precursor sequences are in Suppl. Table S2. Log 2 values of all normalized mirna reads are listed in Suppl. Table S3. Taqman Real-Time PCR Quantitative real-time PCR (qpcr) for mirnas was performed using Taqman probes (Applied Biosystems, Grand Island, NY). qpcr was performed in triplicate on the same RNA samples as the high throughput sequencing. qpcr was performed for mir- 21, mir-133b, mir-7a, and relative mirna levels were determined using the DDCt method and normalized to the levels of mir-9. The qpcr was performed on a Biorad CFX 96 Real time system. To confirm that the morpholinos knocked down mirna expression, either lissamine-tagged Standard Control morpholino, mir-7a morpholino or mir-146a morpholino was individually injected into zebrafish embryos at the 1 4-cell stage. At 24h postfertilization, RNA was collected from lissamine-positive embryos and analyzed by qpcr as described above in biological and technical triplicates. The mir-7a and mir-146a morphants