UHRF1 regulation of Dnmt1 is required for pre-gastrula zebrafish development

Abstract

Landmark epigenetic events underlie early embryonic development, yet how epigenetic modifiers are regulated to achieve rapid epigenome re-patterning is not known. Uhrf1 and DNA methyltransferase 1 (Dnmt1) are known to largely mediate maintenance DNA methylation and Uhrf1 is also required for both Dnmt1 localization and stability. Here, we investigate how these two key epigenetic modifiers regulate early zebrafish development and characterize the developmental consequences of disrupting their homeostatic relationship. Unlike Uhrf1 knockdown, which causes developmental arrest and death prior to gastrulation, overexpression of human UHRF1 (WT-UHRF1) caused asymmetric epiboly, inefficient gastrulation and multi-systemic defects. UHRF1 phosphorylation was previously demonstrated as essential for zebrafish embryogenesis, and we found that penetrance of the asymmetric epiboly phenotype was significantly increased in embryos injected with mRNA encoding non-phosphorylatable UHRF1 (UHRF1^S661A). Surprisingly, both WT-UHRF1 and UHRF1^S661A overexpression caused DNA hypomethylation. However, since other approaches that caused an equivalent degree of DNA hypomethylation did not cause the asymmetric epiboly phenotype, we conclude that bulk DNA methylation is not the primary mechanism. Instead, UHRF1^S661A overexpression resulted in accumulation of Dnmt1 protein and the overexpression of both WT and a catalytically inactive Dnmt1 phenocopied the WT-UHRF1 overexpressing embryos. Dnmt1 knockdown suppressed the phenotype caused by UHRF1^S661A overexpression, and Uhrf1 knockdown suppressed the effect of Dnmt1 overexpression. Therefore, we conclude that the interaction between these two proteins is the mechanism underlying the observed developmental defects. This indicates that Dnmt1 stability requires UHRF1 phosphorylation and that crosstalk between the proteins is essential for the function of these two important epigenetic regulators during gastrulation.

Introduction

Covalent, reversible, epigenetic modification of the chromatin is a fundamental mechanism that patterns the genome and regulates gene expression during embryonic development. Dynamic changes in the patterns of several epigenetic marks, including DNA methylation, have been reported in mouse and zebrafish embryos during the critical stages of early development when the parental genomes become integrated and re-patterned and the zygotic genome is activated (Andersen et al., 2012; Borgel et al., 2010; Mhanni and McGowan, 2004; Rai et al., 2008; 2010; 2006; Reik, 2007; Smallwood et al., 2011; Smith et al., 2012; Wu et al., 2011; Xie et al., 2012). Ubiquitin-like, containing PHD and RING finger domains, 1 (UHRF1) and DNA methyltransferases (DNMTs) are the major components of the DNA methylation machinery. Loss of UHRF1 or DNMT1 is catastrophic for embryos, as null mutation in either gene causes early embryonic lethality (Bostick et al., 2007; Jacob et al., 2015; Lei et al., 1996; Muto et al., 2002; Sadler et al., 2007; Sharif et al., 2007) and knockdown of both maternal and zygotic Uhrf1 in zebrafish embryos causes death prior to gastrulation (Chu et al., 2012). However, hypomorphic mutations cause post-natal defects and cancer in mice (Gaudet et al., 2004) and multi-systemic defects resulting in lethality late in zebrafish development (Anderson et al., 2009; Jacob et al., 2015; Sadler et al., 2007; Tittle et al., 2011). Therefore, while it is clear that epigenetics underlies nearly every aspect of embryogenesis, whether the requirement for Uhrf1 and Dnmt1 during development can be solely ascribed to their role in DNA methylation has not been addressed.

UHRF1 and DNMT1 have been well studied and the molecular basis for their interaction has been established in tissue culture cells and in vitro, however, less is known about their function in vivo. UHRF1 recruits chromatin-modifying enzymes to the chromatin to mediate the deposition of repressive epigenetic modifications during S phase of the cell cycle. UHRF1 is conserved across vertebrate species. Its best-characterized function is to recognize hemimethylated cytosine residues in a CpG dinucleotide context via its SET and RING associated (SRA) domain. UHRF1 then recruits DNMT1 to methylate newly synthesized cytosine residues of the daughter strand during DNA replication (Arita et al., 2008; Avvakumov et al., 2008; Bostick et al., 2007; Hashimoto et al., 2008; Qian et al., 2008; Sharif et al., 2007). In all cases studied, loss of either UHRF1 (Feng et al., 2010) or DNMT1 (Jackson-Grusby et al., 2001; Jacob et al., 2015) result in global DNA hypomethylation. In addition, UHRF1 is also essential for DNMT1 specificity, activity, and protein stability. UHRF1 possesses a carboxy-terminal RING domain that mediates its E3 ubiquitin ligase activity. UHRF1 can ubiquitinate and target both itself and DNMT1 for destruction by the proteasome (Bashtrykov et al., 2014; Berkyurek et al., 2014; Du et al., 2010; Qin et al., 2011). In the developing zebrafish liver, UHRF1 overexpression destabilizes and delocalizes Dnmt1, thus serving to inhibit DNA methylation (Mudbhary et al., 2014). This is important, as UHRF1 is an oncogene and high levels of Uhrf1 are detrimental (Mudbhary et al., 2014). We hypothesize that interaction with Dnmt1 is a primary mechanism by which UHRF1 overexpression causes cancer.

DNA hypomethylation can lead to apoptosis, cellular senescence, disrupted differentiation programs of progenitor cells and cancer (Anderson et al., 2009; Jackson-Grusby et al., 2001; Rai et al., 2010), and the pre-gastrula lethality resulting from loss of uhrf1 and dnmt1 has largely been attributed to the failure to write the embryonic DNA methylome, as both the mouse and zebrafish uhrf1 mutants display genome-wide DNA hypomethylation (Feng et al., 2010). While DNA methylation is undisputedly an important epigenetic mark, whether alterations in DNA methylation account for all of the developmental phenotypes associated with uhrf1 and dnmt1 loss is not clear. Since both Uhrf1 and Dnmt1 have been reported to carry out functions in addition to DNA methylation (Espada et al., 2011; Fukagawa et al., 2015; Gelato et al., 2014; Hervouet et al., 2010; Robertson et al., 2000; Tauber and Fischle, 2015) it is possible that development may also require other roles of these two important proteins.

We previously reported that human UHRF1 possesses a conserved cyclin-dependent kinase 2 (CDK2) phosphorylation site and is phosphorylated by CDK2 when complexed with cyclin A2 (CCNA2) at a serine sitting in between the SRA and the RING domains (serine-661 in the human protein and serine-648 in zebrafish) (Chu et al., 2012). Mutation of this phospho-acceptor site via site-directed mutagenesis disrupts UHRF1 function as evidenced by its failure to rescue embryonic lethality upon uhrf1 knockdown, suggesting that UHRF1 phosphorylation is necessary for early embryogenesis. The increase in Ccna2 during S-phase suggests a model whereby phosphorylation of UHRF1 during DNA replication could regulate its function.

