The relative importance of DNA methylation and Dnmt2-mediated epigenetic regulation on Wolbachia densities and cytoplasmic incompatibility

Fly rearing and dissections

All flies were reared on a cornmeal and molasses-based media at 25 °C. The Dnmt2 loss-of-function mutant has been previously described (Goll et al., 2006). Briefly, the mutant contains a 28bp insertion with multiple stop codons as well as a frameshift within the coding region of Dnmt2. Overexpressing flies were created through the Gal4-UAS system. Crosses were performed between virgin nos-Gal4 driver females (y¹w^*; P{w[ + mC] = GAL4-nos.NGT}40 (either Wolbachia-infected or uninfected)) and 5–6 uninfected UAS-Dnmt2 (Kunert et al., 2003), UAS-GFP or W¹¹¹⁸ males. Crosses for Fig. S1 were conducted between virgin Act5c-Gal4 driver females (y¹w^*; P{w[ + mC] = Act5C-GAL4}25FO1/CyO, y⁺, Wolbachia infected or uninfected depending on desired progeny) and UAS-Dnmt2 males. For Act5c-Gal4 crosses, straight-winged progeny were assumed to be overexpressing Dnmt2 while CyO expressing lines were used as the wild-type expressing lines. Wolbachia-uninfected lines were created through tetracycline treatment (20 ug/mL for 3 generations) and infection status was confirmed through PCR using the following primers: WolbF (GAAGATAATGACGGTACTCAC) and WolbR3 (GTCACTGATCCCACTTTAAATAAC) which target the 16S rRNA gene of Wolbachia. These lines were further reared for at least three generations on undrugged media before experimentation to avoid detrimental paternal effects seen in other systems (Zeh et al., 2012).

The wAu, wNo, and wRi (also known as wRi Agadir) strains of Drosophila simulans were kindly provided by Charlat Sylvain (University of Lyon, France). All testes and ovary dissections were performed in cold phosphate buffered saline (PBS). Males were dissected within 24 h of emergence while females were aged 3–4 days before dissections. Testes samples consisted of tissue obtained from a minimum of 20 males while ovary samples were pooled from 10 females each. Tissues were frozen and stored at −80 °C before analysis.

Hatch rate assays

Assays were performed using a grape juice/agar media in 30 mm plates for egg laying. For each cross 32–48 individual crosses of one male and one female were set up in separate mating chambers with individual grape juice plates. A minimal amount of a 1:2 dry yeast and water mix was added to each plate and the parents were allowed to mate for 16 h before the grape juice plates were discarded. Fresh plates were then used for 24 h, removed, and the number of eggs laid was counted for each cross. The number of unhatched eggs was counted again at 36 h after the plates had been removed to determine hatch rates.

MethylFlash quantification of DNA methylation

Genomic levels of cytosine methylation (5-mC) were measured using the MethylFlash kit (Epigentek, Farmingdale, NY, USA). 8–10 replicate sets of testes (20–40 testes pairs each replicate) were dissected and DNA was isolated using the Puregene Tissue kit (Qiagen, Venlo, Netherlands). 100 ng of genomic DNA from each sample was used and each sample was analyzed in duplicate on a BMG LabTech FLOUstar OPTIMA plate reader (Ortenberg, Germany) according to manufacturer instructions.

Wolbachia density

Eight replicates each of whole animals (pools of 3), testes (pools of 20 pairs), and ovaries (pools of 10 pairs) were collected and DNA was isolated. All males were less than 24 h old while females had been aged 3–4 days. Quantitative PCR was performed on a Bio-Rad CFX96 Real-Time System using iTaq Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA). groEL copy number, determined against a standard curve, was compared to counts for the host gene Actin, also determined against a standard curve. It was assumed that one copy of groEL was present in each Wolbachia genome and 1 or 2 copies of Act5c (for males and females, respectively, as the gene is on the X chromosome) in each Drosophila genome. Primers: Act5c (231bp product, Forward: ATGTGTGACGAAGAAGTTGCT Reverse: GTCCCGTTGGTCACGATACC), groEL (97bp product, Forward: CTAAAGTGCTTAATGCTTCACCTTC Reverse: CAACCTTTACTTCCTATTCTTG). qPCR conditions: 50°10 min, 95°5 min, 40×(95°10 s, 55°30 s), 95°30 s. Followed by melt curve analysis (0.5°steps from 65–95°for 5 s each).

Gene expression

Quantitative PCR was performed on a Bio-Rad CFX96 Real-Time System using iTaq Universal SYBR Green Supermix (Bio-Rad, Hercules, CA, USA). RNA was isolated from 8 sets of testes (20 pairs each) using the RNeasy Mini kit (Qiagen, Venlo, Netherlands) and DNA was removed with the TURBO DNA-free DNase kit (Ambion, Grand Island, NY, USA). cDNA was synthesized using a SuperScript III First-Strand kit (Invitrogen, Grand Island, NY, USA) and diluted 1:20. All calculations were done using delta delta Ct with Rp49 expression used for normalization of results. Primers: Dnmt2 (150bp product, Forward: CCGTGGCGTGAAATAGCG Reverse: ACACCGCTTTCGGAGGACG), Rp49 (154bp product, Forward: CGGTTACGGATCGAACAAGC Reverse: CTTGCGCTTCTTGGAGGAGA). qRT-PCR conditions are the same as used in qPCR for Wolbachia densities.

