ABSTRACT
Maternally transmitted Wolbachia bacteria infect about half of all insect species. Many Wolbachia cause cytoplasmic incompatibility (CI), reduced egg hatch when uninfected females mate with infected males. CI produces a frequency-dependent fitness advantage that leads to high equilibrium Wolbachia frequencies but does not aid Wolbachia spread from low frequencies. The fitness advantages that produce initial Wolbachia spread and maintain non-CI Wolbachia remain elusive. Wolbachia variant wMau that infects Drosophila mauritiana does not cause CI, despite being very closely related to the CI-causing Wolbachia wNo in D. simulans (0.068% sequence divergence over 682,494 bp), suggesting that CI was recently lost. Using draft Wolbachia genomes, we identify a deletion in a WO phage gene that may disrupt CI. We find no evidence that wMau increases host fecundity. We report intermediate and apparently stable wMau infection frequencies on the Indian Ocean island of Mauritius, consistent with no CI and no appreciable positive effect on fecundity. Our data indicate that wMau frequencies reflect a balance between unknown, weak positive effects on fitness and imperfect maternal transmission. Phylogenomic analyses suggest that Supergroup B Wolbachia, including wMau, diverged from Supergroup A Wolbachia, including wMel from D. melanogaster, 6–46 million years ago, later than previously estimated.
INTRODUCTION
Maternally transmitted Wolbachia infect about half of all insect species (Werren and Windsor 2000; Zug and Hammerstein 2012; Weinert et al. 2015), as well as other arthropods (Jeyaprakash and Hoy 2000; Hilgenboecker et al. 2008) and nematodes (Taylor et al. 2013). Host species may acquire Wolbachia from common ancestors, from sister species via hybridization and introgression, or horizontally (O'Neill et al. 1992; Rousset and Solignac 1995; Huigens et al. 2004; Baldo et al. 2008; Raychoudhury et al. 2009; Gerth and Bleidorn 2016; Schuler et al. 2016; Turelli et al. 2018). Many Wolbachia cause cytoplasmic incompatibility (CI), reduced egg hatch when uninfected females mate with Wolbachia-infected males. Three parameters usefully approximate the frequency dynamics and equilibria of CI-causing Wolbachia: the relative hatch rate of uninfected eggs fertilized by infected males (H), the fitness of infected females relative to uninfected females (F), and the proportion of uninfected ova produced by infected females (u) (Caspari and Watson 1959; Hoffmann et al. 1990). To spread deterministically from low frequencies, Wolbachia must produce F(1 − μ) > 1, irrespective of CI. Once they become sufficiently common, Cl-causing infections, such as wRi-like Wolbachia in Drosophila simulans and several other Drosophila (Turelli et al. 2018), spread to high equilibrium frequencies, dominated by a balance between CI and imperfect maternal transmission (Turelli and Hoffmann 1995; Kreisner et al. 2016). In contrast, non-CI-causing Wolbachia, such as wAu in D. simulans, typically persist at lower equilibrium frequencies, maintained by a balance between positive Wolbachia effects on host fitness and imperfect maternal transmission (Hoffmann and Turelli 1997; Kreisner et al. 2013). When H < F(1 − μ) < 1, “bistable” dynamics result, producing stable equilibria at 0 and at a higher frequency denoted ps, where 0.50 < ps ≤ 1 (Turelli and Hoffmann 1995). Bistability explains the pattern and rate of spread of wMel transinfected into Aedes aegypti to suppress the spread of dengue, Zika and other human diseases (Hoffmann et al. 2011; Barton and Turelli 2011; Turelli and Barton 2017; Schmidt et al. 2017).
In contrast to the bistability observed with wMel transinfections, natural Wolbachia infections seem to spread via “Fisherian” dynamics with F(1 − μ) > 1 (Fisher 1937; Kriesner et al. 2013; Hamm et al. 2014). Several Wolbachia effects could generate F(1 − μ) > 1, but we do not yet know which ones actually do. For example, wRi has evolved to increase D. simulans fecundity in only a few decades (Weeks et al. 2007), wMel seems to enhance D. melanogaster fitness in high and low iron environments (Brownlie et al. 2009), and several Wolbachia including wMel protect their Drosophila hosts from RNA viruses (Hedges et al. 2008; Teixeira et al. 2008; Martinez et al. 2014). However, it remains unknown which if any these potential fitness benefits underlie Wolbachia spread in nature. For instance, wMel seems to have little effect on viral abundance in wild-caught D. melanogaster (Shi et al. 2018).
D. mauritiana, D. simulans and D. sechellia comprise the D. simulans clade within the nine-species D. melanogaster subgroup of Drosophila. The D. simulans clade diverged from D. melanogaster approximately three million years ago (mya), with the island endemics D. sechellia (Seychelles archipelago) and D. mauritiana (Mauritius) thought to originate in only the last few hundred thousand years (Lachaise et al. 1986; Ballard 2000a; Dean and Ballard 2004; McDermott and Kliman 2008; Garrigan et al. 2012; Brand et al. 2013; Garrigan et al. 2014). D. simulans is widely distributed around the globe, but has never been collected on Mauritius (David et al. 1989; Legrand et al. 2011). However, evidence of mitochondrial and nuclear introgression supports interisland migration and hybridization between these species (Ballard 2000a; Nunes et al. 2010; Garrigan et al. 2012), which could allow introgressive Wolbachia transfer (Rousset and Solignac 1995).
D. mauritiana is infected with Wolbachia denoted wMau, likely acquired via introgression from other D. simulans-clade hosts (Rousset and Solignac 1995). Wolbachia variant wMau may also infect D. simulans in Madagascar and elsewhere in Africa and the South Pacific (denoted wMa in D. simulans) (Ballard 2000a; Ballard 2004). wMau does not cause CI in D. mauritiana or when transinfected into D. simulans (Giordano et al. 1995), despite being very closely related to wNo strains that do cause CI in D. simulans (Merçot et al. 1995; Rousset and Solignac 1995; James and Ballard 2000). (Also, D. simulans seems to be a “permissive” host for CI, as evidenced by the fact that wMel, which causes little CI in its native host, D. melanogaster, causes intense CI in D. simulans [Poinsot et al. 1998].) In contrast to the lack of CI, Fast et al. (2011) reported that a wMau variant increased D. mauritiana fecundity four-fold. This fecundity effect occurred in concert with wMau-induced alternations of programmed cell death in the germarium and of germline stem cell mitosis, providing insight into the mechanisms underlying increased egg production (Fast et al. 2011). However, the generality of these effects across wMau variants, and whether such extreme effects are plausible mechanisms of wMau spread, remains unknown.
Here, we assess the genetic and phenotypic basis of wMau frequencies in D. mauritiana on Mauritius by combining analysis of wMau draft genomes with analysis of wMau effects on host fecundity and egg hatch. We identify a single mutation that we predict disrupts CI in wMau, relative to its apparent sister wNo. The loss of CI in the wMau lineage is consistent with theory demonstrating that selection within host species does not act to increase or maintain the level of CI (Prout 1994; Turelli 1994; Haygood and Turelli 2009), but instead acts to increase F(1 - u), the product of Wolbachia effects on host fitness and maternal transmission efficiency (Turelli 1994). The loss of CI helps explain the intermediate wMau frequencies on Mauritius, reported by us and Giordano et al. (1995). We find no wMau effects on host fecundity, suggesting that the results of Fast et al. (2011) may be anomalous. Indeed, theoretical analyses show that even a two-fold fecundity increase cannot be reconciled with the observed intermediate population frequencies, unless maternal wMau transmission is exceptionally unreliable. Finally, we report the unexpected maintenance of two distinct classes of mtDNA haplotypes among Wolbachia- uninfected D. mauritiana and present theoretical analyses illustrating why this is unexpected.