Here, we investigate the mechanism by which altering Uhrf1 levels regulates early zebrafish development by defining its effects on DNA methylation and Dnmt1 levels. The early lethality associated with uhrf1 knockdown in zebrafish embryos can be rescued by wild-type (WT) human UHRF1 mRNA, indicating that UHRF1 function is conserved across species (Chu et al., 2012). We found that overexpression of both human and zebrafish WT and phospho-deficient UHRF1 causes a profound defect in epiboly. Despite the important role for Uhrf1 in DNA methylation, this defect appears to be independent of DNA methylation, as other mechanisms of blocking bulk DNA methylation do not phenocopy the unique effect of UHRF1 overexpression on epiboly. Instead, overexpression of the non-phosphorylatable mutant (UHRF1^S661A) causes a significant increase in Dnmt1 protein levels. Genetic epistasis experiments demonstrate that excess Dnmt1 is the mechanism by which UHRF1 overexpression disrupts epiboly and gastrulation. Since the effects of Dnmt1 overexpression require Uhrf1 and, vise versa, we conclude that a high level of Dnmt1 protein is not tolerated during gastrulation and that these two proteins act together to allow embryos to proceed through gastrulation.

Materials and Methods

Zebrafish maintenance

WT (ABNYU, TAB14) adult zebrafish were maintained on a 14:10 light:dark cycle at 28°C in accordance with the policies of the Mount Sinai Institutional Animal Care and Use Committee. Fertilized eggs were collected following natural spawning and cultured at 28°C in egg water (0.6 g/L Crystal Sea Marinemix; Marine Enterprises International, Baltimore, MD; 0.002 g/L of methylene blue) at the same light:dark cycle.

Embryo treatments

Morpholinos targeting the ATG start site of dnmt1 (5’ACAAT GAGGT CTTGG TAGGC ATTTC) (Rai et al., 2006) or uhrf1 (MO1-ATG: 5’CACCT GAATC CACAT GGCGG CAAAC) (Chu et al., 2012) and immediate upstream region of the 5’UTR (MO2-UTR: 5’CGCAG CACTT TTGTT AGAGT TCAGT) or standard morpholino (5’CCTC TTAC CTCA GTTA CAAT TTATA) were purchased from Gene Tools, LLC and diluted in RNase-free water to a 1 mM stock and stored at room temperature.

mRNA was generated using mMessage mMachine (Life Technologies Corporation) according to the manufacturers protocol. RNA encoding human UHRF1 or the phospho-deficient mutant (UHRF1^S661A) with a N-terminal 6X-Myc tag described by (Chu et al., 2012), zebrafish WT-dnmt1 and a catalytically inactive mutant (dnmt1^C1109S) (Rai et al., 2006) were generously provided by D. Jones (Oklahoma Research Institute) and EGFP from the pcGlobin2 vector (Ro et al., 2004) were diluted to the indicated stock concentrations in RNAse free water.

All injections were carried out prior to the 2-cell stage of development, and, on average, a total of 1 nanoliter was injected. The concentrations of the stock solutions that were injected in each experiment are indicated in the text and figures. For co-injected embryos, mRNAs were mixed with morpholinos immediately prior to injection.

Embryos were treated with 5-Azacytidine (Sigma) concentrations ranging from 50-1,000 µM in egg water from fertilization and observed through Prim-5.

Phenotype Analysis

Individual embryos were placed into 12-well plates with approximately 1 mL of egg water and followed through the first 24 hours of development at selected developmental stages including the high stage (3.3 hours post fertilization (hpf)), sphere stage (4.0 hpf), 50% epiboly (5.25 hpf), shield stage (6.0 hpf), bud stage (10.0 hpf), 7–8 somite stage (12.0 hpf), and the prim 5 stage (24.0 hpf). Phenotypes were scored as either normal, developmental delay, or presentation of either embryonic arrest associated with uhrf1 knockdown and 5-AZA treatment or the asymmetric epiboly phenotype associated with WT-UHRF1, UHRF1-S661A and dnmt1 overexpression. These numbers were quantified and totaled to determine the proportion of each phenotype seen per treatment.

DNA methylation analysis

5-methylcytosine (5-MeC) levels were measured in total genomic DNA isolated from whole embryos as described (Mudbhary et al., 2014). In brief, DNA was denatured using 0.4M NaOH, neutralized using cold 2M C₂H₃O₂NH₄ and 0.1 – 1 µg of genomic DNA was blotted in duplicate onto nitrocellulose membrane using a slot blot apparatus (BioRad Laboratories, Inc.), and one membrane was incubated in anti-5MeC (1:1000; Eurogentech) followed by HRP conjugated anti-mouse secondary antibody (Promega 1:2000) and visualized via chemiluminescence kit (Roche). The other membrane was stained with methylene blue or was probed with anti-double-stranded DNA primary antibody (Abcam 1:2000) followed by HRP conjugated anti-mouse secondary antibody (1:2000; Abcam) for total gDNA detection. Bands were quantified using GelAnalyzer software and 5-MeC staining was normalized to total gDNA stained with methylene blue or blotted with anti-dsDNA for each paired sample. Ratios were compared to controls to determine the degree of DNA methylation for each treatment.

RNA isolation and PCR

RNA was extracted in Trizol (Ambion) using standard Phenol/Chloroform isolation procedures and cDNA was prepared for qRT-PCR analysis as described (Mudbhary et al., 2014). More specifically, RNA was reverse transcribed using poly A primers and qScript cDNA SuperMix (Quanta). All primers are listed in Table S1. qRT-PCR was carried out in Light Cycler 480 (Roche) using gene-specific primers and PerfeCTa SYBRGreen Fastmix (Quanta). Reactions were run in triplicate and target gene expression was calculated using the cycle threshold (C_T) method (2^{−Ct(target)}/ 2^−Ct(rpp0)) utilizing rpp0 C_T values as reference of gene expression. RT-PCR and gel electrophoresis was carried to observe relative changes in uhrf1, dnmt1, and rpp0 mRNA levels at each indicated stage of wild-type zebrafish development.

Immunodetection

Protein was isolated from embryos collected in 2 µL of protein lysis buffer (50 mM Tris pH 6.8, 2% SDS, 5% Glycerol, 1% 2-Mercaptoethanol) containing protease inhibitor cocktail per embryo and homogenized by sonication as described (Jacob et al., 2015). Soluble protein extract volumes from 6 hpf embryos equivalent to 2 embryos were loaded and run on 8% – 12% SDS-PAGE gels, transferred to nitrocellulose or PVDF membrane and blotted with antibodies to Myc-UHRF1 (1:1000 anti-Myc, Sigma-Aldrich), Dnmt1 (1:2000, Santa Cruz), and α-Tubulin (1:4000, Developmental Studies Hybridoma Bank). Bands were detected via chemiluminescence (Roche) and quantified with GelAnalyzer software and were normalized to α-Tubulin signal.

