Genomic islands in Wolbachia prophages drive amplification and diversification of cytoplasmic incompatibility genes in Culex pipiens

Abstract

Wolbachia are maternally inherited endosymbiotic bacteria widespread among arthropods. They manipulate their host reproduction to enhance their prevalence in host populations. The most common manipulation is cytoplasmic incompatibility (CI), which causes embryonic death in crosses between infected males and uninfected females, or between individuals carrying incompatible Wolbachia strains. CI patterns are highly complex in the mosquito Culex pipiens, where the causal genes cidA and cidB are amplified and diversified, forming a “cid repertoire” within each Wolbachia wPip genome. Despite their central role in CI, the genomic mechanisms underlying such cid amplification and diversification remain poorly understood. This knowledge gap is largely due to the difficulty of assembling wPip’s genomes due to highly repeated genes and mobile elements, especially in WO prophages. Here, we directly annotated Illumina-corrected Nanopore sequences to investigate the genomic flanking context of cid genes in 3 distinct wPip lineages. We assembled WO prophage regions of substantial length containing the entire cid repertoire previously described in these bacterial lineages. Within these WO regions, cid genes are consistently embedded in modular and rearrangeable islands composed of MutL, rnhA, and small mobile elements, all displaying hyperconserved nucleotide identity across islands. These genomic islands are probably drivers for major rearrangement and recombination events responsible for the amplification and diversification of cid^wPip genes within and between the wPip genomes, leading to CI complexity in C. pipiens.

Wolbachia are widespread bacteria that manipulate arthropod reproduction through cytoplasmic incompatibility (CI). In the mosquito Culex pipiens, crossing patterns are complex due to the diversification of a CI factor gene family called cid. Here, Amoros et al. investigated the genomic drivers of this diversification by assembling and analyzing their genomic environment using Nanopore sequencing. The authors found that cid genes are embedded in highly conserved, modular genomic islands within WO prophages. Their results show that these islands drive cid gene diversification, providing new insights into the genomic basis of CI complexity

Introduction

Wolbachia are highly prevalent endosymbiotic bacteria infecting arthropods and nematodes (Zug and Hammerstein 2012; Weinert et al. 2015). The best-known Wolbachia phenotype is cytoplasmic incompatibility (CI), a form of reproductive parasitism that reduces egg hatching rate when Wolbachia-infected males mate with uninfected females, or between individuals carrying incompatible Wolbachia strains (Yen and Barr 1973; Breeuwer and Werren 1990; O’Neill and Karr 1990; Bordenstein et al. 2001; Duron et al. 2006; Atyame et al. 2014). CI is primarily induced by CI-inducing factor (Cif) proteins encoded by cifA-cifB genes arranged in tandems inside Wolbachia genomes (Beckmann et al. 2017; LePage et al. 2017; Lindsey et al. 2018). To date, 10 phylogenetic groups of cif genes have been identified, named cif types I to X (Lindsey et al. 2018; Martinez et al. 2021; Tan et al. 2024; Amoros et al. 2025). Among them, type I (from wMel and wPip), type III (from wNo), and type IV (from wPip) were functionally proven to induce and rescue CI through transgenic expression in Drosophila (Beckmann et al. 2017; LePage et al. 2017; Chen et al. 2019; Sun et al. 2022). These genes can also be classified according to their functional domains: cif type I genes containing an active deubiquitinase (DUB) domain are functionally classified as cid genes (for CI-inducing DUB), while cif types II, III, and IV genes, containing active nuclease domains (Martinez et al. 2021), are functionally classified as cin (CI-inducing nuclease). Transgenesis experiments in Drosophila and yeast cells demonstrated that the B genes encode toxins preventing the normal cell culture growth, while the A genes encode antidotes which restore normal growth (Beckmann et al. 2017; Chen et al. 2019; Horard et al. 2022; Sun et al. 2022; Terretaz et al. 2023; Namias et al. 2025).

