Natural Wolbachia infections are common in the major malaria vectors in Central Africa

Abstract During the last decade, the endosymbiont bacterium Wolbachia has emerged as a biological tool for vector disease control. However, for long time, it was believed that Wolbachia was absent in natural populations of Anopheles. The recent discovery that species within the Anopheles gambiae complex host Wolbachia in natural conditions has opened new opportunities for malaria control research in Africa. Here, we investigated the prevalence and diversity of Wolbachia infection in 25 African Anopheles species in Gabon (Central Africa). Our results revealed the presence of Wolbachia in 16 of these species, including the major malaria vectors in this area. The infection prevalence varied greatly among species, confirming that sample size is a key factor to detect the infection. Moreover, our sequencing and phylogenetic analyses showed the important diversity of Wolbachia strains that infect Anopheles. Co‐evolutionary analysis unveiled patterns of Wolbachia transmission within some Anopheles species, suggesting that past independent acquisition events were followed by co‐cladogenesis. The large diversity of Wolbachia strains that infect natural populations of Anopheles offers a promising opportunity to select suitable phenotypes for suppressing Plasmodium transmission and/or manipulating Anopheles reproduction, which in turn could be used to reduce the malaria burden in Africa.

to the massive use of bed nets (Pates & Curtis, 2005) might challenge malaria eradication in the coming decades. Therefore, it is vital to develop alternative and non-insecticide-based control strategies for malaria control, at it has been promoted by the Global Technical Strategy form Malaria 2016-2030, which look for "reducing global malaria incidence and mortality rates by at least 90% by 2030" (Newby et al., 2016;WHO, 2015).
Several methods have been proposed to accompany or replace the use of synthetic insecticides (McGraw & O'Neill, 2013). Among them, the use of the maternally inherited Wolbachia bacteria (α-proteobacteria, Anaplasmataceae family) has emerged as a promising alternative biological tool for fighting malaria and other vector-borne diseases (Bourtzis et al., 2014;Hoffmann, Ross, & Rasic, 2015;Iturbe-Ormaetxe, Walker, & Neill, 2011;Kambris, Cook, Phuc, & Sinkins, 2009;McGraw & O'Neill, 2013). This bacterium exhibits a large spectrum of interactions with its hosts: from mutualism and commensalism to parasitism (Werren, Baldo, & Clark, 2008). Moreover, Wolbachia can invade mosquito populations and/or prevent vectorborne infections in some of the most important mosquito vectors Hoffmann et al., 2015;Iturbe-Ormaetxe et al., 2011). Indeed, Aedes aegypti populations that were artificially infected with Wolbachia have been successfully used to suppress dengue transmission in laboratory conditions and have been released in natural populations of this mosquito Schmidt et al., 2017). Similarly, laboratory studies showed that infection of Anopheles (the vector of human malaria) with Wolbachia strains has a negative impact on the transmission of Plasmodium parasites (Bian et al., 2013;Hughes, Koga, Xue, Fukatsu, & Rasgon, 2011;Kambris et al., 2010), providing a relevant alternative for malaria control.
Unfortunately, only one stable transfected Wolbachia colony has been described in Anopheles stephensi (Bian et al., 2013). Therefore, data on the use Wolbachia for Anopheles control remain scarce and mainly concern experimental studies in laboratory conditions (Bian et al., 2013;Hughes, Vega-Rodriguez, Xue, & Rasgon, 2012), due to technical (i.e., egg microinjection) and biological (i.e., competitive exclusion with the bacterium Asaia) difficulties in carrying out transinfections in Anopheles, despite multiple assays Jeffries, Lawrence, et al., 2018;Rossi et al., 2015). For a long time, it was assumed that Wolbachia was absent in natural populations of Anopheles . However, in the last few years, three studies reported that Anopheles gambiae, Anopheles coluzzii and Anopheles arabiensis (three major malaria vectors) populations from Burkina Faso and Mali (West Africa) are naturally infected by Wolbachia (Baldini et al., 2014;Gomes et al., 2017;Shaw et al., 2016). Notably, they showed a negative correlation between Wolbachia infection and Plasmodium development (Gomes et al., 2017;Shaw et al., 2016). Moreover, a very recent report suggests that other Anopheles species also are infected with Wolbachia Jeffries, Lawrence, et al., 2018). These findings support the development of novel vector control strategies based on Wolbachia-Anopheles interactions. However, although Wolbachia naturally infects 40%-60% of arthropods (Duron et al., 2008;Zug & Hammerstein, 2012), infection of Anopheles species is still not well documented. Moreover, during the last decade, screens in many other malaria mosquito species worldwide (n = 38) did not bring any evidence of Wolbachia infection (Bourtzis et al., 2014;Hughes, Dodson, et al., 2014;Osei-Poku, Han, Mbogo, & Jiggins, 2012).