MATERIALS AND METHODS
Drosophila Husbandry and Stocks
The D. mauritiana isofemale lines used in this study (N =32) were sampled from Mauritius in 2006 by Margarita Womack and kindly provided to us by Prof. Daniel Matute at the University of North Carolina in Chapel Hill. Stocks were maintained on modified version of the standard Bloomington-cornmeal medium (Bloomington Stock Center, Bloomington, IN) and were kept at 25°C, 12 light: 12 dark photoperiod prior to the start of our experiments.
Determining Wolbachia infection status and comparing infection frequencies
DNA was extracted from each isofemale line using a standard ‘squish’ buffer protocol (Gloor et al. 1993), and infection status was determined using a polymerase chain reaction (PCR) assay (Simpliamp ThermoCycler, Applied Biosystems, Singapore). We used the GoTaq™ DNA Polymerase (Promega™, Wisconsin, USA) master mix, and PCR began with 3 minutes at 94°C, followed by 34 rounds of 30 seconds at 94°C, 30 seconds at 55°C, and 1 minute and 15 seconds at 72°C. The profile finished with 8 minutes at 72°C. We amplified the Wolbachia-specific wsp gene and a nuclear control region of the 2L chromosome; primers are listed in Supplementary Table 1. PCR products were visualized using 1% agarose gels that included a molecular-weight ladder. Assuming a binomial distribution, we estimated exact 95% binomial confidence intervals for the infection frequencies on Mauritius. Using Fisher’s Exact Test, we tested for temporal variation in wMau frequencies by comparing our frequency estimate to a previous estimate (Giordano et al. 1995). All analyses were performed using R version 3.5.1 (R Team 2015).
We used quantitative PCR (qPCR) (MX3000P, Agilent Technologies, Germany) to confirm that tetracycline-treated flies were cleared of wMau. DNA was extracted from D. mauritiana flies after four generations of tetracycline treatment, as described below. Our qPCR used a PowerUp™ SYBR™ Green Master Mix (Applied Biosystems™, California, USA). qPCR began with 2 minutes at 50°C and 10 minutes at 95°C, followed by 40 cycles of 10 seconds at 95°C and 30 seconds at 60°C. We finished with 15 seconds at 95°C, 15 seconds at 55°C, and 15 seconds at 95°C. Primers are listed in Supplementary Table 1.
Wolbachia DNA extraction, library preparation, and sequencing
We sequenced wMau-infected R9, R29, and R60 D. mauritiana genotypes. Tissue samples for genomic DNA were extracted using a modified CTAB Genomic DNA Extraction protocol. Pools of flies (N =5), including males and females, were homogenized in 200 ml CTAB solution (100mM Tris HCl pH 8.0, 10nM EDTA, 1.4M NaCl, CTAB 2% (w/v)) using motorized pestle. We incubated each sample for 10 min at 65°C, after which we added 200 μl of chloroform; samples were then centrifuged at 13,000 rpm for 5 min at 4°C. We transferred the aqueous phase into a new tube, added 1:1 volume isopropanol, incubated samples on ice for 1 hr, and centrifuged them at 13,000 rpm for 30 min at 4°C. We discarded the supernatant, added 500 μl 70% EtOH, and then inverted the samples 2–3 times to re-suspend the pellets. Finally, we centrifuged the samples at 13,000 rpm for 15 min at 4°C, incubated at room temperature for 15 min, and centrifuged again for 5 min. We discarded the ethanol and repeated these final steps. The resulting pellet was air dried at room temperature for 5 min and dissolved in 40ul 1 × TE Buffer.
DNA quantity was tested on an Implen Nanodrop (Implen, München, Germany) and total DNA was quantified by Qubit Fluorometric Quantitation (Invitrogen, Carlsbad, California, USA). DNA was cleaned using Agencourt AMPure XP beads (Beckman Coulter, Inc., Brea, CA, U.S.A), following manufacturers’ instructions, and eluted in 50 μl 1× TE Buffer for shearing. DNA was sheared using a Covaris E220 Focused Ultrasonicator (Covaris Inc., Woburn, MA) to a target size of 400 bp.
We prepared libraries using NEBNext® Ultra™ II DNA Library Prep with Sample Purification Beads (New England BioLabs, Ipswich, Massachusetts). End Preparation, Adaptor Ligation, Bead Clean-up with size selection, PCR Enrichment of Adaptor-Ligated DNA, and Clean-up of PCR Reaction were completed following the manufacturer’s protocols. Final fragment sizes and concentrations were confirmed using a TapeStation 2200 system (Agilent, Santa Clara, California). We indexed samples using NEBNext® Multiplex Oligos for Illumina® (Index Primers Set 3 & Index Primers Set 4), and 10 μl of each sample was shipped to Novogene (Sacramento, CA) for sequencing using Illumina HiSeq 4000 (San Diego, CA), generating paired-end, 150 bp reads.
Wolbachia assembly
We obtained published reads (N =6) from Garrigan et al. (2014), and assembled these genomes along with the R9, R29, and R60 genomes that we sequenced. Reads were trimmed using Sickle v. 1.33 (Joshi and Fass 2011) and assembled using ABySS v. 2.0.2 (Jackman et al. 2017). K values of 41, 51, 61, and 71 were used, and scaffolds with the best nucleotide BLAST matches to known Wolbachia sequences with E-values less than 10−10 were extracted as the draft Wolbachia assemblies. We deemed samples infected if the largest Wolbachia assembly was at least 1 million bases and uninfected if the largest assembly was fewer than 100,000 bases. No samples produced Wolbachia assemblies between 100,000 and 1 million bases. Of the six sets of published reads we analyzed (Garrigan et al. 2014), only lines R31 and R41 were wMau-infected.
To assess the quality of our draft assemblies, we used BUSCO v. 3.0.0 to search for orthologs of the near-universal, single-copy genes in the BUSCO proteobacteria database (Simao et al. 2015). As a control, we performed the same search using the reference genomes for wRi (Klasson et al. 2009), wAu (Sutton et al. 2014), wMel (Wu et al. 2004), wHa (Ellegaard et al. 2013), and wNo (Ellegaard et al. 2013).