In the mosquitoes of the Culex pipiens complex, in which > 99% individuals are infected with wPip Wolbachia (Klasson et al. 2008; Dumas et al. 2013), crossing between dozens of lineages revealed highly complex CI patterns, including both unidirectional and bidirectional incompatibilities (Laven 1967; Duron et al. 2006; Atyame et al. 2014) (Supplementary Table 1). A Multilocus Sequence Typing (MLST) approach, specific to wPip, identified 5 distinct clades named group wPip-I to wPip-V (Atyame et al. 2011; Dumas et al. 2013) (Fig. 1a), which strongly correlate with CI patterns. Indeed, mosquitoes infected with wPip from the same group are generally compatible, while compatibility in crosses between individuals infected with distinct groups is less predictable (Atyame et al. 2014). These results, along with backcross experiments, demonstrated that genetic variations in wPip, but not in the host genetic background, are responsible for CI patterns’ complexity in C. pipiens (Duron et al. 2005; Atyame et al. 2011, 2014; Sicard et al. 2021). Before the discovery of cif genes, both theoretical and experimental studies had concluded that each wPip genome must carry multiple and polymorphic genes responsible for CI to explain complex CI patterns in C. pipiens (Nor et al. 2013; Atyame et al. 2014). Since then, cid^wPip (i.e. type I cif) genes were indeed found to be amplified and diversified in all wPip genomes studied and are now recognized as the main drivers of the Culex-CI complexity (Bonneau et al. 2018a, 2019; Sicard et al. 2021; Namias et al. 2024a, 2025) (Table 1). Each wPip genome carries up to 6 distinct cidA–B pairs, with 2 forms of cidB^wPip co-occurring: (i) a full-length cidB^wPip variant of around 3.5 kb, bearing a DUB domain and demonstrated as key for CI phenotypes (Beckmann et al. 2017; Beckmann et al. 2019a; Horard et al. 2022; Namias et al. 2025), and (ii) several truncated cidB^wPip variants, lacking part or all of the DUB domain that are less stable (Namias et al. 2025). This arsenal of cid^wPip genes, composed of distinct cidA^wPip and cidB^wPip variants in 1 genome, was called a “cid repertoire” (Bonneau et al. 2018a; Namias et al. 2024a, 2025). In addition to these cidA–B^wPip pairs, all the wPip genomes studied so far present cif genes from the type IV, called cinA–B^wPip. As these cin genes are present in a single copy and exhibit no polymorphism between Culex lines, their implication in CI complexity in Culex was ruled out (Bonneau et al. 2018a; Martinez et al. 2021) (Table 1).

Fig. 1.

WO prophages from publicly available wPip genomes (wPip-Pel, JHB, and wPip Mol). a) Maximum likelihood phylogeny of the 5 wPip groups (I to V) based on 5 genes (pk1, pk2, MutL, GP12, and GP15) using 1,000 replicates (HKY + I + G4 substitution model, 4,012 nucleotides). Only bootstrap values greater than 60 are shown. Red stars indicate the 3 available public wPip draft genomes, while black stars represent the wPip genomic data analyzed in this study. b) Genomic location of WO prophages and cif pairs (cid and cin) on the wPip-Pel, JHB, and wPip-Mol chromosomes. The dark grey areas correspond to WO prophages, with the cid pairs highlighted in pink, and the cin pairs in black. The dotted lines indicate assembly gaps. For wPip-JHB and wPip-Mol, WO prophages are located in unassembled contigs. c) Detailed (re)annotations of the WO prophages in the 3 public wPip genomes. The conventional separation between WOPip1 and WOPip2 is marked by a black arrowhead, as these 2 prophages occur sequentially in the chromosome. DIcidB refers to the isolated cidB fragments containing a deubiquitinase domain in wPip-Pel (WP1290 to 1291). The length of each sequence in kb is indicated at the end of each prophage region.

Table 1.