In this study, we investigated the presence of Wolbachia in 25
Anopheles species in Gabon, Central Africa. We sampled mosquitoes across the country and in a variety of ecological settings, from deep rainforest to urban habitats. By using a molecular approach, we

| Mosquito sampling and DNA extraction
Mosquitoes were collected in eight sites across Gabon, Central Africa, from 2012 to 2016 ( Figure 1, Table 1, Appendix S1). These sites included sylvatic (national parks) and domestic habitats (villages and cities). Adult females were collected using Center for Disease Control () light traps, BioGents (BG) traps and HLC. Overall, CDC and BG were used in sylvatic and HLC in domestic sites (see Figure 1, Table S1). Collected specimens were taxonomically identified according to standard morphological features (Gillies & Coetzee, 1987;Gillies & de Meillon, 1968). Then, they were individually stored in 1.5 ml tubes at −20°C and sent to Centre International de Recherches Scientifiques de Franceville for molecular analysis. When possible, at least 30 mosquitoes (from 1 to 58) for each Anopheles species from different sites were selected for genomic analysis. Total genomic DNA was extracted from the whole body using the DNeasy Blood and Tissue Kit (Qiagen), according to the manufacturer's instructions. Genomic DNA was eluted in 100 μl of TE buffer. Specimens belonging to the An. gambiae complex, An. funestus group, An. moucheti complex and An. nili complex were molecularly identified using PCR-based diagnostic protocols (Cohuet et al., 2003;Fanello, Santolamazza, & della Torre, 2002;Kengne et al., 2007;Kengne, Awono-Ambene, Nkondjio, Simard, & Fontenille, 2003;Santolamazza et al., 2008).

| Wolbachia screening and multilocus sequence typing analysis
Wolbachia infection in adult females was detected by nested PCR amplification of a Wolbachia-specific 16S rDNA fragment (~400 bp) using 2 μl of host genomic DNA, according to the protocol developed in Catteruccia's laboratory (Shaw et al., 2016). Amplification of this 16S rDNA fragment in infected Aedes albopictus and Culex pipiens genomic DNA (data not shown) confirmed the performance of this nested PCR protocol to detect Wolbachia in many different mosquito species (Shaw et al., 2016). To detect potential contaminations, Ae. albopictus and Culex quinquefasciatus from Gabon were used as positive controls, and water and Ae. aegypti as negative controls. Moreover, PCR amplifications for each species were carried out independently and on different days. The amplicon size was checked on 1.5% agarose gels, and amplified 16S rDNA fragments were sent to Genewiz (UK) for sequencing (forward and reverse) to confirm the presence of Wolbachia-specific sequences. The DNA quality of all samples was confirmed by the successful amplification of a fragment (~800 bp) of the mitochondrial gene COII in all the Anopheles species under study (Ndo et al., 2010;Rahola et al., 2014). PCR products were run on 1.5% agarose gels, and COII fragments from 176 mosquito specimens of the 25 species were sequenced (forward and reverse) by Genewiz (UK) for the Anopheles phylogenetic studies. Wolbachia-positive genomic DNA samples (2 μl/sample) were then genotyped by multilocus sequence typing (MLST) using three loci, coxA (~450 bp) ftsZ (~500 bp) and fbpA (~460 bp) (Baldo et al., 2006), and according to standard conditions (Baldo et al., 2006). If the three fragments could not be amplified, a newly developed nested PCR protocol was used. Specifically, after the first run with the standard primers, 2 μl of the obtained product was amplified again using internal primers specific for each gene: coxA (coxA_NF-2: 5′-TTTAACATGCGCGCAAAAGG-3′;  (Table S1). Unfortunately, the other three MLST genes (gatB, wsp and hcpA) could not be amplified, due to technical problems (i.e., multiple bands).

