Cytoplasmic Incompatibility Variations in Relation with Wolbachia cid Genes Divergence in Culex pipiens

Culex pipiens mosquitoes are infected with wPip. These endosymbionts induce a conditional sterility called CI resulting from embryonic deaths, which constitutes a cornerstone for Wolbachia antivectorial methods. ABSTRACT In arthropods, Wolbachia endosymbionts induce conditional sterility, called cytoplasmic incompatibility (CI), resulting from embryonic lethality. CI penetrance (i.e., embryonic death rate) varies depending on host species and Wolbachia strains involved. All Culex pipiens mosquitoes are infected by the endosymbiotic alphaproteobacteria Wolbachia wPip. CI in Culex, characterized as a binary “compatible/incompatible” phenomenon, revealed an unparalleled diversity of patterns linked to the amplification-diversification of cidA and cidB genes. Here, we accurately studied CI penetrance variations in the light of cid genes divergence by generating a C. pipiens compatibility matrix between 11 lines hosting different phylogenetic wPip groups and exhibiting distinct cid gene repertoires. We showed, as expected, that crosses involving wPip from the same group were mostly compatible. In contrast, only 22% of the crosses involving different wPip groups were compatible, while 54% were fully incompatible. For the remaining 24% of the crosses, “intermediate” compatibilities were reported, and a cytological observation of the first zygotic division confirmed the occurrence of “canonical” CI phenotypes in a fraction of the eggs. Backcross experiments demonstrated that intermediate compatibilities were not linked to host genetic background but to the Wolbachia strains involved. This previously unstudied intermediate penetrance CI was more severe and frequent in crosses involving wPip-IV strains exhibiting cid variants markedly divergent from other wPip groups. Our data demonstrate that CI is not always a binary compatible/incompatible phenomenon in C. pipiens but that intermediate compatibilities putatively resulting from partial mismatch due to Cid proteins divergence exist in this species complex. IMPORTANCE Culex pipiens mosquitoes are infected with wPip. These endosymbionts induce a conditional sterility called CI resulting from embryonic deaths, which constitutes a cornerstone for Wolbachia antivectorial methods. Recent studies revealed that (i) two genes, cidA and cidB, are central in Wolbachia-CI mechanisms, and (ii) compatibility versus incompatibility between mosquito lines depends on the wPip phylogenetic groups at play. Here, we studied CI variations in relation to wPip groups and cid genes divergence. We showed, as expected, that the crosses involving wPip from the same group were compatible. In contrast, 78% of the crosses involving different wPip groups were partially or fully incompatible. In such crosses, we reported defects during the first zygotic division, a hallmark of CI. We showed that CI was more severe and frequent in crosses involving wPip-IV strains exhibiting cid variants, which markedly diverge from those of other wPip groups.

the latter situation, incompatibility can be either unidirectional (one cross direction is incompatible, while the reciprocal cross is compatible) or bidirectional (both cross directions are incompatible) (32,35,36). Reconstruction of wPip phylogeny revealed that mosquitoes infected with strains from the same group are more likely to be compatible with each other, while the compatibility between host-harboring wPip strains from different groups is mostly unpredictable (31). Moreover, specific variations in cidB repertoires harbored by males correlated with compatibility/incompatibility variations between C. pipiens lines, suggesting that some specific variants may play a strong role in this "yes-or-no" CI (19,27). However, few cases were also reported with intermediate HR, i.e., 10% # mean HR # 80%, without knowing if those intermediate HR were linked to the Wolbachia strains involved or other factors such as nuclear incompatibilities (30,(37)(38)(39)(40)(41)(42)(43). Indeed, at the time of these intermediate HR observations, no diversity between wPip strains was discovered, and it was not possible to decipher the part of nuclear genetic background versus Wolbachia in the observed intermediate HR.
