Wolbachia cifB induces cytoplasmic incompatibility in the malaria mosquito vector

Wolbachia, a maternally inherited intracellular bacterial species, can manipulate host insect reproduction by cytoplasmic incompatibility (CI), which results in embryo lethality in crosses between infected males and uninfected females. CI is encoded by two prophage genes, cifA and cifB. Wolbachia, coupled with the sterile insect technique, has been used in field trials to control populations of the dengue vector Aedes albopictus, but CI-inducing strains are not known to infect the malaria vector Anopheles gambiae. Here we show that cifA and cifB can induce conditional sterility in the malaria vector An. gambiae. We used transgenic expression of these Wolbachia-derived genes in the An. gambiae germline to show that cifB is sufficient to cause embryonic lethality and that cifB-induced sterility is rescued by cifA expression in females. When we co-expressed cifA and cifB in male mosquitoes, the CI phenotype was attenuated. In female mosquitoes, cifB impaired fertility, which was overcome by co-expression of cifA. Our findings pave the way towards using CI to control malaria mosquito vectors.

High expression levels of cifA in females rescues CI in An. gambiae. We speculated that the lack of fertility rescue by zpg-cifA could be due to insufficient expression of cifA in females, as the rescue effect has been shown to be promoter-dependent 7 . To test this possibility, we engineered transgenic expression of cifA from the vasa promoter 38 (Fig. 1a) because the vasa promoter has considerably higher expression levels in the female germline than the zpg promoter (Fig. 2a). In D. melanogaster, in addition to different expression levels, vasa has a different expression pattern compared with zpg as it is also expressed in somatic gonadal precursors 39 , although it is not known whether this is true in Anopheles 38 . When mated to zpg-cifA;B males, high levels of infertility were observed in both  zpg-cifA and WT females as above, but in this case fertility was fully restored in crosses with vasa-cifA females, demonstrating effective rescue by this transgene (Fig. 2b). Combined, these results reveal that CI can be recapitulated in An. gambiae mosquitoes by transgenic expression of cifA and cifB. We also attempted co-injections of vasa-cifA and vasa-cifB constructs (Fig. 1a), but failed to isolate any cifB-expressing progeny.
Expression of cifB alone in males induces CI. Next, we investigated whether cifB alone can induce CI. Although we could not maintain a zpg-cifB colony in the absence of cifA, we were able to isolate, by fluorescent screening, a limited number of F1 zpg-cifB males from natural colony matings between heterozygous mosquitoes. We found that zpg-cifB males induced high infertility when mated to WT females, at a rate that was statistically indistinguishable from the infertility levels induced by zpg-cifA;B males ( Fig. 3a). In contrast, we found that progeny sired by zpg-cifA males were fully fertile (Fig. 3a). CI induction did not differ whether zpg-cifB males were isolated from zpg-cifA;B or vasa-cifA;zpg-cifB colonies (denoted (z)zpg-cifB or (v)zpg-cifB, respectively) (Fig. 3b). We also showed that vasa-cifA expression in females was sufficient to completely rescue sterility caused by zpg-cifB males, ruling out CI-independent effects (Fig. 3b). Cytology of 69 embryos confirmed the results obtained with zpg-cifA;B males, revealing the canonical features of CI (Extended Data Fig. 2). These findings show that conditional sterility can be induced by cifB alone in mosquitos.
