dnmt1 function is required to maintain retinal stem cells within the ciliary marginal zone of the zebrafish eye

The ciliary marginal zone (CMZ) of the zebrafish retina contains a population of actively proliferating resident stem cells, which generate retinal neurons throughout life. The maintenance methyltransferase, dnmt1, is expressed within the CMZ. Loss of dnmt1 function results in gene misregulation and cell death in a variety of developmental contexts, however, its role in retinal stem cell (RSC) maintenance is currently unknown. Here, we demonstrate that zebrafish dnmt1s872 mutants possess severe defects in RSC maintenance within the CMZ. Using a combination of immunohistochemistry, in situ hybridization, and a transgenic reporter assay, our results demonstrate a requirement for dnmt1 activity in the regulation of RSC proliferation, gene expression and in the repression of endogenous retroelements (REs). Ultimately, cell death is elevated in the dnmt1−/− CMZ, but in a p53-independent manner. Using a transgenic reporter for RE transposition activity, we demonstrate increased transposition in the dnmt1−/− CMZ. Taken together our data identify a critical role for dnmt1 function in RSC maintenance in the vertebrate eye.


Results
dnmt1 mutants possess defects in the ciliary marginal zone. Previously, we identified a requirement for dnmt1 in maintaining lens epithelial cell viability using dnmt1 s872 mutant zebrafish 27 . During these previous studies, we also detected photoreceptor layer abnormalities, similar to those documented in Dnmt1 −/− conditional knockout mice 45,46 , and an apparent defect in the CMZ. With an interest in the role that dnmt1 plays in maintaining RSCs in vivo, here, we focused further on the CMZ phenotype. Using DAPI to label and count retinal nuclei, we confirmed a progressive degeneration of CMZ morphology beginning at 4 days post fertilization (dpf; Fig. 1A-F) and a significant decline in retinal cell numbers through 5dpf (Fig. 1G). The total number of cells present within central retina sections are equivalent between dnmt1 −/− and sibling larvae at 3dpf; however, numbers in dnmt1 −/− larvae diminish significantly between 4 and 5dpf (18.8% and 26.6% reduction respectively; p < 0.0005; Fig. 1G). Additionally, we compared the proportions of nuclei within the ganglion cell layer (GCL), inner nuclear layer (INL), outer nuclear layer (ONL), and CMZ between dnmt1 −/− larvae and siblings from 3 to 5dpf (Fig. 1H). Interestingly, the proportions of cells in all three retinal laminae (GCL, INL, and ONL) remained equivalent over time in dnmt1 −/− larvae when compared to siblings, with only a slight increase in the ONL at 4dpf (Fig. 1H and Supplementary Figure 1A-C; p < 0.005). In contrast, the CMZ proportion decreased significantly from 3 to 5dpf suggesting that dnmt1 function in the retina is required within the CMZ to maintain the RSC population ( Fig. 1H and Supplementary Figure 1D; p < 0.0005).
To quantify these findings and further assess gene expression changes in dnmt1 −/− larvae, we conducted quantitative PCR analysis of the expression of cell cycle, cell death, and immune response genes using whole larval samples (Fig. 4U). RNA was isolated from 4dpf sibling and dnmt1 −/− larvae (n = 16 each) in three biological replicates, converted into cDNA, and analyzed for gene expression levels. Overall, the cell cycle progression www.nature.com/scientificreports/ genes (ccna2, ccnb1, ccnd1, ccne, cdk1, cdk2, and cdk4) displayed reduced expression levels while cell arrest genes (caspa, caspb, mdm2, p53, and ripk1) were equivalent or slightly increased in dnmt1 −/− larvae compared to sibling controls. Additionally, dnmt1 −/− larvae showed increased levels of immune response genes (tnfα and il-1β) consistent with previous reports 37 . While these qPCR data correlate with in situ hybridization data for CMZspecific expression, the changes were not statistically significant when assessed by 2-way ANOVA analysis. This is not surprising since whole larvae were used for qPCR and each of these genes is expressed in numerous larval regions outside of the CMZ; this non-ocular expression likely masks changes in the CMZ. Nonetheless, the trends are consistent with apparent loss or reduction of expression in the CMZ of dnmt1 −/− larvae detected by in situ hybridization. Taken together, these data suggest that RSCs are present at the onset of morphological defects in the dnmt1 −/− CMZ, but could be impaired in their ability to progress through the cell cycle and self-renew.