Bisulfite sequencing

One hundred testes were dissected in PBS from Wolbachia infected (y¹w^*) and uninfected males and flash frozen. gDNA was then isolated using the Puregene kit (Qiagen, Venlo, Netherlands) and fragmented by Covaris shearing. gDNA was submitted to Vanderbilt Technologies for Advanced genomics (VANTAGE) where the PE-75 bp library was generated using the TruSeq sample preparation kit (with methylated adapters), bisulfite treated, PCR amplified (EpiMark and ZymoTaq) and sequenced (Illumina HiSeq 2000, 86bp PE read). Sequences with ≥10× coverage were analyzed using Bismark (Krueger & Andrews, 2011) and cytosines which were methylated in at least one read were counted.

Wolbachia wMel increases levels of testes DNA methylation

MethylFlash analysis of host DNA from testes revealed that infection with the Wolbachia strain wMel in Drosophila melanogaster increases levels of genome-wide cytosine methylation (Fig. 1A). More importantly, this methylation is specific to the host testes (55% increase, P = 0.0015, Mann Whitney U test) and is not observed in the ovaries, consistent with the prediction that only the paternal genome is modified during cytoplasmic incompatibility. The overall levels of methylation are extremely low, which is consistent with previously reported levels of methylation in Drosophila melanogaster (Lyko, Ramsahoye & Jaenisch, 2000; Kunert et al., 2003). Conflicting reports over the strength and prevalence of DNA methylation in D. melanogaster (Lyko, Ramsahoye & Jaenisch, 2000; Raddatz et al., 2013; Schaefer & Lyko, 2010) led us to test the validity of our initial results with genome-wide bisulfite sequencing. Results indicate that, contrary to most other species, DNA methylation in Drosophila melanogaster is not CpG specific and is evenly distributed over cytosine residues (Fig. 1B and Table S1). Sequencing results also mirror those of MethylFlash and show that infection with wMel increases testes DNA methylation 46% across all cytosine residues with a range of 43–54% depending upon the type of cytosine residue (CpG, CHG, or CHH) (Fig. 1B and Table S1). The minor discrepancies between MethylFlash and bisulfite sequencing (55% and 46% increase in methylation, respectively) are likely due to the sensitivity of the MethylFlash system on such low quantities of methylation. A more thorough investigation of the bisulfite sequencing, including changes in promoter and gene body methylation, is ongoing.

Overexpression of Dnmt2 neither induces nor strengthens CI

Drosophila melanogaster possess just one canonical DNA methyltransferase, Dnmt2, and overexpression of this enzyme in fruit flies has previously been shown to increase levels of DNA methylation (Kunert et al., 2003; Schaefer, Steringer & Lyko, 2008). Utilizing the Gal4-UAS expression system, we overexpressed Dnmt2 in uninfected males to test if an increase in host methylation alone could induce the CI defect of reduced embryo hatching rates. Figure 2 shows that there was no discernable difference in hatching rates with uninfected males expressing increased or wild type levels of Dnmt2 (P = 0.91, MWU). The result was confirmed using an Actin-based driver that again yielded no discernable differences in hatch rates compared to wild type flies (Fig. S1, P = 0.83, MWU). These findings specify that amplified levels of Dnmt2-mediated epigenetic regulation are not sufficient to recapitulate cytoplasmic incompatibility.

If multiple factors are responsible for CI, it is possible that overexpression of Dnmt2, while unable to induce CI in uninfected flies, may be able to strengthen the modification in the presence of Wolbachia. To test this hypothesis, we overexpressed Dnmt2 in Wolbachia-infected males that were then mated to uninfected virgin females. Surprisingly, Dnmt2 overexpression in males decreased the level of cytoplasmic incompatibility by an average of 17.4% (Fig. 3). This effect is not dependent on the Dnmt2 expression status of the female and suggests that increased methylation of host DNA can diminish the penetrance of cytoplasmic incompatibility.