Statistical Analysis

Standard error of the mean was calculated using the standard deviation and the number of samples per experiment type in Microsoft Excel. Either paired 2-tailed Student’s t-test or Fisher’s Exact Test was used to analyze the significance of phenotype distribution in the genetic experiments. The log-rank test was used to assess survival and onset of phenotype by the 50% epiboly stage of development. One-way ANOVA was used to determine the significance between of the asymmetric epiboly in the co-injection experiments. To analyze differences in the amount of protein present (Dnmt1, exogenous UHRF1, etc.), the protein of interest was normalized to an internal loading control (assuming relatively standard levels of Tubulin or Histone H3) and a 2-tailed Student’s t-test was used to determine the statistical significance of the change in protein level. A p-value of less than 0.05 was considered a statistically significant observation. A Fisher’s exact test (2 × 3 contingency table) was used to analyze the distribution of the percent of embryos that were normal, percent of embryos that were developmentally delayed, and percent of embryos that displayed either pre-gastrula arrest or the asymmetric epiboly phenotype at each indicated stage and condition.

RESULTS

The nomenclature used reflects the species of origin of the factor under discussion. Experiments where human UHRF1 is injected, human nomenclature is used to refer to UHRF1 but when the endogenous gene, mRNA or protein are manipulated or discussed, zebrafish nomenclature is used.

UHRF1 overexpression induces asymmetric epiboly

uhrf1 is a maternally provided and abundant mRNA present throughout the early stages of development up to the prim 5 stage at 24 hpf (Figure 1A and (Jacob et al., 2015)). The observation that both reduction and overexpression of Uhrf1 in zebrafish embryos results in significant mortality by gastrulation (Chu et al., 2012) suggests that early development requires tight regulation of Uhrf1 levels.

Figure 1. Distinct gastrulation defects in UHRF1 overexpressing embryos during early zebrafish embryogenesis.

(A) Endogenous uhrf1 and dnmt1 are maternally provided. RT-qPCR and RT-PCR (inset, gel electrophoresis) display the relative transcript levels of uhrf1 in unfertilized eggs, shield stage embryos (6 hpf), prim 5 stage embryos (24 hpf), 3 dpf larvae, and 5 dpf larvae. (B) Progression of embryonic development in uninjected wild-type embryos and UHRF1 overexpressing embryos at each indicated stage. At high stage (3.3 hpf) all embryos appear normal and at 50% epiboly (~5.25 hpf) and shield stage (6.0 hpf), arrows indicate asymmetric epiboly and delay in dorsal organizer formation. At bud stage (10.0 hpf), arrows indicate failure to complete epiboly and premature development of emerging tail bud and at 7-8-Somites stage (~12.0 hpf) arrows depict a delay in somite formation, pinched off yolk and, in some cases, a lack of a developing tail. At Prim-5 stage (24.0 hpf) arrows indicate morphological defects such as under-developed eyes and head, and defects in tail extension. Scale bar = 500 micrometers. (C) The percent of embryos displaying asymmetric epiboly at 6 hpf in those injected with control RNA (EGFP and mpi) and WT-UHRF1. Significance of the percentage of the asymmetric epibly was determined by Fisher’s exact test. Asterisks (*) indicate p < 0.0001. The number of embryos and number of clutches for each condition is indicated.

To identify the developmental defects caused by WT-UHRF1 overexpression, 100 pg of mRNA encoding the human WT-UHRF1 mRNA with an N-terminal 6-repeat Myc tag (6X-Myc-UHRF1, hereafter called WT-UHRF1; Figure 1B) or EGFP as a control (Figure S1A) was injected into embryos. Individual embryos from three separate clutches were tracked from the high stage (3.3 hpf; mid-blastula transition) through the prim-5 stage (Figure 1B and Figure S1A). Over 20% of embryos injected with WT-UHRF1 that were individually tracked had a distinct phenotype that was first identifiable at the dome stage (~4.5 hpf) and became most pronounced at the 50% epiboly stage (~5.25 hpf). This was characterized by a failure of cells to migrate and intercalate on the dorsal side of the embryo, so that epiboly movements appeared to be asymmetrical (black arrow and inset in Figure 1B). Many of the embryos with asymmetric epiboly failed to form an organizer (Figure 1B, arrow head) and 17% of embryos developed severe developmental phenotypes by the bud stage (10.0 hpf), including a failure to complete epiboly prior to bud formation and pinching off of the yolk cell (Figure 1B). Although affected embryos were able to establish body axes, form somites and neural tissue, their development was severely delayed and embryos displayed defects along the anterior-posterior (A–P) axis with severely underdeveloped head and tail morphologies by the prim-5 stage (8/76 embryos; ~11%). All epiboly defective embryos as well as over 20% of the embryos that appeared normal at shield stage displayed other severe defects during somitogenesis (Figure 1B). The same phenotype was observed in the embryos injected with the zebrafish uhrf1 mRNA sequence (Figure S1B) but was never observed in uninjected embryos or those injected with similar concentrations of mRNA encoding EGFP (Figure 1C) or an unrelated mRNA encoding mannose phosphate isomerase (Mpi; Figure 1C; Chu et al., 2013).

In an average of three experiments where individual embryos were tracked through developmental stages, 21% of WT-UHRF1 injected embryos presented with the asymmetric epiboly phenotype at the time when controls had reached 50% epiboly; by Prim5, nearly half the injected embryos were abnormal (Figure 1B), with a cumulative mortality of 14% (Figure S1C). The progressive decline in the presentation of severely affected embryos through developmental time (Figure 1B) is due to the death of embryos as development progressed through gastrulation. Although there was clutch to clutch variability (Figure S1D), on average across 24 individual clutches with over 500 embryos injected with 100 pg of mRNA, 36% presented with the asymmetric epiboly phenotype. In contrast, this phenotype was never observed in 962 of the uninjected sibling controls (Figure 1C and S1C). Injecting higher concentrations of WT-UHRF1 mRNA did not increase the phenotype penetrance further, but did significantly increase mortality (not shown). We conclude that early embryos are highly sensitive to changes in Uhrf1 levels, and we reasoned that Uhrf1 is tightly regulated at this stage.

Phosphorylation-deficient UHRF1 (UHRF1-S661A) is a potent inducer of asymmetric epiboly

We previously demonstrated that Uhrf1 is phosphorylated on a conserved serine (661 in humans and 648 in zebrafish) by CyclinA2/Cdk2 and that a non-phosphorylatable UHRF1 mutant (UHRF1^S661A) is unable to rescue the early embryonic lethality caused by Uhrf1 depletion (Chu et al., 2012). We therefore asked whether UHRF1 phosphorylation is required for inducing the asymmetrical epiboly phenotype by tracking individual embryos injected with 6X-Myc-UHRF1^S661A (called UHRF1^S661A from hereon) from high stage through the prim-5 stage (Figure 2A). Injecting 100 pg of human UHRF1^S661A mRNA caused the same phenotype as WT-UHRF1: developmental delay (~10%; Figure 2C), mortality (16% cumulative by prim-5; Figure S2C) and asymmetric epiboly (arrowhead Figure 2A, inset). The same phenotype was observed in embryos injected with mRNA encoding the zebrafish non-phosphorylatable mutant (uhrf1^S648A) (figure S1B). Compared to embryos injected with mRNA encoding WT-UHRF1, significantly more embryos were affected by injection with RNA encoding the non-phosphorylatable mutant mRNA (Figure 2B–C and Figure S1A–B).By the time controls had reached 50% epiboly, over half of UHRF1^S661A injected embryos were abnormal (delayed or asymmetric epiboly) compared to 21% in the WT-UHRF1 injected siblings (Figure 2B). The asymmetric epiboly phenotype, but not developmental delay, was directly correlated with RNA concentration (Figure 2C). While penetrance varied across clutches (Figure S3A), at every concentration tested, the percent of affected embryos was higher in the cohort injected with mRNA encoding UHRF1^S661A (45%) compared to WT-UHRF1 (36%; p<0.01; Figure 2C and Figure S2B). These differences were observed despite equivalent levels of UHRF1 protein expressed in embryos injected with each mRNA (Figure 2D and (Chu et al., 2012)). Thus, the phenotype caused by WT-UHRF1 overexpression in early embryos is enhanced when UHRF1 is in the non-phosphorylated state. This is consistent with the hypothesis that UHRF1 phosphorylation on serine 661 is required for its function in early embryos.