Loss of dnmt1 activity results in decreased RSC proliferation. RSCs within the teleost CMZ remain proliferative throughout the lifespan of the animal 3,55,56 and Dnmt1 is known to be required for cell cycle progression within stem cells of various tissue types 24,25,57 . Based on the significant loss of RSCs in dnmt1 −/− larvae between 3 and 5dpf ( Fig. 1) and the inability of dnmt1 −/− RSCs to maintain expression of cell cycle genes (Fig. 4), we hypothesized that dnmt1 −/− RSCs would be defective in their proliferative capacity. To test this hypothesis, larvae were incubated for 2 h in BrdU at 3, 4, and 5dpf, fixed immediately thereafter, and immunolabeled for BrdU and phosphohistone-H3-serine10 (pH3) to identify RSCs in late G2/M. dnmt1 siblings maintained a constant proportion of BrdU + cells within the CMZ between 3-5dpf (   28 . To test this hypothesis, we performed a BrdU birth-dating assay 58 . Our aim was to saturate RSCs with BrdU for a 12-h period (3-3.5dpf) and quantify the average starting number of proliferating cells at 3.5dpf and determine the final position of daughter cells at 5dpf, once they incorporated www.nature.com/scientificreports/ into the retina (Fig. 6A). Initial analysis of these samples revealed that most BrdU + nuclei in both sibling and dnmt1 −/− larvae were located within the CMZ after the 12 h incubation (Fig. 6C, E, G). However, there were a few BrdU + cells that had incorporated into the neural retina at this time (Fig. 6G). By comparing the number of BrdU + nuclei of each retinal domain (CMZ, GCL, INL, ONL, Fig. 6B) to the total number of BrdU + nuclei (Fig. 6H) at 3.5dpf, we noted a significant increase in the proportion of BrdU + nuclei in the dnmt1 −/− CMZ (79.5%, p < 0.05) compared to controls (71.7%, Fig. 6G; Supplemental Fig. S3A). Additionally, we found that the  At 5dpf, all BrdU + cells in the sibling controls had exited the cell cycle and incorporated into the neural retina (Fig. 6D, G), whereas dnmt1 −/− larvae retained 19.8% (p = 0.05) of BrdU + nuclei within the CMZ and had fewer BrdU + cells overall within the retina (Fig. 6F, G). Additionally, there was a significant decrease in the proportion of BrdU + nuclei in the GCL (9.8%, p < 0.0005) ( Fig. 6G; Supplemental Fig S3B) compared to controls (21.4%). Surprisingly, among the cells that remained in the 5dpf dnmt1 −/− CMZ, there was an increase in the BrdU + proportion when compared to siblings (19.76% vs. 0.9% respectively, p = 0.05; Fig. 6G; Supplemental Fig S3B) suggesting an inability for some RSCs to either successfully complete the cell cycle or to integrate into retinal laminae. These data also show that daughter cells produced from the dnmt1 −/− CMZ proportionally incorporate into the INL and ONL at similar levels to those detected in controls ( Fig. 6G; Supplemental Fig S3B) supporting the notion that dnmt1 −/− RSCs are still capable of producing neurons that can successfully integrate into these two layers of the retina.