Overview of previous data obtained on cif genes in the wPip genomes analyzed in this study.

wPip name Accession number	wPip group	Mosquito host species	Geographical origin of host Culex mosquito	Mosquito lineage type	Estimated number of cidA/cidB operon per genome	Estimated number of different cidA/cidB variants per genome	Estimated number of cinA/cinB operon per genome	Estimated number of different cinA/cinB variants per genome
Pel AM999887.1	wPip-I	Culex pipiens quinquefasciatus	Sri Lanka	Colony	1/1a	1/1a	1/1a	1/1a
JHB NZ_ABZA00000000.1	wPip-I	Culex pipiens quinquefasciatus	South Africa	Unknown	2/∼3a	2/∼3a	1/1a	1/1a
Tunis	wPip-I	Culex pipiens pipiens	Tunisia	Isofemale line	6/6b	3–4/3–5c*	1/1d	1/1d
Slab	wPip-III	Culex pipiens quinquefasciatus	United States	Isofemale line	5/5b	3–4/2c*	—	—
Mol NZ_CTEH00000000.1	wPip-IV	Culex pipiens molestus	China	Unknown	0/0a	0/0a	1/1a	1/1a
Harash	wPip-IV	Culex pipiens pipiens	Algeria	Isofemale line	—	4/3c	—	1/1d

^aAssembled draft genomes (Klasson et al. 2008; Salzberg et al. 2009; Pinto et al. 2013).

^bNumber quantified through qPCR, taking the wsp monocopy gene as reference (Bonneau et al. 2018b).

^cDiversity assessed using Nanopore sequencing of cidA and cidB PCR amplicons on a single individual (Namias et al. 2025).

^dAnalysis of depth of coverage and polymorphism by mapping cinA/cinB Illumina reads on wPip-Pel genome (Bonneau et al. 2018a);—missing information, ∼ partial information.

^*Number depending on the individual analyzed.

Altogether, these findings significantly advanced our understanding of the genetic determinism of unprecedented CI complexity induced by wPip in Culex. However, the diversity of available wPip genomes is currently too scarce to allow the study of genomic drivers that could explain such unprecedented diversification of cid genes between and within wPip groups. To date, 3 wPip draft genomes are currently available on public databases (wPip-Pel, JHB, and Mol) (Klasson et al. 2008; Salzberg et al. 2009; Pinto et al. 2013), Pel being the most extensively studied and the currently accepted reference genome, since it is made of a single non-circularized contig, and Mol the most incomplete and fragmented one (Fig. 1b). Since both wPip-Pel and JHB belong to the wPip-I group (Fig. 1a), these genomes only capture a limited part of wPip group diversity. Moreover, cid^wPip gene copy numbers and diversity reported in these genomes do not match recent genetic datasets acquired on a broad range of wPip (Bonneau et al. 2018a; Namias et al. 2023, 2024a, 2025) (Table 1). Indeed, the wPip-Pel genome only contains a full-length cidB^wPip copy and no truncated ones while wPip-JHB has only 2 truncated cidB^wPip copies (Fig. 1b). Strikingly, cid^wPip genes are entirely absent from wPip-Mol contigs (Fig. 1b). This apparent discrepancy between recent cif^wPip genetic data (Bonneau et al. 2018a; Namias et al. 2023, 2024a, 2025) and the 3 wPip draft genomes (Klasson et al. 2008; Salzberg et al. 2009; Pinto et al. 2013) can be explained by the fact that none of these 3 genomes are fully circularized. This difficulty in robustly assembling wPip genomes results from the highly repetitive nature of WO prophage genes (Bordenstein and Wernegreen 2004; Bordenstein and Bordenstein 2022; Vancaester and Blaxter 2023), which represent 20% of wPip-Pel and wPip-JHB genomes (Klasson et al. 2008; Salzberg et al. 2009) (Fig. 1b and c).