Overexpression of Dnmt2 reduces Wolbachia titers in host testes

Previous work suggested that Dnmt2 is detrimental to Wolbachia proliferation in mosquitoes. In fact, Wolbachia strain wMel Pop-CLA utilizes a host miRNA to downregulate Dnmt2 expression when infecting Aedes aegypti (Zhang et al., 2013). We observed no differences in Dnmt2 expression between wMel infected and uninfected D. melanogaster testes (data not shown) but hypothesized that overexpression of Dnmt2 in the host may adversely affect Wolbachia titers. In support of this prediction, we found that Wolbachia density (as measured by the ratio of Wolbachia groEL gene copy number/Drosophila Actin gene copy number) decreased by 17.3% in adult testes overexpressing Dnmt2 transcripts by 9.6% (Fig. 4 and Fig. S2, respectively). The low level of transcript overexpression could be specific to the developmental stage of the experimental sample or due to usage of a pUAST vector for germline expression instead of the more efficient pUASP (Kunert et al., 2003). While the upregulation of Dnmt2 in infected males is not statistically significant, it remains possible that actual protein levels are much higher than those represented by RNA transcripts. Strong protein expression, as measured by Western blot, was seen in uninfected ovaries (data not shown). Nevertheless, the 17.3% decrease in Wolbachia titers compares well with the 17.4% reduction in CI penetrance reported above. Expression of the negative control green fluorescent protein (GFP) did not reduce Wolbachia titers, as expected (Fig. S3). As the bacterial and phage density models of CI specify that Wolbachia titers in the testes are linked to the strength of CI (Breeuwer & Werren, 1993; Bordenstein et al., 2006), we conclude that the reduction of CI observed in Dnmt2-overexpressing males is likely due to reduced Wolbachia density.

Even though we do not observe any change in Dnmt2 mRNA levels after Wolbachia infection, we cannot rule out that Wolbachia may be affecting intracellular Dnmt2 localization rather than levels of gene expression. An increase in localization of Dnmt2 to the nucleus would not only protect the cytosolic Wolbachia but also explain the additional genomic methylation associated with infection. In this scenario, the testes-specific increase in host methylation initially observed would simply be a by-product of high Wolbachia activity. Additionally, an immunomodulatory role for Dnmt2 in Drosophila has already been documented in protection against RNA viruses (Durdevic et al., 2013) though we believe the findings in this report are the first evidence for a putative antibacterial role for Dnmt2 in fruit flies.

Hosts defective in DNA methylation still exhibit CI

As Dnmt2 overexpression did not induce nor increase cytoplasmic incompatibility, we next tested the strength of CI in hosts defective in the methyltransferase pathway. Knockout mutants for Dnmt2 characterized by Goll et al. (2006) were acquired and found by PCR and amplicon sequencing to be infected by the wMel strain of Wolbachia. The strain is hereafter referred to as Mut and was tetracycline treated for three generations to create the uninfected line MutT. We show by MethylFlash that the increase in host DNA methylation induced by Wolbachia infection is abolished in the knockout Mut background (Fig. S4) and is thus Dnmt2-dependent. However, loss of this crucial enzyme in the DNA methylation pathway has no effect on the penetrance of CI (Fig. 5), as shown in comparisons between mutant and wild type males mated to uninfected females (P = 0.13, MWU). The low level of DNA methylation still present in mutants has recently been observed by others (Boffelli, Takayama & Martin, 2014) and suggests a possible mechanism of DNA methylation in Drosophila that is independent of canonical DNA methyltransferases. Thus, it is possible that CI could be induced by alterations in genomic methylation but in a Dnmt2-independent manner.

Curiously, despite the previously observed role for Dnmt2 in host immunity (Zhang et al., 2013; Durdevic et al., 2013), the mutants observed here exhibit no increase in Wolbachia titers within any of the tissues tested (Fig. S5). It is interesting to note that Dnmt2 mutant Drosophila, derived from the W¹¹¹⁸ background line, have titers that are, on average, half of those seen in y¹w^* background lines (see Fig. 4). This difference has been observed several times in our experiments and suggests either a differing ability of the host lines to control Wolbachia titers or an as yet unclassified difference in the wMel strains infecting these flies.

Host levels of DNA methylation do not correlate with strength of CI

To substantiate the claim that DNA methylation is not involved in the induction of CI for other Wolbachia strains and/or host species, we tested the DNA methylation status of testes DNA from Drosophila species infected with various strains of Wolbachia. These taxa include D. simulans infected with strains wRi, wNo, and wAu, which express strong, moderate, and no CI, respectively. We also tested a different D. melanogaster-infecting strain wMel derived from the W¹¹¹⁸ background strain instead of y¹w^*. As previously mentioned, while the W¹¹¹⁸ line induces strong CI, Wolbachia titers are much lower in these animals compared to the infection found in y¹w^*.

Results show that methylation status of the infected host testes is random with regards to the strength of CI (Fig. 6). While infection with the high CI-inducer wRi exhibits higher methylation in infected testes as compared to uninfecteds, this effect is marginally insignificant (P = 0.072, MWU) and is countered by data from the wAu strain, which causes no CI but still significantly increases host DNA methylation in testes (P = 0.0047, MWU). Furthermore, infection with the wNo strain of Wolbachia, which causes moderate CI, actually has less methylation in host testes. Finally, a low-titer infection of wMel (W¹¹¹⁸), while still inducing CI, does not induce the same level of DNA methylation associated with a high-density infection (y¹w^*).