Figure 2. Overexpression of phosphorylation-deficient mutant UHRF1 (UHRF1^S661A) phenocopies UHRF1 overexpressing embryos eliciting dramatic gastrulation defects.

(A) Progression of embryonic development in uninjected and embryos injected with 100 pg of UHRF1^S661A mRNA . At 50% epiboly (5.25 hpf), arrows indicate asymmetric epiboly. At Shield stage (6.0 hpf) arrows point to the missing shield and cells that did not undergo proper epiboly cell movements. At bud stage (10.0 hpf) arrows indicate failure to complete epiboly and premature development of tail bud. At 7-8-Somite stage (~12.0 hpf) arrows indicate disrupted anterior-posterior axis, somite formation, and irregular tail formation. At Prim-5 stage (24.0 hpf) arrows indicate defects in head and tail. Scale bar = 500 micrometers. (B) Curve displays the percent of embryos injected with 100 pg of WT-UHRF1 or UHRF1^S661A or uninjected controls that were scored as normal; the stages set by the progression of uninjected control embryos (top) and hpf (bottom). Significance was determined by Log-rank Test (p < 0.0001). (C) Distribution of phenotypes of embryos injected with 25, 50 and 100 pg of WT-UHRF1 and UHRF1^S661A mRNA. Penetrance of the asymmetric epiboly phenotype is statistically significant compared to uninjected controls at each indicated concentration by Fisher’s exact test. Asterisks (*) indicate p < 0.001. n.s. = not significant. Number of biological replicates (clutches), number of embryos per treatment, and percent of embryos displaying the asymmetric epiboly phenotype are indicated. (D) Western blot using the Myc tag to detect UHRF1 in WT-UHRF1 and UHRF1^S661A overexpressing embryos. Mean levels of Myc-UHRF1 protein per treatment are quantified relative to tubulin and were determined to be equivalent by Student’s t-test (p = 0.36). n.s. = not significant.

DNA hypomethylation does not cause asymmetric epiboly

We reasoned that the phenotype caused by WT-UHRF1 overexpression could be attributed to a dominant negative effect. To test this, we compared the asymmetric epiboly phenotype caused by WT-UHRF1 overexpression to loss of endogenous Uhrf1 in early embryos. Uhrf1 is maternally provided (Figure 1A) and thus zygotic uhrf1 mutants progress through early developmental stages, but begin to show defects during mid-organogenesis and die during development starting at 7 days post fertilization (dpf) (Jacob et al., 2015; Sadler et al., 2007; Tittle et al., 2011). Thus, in order to investigate the effects of uhrf1 loss on early development, we utilized two non-overlapping morpholinos targeting maternal and zygotic uhrf1 mRNA (MO1-ATG blocks the ATG and downstream nucleotides in the coding region, and MO2-5’-UTR binds directly upstream of the translation start site). Both induced the phenotype we previously described (Chu et al., 2012): developmental arrest at high stage (3.3 hpf), prior to epiboly (12 – 19% with MO1-ATG; p<0.001 and 9–32% arrest at high stage with MO2-5’-UTR; p<0.001; Figure 3A and Figure S3A). All embryos that had this arrest phenotype died by the time their siblings had undergone gastrulation (not shown). Surviving uhrf1 morphants showed abnormalities during epiboly including yolk extrusion, bud formation prior to completing epiboly, and despite forming the body axes, uhrf1 morphants did not form all of their somites. Morphant embryos also failed to form most neural structures, and were highly abnormal by the time their uninjected siblings reached prim-5 (Figure 3A).

Figure 3. uhrf1 knockdown does not phenocopy UHRF1 overexpression.

(A) Developmental progression of uninjected embryos and uhrf1 morphants (MO2-5’-UTR) at each indicated stage. At high stage (~3.3 hpf) both uninjected embryos and uhrf1 morphants appear normal and at 50% Epiboly (~5.25 hpf), arrow indicates arrest of uhrf1 morphant that have progressed to yolk extrusion. At shield stage (~6.0 hpf), arrow indicates significant developmental delay in uhrf1 morphants that did not arrest at high stage. Morphants that did arrest at high stage are dead by the time uninjected embryos have reached shield stage. At bud stage (~10.0 hpf), arrow indicates tail bud formation prior to the completion of epiboly in uhrf1 morphants. At 7-8 Somite stage (~12.0 hpf), arrows indicate disrupted CNS, somites and tail bud. At Prim 5 stage (~24.0 hpf), arrows indicate that all surviving uhrf1 morphants display CNS and tail defects. Scale bar = 500 micrometers. (B) Embryos were scored at 6 hpf for asymmetric epiboly, high stage arrest or delay. The depletion of endogenous Uhrf1 by injection of 13 ng 5’UTR uhrf1 morpholino does not affect the asymmetric epiboly phenotype caused by injection of 100 pg of WT-UHRF1 or UHRF1^S661A mRNA. The co-injection of WT-UHRF1 or UHRF1^S661A mRNA additively increases the percentage of the embryos affected by the asymmetric epiboly phenotype. The treatments, the number of biological replicates, number of embryos and the proportion of embryos displaying the asymmetric epiboly phenotype are indicated.

The specificity of this phenotype was confirmed by the findings that it was (i) concentration dependent (Figure S3A), (ii) rescued by co-injecting WT-UHRF1 mRNA (Figure 3B and (Chu et al., 2012), (iii) was unaffected by co-injection with a p53 morpholino (Chu et al., 2012) and (iv) was not observed with a morpholino targeting a different gene that is unrelated to DNA methylation (mpi morpholino; Figure S3B and Chu et al., 2013). Despite these profound phenotypes, none of the uhrf1 morphants presented with the asymmetric epiboly phenotype.

Finally, co-injection of WT-UHRF1 or UHRF1^S661A mRNA and uhrf1 MO2–5’-UTR morpholino did not suppress the asymmetric epiboly phenotype induced by UHRF1 overexpression (Figure 3B), but coinjecting WT-UHRF1 or UHRF1^S661A mRNA additively enhanced the penetrance of the asymmetric epiboly phenotype. These findings suggest that the phenotype is primarily caused by an accumulation of unphosphorylated UHRF1, which would occur even if embryos are injected with WT-UHRF1. This overabundance could saturate Cdk2/Ccna2 kinase capacity and effectively increase the total amount of unphosphorylated UHRF1 in the embryo.