Our recent reconstruction of wPip phylogenetic groups (29,31) and discovery of cid genes' amplification and diversification led us to correlate cid and "yes-or-no" CI diversities in C. pipiens (19,27). In the present study, we accurately monitored CI penetrance variations in the light of cid genes divergence by generating a C. pipiens compatibility matrix involving 11 lines harboring Wolbachia strains belonging to different wPip groups (wPip-I to wPip-IV) and all harboring different cid repertoires (9,19). This compatibility matrix is composed of estimated HR obtained from (i) 11 intraline crosses (INTRA), (ii) 12 crosses between lines harboring wPip strains from the same group (INTER-INTRA), and (iii) 83 crosses between lines harboring wPip from different groups (INTER-INTER). We showed, as expected, that all INTRA and INTER-INTRA (except two) crosses were fully compatible. Among the INTER-INTER crosses, 54% were totally incompatible, displaying no hatching, and 22% were considered fully compatible, while 24% of the crosses exhibited mean HRs that can be qualified as intermediate. Backcross experiments demonstrated that such intermediate HRs were not linked to host genetic background but to the Wolbachia strains involved. Moreover, we showed that intermediate HR values were particularly low within crosses involving wPip-IV strains that also present marked phylogenetic difference in their cid repertoires from other wPip groups (19). To visualize the developmental defects responsible for intermediate HR, we monitored the embryonic development and found defects during the first zygotic division and subsequent developmental arrest, which are typical hallmarks of "canonical CI" (9,14). Altogether, our data demonstrate that CI is not always a "yes-or-no" phenomenon in C. pipiens but that subtle CI variations, referred to as "cryptic CI," putatively resulting from partial mismatch due to Cif protein divergence, exist in this species complex.

RESULTS
HR in fully compatible crosses. Mean HR of the 11 INTRA crosses were comprised between 0.78 and 0.95, showing that an important part of the eggs (up to 22%) failed to develop even in INTRA crosses. Intermediate HR can thus only refer to crosses with mean HR #78% ( Fig. 1; Table S1 in the supplemental material; Data Set S1).
Depriving lines from Wolbachia did not influence INTRA HR. To test for the effect of presence/absence of Wolbachia, two C. pipiens lines were tetracycline treated (SlabTC and IstanbulTC). For these "cured lines", mean HRs were not significantly different from HRs of the corresponding INTRA crosses with infected lines (Wilcoxon W = 356, P = 0.168; and W = 344, P = 0.119 for Slab/SlabTC and Istanbul/IstanbulTC, respectively) (Table S1; Data Set S2).
INTER-INTER crosses exhibit significantly reduced HR. The full distribution of HR per egg raft for all the crosses is presented in Fig. 1A. The mean HR (i.e., calculated on   Table S1).
The lowest HRs were observed in INTER-INTER crosses involving wPip-IV strains. For the 38 INTER-INTER crosses which were not fully incompatible, global models did not reveal any significant effect of the wPip group hosted by either female or male lines (GLMM; x 2 = 0.268, df = 3, P = 0.966; x 2 = 2.742, df = 3, P = 0.433, respectively) but pointed out a significant interaction effect between the wPip groups involved in the crosses (generalized linear models [GLM]; x 2 = 113.764, df = 13, P , 0.001; for detailed statistics, see Text S1). Careful inspection of the HR matrix revealed that 8 INTER-INTER crosses out of 38 showed a mean HR below 60%, here called low HR ( Fig. 1B; Table S1). All these eight INTER-INTER crosses with low HR involved wPip-IV strains (see HR per egg raft full distribution in Fig. 1B; pink dots show HR obtained in crosses involving males infected with wPip-IV strains). INTER-INTER crosses with backcrossed line Sl(wPip-IV-Harash) did not differ from crosses involving Harash lines (GLM; x 2 = 0.0137, df = 1, P = 0.907), demonstrating that it was the wPip-IV strain harbored in the cytoplasm and not the host genetic background that explained such a low HR.
Intermediate HR results from cryptic but canonical CI. As low HRs (mean HR under 0.6) were only observed in INTER-INTER crosses involving wPip-IV strains, we (i) studied the first zygotic division resulting from these crosses, and (ii) in an attempt to quantify putative CI defects, compared them with INTER-INTRA and INTRA crosses at 5 h (Table 1 and Table S2). To verify whether intermediate HRs were due to previously described canonical CI cellular mechanisms (5-9), we visualized the first zygotic division with paternal and maternal chromatin labeled in green/yellow and red, respectively. In INTER-INTER crosses with intermediate HR, an important proportion of eggs normally hatched. Such normal embryogenesis, as documented in Fig. 2, is similar to what was observed for all INTRA embryos previously documented (9). After   (9). As it was the green-labeled chromatin that exhibited such defects, it can be concluded that paternal chromatin is affected (Fig. 3A, A9, B, C, and D). For only three crosses involving wPip-I and wPip-IV strains (one INTRA between Tunis wPip-I infected individuals; one INTER-INTRA between Ichkeul-13 wPip-IV males and Harash wPip-IV strain females; one INTER-INTER between Harash wPip-IV males and Tunis wPip-I females), we were able to produce enough observable embryos to assess the proportion of embryos with abnormal development 5 h postoviposition as presented in Table 1. At this time, embryos should have reached the syncytial blastoderm stage (;3,200 "normal" nuclei; Fig. 4A and C), while embryos considered "abnormal" only presented few nuclei (less than 50; Fig. 4B). Moreover, atypical mitotic features were observed in these abnormal embryos ( Fig. 4D and E)  cid variants from wPip-IV repertoires are divergent from those of other wPip groups. The phylogenetic cidA and cidB networks constructed with wPip strains repertoires showed that wPip strains from the wPip-IV group exhibited markedly divergent cidA and cidB variants. For both cidA and cidB variants, wPip-IV variants clustered remotely from other groups' variants ( Fig. 5; Tables S3 and S4). Two well-separated clusters of wPip-IV cidA variants appeared on the network, while all cidB variants clustered altogether (Fig. 5). For other wPip groups, no clear wPip-group-based clustering was observed (Fig. 5).