WO prophages are basically composed of both core phage genes and Eukaryotic Association Module (EAM) genes (Bordenstein and Bordenstein 2016, 2022) (Fig. 1b and c). They have been proposed to act as genomic hotspots that facilitate horizontal gene transfer among Wolbachia strains (Bordenstein and Wernegreen 2004; Kent et al. 2011; Bordenstein and Bordenstein 2016, 2022). In the 3 available wPip draft genomes, both cid^wPip and cin^wPip genes are located within WO prophage regions (Bordenstein and Bordenstein 2022) (Fig. 1c). If such locations suggest that WO prophages might be involved in cid^wPip gene extreme amplification and diversification within and between wPip genomes, the genomic mechanism(s) at play is not yet understood. A first putative mechanism is phage transduction, often considered the most straightforward explanation of horizontal transfer between Wolbachia genomes (Bordenstein and Reznikoff 2005; Kent et al. 2011; Bordenstein and Bordenstein 2016). While phage particles could theoretically promote cid^wPip transfer between wPip genomes, to date, the genomes in virus particles attributed to WO phages have only been sequenced in wVitA-infected Nasonia giraulti and wSoc-infected Allonemobius socius, and none of their genomes contained the cif genes (Bordenstein and Bordenstein 2016; Kupritz et al. 2021). An alternative mechanism to transduction is that the many gene repeats in WO prophages affect the cid gene copy number and diversity by recombination and rearrangement during DNA replication (Slack et al. 2006; Lovett 2017). In addition, small mobile genetic elements, such as transposons and retrotransposons, within WO prophages have also been pointed out as major sources of Wolbachia genomic diversification (Cerveau et al. 2011; Leclercq et al. 2011; Cooper et al. 2019; Tan et al. 2024; Oswalt et al. 2025).

Here, by directly annotating Illumina-corrected Nanopore sequences, we investigated the immediate genomic environments of cif^wPip genes (both cid^wPip and cin^wPip) and the diversity of these environments among 3 different wPip from 3 distinct groups, known to induce distinct CI patterns in their mosquito hosts: wPip-I Tunis, wPip-III Slab, and wPip-IV Harash (Atyame et al. 2014; Sicard et al. 2021) (Supplementary Table 1). We found that cid^wPip genes reside in modular and rearrangeable genomic islands within highly variable WO prophages, while the cin^wPip genes are found in WO syntenic regions among the different wPip. The cid^wPip genomic islands harbor multicopy genes involved in horizontal transfer, recombination, and DNA repair, which display hyperconserved nucleotide identity. These genomic islands are probably drivers for major rearrangement and recombination events responsible for the amplification and diversification of cid^wPip genes within and between the wPip genomes, leading to CI complexity in C. pipiens.

Materials and methods

Nanopore and Illumina sequencing of mosquito lines

Three mosquito isofemale lines, Tunis, Slab, and Harash were used in this study (Table 1). Each isofemale line was founded with the progeny of a single egg raft from a single female. Mosquitoes were reared in 65 dm³ screened cages housed in a single room maintained at 25 °C with 70% relative humidity and a 12:12 light–dark cycle. Larvae were fed with a mixture of shrimp powder and rabbit pellets, while adults were provided with a honey solution. Females were blood-fed on turkey blood using a Hemotek membrane feeding system (Discovery Workshops, UK) to allow egg production. Each mosquito line is infected with a single wPip type, previously characterized by specific MLST (Atyame et al. 2011, 2014; Dumas et al. 2013): Tunis (wPip-I), Slab (wPip-III), or Harash (wPip-IV) (Fig. 1a). Every 3 months, to check for contamination, a pool of a hundred larvae per line was collected, the DNA extracted, and the wPip group determined using the PK1 marker, as described by Namias et al. (2024b).

High molecular weight genomic DNA was extracted from 10 to 20 females for each isofemale line. For the Harash and Tunis lines, DNA was extracted using the Qiagen Genomic-tip 20G kit following the manufacturer's protocol for insects (Namias et al. 2025), while for the Slab isofemale line, DNA was extracted using the phenol-chloroform extraction protocol with no vortexing to avoid DNA shearing. All DNA libraries were prepared using the Ligation Sequencing Kit (SQK-LSK109) and Native Barcoding Expansion Kit (EXP-NBD104), and then sequenced on a Minion flow cell with R9.4.1 or R10.4.1 chemistries. Short read DNA libraries were prepared in accordance with the Illumina DNA Prep Library guide and sequenced in NextSeq 550 Flowcell using the Illumina chemistry V2, 2 × 150 bp paired-end run. Reads from the mosquito host were retrieved from the dataset by mapping the C. pipiens reference genomes “GCA_041146695.1” and “GCA_015732765.1” on the trimmed long reads with Minimap2 v2.24 (Li 2018). Nanopore longreads only and hybrids (Illumina + Nanopore reads) de novo assemblies were attempted to obtain circular chromosomes using Flye v2.9.6 (Kolmogorov et al. 2019) and Unicycler v0.5.1 (Wick et al. 2017). Assembly graphs were inspected with Bandage v0.8.1.