Since none of the WT-UHRF1 overexpressing embryos presented a pre-gastrula arrest phenotype at high stage, we conclude that overexpression and knockdown of Uhrf1 have different developmental consequences, and thus the phenotype caused by WT-UHRF1 overexpression is not attributed to a dominant negative effect. We investigated whether the differences in phenotype could be attributed to differences in genome-wide DNA methylation levels. As expected, the level of global 5-methylcytosine (5-MeC) levels in uhrf1 morphants were significantly reduced at 3 and 6 hpf, with average residual levels of DNA methylation being 36% and 60%, respectively (Student’s t-test, p < 0.05; Figure 4A and 4B). Consistent with our previous findings that WT-UHRF1 overexpression in zebrafish hepatocytes reduces DNA methylation (Mudbhary et al., 2014), embryos overexpressing both WT-UHRF1 and UHRF1^S661A had similarly lower levels of bulk DNA methylation at both 3 hpf and 6 hpf (Figure 4A–B; Figure 3 hpf: Student’s t-test, p < 0.05; 6hpf: Student’s t-test, p < 0.05).

Figure 4. Knockdown of endogenous uhrf1 and overexpression of UHRF1 results in equivalent genome-wide DNA hypomethylation.

Slot blot analysis and quantification of methylated genomic DNA of (A) embryos injected with 100 pg of WT-UHRF1 or UHRF1^S661A mRNA, 3.4 ng of 5’UTR uhrf1 morpholino, and 0.43 pg of dnmt1 morpholino were normalized to their uninjected siblings at 3 hpf. (B) The same samples were assessed at 6 hpf, with the addition of samples exposed to 70 or 100 µM 5-AZA from 0.5 hpf normalized to uninjected, untreated siblings. Student’s t-test assessed determine significance between each respective treatment (p < 0.05). n.s. = not significant. Number of biological replicates are indicated.

The juxtaposition of similar bulk levels of DNA hypomethylation and distinct phenotypes of the WT-UHRF1 overexpressing embryos and uhrf1 morphants led us to conclude that DNA hypomethylation was not the underlying mechanism of either phenotype. To test this further, we examined whether other means of inducing DNA hypomethylation would recapitulate the asymmetric epiboly phenotype caused by WT-UHRF1 overexpression.

Varying concentrations of dnmt1 morpholino injection, and treatment with the Dnmt1 inhibitor 5-azacytidine (5-AZA), were optimized to achieve a similar degree of DNA hypomethylation at 6 hpf as obtained by WT-UHRF1 overexpression (Figure 4A–B). However, no phenotype was observed across the range of dnmt1 morpholino concentrations tested (Figure 5A–B) despite significant reduction in Dnmt1 protein by 3 hpf (Figure 5C, Student’s t-test, p = 0.02 at 0.43ng dnmt1 MO and p = 0.05 at 0.85ng dnmt1 MO). Two distinct phenotypes were observed in embryos exposed to 5-AZA immediately following fertilization: (i) they proceeded through high stage similar to control embryos, then arrested at sphere stage (70 µM 5-AZA: 37% arrested, Figure S4A), or (ii) were delayed, and attempted epiboly when their untreated control siblings were at bud stage (Figure S4A–B). The percent of embryos arrested at sphere stage was decreased at lower 5-AZA concentrations (70 µM 5-AZA: 17% sphere stage arrest, p < 0.05; 100 µM 5-AZA: 86% sphere stage arrest, p < 0.05; Figure S4B); however, all concentrations tested resulted in 100% mortality during mid-somitogenesis by 12 hpf (Figure S4C). Importantly, by 6 hpf, 70 µM and 100 µM 5-AZA treatment resulted in 40% and 57% residual DNA methylation, respectively (Student’s t-test, p < 0.05; Figure 4B). Since the phenotypes associated with uhrf1 knockdown, dnmt1 knockdown, and 5-AZA treatment were all markedly different, yet each achieved similar degrees of genomic DNA hypomethylation (Table 1), we conclude that the asymmetric epiboly caused by WT-UHRF1 and UHRF1^S661A overexpression is not due to bulk DNA hypomethylation, although it is possible that site-specific differences in DNA methylation could contribute to their unique outcomes.

Figure 5. dnmt1 knockdown does not phenocopy uhrf1 morphants or UHRF1 over-expressing embryos.

(A) Developmental progression of uninjected embryos and those injected with 0.43 ng of dnmt1 morpholino at each indicated stage. Scale bar = 500 micrometers. (B) Quantification of developmental delay at 6 hpf in embryos injected with 0.43, 0.85, 1.28 and 1.7 ng of dnmt1 morpholino and uninjected control embryos. Fisher’s exact test was used to determine statistical significance. n.s. = not significant. (C) Representative western blot and quantification of Dnmt1 protein knockdown at indicated amount of dnmt1 morpholino. Student’s t-test was used to calculate the indicated p-values compared dnmt1 morpholino injected embryos to uninjected controls. Values are average of 3 biological replicates, error bars are standard deviation.

Table 1. Different mechanisms of inducing DNA hypomethyltion cause distinct phenotypes in early zebrafish embryos.

Embryos were injected with the indicated morpholino (MO) or RNA immediately after fertilziation or were treated with 5-aza before the 16 cell stage. Methylation levels were determined by slot blot analysis with anti-5MeC. Dnmt1 protein was determined by Western blot. Not done (ND).

Genetic Approach	Change in DNA Methylation	Embryonic Phenotype	Dnmt1 Protein
uhrf1 Knockdown (MO)	60% Residual DNA Methylation	High Stage Arrest	Increased
dnmt1 Knockdown (MO)	52% Residual DNA Methylation	None	Decreased
5-Azacytidine Treatment	57% Residual DNA Methylation	Sphere Stage Arrest	Inhibited
WT-UHRF1 Over- expression (RNA)	61% Residual DNA Methylation	Asymmetric Epiboly	Slightly Increased
UHRF1S661A Over- expression (RNA)	43% Residual DNA Methylation	Asymmetric Epiboly	Increased
WT-dnmt1 Over- expression (RNA)	32% Residual DNA Methylaton	Asymmetric Epiboly	ND
dnmt1C1109S Over- expression (RNA)	ND	Asymmetric Epiboly	ND

UHRF1 overexpression increases Dnmt1 protein levels

Our finding that the asymmetric epiboly phenotype was due to a DNA methylation-independent function of UHRF1 led us to examine other functions of UHRF1 as an underlying mechanism. UHRF1 displays E3 ligase activity through its C-terminal RING domain and is known to ubiquitinate DNMT1 and target the protein for proteasomal degradation (Bashtrykov et al., 2014; Berkyurek et al., 2014; Felle et al., 2011; Mudbhary et al., 2014; Qin et al., 2011). We previously demonstrated that transiently increasing WT-UHRF1 in developing zebrafish embryos at 48 hpf decreased levels of Dnmt1 in the zebrafish (Mudbhary et al., 2014) and here we tested whether a similar mechanism accounted for DNA hypomethylation in early embryos that overexpress WT-UHRF1. There was no significant change in the mRNA transcript level of dnmt1 compared to uninjected WT siblings (Figure 6A); however, Dnmt1 protein was increased in WT-UHRF1 overexpressing embryos (Student’s t-test, p = 0.08; Figure 6B), and significantly increased in UHRF1^S661A overexpressing embryos (Student’s t-test p = 0.0002; Figure 6B). At 6 hpf UHRF1^S661A overexpressing embryos displayed a 47% increase in Dnmt1 protein compared to controls and 41% more than WT-UHRF1 injected embryos (Figure 6B; Student’s t-test, p = 0.005), despite similar levels of UHRF1 protein (Figure 2D and 6B). Together, these data indicate that UHRF1 overexpression in early embryos does not cause Dnmt1 degradation, as found later in development, but instead suggests that abundance of non-phosphorylated UHRF1 increases Dnmt1 protein levels.