Wolbachia gene extraction and phylogenetics
To determine phylogenetic relationships and estimate divergence times, we obtained the public Wolbachia Supergroup B genomes of: wAlbB that infects Aedes albopictus (Mavingui et al. 2012), wPip_Pel that infects Culex pipiens (Klasson et al. 2008), wPip_Mol that infects Culex molestus (Pinto et al. 2013), wNo that infects Drosophila simulans (Ellegaard et al. 2013), and wVitB that infects Nasonia vitripennis (Kent et al. 2011), in addition to outgroup Supergroup A wMel that infects D. melanogaster (Wu et al. 2004). These previously published genomes and the five wMau-infected D. mauritiana genomes were annotated with Prokka v. 1.11, which identifies orthologs to known bacterial genes (Seemann 2014). To avoid pseudogenes and paralogs, we used only genes present in a single copy, and with no alignment gaps, in all of the genome sequences. Genes were identified as single copy if they uniquely matched a bacterial reference gene identified by Prokka v. 1.11. By requiring all orthologs to have identical length in all of the draft Wolbachia genomes, we removed all loci with indels. 143 genes, a total of 113,943 bp, met these criteria when comparing all of these genomes. However, when our analysis was restricted to the five wMau genomes, our criteria were met by 694 genes, totaling 704,613 bp. Including wNo with the five wMau genomes reduced our set to 671 genes with 682,494 bp. We calculated the percent differences for the three codon positions within wMau and between wMau and wNo.
We estimated a Bayesian phylogram of the Wolbachia sequences with RevBayes 1.0.8 under the GTR + Γ model, partitioning by codon position (Höhna et al. 2016). Four independent runs were performed, which all agreed.
We estimated a chronogram from the Wolbachia sequences using the absolute chronogram procedure implemented in Turelli et al. (2018). Briefly, we generated a relative relaxed clock chronogram with the GTR + Γ model with the root age fixed to 1 and the data partitioned by codon position. The relaxed clock branch rate prior was Γ(2,2). We used substitution-rate estimates of Γ(7,7) × 6.87 × 10−9 substitutions/3rd position site/year to transform the relative chronogram into an absolute chronogram. This rate estimate was chosen so that the upper and lower credible intervals matched the posterior distribution estimated by Richardson et al. (2012), assuming 10 generations/year, normalized by their median estimate of 6.87 × 10−9 substitutions/3rd position site/year. (To assess the robustness of our conclusions to model assumptions, we also performed a strict-clock analysis and a relaxed-clock analysis with branch rate prior Γ(7,7).) For each analysis, four independent runs were performed, which all agreed. Our analyses all support wNo as sister to wMau.
We also estimated a relative chronogram for the host species using the procedure implemented in Turelli et al. (2018). Our host phylogeny was based on the same 20 nuclear genes used in Turelli et al. (2018): aconitase, aldolase, bicoid, ebony, enolase, esc, g6pdh, glyp, glys, ninaE, pepck, pgi, pgm, pic, ptc, tpi, transaldolase, white, wingless and yellow.
Analysis of Wolbachia and mitochondrial genomes
We looked for copy number variation (CNV) between wMau and its closest relative, wNo across the whole wNo genome. Reads from the five infected wMau lines were aligned to the wNo reference (Ellegaard et al. 2013) with bwa 0.7.12 (Li and Durbin 2009). We calculated the normalized read depth for each alignment over sliding 1,000-bp windows by dividing the average depth in the window by the average depth over the entire wNo genome. The results were plotted and visually inspected for putative copy number variants (CNVs). The locations of CNVs were specifically identified with ControlFREEC v. 11.5 (Boeva et al. 2012), using a ploidy of one and a window size of 1,000. We calculated P-values for each identified CNV with the Wilcoxon Rank Sum and the Kolmogorov-Smirnov tests implemented in ControlFREEC.
We used BLAST to search for pairs of WO phage loci in wMau and wNo genomes that are known to cause and rescue CI. These include the orthologous loci pairs WD0631-632 and wPa_0282-283 that cause CI in wMel and in wPip, respectively (Beckmann and Fallon 2013; Beckmann et al. 2017; LePage et al. 2017). We also searched for the related wPip wPa_0294-295 pair (Beckmann et al. 2017). WD0631-632 and wPa_0282-283 orthologs have been referred to as Type I cifA-B pairs, and wPa_0294-295 as a Type IV cifA-B pair (LePage et al. 2017; Lindsey et al. 2018). Beckmann et al. (2017) refer to wPa_0282-283 as cidA-B and wPa_0294-295 as cinA-B based on their predicted function. wNo_RS01055 and wNo_RS01050 have been identified as a Type III cifA-B pair in the wNo genome (LePage et al. 2017; Lindsey et al. 2018). No_RS01055 and wNo_RS01050 are highly diverged from WD0631-632/wPa_0282-283 orthologs and from wPa_0294-295; however, this wNo pair is more similar to wPa_0294-295 in terms of protein domains, lacking a ubiquitin-like protease domain (Lindsey et al. 2018).
We found only the wNo_RS01050/wNo_RS01055 pair in wMau genomes. The orthologs for wNo_RS01050 and wNo_RS01055 were extracted from our draft wMau assemblies and aligned with MAFFT v. 7 (Katoh and Standley 2013). We compared these orthologs to wNo_RS01050/wNo_RS01055 in wNo and checked for single nucleotide variants (SNVs) among our wMau assemblies.
D. mauritiana carry either the maI mitochondrial haplotype, associated with wMau infections, or the maII haplotype (Ballard 2000a; James and Ballard 2000). To determine the mitochondrial haplotype of each D. mauritiana line, we assembled the mitochondrial genomes by down-sampling the reads by a factor of 100, then assembling with ABySS 2.0.2 using a K value of 71 for our data (150 bp reads) and 35 for the published data (76 bp reads) (Garrigan et al. 2014). Down-sampling reads prevents the nuclear genome from assembling but does not inhibit assembly of the mitochondrial genome, which has much higher coverage. We deemed the mitochondrial assembly complete if all protein-coding genes (N =13) were present on the same contig and in the same order as in D. melanogaster. If the first attempt did not produce a complete mitochondrial assembly, we adjusted the down-sampling fraction until a complete assembly was produced for each line.
Annotated reference mitochondrial sequences for the D. mauritiana mitochondrial haplotypes maI and maII were obtained from Ballard et al. (2000b), and the thirteen proteincoding genes were extracted from our assemblies using BLAST and aligned to these references. The maI and maII reference sequences differ at 343 nucleotides over these protein-coding regions. We identified our lines as carrying the maI haplotype if they differed by fewer than five nucleotides from the maI reference and as maII if they differed by fewer than five nucleotides from the maII reference. None of our assemblies differed from both references at five or more nucleotides.
wMau phenotypic analyses
Previous analyses have demonstrated that wMau does not cause Cl (Giordano et al. 1995). To check the generality of this result, we reciprocally crossed wMau-infected R31 D. mauritiana with uninfected R4 and measured egg hatch. Flies were reared under controlled conditions at 25°C for multiple generations leading up to the experiment. We paired 1-2-day-old virgin females with 1-2-day-old males in a vial containing spoons with cornmeal media and yeast paste. After 24 hr, pairs were transferred to new spoons, and this process was repeated for five days. Eggs on each spoon were given 24 hr at 25°C to hatch after flies were removed. To test for Cl, we used nonparametric Wilcoxon tests to compare egg hatch between reciprocal crosses that produced at least 10 eggs. All experiments were carried out at 25°C with a 12 light: 12 dark photoperiod.