cifA expression at high levels in males attenuates CI. Given that vasa-cifA rescues inviability caused by cifB in the embryo while zpg-cifA does not, we next asked whether expressing cifA under the vasa promoter in males may impact the strength of CI. To this end, we compared fertility of crosses between males expressing either zpg-cifA;B or vasa-cifA;zpg-cifB and WT females (Fig. 3c). Intriguingly, vasa-cifA;zpg-cifB males were considerably more fertile (median of 48% hatched embryos) compared with zpg-cifA;B males (median of 0% hatched embryos) (Fig. 3c). Consistent with female data, cifA expression was higher in vasa-cifA;zpg-cifB males compared with zpg-cifA;zpg-cifB males, while cifB expression levels were similar (Fig. 3d). Also in this case, the intermediate sterility effects caused by vasa-cifA;zpg-cifB males were rescued when females expressed vasa-cifA (Extended Data Fig. 3). Further, vasa-cifA expression in males did not induce embryonic lethality, supporting the idea that vasa-cifA does not contribute to CI in males and acts solely as a rescue factor (Extended Data Fig. 3). Higher expression of cifA in males (and/or a difference in localization or timing of expression compared with those achieved by the zpg promoter) thus reduces CI penetrance rather than favoring it, possibly either by limiting CifB activity within the male germline, or by rescuing CifB toxicity in the embryo following transfer of CifA in sperm 40 .
cifB expression in females disrupts fertility and fecundity. Our finding that cifB expression in males is sufficient to induce significant sterility prompted us to investigate toxicity of this factor in females. We designed crosses between WT males and either zpg-cifA;B or vasa-cifA;zpg-cifB females (Fig. 4a) and then characterized egg development and fertility of the zpg-cifB F1 female progeny after mating to WT males. We noticed that, in contrast to males (Fig. 3b), cifB-mediated effects in females were dependent on the colony of origin. When derived from zpg-cifA;B mothers, most F1 zpg-cifB females (called (z mat )zpg-cifB) failed to develop eggs following a blood meal, and only a few females yielded fertile progeny (Fig. 4b,c). Additionally, morphological analysis of the ovaries before and after ingestion of a blood meal showed that follicles were largely absent, suggestive of defects in germline development (Fig. 4d,e). When derived from vasa-cifA;zpg-cifB mothers, F1 females ((v mat )zpg-cifB) showed intermediate phenotypes with substantial follicle development, although both fecundity and fertility were reduced compared with WT females (Fig. 4b-e). However, when the cifB transgene was inherited from vasa-cifA;zpg-cifB fathers (Fig. 4a), most F1 females ((v pat )zpg-cifB) had ovaries similar to those of (z mat )zpg-cifB females, showing remarkably reduced follicle development (Fig. 4d,e). As the zpg-cifB insertion site and promoter is the same in all these groups, these results reveal rescue effects possibly caused by maternal deposition of cifA (as either mRNA of protein) from vasa-cifA-expressing mothers, although we cannot rule out a difference in other host factors. This is consistent with data showing that transgenes expressed under the vasa promoter, but not the zpg promoter, are maternally deposited 38,41 . cifB expression is therefore highly deleterious in the female germline when unchecked by the presence of cifA, and it seems to function during the early stages in germline development based on the capacity for maternally derived cifA to rescue these defects.

Discussion
Using cif genes from wPip in An. gambiae, we show that cifB expression is sufficient to induce embryonic lethality via CI. We show that it is possible to induce and rescue CI in An. gambiae, suggesting that it may be feasible to apply Wolbachia or Wolbachia-derived genes for anopheline vector control 13 . However, the reproductive toxicity observed in both sexes upon cifB expression may partially explain why infections using CI-inducing Wolbachia strains have been difficult to establish in laboratory colonies of these mosquitoes. Previous efforts to generate cifB wPip -expressing D. melanogaster were unsuccessful 6 , consistent with our own difficulties in isolating cifB-expressing individuals using two different promoters unless cifA was also co-expressed, and with our results demonstrating cifB toxicity. Our findings are in contrast with results reported for CI in D. melanogaster where both cifA and cifB from wMel were required to induce CI, and where a cifB transgenic line was isolated in the absence of cifA 5 . Further, cifB wMel females showed no defects in fertility, contrary to our results 5 . Many possible reasons could explain these discrepancies, ranging from different promoters and transgene insertion sites to specific differences in CifB function in its natural host (such as in the case of cifB from wMel in D. melanogaster) compared to a novel host (cifB from wPip in An. gambiae). Additionally, the observation that wMel causes weak CI in its natural host D. melanogaster 42,43 (though it induces strong CI in Drosophila simulans and Ae. aegypti 44,45 ), while wPip causes strong CI in its natural host C. pipiens 46 , highlights the possibility for host-dependent and strain-dependent differences. In future studies, it will be interesting to determine whether wMel cifB can induce embryonic lethality in An. gambiae.