Loss of dnmt1 activity leads to altered Long Terminal Repeat retroelement expression within the cMZ. Half of the zebrafish genome is comprised of endogenous viral elements known as transposons 59,60 , and dnmt1 is required for repressing the retroelement (RE) lineage of transposons 37,61-63 . Though many REs have lost their ability to "jump" throughout evolution, some still retain this ability 64,65 . These studies led us to hypothesize that aberrant DNA methylation resulting from loss of dnmt1 activity in RSCs would result in upregulation of RE expression within the dnmt1 −/− CMZ. To identify RE expression within the CMZ, we performed in situ hybridizations targeting several REs that belong to the Long Terminal Repeat (LTR) class of retrotransposons, specifically Bel20, ERV1, ERV1-N5, ERV4, and Gypsy10 LTRs. We noted endogenous expression of Bel20, ERV4, and Gypsy10 REs within the CMZ but not the neural retina of control larvae at 4dpf (Fig. 7A, D, E). This result was unexpected since REs can be deleterious to cellular function 37,66-68 . However, not all of the LTR REs were detected within control CMZs; ERV1 and ERV1-N5 expression was not detected in the CMZ of siblings (Fig. 7B, C), but rather ERV1-N5 seemed to be expressed within the ONL of some control larvae (Supplemental Fig. S4O). Remarkably, dnmt1 −/− larvae displayed patches of ERV1-N5 expression in the CMZ and within the overlying retinal pigmented epithelium (Fig. 7H). The distributions of Bel20 and ERV4 were also expanded beyond the CMZ into the neural retina of dnmt1 −/− larvae (Fig. 7F, I) when compared to controls. Of note, we also identified several non-ocular tissues that displayed altered RE expression between dnmt1 −/− and sibling control larvae (Supplemental Fig S4). Interestingly, these LTR RE expression patterns were larvae-dependent, suggesting that not all RSCs respond uniformly to loss of dnmt1 function.
A L1RE3-EGFP transgene reports increased LINE1 retrotransposition activity in dnmt1 −/− cMZ. To expand our analysis of RE expression in dnmt1 −/− RSCs, and more specifically, visualize retrotransposition activity in vivo, we generated a non-LTR, LINE1 element transgenic reporter line by modifying the pLRE3-EGFP plasmid 69,70 (referred to as L1RE3-EGFP for the remainder of this study). The L1RE3-EGFP www.nature.com/scientificreports/ construct contains a human-derived LINE1 RE sequence that requires retrotransposition for EGFP to be expressed and translated into a functional protein 69 . p53 is known to repress REs and when used transiently in p53 −/− zebrafish, L1RE3-EGFP was shown to have increased transposition activity and EGFP expression 68 . We validated the stability and effectiveness of the L1RE3-EGFP transgenic using again p53 mutants 48,68 and immunolabeling for EGFP (Supplemental Fig S5). When L1RE3-EGFP was incorporated into the dnmt1 s872 genetic background, ectopic EGFP expression could be seen within the dnmt1 −/− eye when compared to control siblings (Supplemental Fig S5B,C). Notably, we were able to detect ectopic EGFP expression within the dnmt1 −/− CMZ at both 3dpf (Fig. 8C) and 4dpf (Fig. 8D) timepoints when compared to controls (Fig. 8A, B). However, similar to RE expression patterns, clonal EGFP expression patterns were variable, both within and between sibling controls and dnmt1 −/− larvae, again suggesting that the effects of dnmt1 loss is variable from cell to cell and larva to larva.

Discussion
The zebrafish, with its lifelong, actively cycling RSCs within the CMZ, is a powerful model through which we can address how epigenetic regulators function to maintain these stem/progenitor cell populations in vivo. This study focused on the role of the DNA maintenance methyltransferase, dnmt1, within the CMZ, with the goal of determining how dnmt1 activity facilitates RSC maintenance. Previous work has shown that loss of dnmt1 function results in ocular defects 27,45,46,52 , but no studies have yet analyzed RSC populations and determined whether dnmt1 activity modulates their behavior.
Here, we demonstrate that dnmt1 is essential for RSC homeostasis by maintaining CMZ-specific gene expression (Fig. 4), facilitating cell cycle progression (Fig. 5), and incorporation of CMZ-derived cells into the retina (Fig. 6). These data are consistent with Dnmt1 functions described in other in vivo progenitor models such as the lens 27 , hippocampus 50 , kidney 62 , pancreas 30 and intestine 51 . RSCs in S-and G2/M-phases of the cell cycle were detected in reduced proportions in the dnmt1 −/− CMZ and this correlated with a reduction in CMZ expression of genes encoding proteins that function in cell cycle progression, namely ccnD1 (Fig. 4G, H) and www.nature.com/scientificreports/ cdkn1ca (Fig. 4O, P). Defects in cell cycle progression may also contribute to aberrant daughter cell integration into retinal laminae detected in dnmt1 −/− larvae (Fig. 6).