To determine if wMau generally enhances D. mauritiana fecundity, we assayed the fecundity of wMau-infected and uninfected D. mauritiana genotypes generated from reciprocal crosses. Two wMau-infected isofemale lines (R31 and R41) were each reciprocally crossed to an uninfected line (R4) to generate infected and uninfected F1 females with similar genetic backgrounds. The wMau-infected and uninfected F1 females were collected as virgins and placed in holding vials. We paired 3-7-day-old females individually with an uninfected-R4 male (to stimulate oviposition) in vials containing a small spoon filled with standard cornmeal medium and coated with a thin layer of yeast paste. We allowed females to lay eggs for 24 hours, after which pairs were transferred to new vials. This was repeated for five days. At the end of each 24 hr period, spoons were frozen until counted. All experiments were carried out at 25°C with a 12 light: 12 dark photoperiod.
We also assessed whether D. mauritiana genomic backgrounds modify wMau fecundity effects. We backcrossed R31 wMau into the R41 D. mauritiana nuclear background by crossing R31 females with R41 males. F1 females were then backcrossed to R41 males—this was repeated for four generations to generate the R41R31 genotypes (wMau variant denoted by superscripts). A similar approach was taken to generate R31R41 genotypes. Each of these genotypes were then reciprocally crossed to the uninfected R4 line to generate infected and uninfected F1 females with controlled nuclear-genetic backgrounds. Upon emergence the wMau-infected and uninfected F1 females were collected as virgins and placed in holding vials. 3-7-day-old females were then paired individually with an uninfected-R4 male to stimulate oviposition, and egg-lay experiments were carried out over five days as described above.
To determine if wMau fecundity effects depend on host age, we measured egg lay of wMau-infected and uninfected R31 genotypes over 24 days. Fast et al. (2011) carried out their experiments over several weeks, which encouraged us to do the same. In an effort to more closely mimic their approach, we generated tetracycline-treated R31 genotypes (R31-tet) to compare to infected R31 genotypes. We fed flies 0.03% tetracycline concentrated medium for four generations. We screened F1 and F2 individuals for wMau using PCR, and we then fed flies tetracycline food for two additional generations. In the fourth generation, we assessed wMau titer using qPCR to confirm that each genotype was cleared of wMau infection. Following tetracycline treatment, we placed R31D. mauritiana males in food vials and allowed them to eat and defecate for 48 hours. We removed males, and R31-tet individuals were then placed in these same vials to reestablish the gut microbiome. Flies were given at least three more generations to avoid detrimental effects of tetracycline treatment on mitochondrial function (Ballard and Melvin 2007). To assay age effects on fecundity, we paired individual 6-7-day-old virgin R31 (N =30) and R31-tet (N =30) females in bottles on yeasted apple-juice agar plates with an R4 male to stimulate oviposition. Pairs were placed on new egg-lay plates every 24 hrs. After two weeks, we added one or two additional R4 males to each bottle to replace any dead males and to ensure that females were not sperm limited as they aged.
To test for effects of wMau on host fecundity, we used nonparametric Wilcoxon tests to compare the number of eggs laid between infected and uninfected females produced by reciprocally crossing R31, R41, R3 R41, and R41R31 with uninfected R4. We also used Wilcoxon tests to assess variation in fecundity between infected-R31 and uninfected-R31-tet genotypes across all 24 days of egg lay in our final experiment. We then assessed variation in R31 and R31- tet egg lay across several ages to identify fecundity effects that vary with female age. Finally, we estimated the fitness parameter F in the standard discrete-generation model of CI to determine if wMau fecundity effects underlie low frequency spread (Hoffmann et al. 1990; Turelli 1994). Pairs that laid fewer than 10 eggs across each experiment were excluded from analyses, but our results are robust to this threshold.
RESULTS
Wolbachia infection status
Out of 32 D. mauritiana lines that we analyzed, 11 were infected with wMau Wolbachia (infection frequency =0.34; binomial confidence interval: 0.19, 0.53). This new frequency estimate is not statistically different from a previous estimate (Giordano et al. 1995: infection frequency, 0.46; binomial confidence interval, (0.34, 0.58); Fisher’s Exact Test, P =0.293). These relatively low infection frequencies are consistent with theoretical expectations given that wMau does not cause CI (Giordano et al. 1995; our data reported below), as Wolbachia that cause significant CI produce high equilibrium infection frequencies (Kriesner et al. 2016). Intermediate and apparently stable wMau frequencies on Mauritius suggest that wMau persists at a balance between positive effects on host fitness and imperfect maternal transmission.
Quantitative predictions, based on the idealized model of Hoffmann and Turelli (1997), are discussed below. The maintenance of wMau is analogous to the persistence of other non-CI-causing Wolbachia, specifically wSuz in D. suzukii and wSpc in D. subpulchrella (Hamm et al. 2014; Conner et al. 2017; Turelli et al. 2018) and wAu in some Australian populations of D. simulans (Hoffmann et al. 1996; Kriesner et al. 2013).
Draft wMau genome assemblies and comparison to wNo
The five wMau draft genomes we assembled were of very similar quality (Supplemental Table 2). N50 values ranged from 60,027 to 63,676 base pairs, and our assemblies varied in total length from 1,266,004 bases to 1,303,156 bases (Supplemental Table 2). Our BUSCO search found exactly the same genes in each draft assembly, and the presence/absence of genes in our wMau assemblies was comparable to those in the complete genomes used as controls (Supplemental Table 3). In comparing our five wMau draft genomes over 694 single-copy, equal-length loci comprising 704,613 bp, we found only one SNP. Four sequences (R9, R31, R41 and R60) are identical at all 704,613 bp. R29 differs from them at a single nucleotide, a nonsynonymous substitution in a locus which Prokka v. 1.11 annotates as “bifunctional DNA-directed RNA polymerase subunit beta/beta.”
Comparing these five wMau sequences to the wNo reference (Ellegaard et al. 2013) over 671 genes with 682,494 bp, they differ by 0.068% overall, with equivalent divergence at all three codon positions (0.067%, 0.061%, and 0.076%, respectively).
Wolbachia phylogenetics
As expected from the sequence comparisons, our Supergroup B phylogram places wMau sister to wNo (Figure 1A). This is consistent with previous analyses using fewer loci that placed wMau (or wMa in D. simulans) sister to wNo (James and Ballard 2000; Zabalou et al. 2008; Toomey et al. 2013). Our chronogram (Figure 1B) estimates the 95% credible interval for the split between thee Supergroup B versus Subgroup A Wolbachia strains as 6 to 36 mya (point estimate =16 mya). Reducing the variance on the substitution-rate-variation prior by using Γ(7,7) rather than Γ(2,2), changes the credible interval for the A-B split to 8 to 46 mya (point estimate =21 mya); in contrast, a strict clock analysis produces a credible interval of 12 to 64 mya (point estimate =31 mya). These estimates are roughly comparable to an earlier estimate that used a general estimate for the synonymous substitution rate in bacteria (Ochman and Wilson 1987) and only the ftsZ locus (59-67 mya, Werren et al. 1995). However, they are much lower than a recent estimate based on comparative genomics (217 mya, Gerth and Bleidorn 2016). We discuss this discrepancy below.
The observed divergence between wNo and wMau is consistent across all three codon positions, similar to other recent Wolbachia splits (Turelli et al. 2018). Conversely, observed divergence at each codon position generally varies across the chronogram, leading to inflation of the wNo-wMau divergence point estimate of approximately 181,000 years (credible interval =28,000 to 701,000 years; Figure 1B). To obtain a more accurate estimate, we calculated wNo-wMau divergence using the observed third position pairwise divergence (0.077%, or 0.039% from tip to MRCA) and Richardson et al. (2012)’s rate estimate. This approach produces a point estimate of 57,000 years, with a credible interval of 30,000 to 135,000 years for the wNo-wMau split, which seems plausible.