Other studies have shown infertility induced by cifB alone in D. melanogaster, induced by wPip's Type IV cifB homologue (also called cinB) and by the Type I cifB homologue from wRec, a CI-inducing strain found in Drosophila recens 47,48 . Neither study was able to demonstrate rescue of these effects and thus could not conclude that they were CI related 47,48 . However, when cifA was expressed alongside cifB wRec in males, very little embryonic lethality was observed, reminiscent of our results showing that high cifA expression in males can attenuate CI 47 .
Combined with the data we present indicating that high levels of cifA can rescue CI in females but attenuate cifB activity in males, it is possible that Wolbachia may need to fine-tune the relative expression   of cifA and cifB in males and females to induce CI in Anopheles mosquitoes without causing lethal toxicity. Such a balancing act might make maintenance of CI-inducing Wolbachia strains in anopheline insects difficult and one outcome could be silencing of the toxic cifB gene by mutation. Interestingly, cifB nonsense mutations were identified by sequencing wAnM and wAnD strains recently discovered in some Anopheles species, although one cifB homologue in wAnD seems to be intact and it will be interesting to learn if this strain can induce CI 27 . Of note, cifB pseudogenization is not uncommon and is consistent with early evolutionary models that predict male incompatibility not to be selected for within a host lineage 49,50 . It may be possible to enable stable colonization of Anopheles by Wolbachia by limiting cifB toxicity using germline expression of cifA. This would create a route to screening for Wolbachia strains that can block transmission of Plasmodium parasites and pave the way to using Wolbachia endosymbiosis in population replacement strategies for malaria control. Plus, the remarkable sterility induced by cifB or cifA;B co-expression could be utilized for sterile male releases to suppress Anopheles populations even in the absence of Wolbachia infection, similar to the IIT programmes implemented in Aedes mosquito control 16,17 . Due to the difficulty in rearing both cifB and cifA;B mosquitoes, conditional expression of these genes may be required if mass releases were to occur. At a time when novel malaria control strategies are urgently needed, our data presents a step towards utilizing Wolbachia, or Wolbachia-derived genes, in control programmes targeting Anopheles mosquitoes.

Methods
Generation of constructs. The amino acid sequences for cifA (wPa_0282) and cifB (wPa_0283) coding regions from the published wPipI Pel strain of wPip from C. pipiens 51 were codon-optimized for expression in An. gambiae using published codon bias information 52 . Gene blocks were ordered from Integrated DNA Technologies, using custom gene synthesis to create the desired DNA fragments (accession codes OK352257 (cifA) and OK352258 (cifB)). Transgenesis constructs were engineered to express the wPip CI genes cifA and cifB under the control of the germline-specific promoters zpg (zpg, AGAP006241) 35 and vasa (vasa2, AGAP008578) 38,53 . The constructs also express a fluorescent marker under control of the ubiquitous actin promoter to enable selection of transgenic mosquitoes. Integration into the mosquito genome was mediated by piggyBac transposition and rearing lines to homozygosity was accomplished through pupae sorting via fluorescence intensity.
Embryonic micro-injection. PiggyBac transgenic construct pairs corresponding to each germline promoter (zpg-cifA-EYFP and zpg-cifB-DsRed; or vasa-cifA-EYFP and vasa-cifB-DsRed) were co-injected into the posterior of freshly laid embryos from An. gambiae (vasa n = 1,434, zpg n = 512) at a concentration of 250 ng µl −1 . Pupae that survived injection were separated according to sex, reared to adulthood, and backcrossed to wild-type G3 to identify and isolate transgenics. A total of 17 EYFP/DsRed double positive F1 transgenics were recovered from the zpg promoter-driven CI constructs injections. In contrast, only vasa-cifA-EYFP positive transgenics were recovered from the vasa-cifA/cifB co-injections. Irrespective of germline promoter, no cifB transgenic mosquitoes were identified post-injection.