Wolbachia sequences with cif genes and identification of cidA–cidB variants

Given the inability to fully assemble a circular genome with help of full genomic data (Illumina + Nanopore reads; Supplementary Tables 2 and 3), particularly in prophage regions (Supplementary Fig. 1), we proceeded with a manual inspection and annotation of a subset of the long read sequences which were corrected with Illumina reads (2 × 150 bp) using Ratatosk v0.9.0 (Holley et al. 2021). We identified the Wolbachia sequences containing cif^wPip genes (cid or cin) by mapping them, respectively, against cid and cin sequences from wPip-Pel reference genomes using Minimap2 v2.24 (Li 2018). Then, only a subset of sequences ≥ 10 kb containing cif genes was retained for further analyses to generate contigs (Supplementary Table 4). Cif gene types (from I to X) were determined using the methodology described by Amoros et al. (2025). In sequences above 10 kb containing the cid genes, the cidA and cidB variants were characterized based on the cid references and nomenclature established by Namias et al. (2023) and Namias et al. (2025). To achieve this, sequences ≥ 10 kb were independently aligned to cidA and cidB sequences from Namias et al. (2025) using MAFFT v7.505 (Katoh and Standley 2013).

Annotations and assembly of cif gene environments in wPip

Individual annotations of sequences containing cif^wPip genes (cid or cin) and ≥ 10 kb were performed using BLASTn (megablast program) against the wPip-Pel draft genome (accession number: AM999887.1) using version 6.9 of the NCBI Prokaryotic Genome Annotation Pipeline (PGAP). Only nucleotide identity above 75% with wPip-Pel was selected. Gene functions obtained from the wPip-Pel genome were then updated according to the new WO prophage gene annotations proposed by Bordenstein and Bordenstein (2022). To classify transposases, we conducted BLAST searches against ISfinder (https://isfinder.biotoul.fr/blast.php; Siguier et al. 2006) and ran HHpred (Zimmermann et al. 2018) with the SCOPe70 v.2.08, Pfam-A v37.0, SMART v6.0, and COG/KOG v1.0 databases. Group II intron retrotransposons (maturases) were identified using Zbase (Candales et al. 2012). Then, the sequences sharing identical cidA–cidB variants and exhibiting the same annotated gene composition (with ≥ 99% nucleotide sequence identity) were aligned using MAFFT and manually assembled to form contigs (Supplementary Fig. 2; Supplementary Table 4). The quality of each contig was assessed by mapping the raw Nanopore and Illumina reads to estimate coverage using Minimap2 v2.24 and SAMtools v1.9 (Supplementary Tables 2 and 3; Supplementary Fig. 3).

Comparison of new WO prophage regions with reference ones

The gene compositions of newly identified WO prophages in draft wPip genomes JHB and Mol, as well as those carrying cif genes in Tunis, Slab, and Harash, were compared to the gene compositions of the 5 reference prophages of wPip-Pel (WOPip1 to WOPip5; Fig. 1c). Each reference prophage displays distinctive specific genes that enable their identification: (i) WOPip1 is characterized by the BRO toxin (annotated as WP0260), replication/repair genes located near an IS630 and patatin-like putative lysis genes (Fig. 1c), (ii) WOPip2 by long genes containing ankyrin repeats and a helicase gene (WP0319), and (iii) WOPip3 is the only one containing both ResIII and calponin-like genes (WP0350 and WP0351). Both WOPip4 and WOPip5 share structural modules such as tail and tail fiber genes, but differ in accessory content: WOPip4 presents an RDD transporter (WP0409), a membrane protein (WP0428), a RhuM toxin (WP0431), an acetyltransferase (WP0463 to 0464), and a LigA gene (WP0465), whereas WOPip5 harbors a COG5558 transposase (WP1294), WO-PC1 genes (WP1319 to 1320), and the octomom-like module (WP1341 to 1350). The presence/absence of these specific genes, as well as the full gene composition of the reference prophages, were used to characterize the newly obtained WO prophage regions.