| Phylogenetic and statistical analysis
All Wolbachia sequences for the 16S, coxA, fbpA and ftsZ gene fragments and for Anopheles COII were manually corrected using Geneious R10 (Kearse et al., 2012). The resulting consensus sequences for each gene were aligned with sequences that represent the main known Wolbachia supergroups obtained from GenBank (see Table S1). Only unique haplotypes for each species were included in the analysis (haplotype was defined as a unique allelic profile for each examined locus). Inference of phylogenetic trees was performed using the maximum likelihood (ML) method and RAxML (Stamatakis, 2014) with a substitution model GTR + CAT (Stamatakis, 2006) and 1,000 bootstrapping replicates. Finally, all MLST Wolbachia sequences were used to build phylogenetic trees using RAxML (GTR + CAT model, 1,000 bootstrapping replicates).
To quantify the accuracy of the observed Wolbachia infection prevalence, the influence of sample size on its estimation was assessed. For this, it was assumed that Wolbachia prevalence within a host species followed a beta binomial distribution (Zug & Hammerstein, 2012) yielding many species with a low or a high TA B L E 1 Summary of the Anopheles species screened in this study

| Wolbachia naturally infects a large number of Anopheles species from Gabon
In this study, we screened 648 mosquitoes from eight sites in Gabon ( Figure 1, Table 1, Table S1). On the basis of their morphological traits (Gillies & Coetzee, 1987) and molecular analysis results (Cohuet et al., 2003;Kengne et al., 2007Kengne et al., , 2003Rahola et al., 2014;Santolamazza et al., 2008), we identified 25 Anopheles species (Appendix S1). Our sampling included all the species in which the presence of Wolbachia was previously investigated in Africa (An. gambiae, An. coluzzii, An. funestus and Anopheles coustani), with the exception of An. arabiensis that is absent in Gabon (Table 1) . By PCR amplification of a 16S rRNA fragment (Shaw et al., 2016), we found 70 Wolbachia-positive specimens that belonged to 16 different Anopheles species, distributed throughout the country (Figure 1, Table S1). When considering only species with more than 10 screened individuals, we observed that Wolbachia infection was commonly lower than 15% (11/13), as observed in other arthropods (Duron et al., 2008;Zug & Hammerstein, 2012). On the other hand, two species, and moreover major malaria vectors, An. moucheti and An. nili, F I G U R E 2 Circular phylograms of the Wolbachia strains isolated in the 16 Anopheles species. The phylogenetic trees were built with RAxML (Stamatakis, 2014). The names of the Anopheles species from which the Wolbachia-specific sequences were isolated in this study are shown in blue (positive for Wolbachia supergroup B), red (positive for supergroup A) and brown (positive for supergroup C), while the names of mosquitoes species (Diptera) from which the previously published Wolbachia sequences were isolated are in green. Other Wolbachia strains sequences ("others," in grey) were obtained directly from gene sequence repository ncbi (https ://www.ncbi.nlm.nih.gov/). Red dots show branches supporting a bootstrap >70% from 1,000 replicates. (a) Circular phylogenetic tree using the Wolbachia-specific 16S rRNA fragment and Anaplasma marginale as outgroup. Different Wolbachia strains found in the same Anopheles species are connected by pink lines. The pink bar charts indicate the number of identical Wolbachia haplotypes found in each species. Scale bar corresponds to nucleotide substitutions per site. (b) Circular phylogenetic trees based on the coxA, fbpA and ftsZ fragment sequences using Dirofilaria immitis (supergroup C) as outgroup. Specimens with a different supergroup assignation than 16S are marked with asterisks. Only, Anopheles vinckei M002 (purple) oscillated between groups B and A across the four genes exhibited more than 50% of Wolbachia infection (Table 1), as previously reported in other mosquito species where prevalence can be very high (Dumas et al., 2013;Duron et al., 2005).

| Wolbachia is maternally inherited in An. moucheti
Although Wolbachia is mainly maternally transmitted (Werren et al., 2008), horizontal transmission may occasionally occur in natural conditions (Ahmed, De Barro, Ren, Greeff, & Qiu, 2013;Li et al., 2017;Werren, Zhang, & Guo, 1995). To confirm the maternal transmission in the infected mosquito species, we focused on An. moucheti for logistic reasons (i.e., highest Wolbachia prevalence and ease of sampling). Although no laboratory An. moucheti strain is currently available, we obtained eggs from six Wolbachia-infected females. In total, we analysed the infectious status of 79 progeny by PCR amplification of the same 16S rRNA fragment (Shaw et al., 2016) (Table S3) and found that 70 were infected, with an average maternal transmission frequency of 97.54% (range: 90%-100%).