Analysis of Wolbachia and mitochondrial genomes
We looked for CNVs in wMau relative to sister wNo by plotting normalized read depth along the wNo genome. There were no differences in copy number among the wMau variants, but compared to wNo, ControlFREEC identified four regions deleted from all wMau that were significant according to the Wilcoxon Rank Sum and Kolmogorov-Smirnov tests (Figure 2 and Table 1). These deleted regions of the wMau genomes include many genes, including many phage-related loci, listed in Supplementary Table 4.
To test the hypothesis that loci known to underlie CI are disrupted, we searched for pairs of loci known to be associated with CI and found orthologs to the Type III gene pair wNo_RS01050-1055 in each of our draft assemblies, but we did not find orthologs to the WD0631-632/wPip_0282-283 pairs or to the wPip_0294-295 pair. There were no variable sites in the wNo_RS01050-1055 gene regions among our five wMau assemblies. All wMau variants share a one base pair deletion at base 1133 out of 2091 (amino acid 378) in the wNo_RS01050 gene, relative to this same region in the wNo genome. This frameshift introduces over ten stop codons, with the first at amino acid 388, likely making this predicted CI-causing protein nonfunctional. We also identified a nonsynonymous substitution in amino acid 264 of wNo_RS01050 (wNo codon ACA, Thr; wMau codon AUA, Ile) and two SNVs in wNo_RS01055: a synonymous substitution in amino acid 365 (wNo codon GUC, wMau codon GUU) and a nonsynonymous substitution in amino acid 397 (wNo codon GCU, amino acid Ala; wMau codon GAU, amino acid Asp). While one might intuitively expect selection to favor CI, disruption of CI in wMau is consistent with theoretical analyses showing that selection within a host species does not act directly on the level of CI (Prout 1994; Turelli 1994; Haygood and Turelli 2009). Instead, selection should act to increase the product of Wolbachia effects on host fitness and the efficiency of maternal transmission (Turelli 1994).
Of the D. mauritiana lines tested (N =9), one line (uninfected-R44) carries the maII mitochondrial haplotype, while the other eight carry maI. The maI and maII references differ by 343 SNVs across the proteome, and R44 differs from the maII reference by 4 SNVs in the proteome. Four of our maI lines (R23, R29, R32, and R39) are identical to the maI reference, while three (R31, R41, and R60) have one SNV and one (R9) has two SNVs relative to maI reference. One SNV is shared between R9 and R60, but the other three SNVs are unique. Our results agree with past analyses that found wMau is perfectly associated with the maI mitochondrial haplotype (Ballard 2000a; James and Ballard 2000). The presence of maII among the uninfected is interesting. In contrast to maI, which is associated with introgression with D. simulans (Ballard 2000a; James and Ballard 2000), maII appears as an outgroup on the mtDNA phylogeny of the D. simulans clade and is not associated with Wolbachia (Ballard 2000b, Fig. 5; James and Ballard 2000). Whether or not Wolbachia cause CI, if they are maintained by election-imperfect-transmission balance, we expect all uninfected flies to eventually carry the mtDNA associated with infected mothers (Turelli et al. 1992). We present a mathematical analysis of the persistence of maII below.
Analysis of wMau phenotypes
wMau is reported to not cause CI (Giordano et al. 1995), but a single wMau-infected D. mauritiana genotype displayed a four-fold increase in egg lay relative to an uninfected counterpart (Fast et al. 2011). We sought to determine whether wMau can cause CI and whether wMau generally increases D. mauritiana fecundity. In agreement with past analyses, we found no difference between the egg hatch of uninfected females crossed to wMau-infected males (0.34 ± 0.23 SD, N =25) and the reciprocal cross (0.29 ± 0.28 SD, N =24). Contrary to previous results, we find no evidence for wMau effects on D. mauritiana fecundity (Table 2 and Figure 3), regardless of host genetic backgrounds. Across both experiments assessing wMau fecundity effects in their natural backgrounds (R31 and R41), we counted 27,221 eggs and found no difference in the number of eggs laid by infected (mean =238.20, SD =52.35, N =60) versus uninfected (mean =226.82, SD =67.21, N =57) females over the five days of egg lay (Wilcoxon test, W =1540.5, P =0.357). Across both experiments that assessed wMau fecundity effects in novel host backgrounds (R31R41 and R41R31), we counted 30,358 eggs and found no difference in the number of eggs laid by infected (mean =253.30, SD =51.99, N =60) versus uninfected (mean =252.67, SD =63.53, N =60) females over five days (Wilcoxon test, W =1869.5, P =0.719). [The mean number of eggs laid over 5 days, standard deviation (SD), sample size (N), and P-values from Wilcoxon tests are presented in Table 2 for all pairs.]
We sought to determine if wMau fecundity effects depend on host age with a separate experiment that assessed egg lay over 24 days on apple-agar plates, similar to Fast et al. (2011). Across all ages, we counted 9,459 eggs and found no difference in the number of eggs laid by infected (mean =187.70, SD =168.28, N =27) versus uninfected (mean =156.29, SD =138.04, N =28) females (Wilcoxon test, W =409, P =0.608). This pattern was consistent across different age classes, suggesting age has little influence on wMau fecundity effects (Table 3 and Figure 4). (The mean number of eggs laid, SD, N, and P-values from Wilcoxon tests are presented in Table 3 for all age classes.)
With these data, we estimated the fitness parameter F in the standard discrete-generation model of CI (Hoffmann et al. 1990; Turelli 1994). Taking the ratio of replicate mean fecundity observed for wMau-infected females to the replicate mean fecundity of uninfected females in naturally sampled R31 and R41 D. mauritiana backgrounds, we estimated F =1.05 (95% BCa interval: 0.96, 1.16). Following reciprocal introgression of wMau and host backgrounds (i.e., the R31R41 and R41R31 genotypes), we estimated F =1.0 (95% BCa interval: 0.93, 1.09). Finally, across all 24 days of our age-effects experiment, we estimated F =0.83, 95% BCa interval: 0.52, 1.34) for R31, which overlaps with our estimate of F for R31 in our initial experiment (Table 4). Thus, and in support of our other analyses, there is little evidence that wMau increases F to encourage spread. BCa confidence intervals were calculated using the two-sample acceleration constant given by equation 15.36 of Efron and Tibshirani (1993). (Estimates of F and the associated BCa confidence intervals are reported in Table 4 for different strains and age classes.) Our mathematical analysis below show that large fitness advantages associated with wMau are inconsistent with the lack of CI and plausible levels of imperfect transmission.