It is critical to note that while the RSCs are more affected by loss of dnmt1 function than fully differentiated neurons within the GCL, INL, and ONL, we cannot rule out the possibility that any of the surrounding tissues could be contributing to the CMZ phenotype. Indeed, it is known dnmt1 loss can influence cells and tissues through both autonomous 27,71 and non-autonomous 45 mechanisms. There are multiple tissues surrounding the CMZ that influence RSC identity 3,16,72 and these include differentiated neurons in the retina, lens, RPE and vasculature; loss of dnmt1 function in any of these could non-autonomously result in CMZ defects. Future work focused on tissue and/or cell type-specific loss of dnmt1 function will be critical for defining its autonomous and non-autonomous roles in RSC maintenance.
While loss of p53 function in the dnmt1 −/− background significantly rescued cell death within the laminated retina, validating that the p53 zdf1 allele is in fact inhibiting p53-driven apoptosis, loss of p53 in the dnmt1 −/− CMZ had no effect on CMZ cell numbers suggesting a p53-independent cell death pathway is likely modulated by dnmt1 in the CMZ 73 . Recent reports have demonstrated an upregulation of an innate inflammatory response in dnmt1 −/− larvae 37 . Necroptosis, a programmed cell death pathway tightly linked to a cell's innate viral detection system and inflammatory response, also results in DNA fragmentation and, in its later stages, is detected by TUNEL 73 . Indeed, we noted upregulation of the inflammatory genes, tnfα and il-1β, and some cell death pathway markers, p53 and ripk1 (Fig. 4U); however, these data were obtained from whole larvae qPCR and thus are compounded by systemic expression changes. Accordingly, we considered the possibility that dnmt1 −/− RSCs were instead lost via necroptosis. We tested this hypothesis using several chemical inhibitors of necroptosis, some of which have been reported to function in the zebrafish 74,75 ; however, we were unable to replicate necroptotic inhibition nor validate drug efficacy. None the less, we predict that either necroptosis or pyroptosis (a programmed cell death pathway triggered by intracellular bacterial infections 76,77 ) are the most likely mechanisms of cell death in dnmt1-deficient RSCs, but this will require the development of new tools to enable further analysis.
Alterations in RE expression activity the dnmt1 −/− CMZ (Figs. 7, 8) are exciting given Dnmt1's known roles in repressing RE activity [36][37][38][39] . RE expression was aberrant in most dnmt1 −/− CMZs examined (Fig. 7); however, expression changes and levels were variable between larvae, suggesting that the location and extent of genomic hypomethylation resulting from loss of dnmt1 function is inherently variable between cells of each larva. Previous reports demonstrated innate RE activity within somatic neural tissue 64,65,[78][79][80] . Indeed, we detected retrotransposition activity within the larval zebrafish brain (Supplemental Fig S5D-I) of both siblings and dnmt1 −/− larvae from 2 to 4dpf, similar to activity detected in human hippocampal neurons 64,65,80 . However, RE retrotransposition is highly variable between larvae. Further studies will be required to determine what cellular processes might sensitize a cell-or tissue-type to upregulate REs and whether these REs have a mechanistic purpose within the cell.
In conclusion, our results demonstrate that dnmt1 functions to maintain RSC proliferation, gene expression, and integration of RSC daughters into the retina. Additionally, some REs are innately expressed within RSCs, however dnmt1 function is required to maintain tight control of these viral elements. Without dnmt1 activity, LTR expression remains active within the retina and L1RE3-EGFP retrotransposition activity is increased. Interestingly, RE activity within RSCs does not result in p53-mediated apoptosis, supporting a model in which dnmt1 −/− RSCs are lost through another mechanism of cell death. As discussed above, we predict that this increase in RE activity most likely activates necroptotic or pyroptotic cell death pathways, which are both known to result from intracellular responses to invading pathogens 73,76,77 . Regarding the innate LTR expression within dnmt1 +/+ RSCs, in conjunction with previous reports of inherent RE activity within human neural tissue, it is worth considering how RE activity may contribute to neural stem cell biology. It is well known that dysregulation of REs is a hallmark of many human neurodegenerative diseases 67,[81][82][83][84] . Future evaluations regarding the innate cost-tobenefit ratio of RE activity could provide crucial evidence for the development of neurodegenerative therapies.