| Naturally occurring Wolbachia strains in Anopheles reveal high genetic diversity
By sequence analysis of the 16S rRNA fragment PCR amplified from each Anopheles sample (Table 1) Finally, we found that one An. coustani individual was infected by a Wolbachia strain from supergroup C that is known to infect only filarial worms. Therefore, we investigated the presence of filarial nematode DNA in the mosquito by PCR amplification and sequencing of a fragment of the COI filarial gene (Casiraghi, Anderson, Bandi, Bazzocchi, & Genchi, 2001), followed by phylogenetic analysis with RAxML. Our results confirmed the presence of Dirofilaria immitis in this specimen ( Figure S1). This canine filarial parasite hosts Wolbachia and is transmitted by many mosquitoes, including Anopheles (Simon et al., 2012). Therefore, it is not surprising to find an An. coustani specimen infected by this filarial nematode.
To expand our knowledge on the Wolbachia strains that infect natural Anopheles populations, we PCR amplified, sequenced and analysed fragments from three conserved Wolbachia genes (coxA, fbpA and ftsZ) that are commonly used for strain typing and evolutionary studies (Baldo et al., 2006) (Figure 2). We used a new nested PCR protocol (see section 22) for samples that could not be genotyped using the classical MLST primers (Table S1).
Our phylogenetic analyses confirmed the 16S results, assigning most of the species to supergroups A and B. Few samples (asterisks in Figure 2, gene coxA) showed some incongruence relative to the 16S results. They suggest signals of recent recombination between the supergroups A and B, as previously demonstrated (Baldo et al., 2006). Detailed sequence analysis revealed that mosquito species belonging to the same group or complex (i.e., or Anopheles marshallii, while the strain infecting An. nili (wAnni), which evidenced strains variation even in the same locality, was more closely related to those found in other mosquito species, such as Ae. albopictus or Cx. quinquefasciatus (Figures 2 and 3).

An. moucheti and
Conversely, the other haplotypes were associated with one specific host.

| Wolbachia independently evolves in malariatransmitting mosquitoes
As Wolbachia is mainly a maternally inherited bacterium, the host mitochondrial DNA (mtDNA) is a suitable marker to study its evolutionary history in Anopheles (Richardson et al., 2012). Analysis of COII sequences from 176 specimens belonging to the 25 Anopheles species collected in Gabon provided the most exhaustive phylogenetic tree of Anopheles in Central Africa (Figure 3). This analysis highlighted the independent acquisition and apparent loss of Wolbachia across the different Anopheles species clades.

Moreover, the genetic distances of Wolbachia strains and their
Anopheles host were not correlated (Mantel test, p > 0.05; Figure   S2). Nevertheless, mosquitoes from the An. moucheti complex, and therefore genetically very close, shared the same Wolbachia supergroup and haplotypes (Figure 3 and Figure S2). Finally, we investigated how Wolbachia evolved within each Anopheles species (Charlat et al., 2009). Our results revealed that Wolbachia-infected and noninfected mosquitoes shared the same mtDNA haplotype (Figure 3), indicating that infection status and host haplotypes are not associated.
F I G U R E 3 Maximum likelihood phylogeny of the 25 Anopheles species under study and Wolbachia haplotypes. The tree was inferred with RAxML (Stamatakis, 2014) using the sequences of the COII fragment from 176 Anopheles specimens belonging to the 25 species under study and rooted with Anopheles darlingi as outgroup (New World mosquito, diverged 100 Myr ago (Neafsey et al., 2015)). Red dots in branches represent bootstrap values >70% from 1,000 replicates. The shape of each field column represents the 16S (rectangle), coxA (rhombus), fbpA (triangle) and ftsZ (hexagon) genes. The different Wolbachia gene haplotypes (i.e., unique allelic profiles) are indicated with colour codes (all pink = the newly identified wAnmo strain). The bar chart size indicates the number of individuals of the same species with the same haplotype, and the colour represents their infection status: grey, noninfected; blue, infected by the Wolbachia supergroup B; red, infected by supergroup A; brown, infected by supergroup C