Mathematical analyses of Wolbachia frequencies and mtDNA polymorphism
If Wolbachia do not cause CI (or any other reproductive manipulation), their dynamics can be approximated by a discrete-generation haploid selection model. Following Hoffmann and Turelli (1997), we assume that the relative fecundity of Wolbachia-infected females is F, but a fraction μ of their ova do not carry Wolbachia. Given our ignorance of the nature of Wolbachia’s favorable effects, the F represents an approximation for all possible fitness benefits. If F(1 − μ) > 1, the equilibrium Wolbachia frequency among adults is
Imperfect maternal transmisson has been documented for field-collected D. simulans infected with wRi (Hoffmann and Turelli 1988; Turelli and Hoffmann 1995; Carrington et al. 2011), D. melanogaster infected with wMel (Hoffmann et al. 1998) and D. suzukii infected with wSuz (Hamm et al. 2014). The estimates range from about 0.01 to 0.1. In order for the equilibrium Wolbachia frequency to be below 0.5, approximation (1) requires that the relative fecundity of infected females satisfies
Thus, even for μ as large a 0.15, which exceeds essentially all estimates of maternal transmission failure from nature, Wolbachia can increase fitness by at most 43% and produce an equilibrium frequency below 0.5. Conversely, (1) implies that a doubling of relative fecundity by Wolbachia would produce an equilibrium frequency 1 − 2μ If μ ≤ 0.25, consistent with all available data, the predicted equilibrium implied by a fitness doubling significantly exceeds the observed frequency of wMau. Hence, a four-fold fecundity effect, as described by Fast et al. (2011), is inconsistent with the frequency of wMau in natural populations of D. mauritiana. Field estimates of μ for D. mauritiana will provide better theory-based bounds on wMau fitness effects.
As noted by Turelli et al. (1992), if Wolbachia is introduced into a population along with a diagnostic mtDNA haplotype that has no effect on fitness, imperfect Wolbachia maternal transmission implies that all infected and uninfected individuals will eventually carry the Wolbachia-associated mtDNA, because all will have had Wolbachia-infected maternal ancestors. We conjectured that a stable mtDNA polymorphism might be maintained if a Wolbachia- associated mtDNA introduced by introgression is deleterious in its new nuclear background. We refute our conjecture in Appendix 1. We show that the condition for Wolbachia to increase when rare, F(1 − μ) > 1, ensures that the native mtDNA will be completely displaced by the Wolbachia-associated mtDNA, even if it lowers host fitness once separated from Wolbachia.
How fast is the mtDNA turnover, among Wolbachia-uninfected individuals, as a new Wolbachia invades? This is easiest to analyze when the mtDNA introduced with Wolbachia has no effect on fitness, so that the relative fitness of Wolbachia-infected versus uninfected individuals is F, irrespective of the mtDNA haplotype of the uninfected individuals. As shown in Appendix 1, the frequency of the ancestral mtDNA haplotype among uninfected individuals, denoted rt, declines as
Assuming r0 =1, recursion (3) implies that even if F(1 − μ) is only 1.01, the frequency of the ancestral mtDNA haplotype should fall below 10−4 after 1000 generations. A much more rapid mtDNA turnover was seen as the CI-causing wRi swept northward through California populations of D. simulans (Turelli et al. 1992; Turelli and Hoffmann 1995). Thus, it is anomolous that mtDNA haplotype maII, which seems to be ancestral in D. mauritiana (Rousset and Solignac 1995; Ballard 2000a), persists among Wolbachia-uninfected D. mauritiana, given that all sampled Wolbachia-infected individuals carry maI.
DISCUSSION
Understanding the genetic and phenotypic basis of Wolbachia frequencies in nature remains challenging and is particularly interesting for strains like wMau that persist without inducing CI. Our results demonstrate intermediate and apparently stable wMau frequencies on Mauritius, consistent with no CI. The lack of wMau CI probably depends at least in part on the one-base-pair deletion that we identified in the wNo_RS01050 gene. This mutation introduces a frameshift and more than ten stop codons, relative to sister wNo, that we predict disrupt CI in wMau. We also find no wMau fecundity effects, indicating that wMau must persist with unknown positive fitness effects balancing imperfect maternal transmission. We discuss the implications of our findings below.
wMau is sister to wNo and diverged from Supergroup A Wolbachia less than 100 mya
Our phylogenetic analyses place wMau sister to wNo, in agreement with past analyses using fewer data (James and Ballard 2000; Zabalou et al. 2008; Toomey et al. 2013). The relationships we infer agree with those from recently published phylograms (Gerth and Bleidorn 2016; Lindsey et al. 2018) (Figure 1A).
Depending on the prior we use for substitution-rate variation, we estimate that wMau and other Supergroup B Wolbachia diverged from Supergroup A strains about 6-46 mya. This is roughly consistent with a prior estimate using only ftsZ (58 to 67 mya, Werren et al. 1995), but is inconsistent with a recent estimate using 179,763 bases across 252 loci (76-460 mya, Gerth and Bleidorn 2016). There are several reasons why we question the Gerth and Bleidorn (2016) calibration. First, Gerth and Bleidorn (2016)’s chronogram placed wNo sister to all other Supergroup B Wolbachia, in disagreement with their own phylogram (Gerth and Bleidorn 2016, Figure 3). In contrast, our phylogram and that of Lindsey et al. (2018) support wAlbB splitting from all other strains at this node. Second, the Gerth and Bleidorn (2016) calibration estimated the split between wRi that infects D. simulans and wSuz that infects D. suzukii at 900,000 years. This estimate is nearly two orders of magnitude higher than the 11,000 year estimate of Turelli et al. (2018) who found 0.014% third position divergence between wRi and wSuz (i.e., 0.007% along each branch) over 506,307 bases. Raychoudhury et al. (2009) and Richardson et al. (2012) both estimated a rate of about 7 × 10−9 substitutions/3rd position site/year between Wolbachia in Nasonia wasps and within wMel, respectively. An estimate of 900,000 years requires a rate about 100 times slower, 7.8 × 10−11 substitutions/3rd position site/year, which seems implausible. Finally, using data kindly provided by Michael Gerth, additional analyses indicate that the third position rates required for the Wolbachia divergence times estimated by Gerth and Bleidorn (2016) between Nomada flava and N. leucophthalma (1.72 × 10−10), N. flava and N. panzeri (3.78 × 10−10) (their calibration point), and N. flava and N. ferruginata (4.14 × 10−10) are each more than 10 times slower than those estimated by Raychoudhury et al. (2009) and Richardson et al. (2012), which seems unlikely. Our analyses suggest that the A-B group split occurred less than 100 mya.