DISCUSSION
In arthropods in which CI is mainly studied between infected males and uninfected females, including major insect models such as Drosophila and Nasonia, CI penetrance was proved to depend on Wolbachia strains, their densities, host genetic background, age of the males, and environmental factors such as temperature (44)(45)(46)(47)(48)(49)(50)(51)(52)(53)(54)(55)(56)(57)(58)(59). On the contrary, in Culex pipiens s.l. mosquitoes, these factors did not affect CI penetrance (19,(22)(23)(24)33): full CI (hatching rate [HR], 0) is reported between infected males and uninfected females (cured of Wolbachia with antibiotics) whatever their geographical origin, age, or genetic background (9,18,23,33). However, hundreds of crosses between lines infected with Wolbachia revealed unparalleled variations in CI patterns in C. pipiens. Two main opposite outcomes were observed: either the crosses were compatible (mean HR $ 80%) or incompatible, producing almost no larvae (mean HR , 0.01%) (18, 21-23, 31, 32, 34, 40). Early in the study of CI in Culex, backcross experiments demonstrated that the host genome does not influence the outcome of a given cross (9,24,32,60). In the present study, we conducted backcross experiments for two of our lines and also performed crosses between cured individuals from different C. pipiens species, which again confirmed that host genetic background does not impact compatibility. CI Variations in Relation with cid Genes Divergence ® Most crosses and backcrosses showed that CI in C. pipiens is a binary compatible/incompatible phenotype under the sole control of Wolbachia. However, in the numerous articles that presented results of interline C. pipiens crosses from different parts of the world, rare cases of intermediate HR were reported (30,(37)(38)(39)(40)(41)(42)(43). At the time of these publications, all the wPip were considered clonal due to monomorphic genetic markers available (34,61). Intermediate HRs were thus attributed to putative undiscovered Wolbachia variability (including different wPip sublines in the same laboratory line) and most probably to putative host "restorer" nuclear factors counteracting Wolbachia CI induction (40,41,62). In the present paper, we investigated these intermediate HR situations in light of our present knowledge of wPip genomes (19,27,29,31). To that extent, we studied 106 crosses between 11 C. pipiens isofemale lines infected with different wPip strains from different groups (I to IV), each exhibiting different cidA-cidB repertoires (9,19) (Table S3 in

Sicard et al.
® treatment) did not detect any CI typical defects (9). Here, we reported no difference in HR in the same lines with or without Wolbachia, confirming that CI induced by Wolbachia is not responsible for the 5% to 22% of the eggs that did not reach the larval stage. Abortive eggs in INTRA crosses certainly resulted from imperfect fertilization and/or intrinsic mortality during development from eggs to larvae (9,18,40). The 12 INTER-INTRA crosses, involving lines from different locations but harboring the same wPip group, exhibited HRs similar to INTRA crosses, except for two cases of unidirectional incompatibility, again demonstrating that the wPip group is a major predictor of compatibility between C. pipiens lines ( Fig. 1A; Table S1) (31).