Methods
Zebrafish maintenance. Zebrafish (Danio rerio) were maintained at 28.5 °C on a 14 h light/10 h dark cycle. All protocols used within this study were approved by the Institutional Animal Care and Use Committee of The University of Pittsburgh School of Medicine, and conform to the National Institutes of Health Guide for the Care and Use of Laboratory Animals. Mutant alleles used in this study were dnmt1 s872 and tp53 zdf1 . dnmt1 s872 and tp53 zdf1 zebrafish were genotyped using BioRad's CFX Manager 3.1 and Precision Melt Analysis software (v4.0.52.0602). All genotyping primers are listed in Supplemental Table 1. Transgenic Tg(CMV:Has.L1RE3, EGFP, myl7:EGFP) pt701 zebrafish were generated as described 85 using constructs generously provided by Kristen Kwan and Chi-Bin Chien (University of Utah, Salt Lake City).
BrdU labeling. To assess cellular proliferation, larvae were incubated in 10 mM BrdU for either 2 or 12 h, after which the BrdU was washed out and larvae were either collected or used for BrdU pulse-chase experiments.
Immunohistochemistry and fluorescent labeling. Immunohistochemistry performed as described previously 86   Cloning and probe synthesis. CMZ-specific probes have been published previously 14,27 . Retroelement probes were generated using reverse transcription-polymerase chain reaction (RT-PCR) on Trizol-isolated RNA from 24hpf and 5dpf embryos. Primer sequences were kindly provided by Dr. Kirsten Sadler (NYU Abu Dhabi) and PCR products were ligated into pGEM-T-easy vector (Promega Cat# PR-A1360) and verified by Sanger sequencing. Plasmids containing the correct clones were linearized and used as templates to in vitro transcribe digoxigenin-labeled RNA probes (Roche).
in situ hybridization. Hybridizations using digoxigenin labeled antisense RNA probes were performed essentially as described 87 , except that they were pre-incubated with 1 mg/mL Collagenase type 1A (Sigma, C9891) to allow probe diffusion throughout the tissue. All probe primer sequences are listed in Supplemental  Table S1.
RnA isolation and cDnA synthesis. Total  qpcR statistical analyses. Cq values were transformed to linear scale and the normalization factor was calculated as the geometric mean of candidate reference genes included in the dataset as described 88 . Variance analyses between siblings and dnmt1 −/− larvae were performed using 2-way ANOVA test followed by a post-hoc Bonferroni test with significance set to p < 0.05. Graph (Fig. 4U) depicts average relative fold expression levels with 95% confidence intervals of dnmt1 −/− larvae relative to sibling controls.
Microscopy and image processing. For sectioned embryos, imaging was performed with an Olympus FV1200 confocal microscope. Confocal Z-stacks were collected in 1 µm optical sections. Z-stacks were maxprojected using ImageJ (version 1.52r) software (National Institutes of Health) and quantification was conducted using the "Cell Counter" plugin. Figures were prepared using Adobe Illustrator CS6 (Adobe Systems). In situ cryosections were imaged utilizing a Leica DM2500 with a 100X oil immersion objective (NA: 1.25).
Cell counting and quantification. Each data point was collected from an individual larva. Each larva was analyzed using three consecutive 12 µm sections of the central retina using the optic nerve and lens morphology as retinal landmarks. The CMZ domain was defined as the region of cells posterior to the RPE and anterior to the IPL and OPL, using both nuclear and Phalloidin staining as markers. Nuclear morphology was taken into consideration when determining layer-specific cellular locations where CMZ nuclei display an elongated, or ovular, shape in comparison to the spherical nuclei seen in the GCL and INL. Photoreceptor nuclei were defined by elongated morphology and with peripheral phalloidin staining of outer segments. The average of the three consecutive sections was used as a single data point (n ≥ 4 for all datasets). Proportions of retinal domains were calculated by dividing the number of DAPI-labeled nuclei in each domain over the total number of retinal nuclei.

Statistics.
For all statistical analysis, data were imported into GraphPad Prism 8 software. Quantification of nuclei and immunolabeled cells was statistically assessed using Student's two-tailed unpaired T test with p < 0.05 as a significance threshold.