Phylogenetic and recombination analyses

All phylogenetic analyses in this study were conducted using the maximum likelihood method with 1,000 bootstrap replicates using raxml-ng v1.1.0 (Stamatakis 2014). The best-fit evolutionary model was selected using modeltest-ng v0.1.7 (Darriba et al. 2020) based on the corrected Akaike information criterion. All network analyses were generated using the Neighbour-net method, and formal tests of recombination were processed using the PHI-test implemented in SplitsTree v4.19.2 (Huson and Bryant 2024). Sequence alignments were carried out using MAFFT v7.505 (Katoh and Standley 2013), implemented in Unipro UGENE v52.0 (Okonechnikov et al. 2012). The wPip phylogeny was inferred from 5 multilocus genes: pk1, pk2, MutL, GP12, and GP15 (Fig. 1a) (Atyame et al. 2011). To identify the srWO groups of wPip prophages, phylogenetic and networking analyses based on the amino acid sequences of the large serine recombinase were performed as defined by Bordenstein and Bordenstein (2022). Additionally, 7 widely distributed core WO prophage proteins, including structural, replication, and modification functions, were analysed by phylogenetic and network analyses using the same methods as described above. These included the Minor capsid protein (WP0252), Terminase large subunit GpA (WP0255), Tail tape measure protein (WP0448), Tail fiber protein (WP1325), Baseplate protein GpV (WP0246), RepA (WP0433), and Methylase (WP0258).

Results

The contigs with cif genes represent up to 20% of full wPip genome

The Nanopore long reads obtained from wPip-I Tunis, wPip-III Slab, and wPip-IV Harash lines did not allow the de novo assembly of Wolbachia genomes (Supplementary Fig. 1; Supplementary Tables 2 and 3). Moreover, cif genes were mostly located on fragmented contigs, with an important difference between the 2 assembly methods, resulting in prophage regions for which sequence reliability remains very limited. To circumvent difficulties associated with the assembly of repetitive regions, we manually assembled the genomic regions flanking cif genes using a subset of long reads corrected with Illumina sequences. Reads ≥ 10 kb harboring cif genes from wPip-I Tunis, wPip-III Slab, and wPip-IV Harash (Supplementary Table 4) were individually annotated. Sequences with the same gene order and the same cif variant were aligned to allow assembly of contigs harboring cif genes for each wPip line. With this protocol, we reconstructed 4 different contigs for wPip-I Tunis (named Tunisp-a to d, p for prophage), 3 different contigs for wPip-III Slab (named Slabp-a to c), and 4 different contigs for wPip-IV Harash (named Harashp-a to d) (Fig. 2; Supplementary Fig. 2; Supplementary Table 4). Altogether, the contigs represent approximately 200 kb for Tunis, 150 kb for Slab, and 300 kb for Harash, corresponding to 13% to 20% of the wPip genome, which is estimated to be ∼1.5 Mbp. Mapping of raw Nanopore and Illumina reads onto the manually reconstructed contigs showed that, for Tunis and Harash, the 2 lineages for which a substantial proportion of Wolbachia reads was obtained, these regions were covered at a depth equal to or greater than the expected Wolbachia coverage (Supplementary Tables 2 and 3; Supplementary Fig. 3). Some contigs exhibited coverage levels 2 to 3 times higher than the expected depth, consistent with the presence of repetitive genes or repeated regions. In Harash, no major uneven coverage was detected, except for locally increased coverage over several highly repetitive genes. In Tunis, only a small region at the end of Tunisp-b displayed a drop in Nanopore read coverage that was not observed with Illumina reads. Only Tunisp-d contig showed very uneven coverage, suggesting that it may either represent a sequencing artifact or correspond to a low-frequency variant present in only a small subset of clones within this isofemale line. In contrast, coverage in Slab was lower and more uneven, likely due to the reduced proportion of Wolbachia reads.

Fig. 2.