Disruption of the CI phenotype generates relatively low infection frequencies
Across the 171 genes (139,902 bases) included in our phylogenetic analyses, the wMau genomes were identical and differed from wNo by only 0.072%. Across the coding regions we analyzed, we found few SNVs and no CNVs among wMau variants. Our analyses did identify four large deletions shared by all wMau genomes, relative to wNo. Despite the close relationship between wMau and wNo, wNo causes CI while wMau does not (Giordano et al. 1995; Merçot et al. 1995; Rousset and Solignac 1995, our data). We searched for all pairs of loci known to cause CI and found only the wNo_RS01050-1055 pair in both wMau and wNo genomes. All wMau variants share a one-base-pair deletion in wNo_RS01050, relative to this same region in wNo. This mutation introduces a frameshift and more than ten stop codons that we predict disrupt CI in wMau. Disruption of CI in wMau is consistent with the prediction that selection does not directly act on the intensity of CI (Prout 1994; Turelli 1994), and we predict that evidence of such disruption will be found in many other non-CI causing Wolbachia genomes—as has already been reported for a few other strains (Lindsey et al. 2018). Conversely, we predict that facultative Wolbachia lineages displaying loss of CI will be relatively recent. Importantly, non-CI Wolbachia have lower expected equilibrium infection frequencies than do CI-causing variants (Kriesner et al. 2016). This is consistent with a wMau infection frequency of approximately 0.34 on Mauritius (Giordano et al. 1995; our data).
wMau co-occurs with essentially the same mitochondrial haplotype as wMa that infects D. simulans on Madagascar and elsewhere in Africa and the South Pacific (Rousset and Solignac 1995; Merçot and Poinsot 1998; Ballard 2000a; James and Ballard 2000; James et al. 2002; Ballard 2004), suggesting wMau and wMa may be the same strain infecting different host species following introgressive Wolbachia transfer (see below). The CI phenotypes of wMau and wMa are also more similar to one another than to wNo, with only certain crosses between wMa-infected D. simulans males and uninfected D. simulans females inducing CI (James and Ballard 2000). Polymorphism in the strength of CI induced by wMa could result from host modification of Wolbachia-induced CI (Reynolds and Hoffmann 2002; Cooper et al. 2017), or from Wolbachia titer variation that influences the strength of CI and/or the strength of CI rescue by infected females. Alternatively, the single-base-pair deletion in the wNo_RS01050 gene we observe in wMau, or other mutations that influence CI strength, could be polymorphic in wMa. wMa infection frequencies are intermediate on Madagascar (infection frequency =0.25, binomial confidence intervals: 0.14, 0.40; James and Ballard 2000), consistent with no CI, which suggests to us that crosses finding wMa CI are likely anomalous or rare. Inclusion of D. simulans sampled from the island of Reunion in this estimate provides additional support for the conjecture that wMa is non-CI-causing (infection frequency =0.31, binomial confidence intervals: 0.20, 0.45; James and Ballard 2000). Unfortunately, wMa-infected D. simulans were not available for us to genotype and phenotype.
Our genomic data indicate that wMau may maintain an ability to rescue CI, as the wNo_RS01055 gene is intact with only one nonsynonymous substitution relative to the same region in wNo. wNo_RS01055 is orthologous to WD0631 (cifA) in wMel, recently shown to underlie CI rescue (Shropshire et al. 2018). wMa seems to sometimes rescue CI, but conflicting patterns have been found, and additional experiments are needed to resolve this (Rousset and Solignac 1995; Bourtzis et al. 1998; Merçot and Poinsot 1998; James and Ballard 2000; Merçot and Poinsot 2003; Zabalou et al. 2008). Testing for CI rescue by wMau- and wMa-infected females crossed with males infected with wNo or other CI-causing strains, combined with genomic analysis of CI loci in wMa, will help resolve the relationship between these variants and their CI phenotypes.
wMau does not influence D. mauritiana fecundity
While selection does not directly act on the level of CI (Prout 1994; Turelli 1994; Haygood and Turelli 2009), it does act to increase the product of Wolbachia effects on host fitness and the efficiency of maternal transmission (Turelli 1994). Understanding the Wolbachia effects that lead to spread from low frequencies and the persistence of non-CI causing Wolbachia is crucial to explaining Wolbachia prevalence among insects and other arthropods. The four-fold fecundity effect of wMau reported by Fast et al. (2011) in D. mauritiana is inconsistent with our experiments and with the intermediate infection frequencies observed in nature. We find no wMau effects on host fecundity, regardless of host background or female age. Our results are consistent with an earlier analysis that assessed egg lay of a single genotype that found no wMau fecundity effects (Giordano et al. 1995), and with our mathematical analyses that indicate Wolbachia can increase host fitness by at most 50% for reasonable estimates of /u.
Introgressive Wolbachia transfer likely predominates in the D. simulans clade
Hybridization and introgression in the D. simulans clade may have led to introgressive transfer of Wolbachia among host species (Rousset and Solignac 1995), which has also been observed in other Drosophila (Turelli et al. 2018) and in Nasonia wasps (Raychoudhury et al. 2009). While D. mauritiana is singly infected by wMau, D. simulans is infected by several strains that include CI-causing wHa and wNo that often occur as double infections within individuals (O'Neill and Karr 1990; Merçot et al. 1995; Rousset and Solignac 1995). wHa and wNo are similar to wSh and wSn, respectively, that infect D. sechellia (Giordano et al. 1995; Rousset and Solignac 1995). wHa and wSh also occur as single infection in D. simulans and in D. sechellia, respectively (Rousset and Solignac 1995). In contrast, wNo almost always occurs alongside wHa in doubly infected D. simulans individuals (James et al. 2002), and wSn seems to occur only in individuals also infected by wSh (Rousset and Solignac 1995). D. simulans has three distinct mitochondrial haplotypes (siI, siII, siIII) associated with wAu/wRi (siII), wHa/wNo (siI), and wMa (siIII). The siI haplotype is closely related to the se haplotype found with wSh and wSn in D. sechellia (Ballard 2000b). wMa co-occurs with the s/III haplotype, which differs over its protein-coding genes by a single base pair from the maI mitochondrial haplotype carried by wMau-infected D. mauritiana—a second haplotype (maII) is carried by only uninfected D. mauritiana (Ballard 2000a; James and Ballard 2000).
The lack of whole wMa genome data precludes us from confidently resolving the mode of wMau acquisition in D. mauritiana. However, mitochondrial relationships support the proposal of Ballard (2000b) that D. mauritiana acquired wMau and the maI mitochondrial haplotype via introgression from wMa-infected D. simulans carrying siIII. D mauritiana mitochondria are paraphyletic relative to D. sechellia and D. simulans mitochondria (Solignac and Monnerot 1986; Satta and Takahata 1990; Ballard 2000a, 2000b), with maI sister to siIII and maII outgroup to all other D. simulans-clade haplotypes (see Figure 5). Of the nine genomes we assessed, all but one (uninfected-R44) carry the maI haplotype, and genotypes carrying maI are both wMau-infected (N =5) and uninfected (N =3). While wMa-infected D. simulans carry siIII, wNo-infected D. simulans carry siI. We estimate that wMau and wNo diverged about 55,000 years ago, with only 0.068% sequence divergence over 682,494 bp. Nevertheless, it seems implausible that wNo (versus wMa) was transferred directly to D. mauritiana as this requires horizontal or paternal transmission of wNo into a D. mauritiana background already carrying the maI mitochondrial haplotype. Figure 5 shows relative chronograms for nuclear genes (A) and mtDNA haplotypes (B) for the D. simulans clade with D. melanogaster as the outgroup. Although the nuclear result suggests a confident phylogenetic resolution of the D. simulans clade, this is an artifact of the bifurcation structure imposed by the phylogenetic analysis. Population genetic analyses show a complex history of introgression and probable shared ancestral polymorphisms (Kliman et al. 2000) among these three species. Consistent with this, of the 20 nuclear loci we examined, 6 (aconitase, aldolase, bicoid, ebony, enolase, ninaE) supported D. mauritiana as the outgroup within the D. simulans clade, 7 (glyp, pepck, pgm, pic, ptc, transaldolase, wingless) supported D. sechellia as the outgroup, and 7 (esc, g6pdh, glys, pgi, tpi, white, yellow) supported D. simulans. With successive invasions of the islands and purely allopatric speciation, we expect the outgroup to be the island endemic that diverged first. Figure 5 indicates that the maII haplotype diverged from the other mtDNA haplotypes roughly when the clade diverged, with the other haplotypes subject to a complex history of introgression and Wolbachia-associated sweeps, as described by Ballard (2000b).