Figure 6. Dnmt1 protein levels are stabilized in UHRF1^S661A over-expressing embryos.

(A) dnmt1 mRNA expression at 6 hpf in uninjected embryos, WT-UHRF1 overexpressing embryos, and UHRF1^S661A overexpressing embryos. Mean ΔC_T values are indicated. Student’s t-test was used to determine p-value, n.s. indicates not significant (p< 0.05). (B) Western blot of Dnmt1, Myc-WT-UHRF1, and Myc-UHRF1^S661A (top) and quantification of 5 biological replicates are shown with the median as the middle of the box and the span of the box indicating the 75^th and 25^th percentiles with the whiskers indicating the 90^th and 10^th percentiles. Student’s t-test was used to calculate p-values.

Overexpression of WT-dnmt1 and catalytically inactive dnmt1^C1109S phenocopies the developmental defects induced by UHRF1 overexpression

We hypothesized that the increased levels of Dnmt1 protein contributed to the development of the asymmetric epiboly phenotype caused by WT-UHRF1 and UHRF1^S661A overexpression. To test this, we injected embryos with mRNA encoding zebrafish Dnmt1 and followed individual embryos from high stage through Prim-5 (Figure 7A). In this experiment, 9/49 (14%) of dnmt1 mRNA injected embryos presented with the asymmetric epiboly phenotype (Figure 7A). Averaging 3–6 clutches of embryos injected with a range of dnmt1 mRNA concentrations, we found a concentration dependent increase in penetrance of the asymmetric epiboly phenotype, with 15% of embryos affected by injection of 1 and 10 pg, 33% of embryos injected with 50 pg and 44% of embryos injected with 100 pg of dnmt1 mRNA. The dnmt1 antisense transcript did not induce this phenotype at any concentration (Figure 7B). In all experiments, dnmt1 injected embryos appeared identical to those shown in Figure 1A and 2A: they failed to initiate dorsal organizer formation, did not complete epiboly before initiating bud formation, and although body axis formation was achieved, dnmt1 overexpressing embryos displayed severe morphological abnormalities along the anterior-posterior axis by 24 hpf when the controls were at the Prim-5 stage.

Figure 7. Dnmt1 overexpression phenocopies the asymmetric epiboly phenotype of UHRF1 and UHRF1^S661A overexpressing embryos.

(A) Developmental progression of uninjected and WT-dnmt1 overexpressing embryos at indicated stages. At High Stage (~3.3 hpf) all embryos appear normal at 50% epiboly (~5.25 hpf) and Shield stage (6.0 hpf) arrows indicate asymmetric epiboly and lack of dorsal organizer. At Bud stage (10.0 hpf) arrow points to the front of the migrating cells indicating delayed completion of epiboly. At Prim-5 stage (24.0 hpf) arrows indicate severe defects in head and tail with lack of head and tail extension. (B) Embryos were scored at 6 hpf for asymmetric epiboly and delay following injection of 1, 10, 12.5, 25, 50 or 100 pg of WT-dnmt1 mRNA (sense) or 25, 50, or 100 antisense dnmt1 RNA as a negative control. The number of clutches, total number of embryos scored and percent of asymmetric epiboly are indicated. (C) Overexpression of catalytically inactive dnmt1^C1109S (MT-dnmt1) causes asymmetric epiboly. Asterisk indicates statistical significance (p < 0.05) using Fisher’s exact test. The number of embryos and biological replicates are indicated. n.s. = not significant.

Remarkably, this same phenotype was recapitulated in embryos overexpressing catalytically inactive dnmt1^C1109S, where 49% of embryos display the asymmetric epiboly phenotype despite the inability to facilitate DNA methylation (54% of WT-dnmt1 injected embryos and 49% of dnmt1^C1109S injected embryos; Fisher’s Exact Test p < 0.0001; Figure 7C) (Rai et al., 2006). Thus, over-expression of dnmt1 drives the asymmetric epiboly phenotype independent of its role in DNA methylation.

The asymmetric epiboly phenotype can be rescued via dnmt1 knockdown in an UHRF1^S661A background

To determine if increased Dnmt1 protein was the mechanism by which UHRF1 overexpression induced asymmetric epiboly, we knocked down Dnmt1 in embryos injected with WT-UHRF1 and UHRF1^S661A mRNA. Embryos were co-injected with the lowest concentration of dnmt1 morpholino that effectively reduced Dnmt1 protein levels (0.43 ng) and 100 pg of mRNA encoding WT or phospho-deficient UHRF1. While dnmt1 morpholino alone did not result in any significant phenotypes at 6 hpf compared to uninjected controls (Figure 5 and Figure 8A), it significantly reduced the incidence of the asymmetric epiboly phenotype in UHRF1^S661A over-expressing embryos to 17% (p < 0.0001 Fisher’s Exact Test; Figure 8A), but only resulted in minor rescue of the asymmetric epiboly phenotype when co-injected with WT-UHRF1 (Figure 8A). This suggests that the enhanced Dnmt1 protein level induced by UHRF1^S661A over-expression was required for asymmetric epiboly. To test whether Uhrf1 was also necessary for the induction of the asymmetric epiboly phenotype, we knocked down Uhrf1 in embryos and co-injected dnmt1 mRNA. Uhrf1 depletion significantly decreased the penetrance of the asymmetric epiboly phenotype (p < 0.02 Paired t-test; Figure 8B). Thus, Dnmt1 requires Uhrf1 to induce asymmetric epiboly.

Figure 8. UHRF1 and Dnmt1 interact to cause asymmetric epiboly.

(A) Embryos co-injected with 0.43 ng dnmt1 morpholino and either 100 pg WT-UHRF1 or UHRF1^S661AmRNA were scored at 6 hpf for asymmetric epiboly. The number of biological replicates, number of embryos per treatment and the proportion of embryos with asymmetric epiboly are indicated. Significance was determined using Fisher’s exact test; asterisk indicates p < 0.001; n.s. = not significant. (B) Embryos from the same clutch were injected with 3.4 ng of 5’UTR uhrf1 MO or with 100 pg dnmt1 mRNA or a combination of these two and were scored at 6 hpf for asymmetric epiboly. The lines connect the data from each treatment in individual clutches. In all 4 replicates, co-injection suppressed the penetrance of the asymmetric epiboly phenotype. The number of biological replicates, the number of embryos per treatment, and the proportion of embryos displaying each phenotype are indicated. Significance of the asymmetric epiboly phenotype between the embryos injected with dnmt1 and the co-injected embryos was assessed by paired Students t-test. Embryos from the same clutch were injected with 25 pg of WT-dnmt1 mRNA with either 50 pg WT-UHRF1 (C) or UHRF1^S661A (D) mRNA. Embryos were scored at 6 hpf for asymmetric epiboly. In all 6 replicates for WT-UHRF1 injection and 4 replicates for UHRF1^S661A injection, co-injection enhanced the penetrance of the phenotype compared to each mRNA alone. Significance was determined using one-way ANOVA and p-values are indicated. The significance of the percentage of the increase of the asymmetric epiboly observed in the co-injected compared to the single injection treatment was assessed by paired t-test (p-values are indicated, n.s. indicates not significant).