Mosquito lines and rearing. An. gambiae mosquitoes from the G3 strain and transgenic derivatives of the G3 strain were maintained in a 27 °C insectary environment with 70-80% humidity and a 12 h light:12 h dark cycle. Adults were given 10% glucose and water ad libitum and fed on human blood (Research Blood Components). Larvae were fed a mixture of Tetramin fish flakes and pellets.
Crosses and fertility assays. To perform crosses between different transgenic lines, individuals were isolated as pupae from these colonies and their transgenes were identified by their respective fluorescent markers. We did not verify whether individuals were homozygous or heterozygous for their transgenes. Pupae were separated by sex under a dissecting microscope, placed in cages with a male to female ratio between 1:1 and 2:1, and allowed to eclose in small BugDorm cages. Natural mating proceeded and mosquitoes were given ad libitum access to 10% glucose solution and water for 5-7 d before blood-feeding females and allowing oviposition in individual cups lined with filter paper. Once laid, eggs were stimulated daily by spraying water and allowed to hatch for a minimum of 4 d. We then assessed fertility of females by counting and scoring eggs under a Leica M80 dissecting microscope, and additionally noting the presence or absence of hatched larvae. For any female that showed no fertile embryos, mating status was verified by checking microscopically for the presence of sperm in the spermatheca. For egg development experiments, egg counts for all females were included regardless of whether they had mated or oviposited, while only those that were mated and oviposited were included in fertility experiments.
Microscopy and tissue staining. Embryo cytology. Embryos were collected from 10-12 WT females after natural matings with zpg-cifA;B males. Four hours after oviposition, embryos were bleached, washed and dechorionated according to methods by Goltsev et al. 55 , and the endochorion was peeled according to methods by Juhn and James 56 . Embryos were then fixed and stained with 4ʹ,6-diaminidino-2-phenylindole and imaged on a Zeiss Inverted Observer Z1 at ×100 or ×400 magnification.
Brightfield microscopy of embryos. A sample of oviposited embryos were imaged on filter paper at either ×5 or ×7.5 on a Leica M80 dissecting scope.
Brightfield and differential interference contrast imaging of ovaries. Ovaries of 4-7-day-old females were dissected in PBS at either 0 h or 24 h post-blood-meal and imaged with a Leica M80 dissecting scope at ×7.5 magnification for general morphology. After initial imaging, ovaries were fixed in 4% paraformaldehyde and then mounted in Vectashield mounting medium with 4ʹ,6-diaminidino-2-phenylindole counterstain (Vector Laboratories). Ovaries were then imaged using differential interference contrast on a Zeiss Inverted Observer Z1 at ×100 magnification.
Statistical methods. In all comparisons of fertility or egg development, Anderson-Darling normality tests showed that all data were not normally distributed, so non-parametric Kruskal-Wallis tests with Dunn's multiple comparisons were used. Distinct samples were used for comparisons. Tests were performed in GraphPad Prism 8. For all fertility or egg development experiments, 2-4 replicates were performed for all groups. We conducted power analysis in G*Power 3.1 to detect a 50% reduction in fertility, yielding n = 5, non-centrality parameter = 21.3, critical chi-squared = 11.07 and total sample size = 10 to estimate the sample size required for detecting differences. On average, we used far greater sample sizes than this power analysis suggested, as we planned to use more stringent tests for non-parametric data (which cannot be estimated by power analysis). Only one replicate was performed for embryo cytology experiments. For Fig. 4d,e, where representative images were selected, one replicate of dissection and imaging of 5-10 individuals from each group was performed; however, these phenotypes were confirmed by dissections performed in experiments for Fig. 4b,c. For all RT-qPCR data, 2-5 technical replicates with 16 individuals each were performed for each group, with exact n given in figure legends.
Reporting Summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
Source data are provided with this paper. Sequence information can be found in GenBank with the accession numbers OK352257 (codon-optimized cifA) and OK352258 (codon-optimized cifB).