Genomic environments of the cif pairs in wPip-I Tunis, wPip-III Slab, and wPip-IV Harash lines. a) Detailed gene annotations of the immediate genomic environments of the cid and cin pairs in wPip-I Tunis, wPip-III Slab, and wPip-IV Harash. The names of the different cidA and cidB variants are indicated below each corresponding gene. DIcidB refers to the isolated cidB fragments containing a deubiquitinase domain in wPip-Pel (WP1290 to 1291). b) The cidA and cidB variants previously identified in the cid repertoires of wPip-I Tunis, wPip-III Slab, and wPip-IV Harash lines, as described by Namias et al. (2025), with the help of Nanopore sequencing of PCR amplicons and their corresponding code name used in panel (a). The length of each sequence in kb is indicated at the end of each prophage region.

Distinct cif repertoires associated with different CI phenotypes are recovered in our contigs

Each contig obtained from the genomes of wPip-I Tunis, wPip-III Slab, and wPip-IV Harash includes 1 or 2 tandems of cidA–B variants (cif type I) or 1 tandem of monomorphic cinA–B (cif type IV). No other types of cif were detected. Cid variants identification with help of references from Namias et al. (2025) showed that the new contigs contained all the different cidA and cidB variants previously described in wPip-I Tunis, wPip-III Slab, and wPip-IV Harash by Nanopore sequencing of polymerase chain reaction (PCR) products (Namias et al. 2025) (Fig. 2). For example, in Harash lineage, we successfully recovered the variants cidA-IV-alpha(5)-1, cidA-IV-alpha(5)-2, cidA-IV-beta(1)-1, cidA-IV-beta(1)-2, cidB-IV-a-1, cidB-IV-b-1, and cidB-IV-b-3 (Fig. 2). The total number of cid variants found in our contigs also perfectly matched the total number of variants in these lines as investigated by quantitative PCR (Table 1). Indeed, 6 cidA–cidB tandems were identified in the genomes of wPip-I Tunis, 5 tandems in wPip-III Slab and wPip-IV Harash, while 1 cinA–cinB tandem was present in all genomes (Table 1). This suggests that our method captured all the genomic regions that contain cif genes in the wPip genomes studied (Fig. 2). As previously described by Namias et al. (2025), we confirmed that only 1 cidB variant per genome was in a full-length form (i.e. including a full DUB domain). However, the same full-length variant was found in 2 copies in Tunis genome, each located in distinct contigs (Fig. 2). All the other cidB variants were truncated (i.e. without a region including the DUB domain) and located within a palindromic organization (cidA–cidB–cidB–cidA) (Fig. 2). As it has been demonstrated that full-length cidB genes are the main effectors of CI diversity (Namias et al. 2025), the fact that our 3 wPip lines harbor distinct full-length cidB variants (Fig. 2) and exhibit 3 different crossing (Supplementary Table 1) is consistent with this finding.

All the cif genes are located within WO prophage regions in wPip

Annotations of the contigs demonstrated that the different cid variants and the monomorphic cin genes are all located in WO prophages (Fig. 2). Indeed, genomic regions flanking cif were always composed of (i) core phage genes (e.g. head, tail, and replication genes), (ii) WO-protein clusters 1 and 2 (WO-PC1 and WO-PC2), gene clusters with unknown function that was formalized as WO prophage markers by Bordenstein and Bordenstein (2022), and/or (iii) EAM, that include genes with putative eukaryotic origins that are frequently described within WO prophages (Bordenstein and Bordenstein 2016, 2022; Vancaester and Blaxter 2023; Grève et al. 2024) (Fig. 2). The total WO prophage regions hosting cif genes represent a significant part of the 3 wPip genomes investigated here. As cumulative WO prophage regions in both wPip-Pel and wPip-JHB represent ∼20% of the complete genomes, and prophage regions in the new contigs containing cif represent 13% to 19% of the expected wPip genome size (∼1.5 Mb), our data likely encompass a substantial fraction of the WO prophage regions. The phylogeny of the large serine recombinase, previously proposed as a “typing tool” to discriminate clades (Sr1 to Sr4WO) within the Wovirus genus (Bordenstein and Bordenstein 2022), reveals that all WO prophages of wPip belong to the sr3WO clade (Supplementary Figs. 4a and 5a). This suggests that they are phylogenetically closely related and have recently diverged. However, the analysis of other WO prophage markers reveals distinct evolutionary histories for each gene examined, with genes from wPip sometimes clustering with genes of WO prophages from other Wolbachia strains (Supplementary Figs. 4b–h and 5b–h). PHI tests for recombination confirm that many genes in wPip WO prophages are recombination hotspots (Supplementary Table 5).