If increased Dnmt1 levels mediate the asymmetric epiboly phenotype caused by UHRF1 over-expression, then further increasing Dnmt1 should enhance the phenotype. We tested this by co-injecting embryos with dnmt1 mRNA and either WT-UHRF1 or UHRF1^S661A mRNA (Figure 8C–D, Figure S5A). These appear to have an additive effect on the incidence of the asymmetric epiboly phenotype, in 6 clutches, the incidence of asymmetric epiboly caused by WT-UHRF1 injection (50 pg) increased from 45% to 70% upon co-injection with dnmt1 mRNA (25 pg) and in 4 clutches injected with UHRF1^S661A mRNA (50 pg), incidence increased from 39% to 62% (Figure 8C–D). While lower concentrations of dnmt1 mRNA did induce a low incidence of the asymmetric epiboly phenotype, the additive effect was only observed at higher concentrations of dnmt1 (Figure S5A), while mRNA encoding an unrelated protein (Mpi) had no significant effect on the phenotype induced by injection of either WT or phospho-deficient UHRF1 (Figure S5B). This suggests a threshold concentration of UHRF1 or dnmt1 mRNA may be needed to elicit the asymmetric epiboly phenotype. Injecting concentrations of each mRNA, that alone did not elicit any phenotype, were unsuccessful at inducing a phenotype when injected together (not shown).

In summary, our data support the hypothesis that increased levels of Dnmt1 protein is a mechanism by which high UHRF1 levels block gastrulation. We provide evidence that UHRF1 phosphorylation is a mechanism by which UHRF1 controls Dnmt1 levels in the early embryo. Moreover, our findings lead to the surprising conclusion that loss of bulk DNA methylation does not underlie the range of phenotypes caused by alterations in the DNA methylation machinery (Table 1) and suggest that these factors participate in other complexes.

Discussion

UHRF1 is a multi-domain protein that is essential for multiple aspects of epigenomic patterning. Interaction of UHRF1 and DNMT1 is required for maintaining DNA methylation during DNA replication and a logical assumption that the global loss of DNA methylation caused by loss of either factor is the mechanism of early embryonic lethality in mice (Bostick et al., 2007; Lei et al., 1996; Muto et al., 2002; Sharif et al., 2007) and zebrafish (Chu et al., 2012; Jacob et al., 2015; Sadler et al., 2007; Tittle et al., 2011). However, UHRF1 has other functions, as does DNMT1, albeit less well understood. For instance, UHRF1 has been shown to ubiquitinate DNMT1 via its RING domain (Bashtrykov et al., 2014; Berkyurek et al., 2014; Felle et al., 2011; Mudbhary et al., 2014; Qin et al., 2011) and target it for destruction by the proteasome. Thus, UHRF1 is required for both promoting DNMT1 action and for reducing its efficacy. Our previous work demonstrated that UHRF1 is specifically phosphorylated by CCNA2/CDK2 in vivo and in vitro (Chu et al., 2012) and this is among the few modifications of the UHRF1 protein (Tauber and Fischle, 2015) which is essential for function. Serine-661 is located between the SRA and RING domains and our current findings suggest that phosphorylation in this region may induce a conformational change (Tauber and Fischle, 2015) that reduces the ability of Uhrf1 to target Dnmt1 for destruction. Moreover, our finding that the early embryonic phenotype induced by UHRF1 requires Dnmt1 and, vice versa, the effect of Dnmt1 overexpression requires Uhrf1 demonstrates an important, interdependent relationship between these two proteins during the stages of rapid epigenome re-patterning in early development.

Both loss and overexpression of UHRF1 in early zebrafish embryos caused pre-gastrulation lethality and the effects of uhrf1 loss were rescued by re-introduction of either the human or zebrafish WT, but not phosphorylation deficient, UHRF1 (Chu et al., 2012). This indicated that (i) tight regulation of UHRF1 levels is required for embryonic viability and (ii) that phosphorylation on serine 661 in the human protein or serine 648 in the zebrafish homolog was essential for UHRF1 function in early development. We do not understand why the two experiments in Figure 3B did not recapitulate our previous findings that phosphodeficient UHRF1 could not rescue the uhrf1 morpholino induced mortality (Chu et al., 2012), and efforts are currently underway to elucidate this discrepancy. The focus of the present study was to determine why overexpression of UHRF1 results in early embryonic lethality and how phosphorylation alters UHRF1 function. This is not only important for understanding early development, but also for providing insight into cancer onset and progression, as UHRF1 is an oncogene (Jenkins et al., 2005; Mousli et al., 2003; Mudbhary et al., 2014).

Increased levels of either human or zebrafish UHRF1 in early zebrafish embryos caused severe defects in epiboly and gastrulation and blocking UHRF1 phosphorylation enhanced efficacy at inducing this phenotype. We conclude that in both cases, unphosphorylated UHRF1 accumulates, but that this is more severe in the embryos injected with UHRF1^S661A. This was further confirmed by our finding that injecting both WT and non-phosphorylatable UHRF1 resulted in a very high penetrance of the asymmetric epiboly phenotype. The observation that this was phenocopied by Dnmt1 overexpression, suppressed by Dnmt1 knockdown, and additively enhanced by co-injection with Dnmt1, support the model that elevation of Dnmt1 and UHRF1 together block epiboly and gastrulation (Figure 9). Additionally, our data suggest that UHRF1 phosphorylation promotes degradation of Dnmt1.

Figure 9. Model of relationship between uhrf1 expression level dnmt1 expression level, DNA methylation, and progression through gastrulation.

Knockdown of uhrf1 leads to genome-wide DNA hypomethylation and a pre-gastrula arrest phenotype (high stage arrest, 3.3 hpf). Overexpression of WT-UHRF1 and phospho-deficient UHRF1^S661A also induces genome-wide DNA hypomethylation, but acting together with dnmt1, induces an asymmetric epiboly phenotype, ultimately disrupting gastrulation. This may be due in part to increased levels of non-phosphorylated UHRF1 and a subsequent increase in Dnmt1 protein resulting in the asymmetric epiboly phenotype due to a DNA methylation-independent function of Dnmt1.