| D ISCUSS I ON
The present study provides three key findings. First, the genus few other species (Baldini et al., 2018;Jeffries, Lawrence, et al., 2018;Niang et al., 2018). Several hypotheses can be put forward to explain this. First, low infection prevalence or local variations could have hindered the discovery of Wolbachia infections, independently of the sampling effort. In our study, most Anopheles species exhibited a prevalence lower than 15% (Table 1). This pattern is common in many other arthropods (Duron et al., 2008;Zug & Hammerstein, 2012), and it is usually associated with a weak manipulation of the host reproduction and/or imperfect maternal transmission (Engelstadter & Hurst, 2009). In general, our sampling effort was higher than in previous studies (n < 30) (Bourtzis et al., 2014;Osei-Poku et al., 2012), and this could explain why we found more infected species. Our statistical analysis showed that a sample size of 60 individuals per species is needed to quantify correct prevalence rates lower than 15%, with a probability of 95% ( Figure S3). Moreover, local frequency variations among populations could also hinder the detection of Wolbachia infections (Dumas et al., 2013). For instance, we sampled An. coluzzii in three different sites, but we only found Wolbachia-infected mosquitoes at La Lopé ( Figure 1, Table S1). Therefore, sampling in different localities and in different seasons might improve detection rates. Second, it could be difficult to detect low-density Wolbachia infections in Anopheles with the routinely used molecular tools, as previously reported for other arthropods (Arthofer, Riegler, Avtzis, & Stauffer, 2009;Augustinos et al., 2011) and recently in An. gambiae (Gomes et al., 2017). Our results indicate that conventional PCR amplification (wsp-targeting primers (Baldo et al., 2006) (Baldini et al., 2014;Gomes et al., 2017;Jeffries, Lawrence, et al., 2018;Niang et al., 2018;Shaw et al., 2016), there exist the doubt if they are real infections (Chrostek & Gerth, 2018). The Wolbachia sequences found in our specimens were genetically close to those found in other Diptera, and no signal of extensive divergence was detected (Figures 2 and 3) In Central African Anopheles, Wolbachia acquisition seems to be independent of the host phylogeny (Figures 2 and 3). Our results revealed that the genetic distances between Wolbachia and Anopheles are not positively correlated (Mantel test, p > 0.05; Figure S2). The lack of correlation could lead to think that Wolbachia and the host lineage evolved independently. The different larval ecology of these species suggests other ways of lateral transfer (e.g., during nectar feeding (Li et al., 2017) (Dumas et al., 2013) and ants (Tsutsui, Kauppinen, Oyafuso, & Grosberg, 2003), the same species is infected by different Wolbachia strains according to the region.
The availability of whole-genome sequences for Wolbachia strains (Gerth, Gansauge, Weigert, & Bleidorn, 2014) will enlighten the intricate phylogenetic relationships among the different strains in Anopheles.

| CON CLUS IONS
Wolbachia has emerged as a biological tool for controlling vectorborne diseases Schmidt et al., 2017). In this study, we demonstrated the natural presence of this endosymbiont bacterium in a large number of Anopheles species, including the five major malaria vectors in Central Africa. Previously, it has been shown that Wolbachia ability to interfere with pathogen transmission depends on the bacterium strain (Blagrove, Arias-Goeta, Failloux, & Sinkins, 2012;Kambris et al., 2010;Walker et al., 2011). Therefore, our results offer the opportunity to determine whether the different Anopheles-infecting Wolbachia strains affect Plasmodium transmission and/or Anopheles reproduction. Indeed, three major vectors of human and nonhuman malaria (An. moucheti, An. nili and An. vinckei) were infected by Wolbachia Paupy et al., 2013). Therefore, we could investigate both Wolbachia-mediated decreases (Hughes, Rivero, & Rasgon, 2014;Zele et al., 2014) and increases (Shaw et al., 2016) in susceptibility of these natural vectors to Plasmodium. Moreover, the strongest effect on suppression of pathogen transmission or reproductive manipulation has been observed in Wolbachia transinfections (Bian et al., 2013;Bian, Xu, Lu, Xie, & Xi, 2010;Blagrove et al., 2012;Hughes et al., 2011;Joubert et al., 2016;Moreira et al., 2009;Walker et al., 2011). Therefore, the availability of Wolbachia strains that infect natural Anopheles populations offers promising opportunities for experimental and theoretical studies in Anopheles, and also in other mosquito families that are vectors of other diseases, including Ae. aegypti and Ae. albopictus.
In conclusion, our findings are merely the "tip of the iceberg" of Wolbachia research in Anopheles. The selection of suitable phenotypes for suppressing Plasmodium transmission and/or manipulating Anopheles reproduction could greatly participate to reduce the malaria burden across the world.

ACK N OWLED G EM ENTS
We thank the "Agence Nationale de la Preservation de la Nature" (ANPN) and the "Centre National de la Recherche Scientifique et Technologique of Gabon" (CENAREST) that authorized this study and facilitated the access to the national parks of La Lopé, Moukalaba-Doudou and Plateaux Batékés.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y
Data for this study are available at the Dryad digital Repository: https ://doi.org/10.5061/dryad.sn81548 (Ayala et al., 2019). DNA sequences of Wolbachia and Anopheles recovered in this study and of those used as references for phylogenetic analyses are submitted at Genbank (MK755460-MK755837).

S U PP O RTI N G I N FO R M ATI O N
Additional supporting information may be found online in the Supporting Information section at the end of the article.