Heterogeneity in compatibility clearly increased in INTER-INTER crosses (Table S1). Among the 83 performed here, we found that 54% of them were fully incompatible, while the other 46% (38/83) were fertile and exhibited HR comprised between 48% and 96%. Global HR statistical analyses, including all fertile crosses ( (Fig. 1A). Moreover, we found that 53% of the fertile INTER-INTER crosses actually exhibited HRs that were low enough to be characterized as intermediate. We also found that the interaction between the wPip groups infecting the male and female lines significantly influenced HR. Careful inspection of the HR matrix revealed that the crosses with a low HR below 60% (8 crosses out of the 20 with intermediate HR) were only observed in INTER-INTER crosses involving wPip-IV strains (Fig. 1B and Table S1). Atyame et al. (31) had already shown that wPip-IV group-infected C. pipiens lines exhibited markedly different crossing types from lines infected with other wPip groups. Network phylogenetic analyses of all the 34 cidA and 21 cidB different variants characterized in the wPip strains studied here revealed that cid-IV variants (especially cidB) were divergent, gathering in specific clusters, while other wPip groups are mixed altogether. This suggests that Cid proteins that are considered major effectors of CI (15,17) are divergent in wPip-IV strains compared to other wPip groups (Fig. 5).
To investigate whether intermediate HR resulted from canonical CI, i.e., paternal chromatin defects during first zygotic division, we monitored the first stages of embryonic development in embryos from INTER-INTER crosses. In these crosses, even with low HR, many embryos exhibited normal development into larvae (Fig. 2). However, in a few embryos, we were able to document imperfect paternal chromatin segregation during the first zygotic division (Fig. 3). These embryonic defects, which were never observed in INTRA crosses (9), were similar to those reported in fully incompatible crosses (9). Such defects in the first zygotic division likely produced aneuploid nuclei which might disrupt further development or even arrest the embryogenesis. The proportion of embryos that did not reach blastoderm stage 5 h postoviposition, but presented instead few nuclei only, can be considered a quantitative proxy for the occurrence of CI defects during the first division. We observed a larger amount of abnormal developmental stages, 5 h postoviposition, in INTER-INTER crosses  than the INTRA and INTER-INTRA crosses (Table 1). Abnormal embryos, which represented 20% of the embryos in the INTER-INTER crosses studied and 5% in the INTRA one (Table 1), displayed very few (or no) nuclei (Fig. 4B). These observations suggest that embryonic defects during the first division are responsible for the intermediate HR observed in the analyzed INTER-INTER cross ( Table 1). The intermediate HR observed in INTER-INTER crosses could be attributed to cryptic CI (in that it has a weak penetrance) but canonical CI (in that it translates into the same cytological defects).
In the light of the toxin-antidote model of CI, penetrance would depend on the interaction between CidA, CidB, and their specific substrates, eventually leading to paternal chromatin defects or its rescue (15,16,63). In C. pipiens, as all wPip genomes encode a repertoire of several polymorphic variants of CidA and CidB (19,27), full compatibility could result from multiple interactions between different CidA and CidB variants even in INTRA or INTER-INTRA crosses. In every C. pipiens male, several CidB proteins differing in their amino acid sequences might be introduced in the sperm and then in the egg during fertilization where several CidA proteins might also be present. Full compatibilities reported here in some INTER-INTER crosses involving different wPip groups with totally different CidA/CidB repertoires (Fig. 1) suggest (i) that strict specific interactions between cognate variants are not required for full compatibility, and (ii) a potential redundancy in the interaction between CidA/CidB variants. The intermediate HR resulting from cryptic CI in a given INTER-INTER cross can hypothetically result from partial rescue due to imperfect interactions between the CidA and the CidB from the two wPip strains repertoires. Since most of the embryos from intermediate HR crosses developed into living larvae, it certainly means that, in those individuals, CidB toxicity has been efficiently counteracted. On the contrary, in embryos exhibiting CI, CidB toxicity would not have been counteracted properly. This heterogeneity could be explained if embryo rescue depends on one or a few matching CidA variants which might be required in a larger quantity for the rescue to occur. However, it is possible that in certain eggs, the expression of the(se) CidA variant(s) would be too low to counteract the CidB toxicity. This would be especially true for neutralizing CidB proteins encoded by wPip-IV strains that show striking differences in their sequences from other wPip groups (Fig. 5). Less efficacy in the interactions between CidB-IV proteins and CidA from other groups could explain their higher probability to be involved in both (i) full incompatibility as reported in reference 31, and (ii) cryptic CI as reported here.
The interactions between the CidA and CidB repertoires encoded by wPip strains determine the developmental fate of each embryo of a given cross, normal development versus CI. CI penetrance (i.e., the proportion of embryos undergoing CI) in a given cross could then be determined by the diversity of cidA/cidB genes of the different wPip genomes hosted by the different C. pipiens lines, their expression levels, and the affinity between the resulting proteins.