The striking difference between the phenotype of uhrf1 morphants and embryos overexpressing UHRF1 suggested that the phenotype caused by UHRF1 overexpression could not be attributed to a dominant-negative effect, although it is possible that the DNA hypomethylation caused by high UHRF1 levels could be attributed to mislocalization of Dnmt1, as was observed when UHRF1 was overexpressed in zebrafish hepatocytes (Mudbhary et al., 2014). The uhrf1 morphant phenotype is unlikely to be caused by off target effects as it was rescued by co-injection with UHRF1 mRNA, was not rescued by p53 inhibition (Chu et al., 2012), phenocopies the mouse knock-out (Lei et al., 1996; Muto et al., 2002; Sharif et al., 2007) and mimics the cellular (apoptosis; BK unpublished) and molecular (DNA hypomethylation) consequences of uhrf1 mutation (Jacob et al., 2015; Sadler et al., 2007; Tittle et al., 2011). Since UHRF1 overexpression and knockdown caused equivalent levels of DNA hypomethylation but induced different phenotypes, they must be functionally distinct (Figure 9). Indeed, we were surprised to find that each approach to manipulate the DNA methylation machinery resulted in very different phenotypes, despite similar levels of bulk DNA hypomethylation, suggesting that Uhrf1 and Dnmt1 possess DNA methylation-independent functions during embryogenesis. However, these results could also be attributed to differences in the methylome patterns caused by each approach, as differences in the methylation profiles across the genome could explain the emergent differences in phenotype presentation. Comparative methylome analysis is an essential approach to detect differences in the methylome and an important future goal to address this.

An alternative possibility is that different complexes are affected by each approach used to manipulate the DNA methylation machinery. For instance, when Uhrf1 is depleted there is excess Dnmt1 (BK, not shown and (Jacob et al., 2015)), but it cannot efficiently bind to hemimethylated DNA and thus may promiscuously interact with other proteins (Espada et al., 2011) or regulatory RNAs (Di Ruscio et al., 2013). However, since the asymmetric epiboly phenotype does not occur in uhrf1 morphants where Dnmt1 protein is also elevated (not shown), we conclude that this phenotype is caused by interaction between non-phosphorylated UHRF1 and Dnmt1. Our finding that the asymmetric epiboly phenotype caused by WT-UHRF1 is not accompanied by a significant increase in Dnmt1 protein could suggest that the level of non-phosphorylated UHRF1 that is generated by this experimental approach can function with the endogenous levels of Dnmt1 to generate this phenotype. In contrast, 5-AZA treatment does not appear to alter Dnmt1 protein levels, but prevents it from acting, and thus may stall Uhrf1-Dnmt1 complexes on hemi-methylated DNA during S-phase, which could be sensed as either replication stress or DNA damage. Indeed, 5-AZA does cause DNA damage (Covey et al., 1986), and as the 5–AZA induced phenotype emerged after the stage when checkpoints begin to function (Ikegami et al., 1997), it is possible that a cell cycle arrest triggered by DNA damage underlies the distinct 5-AZA induced phenotype.

Intense investigation into the epigenetic regulation of early development and cell specification suggests that DNA methylation may regulate key developmental events. However, our data suggests that loss of bulk methylation does not interfere with development, as the phenotypes induced by overexpression of UHRF1 or UHRF1^S661A, knockdown of uhrf1, knockdown of dnmt1, and treatment with the Dnmt1 inhibitor, 5-AZA were all different, despite similar levels of bulk DNA hypomethylation. Most strikingly, Dnmt1 knockdown resulting in as low as 35% residual DNA methylation at 3 hpf (Figure 4A) without any discernable phenotype in early embryos, suggests that bulk methylation is not required for gastrulation; however, dnmt1 morphants do develop multi-systemic abnormalities later in development (Rai et al., 2006). The result that overexpression of mutant dnmt1^C1109S, which cannot methylate DNA (Rai et al., 2006), causes the asymmetric epiboly phenotype further supports the hypothesis that DNA methylation-independent functions of Dnmt1 are important in early development, as observed in other systems (Espada, 2012). Interestingly, this same Dnmt1 mutant can rescue the phenotypes that occur at late developmental stages in dnmt1 morphants, raising the intriguing possibility that the important role that Dnmt1 plays in development is independent of its function in DNA methylation (Rai et al., 2006). Exciting work from Xenopus embryos suggests that Dnmt1 functions as a transcriptional repressor (Dunican et al., 2008), further supporting this hypothesis.

What is the basis for the asymmetric epiboly phenotype caused by UHRF1-Dnmt1 excess? We found no evidence of abnormal microtubule morphology (not shown), suggesting that the mechanics of cell migration are not grossly affected. One possibility is that there is a block in cell proliferation, as observed when UHRF1 is overexpressed (Mudbhary et al., 2014) or depleted later in development (Jacob et al., 2015; Tittle et al., 2011). Previous reports have demonstrated that Dnmt1 has been found to complex with E2F, pRb, and HDAC1 (Robertson et al., 2000), which have profound implications in cell cycle regulation. While we have found some differences in cell proliferation in UHRF1 overexpressing embryos, these were not confined to the dorsal side of the embryo, where the epiboly defects are most pronounced. Another possibility is suggested by the striking resemblance of the asymmetric epiboly phenotype caused by UHRF1 overexpression to that observed when E-cadherin (cdh1) is knocked down (Babb and Marrs, 2004). Cdh1 is essentialfor the intercalation of the epithelial cell layer during epiboly in zebrafish (Lepage and Bruce, 2010) and is a central player of the epithelial-mesenchymal transition (EMT) in many systems. EMT occurs during epiboly and gastrulation and, interestingly, UHRF1 (Duvall-Noelle et al., 2015) and DNMT1 (Espada et al., 2011) have both been shown to interact with Snail (Fukagawa et al., 2015; Sun et al., 2015) and other transcription factors that regulate cdh1 expression and are required for EMT. DNMT1 has been demonstrated to repress Cdh1 and increase β-catenin activity independent of its DNA methylation activity through its N-terminal DMAP domain (Espada et al., 2011), suggesting that the phenotype induced by high Dnmt1 levels could be caused by Cdh1 repression. While our preliminary studies did not find cdh1 mRNA to be altered in UHRF1 overexpressing embryos (BK, not shown), we did find that expression of other genes important for EMT is deregulated in embryos that overexpress UHRF1, raising the possibility that this is the mechanism underlying the epiboly and gastrulation defects in these embryos.

The implications of this work are relevant to understanding how DNA methylation and the proteins that maintain the methylome regulate early vertebrate development and also to cancer, where DNA methylation (Das and Singal, 2004; Eden et al., 2003; Karpf and Matsui, 2005) and UHRF1 overexpression (Mudbhary et al., 2014) have been shown to be involved in key oncogenic pathways. Moreover, as UHRF1 is now recognized as a central regulator of the methylome and histone modifications (Arita et al., 2012), it is important to understand how its different functions are regulated. Our work combined with studies from other groups showing that phosphorylation (Ma et al., 2012) and other modifications (Gelato et al., 2014; Tauber and Fischle, 2015) of UHRF1 can dramatically alter its function indicates that post-translational modification is a major mechanism of regulation of this important protein. Understanding the molecular mechanisms underlying how phosphorylation of UHRF1 affects Dnmt1 stability and UHRF1 function is an important future goal.

Highlights.

• Tight control over UHRF1 and DNMT1 levels is required for gastrulation.

• UHRF1 phosphorylation decreases DNMT1 stability in developing embryos.

• UHRF1 and DNMT1 regulate gastrulation through DNA methylation-independent functions.