Ballard (2000b) estimated that siIII-maI diverged about 4,500 years ago, which presumably approximates the date of the acquisition of wMau (and siIII, which became maI) by D. mauritiana. This is surely many thousands of generations. As shown by our mathematical analyses (Eq. 3), the apparent persistence of the maII mtDNA among Wolbachia-uninfected D. mauritiana—without its occurrence among infected individuals—is unexpected. More extensive sampling of natural D. mauritiana populations is needed to see if this unexpected pattern persists.
While paternal transmission has been observed in D. simulans (Hoffmann and Turelli 1988; Turelli and Hoffmann 1995), it seems to be very rare (Richardson et al. 2012; Turelli et al. 2018). wNo almost always occurs in D. simulans individuals also infected with wHa, complicating this scenario further. It is possible that horizontal or paternal transmission of wMa or wNo between D. simulans backgrounds carrying different mitochondrial haplotypes underlies the similarities of these strains within D. simulans, despite their co-occurrence with distinct mitochondria. Given the diversity of Wolbachia that infect D. simulans-clade hosts, and known patterns of hybridization and introgression among hosts (Garrigan et al. 2012; Brand et al. 2013; Garrigan et al. 2014; Matute and Ayroles 2014; Schrider et al. 2018), determining relationships among these Wolbachia and how D. mauritiana acquired wMau will require detailed phylogenomic analysis of nuclear, mitochondrial, and Wolbachia genomes in the D. simulans clade.
Conclusions
Disruption of CI contributes to intermediate wMau infection frequencies on Mauritius that must be at balance between positive wMau fitness effects and imperfect maternal transmission. The specific fitness effects underlying Wolbachia persistence remain elusive in this system as with most other non-CI Wolbachia infections. Analysis of fecundity effects and other phenotypes like nutritional supplementation and protection from harmful bacteria (Brownlie et al. 2009; Gupta et al. 2017; Kriesner and Hoffmann 2018), across a diversity of Wolbachia-infected hosts, will shed light of whether these phenotypes generally influence the spread and persistence of Wolbachia. Our mathematical analysis suggests that extreme fitness effects like those reported by Fast et al. (2011) are biologically implausible. Accurate estimates of Wolbachia effects on host fitness across host genomic backgrounds and abiotic environments, combined with estimates of Wolbachia maternal transmission in nature, are needed to explain the global Wolbachia pandemic and to improve the efficacy of transinfected Wolbachia as biocontrol agents.
AUTHOR CONTRIBUTIONS
MM performed the molecular and phenotypic work, participated in the design of the study, and contributed to the writing; WC performed the phylogenetic and genomic analyses and contributed to the writing; SR contributed to the molecular and phenotypic analyses and to the writing; JB performed the library preparation and contributed to the writing; MT contributed to the analyses, data interpretation, and writing; BSC designed and coordinated the study, contributed to the analyses and data interpretation, and drafted the manuscript. All authors gave final approval for publication.
DATA ACCESSIBILITY STATEMENT
All phenotypic data will be archived in DRYAD and all genetic data will be archived in GenBank at the time of acceptance.
ACKNOWLEDGMENTS
We thank Margarita Womack for sampling the D. mauritiana used in this study and Daniel Matute for sharing them. We thank Michael Gerth for sharing Nomada genomic data and Tim Wheeler for taking photos of D. mauritiana. Isaac Humble, Maria Kirby, and Tim Wheeler assisted with data collection. Michael Gerth, Michael Hague, and Amelia Lindsey provided comments that improved earlier drafts of this manuscript. Computational resources were provided by the University of Montana Genomics Core. Research reported in this publication was supported by the National Institute Of General Medical Sciences of the National Institutes of Health (NIH) under Award Number R35GM124701 to B.S.C. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Appendix 1. Mathematical analyses of mtDNA and Wolbachia dynamics
Our analysis follows the framework developed in Turelli et al. (1992), but is simplified by the lack of CI. We suppose that introgression introduces a cytoplasm carrying Wolbachia and a novel mtDNA haplotype, denoted B. Before Wolbachia introduction, we assume the population is monomorphic for mtDNA haplotype A. With imperfect maternal Wolbachia transmission, uninfected individuals will be produced with mtDNA haplotype B. Without horizontal or paternal transmission (which are very rare, Turelli et al. 2018), all Wolbachia-infected individuals will carry mtDNA haplotype B. Once Wolbachia is introduced, uninfected individuals can have mtDNA haplotype A or B. We assume that these three cytoplasmic types (“cytotypes”) differ only in fecundity, and denote their respective fecundities FI, FA and FB. Denote the frequencies of the three cytotypes among adults in generation t by pI, t, pA, t and pB, t, with pI, t + pA, t + pB, t =1. Without CI, the frequency dynamics are
If the uninfected population is initially monomorphic for mtDNA haplotype A, the Wolbachia infection frequency will increase when rare if and only if
Turelli et al. (1992) showed that if a CI-causing Wolbachia is introduced with a cytoplasm that contains a novel mtDNA haplotype B, which has no effect on fitness, Wolbachia-uninfected individuals will eventually all carry haplotype B. This follows because eventually all uninfected individuals have Wolbachia-infected maternal ancestors. This remains true for non-CI-causing Wolbachia that satisfy (A3). However, we conjectured that if the introduced B mtDNA is deleterious in the new host nuclear background, i.e., FA > FB, a stable polymorphism might be maintained for the alternative mtDNA haplotypes. The motivation was that imperfect maternal transmission seemed analogous to migration introducing a deleterious allele into an “island” of uninfected individuals. To refute this conjecture, consider the equilibria of (A1) with
If all three cytotypes are to be stably maintained, we expect each to increase in frequency when rare. In particular, we expect the fitness-enhancing mtDNA haplotype A to increase when the population contains only infected individuals and uninfected individuals carrying the deleterious Wolbachia-associated mtDNA haplotype B. From (A1), pA, t increases when rare if and only if
In the absence of haplotype A, we expect pI to be at equilibrium between selection and imperfect maternal transmission, i.e., with F =FI/FB (Hoffmann and Turelli 1997). Substituting (A6) into (A5) and simplifying, the condition for pA, t to increase when rare is
By assumption (A4), FI > FB; hence (A7) contradicts condition (A3), required for initial Wolbachia invasion. Thus, simple selection on Wolbachia-uninfected mtDNA haplotypes cannot stably maintain an mtDNA polymorphism. The “ancestral” mtDNA haplotype A is expected to be replaced by the less-fit Wolbachia-associated haplotype B.
To understand the time scale over which this replacement occurs, let rt denote the frequency of haplotype A among Wolbachia-uninfected individuals, i.e., rt =pA, t/(pA, t + pB, t). From (A1),
If we assume that the mtDNA haplotypes do not affect fitness, i.e., FA =FB, and that the Wolbachia infection frequency has reached the equilibrium described by (A6), (A8) reduces to with F =FI/FB.