MATERIALS AND METHODS
Culex pipiens lines. Eleven isofemale lines were used (Table S5 in the supplemental material). They differed in (i) their geographical origins, (ii) the species they belong to, (iii) the wPip group (I, II, III, or IV), and (iv) their cid repertoires. The Wolbachia group was checked by performing a pk1 PCR-restriction fragment length polymorphism (RFLP) test (64) on DNA extracted using cetyltrimethylammonium bromide (CTAB) protocol (65). Tetracycline-treated Wolbachia-free lines (TC lines), named SlabTC and IstanbulTC, were obtained from Slab-and Istanbul-infected lines as described in reference 33. The absence of Wolbachia was checked by PCR on a fragment of the wsp gene using the primers designed in reference 66. TC-treated lines were raised at least four generations without tetracycline before experiments. The wPip-I strain from the Tunis line and wPip-IV strain from the Harash line were independently introgressed into the Slab line nuclear genetic background through 8 backcrosses as described in reference 9.
Hatching rate estimations. To test for a putative effect of Wolbachia on basic intraline HR, we performed two intraline crosses between males and females from cured lines (SlabTC and IstanbulTC) and compared them with infected intraline crosses (Slab and Istanbul, respectively). To test for a putative impact of the different wPip groups on HR, we carried out 106 different crosses between males and females, including (i) from the same line (11 intraline crosses, called INTRA crosses), (ii) from two distinct mosquito lines infected with wPip strains from the same group (12 interline-intragroup crosses, called INTER-INTRA crosses), and (iii) from two distinct mosquito lines infected with wPip strains from different groups (83 interline-intergroup crosses, called INTER-INTER crosses).
To test for a potential impact of host genetic background on HR, we performed 23 extra crosses involving the two backcrossed lines [Sl(wPip-I-Tunis) or Sl(wPip-IV-Harash)]. Moreover, to study the putative effect of interspecies crosses (i.e., C. pipiens versus C. quinquefasciatus) on HR, we performed the reciprocal crosses between SlabTC (C. quinquefasciatus) and IstanbulTC (C. pipiens) lines. To perform each of these 135 different crosses, 2-day-old males (n = 50) and females (n = 100) were put together in cages. After 6 days, females were blood fed with turkey blood using a Hemotek feeding system (Discovery Workshops). After 5 days, egg rafts were collected. After the death of all the larvae (i.e., about 5 days after hatching), pictures of both eggs and larvae for 30 egg rafts per cross were taken. Eggs and larvae were counted manually on ImageJ (67). HR was calculated per egg raft as the ratio between the total number of larvae and the total number of eggs.
Cellular study of embryogenesis. To search for putative embryonic defects that might confirm the involvement of canonical CI in INTER-INTER crosses resulting in intermediate HR, several crosses involving males from different lines infected with wPip-IV strains were performed (Table S2). To that extent, cages containing 50 males and 100 females were put into a closet where the day-night cycle was inverted to allow collection of eggs during the day. After 6 days in these cages, females were fed with turkey blood, and waterpots were placed into the cages for 30 min to 1 h to allow females to lay egg rafts. For C. pipiens embryos, at 25°C, the meiosis is approximately completed 30 min postoviposition and the first mitotic nucleus division 15 min after, while 5 h after oviposition, the embryos reach the blastoderm stage (9). Freshly collected eggs (30 min to 5 h) were fixed, dechorionated, and observed as previously described in reference 9.
Statistical analysis. We used generalized linear models (GLM) or generalized linear models with mixed effects (GLMM) with a logit link function (see Text S1). To test for potential impact of Wolbachia presence/absence and host species, Wilcoxon tests were performed (68). To compare the proportion of  (69). All computations were performed using R version 3.4.4 (70).
Phylogenetic networks of the cidA and cidB genes. All the cidA and cidB repertoires of the Wolbachia strains hosted by the 11 crossed lines were already published except for Brazil that has been obtained by PCR cloning followed by Sanger sequencing as previously described in references 9 and 19. Sequenced variants (accession numbers given in Table S4) were aligned using the Muscle algorithm implemented in SeaView 6.4.1 software (71) and then analyzed within a phylogenetic network framework from uncorrected P distances by the neighbor-net method implemented in SplitsTree4 (72) to account for potentially conflicting signals due to recombination.

ACKNOWLEDGMENTS
This work was funded by the French ANR (project CIAWOL, ANR-16-CE02-0006-01). Sequencing data were generated on the GenSeq platform. Confocal microscopy was performed in the MRI-CRBM platform. Embryos optical observations were performed at the CytoEvol facilities.