High diversity of WO prophages hosting cid genes

For each of the 3 phylogenetically distant wPip analyzed here, at least 3 different prophage regions containing cid genes were recovered per isofemale line (Fig. 3). This result clearly differs from the wPip-Pel reference genome, in which only the WOPip1 prophage contains a cid tandem (Fig. 1b and c). Wolbachia in the Tunis isofemale line (wPip-I), shows a full-length cid version in a WOPip1-like region (Tunisp-a-b) with well-conserved gene composition (Fig. 3). Full-length cid genes in Slab and Harash (respectively, in wPip-III and IV groups; Fig. 1a) were found in mosaic prophage regions, exhibiting gene compositions which could not be assigned to any previously described WO prophage from wPip-Pel (Fig. 3). In Tunis, Slab, and Harash, the cid-truncated palindromic variants were all identified in mosaic prophage regions (Fig. 3). All these prophage regions exhibit different genomic architectures (Fig. 3) showing a greater diversity of WO prophages in wPip than previously suspected. The new in-depth annotation of transposases reveals their high diversity within WO prophages of wPip (Figs. 1c and 2a). Interestingly, some cid variants are located in prophagic environments (Slabp-b and Harashp-a) that contain the full arsenal of genes necessary to encode a complete phage particle, including head, tail, tail fiber, recombinase, putative patatin lysis, baseplate-connector, and replication-repair functions (Bordenstein and Bordenstein 2022) (Fig. 2).

Fig. 3.

Diversity of WO prophage harboring cif genes among different wPip genomes. This figure was built by assigning each portion of the newly annotated prophage to WOPip1, WOPip2, WOPip3, WOPip4, and WOPip5 references from wPip-Pel. The cid genomic islands and the type of cidB (long or truncated) are symbolized to allow their quick identification.

Highly conserved WOPip2-like environment for cin genes

In all wPip genomes yet studied, there is a single unique cin pair per genome consistently located in the same syntenic prophagic region composed of genes enriched in ankyrin repeat domains (WP0292, WP0293, and WP0296), a large serine recombinase (WP0297) and a putative gene composed of a domain with unknown function (DUF) (WP0298) (Figs. 1c, 2, and  4a). Such prophagic environments, a part of WOPip2 in wPip-Pel, are conserved both in the 3 genomes investigated here and previously published draft genomes (wPip-Pel, wPip-JHB, and wPip-Mol). If the cinA and cinB genes were found perfectly conserved between all the wPip genomes, nucleotide polymorphism is present in some downstream cin neighboring genes (Fig. 4a).

Fig. 4.

Gene nucleotide identity among genes from the a) conserved cin environment and b) among genes from the cid genomic islands from different wPip.

The “cid genomic islands”: hotspots of recombination for cid genes

What constitutes the “cid genomic islands”?

Four genes were found associated with cid among the different prophage regions annotated in this study: (i) a DNA repair gene (MutL, WP0278), (ii) an RNAse (rnhA, WP0279), (iii) a maturase required for processing of group II intron-retroelements (intII, WP0280), and (iv) a PDDEXK2 transposase composed of a PDDE(X/K) nuclease-like domain (WP0281) (Figs. 4b and  5a). These 4 genes and a cid pair form repeated genomic units that we called “cid genomic islands.” Indeed, these islands are present in multiple copies in distinct WO prophage regions, contrary to what was previously established in the wPip-Pel reference genome, where they were only present in WOPip1 (Figs. 1c and 3).

Fig. 5.

Modularity of the wPip cid genomic islands and WO-PC2. a) Gene compositions and variation of the 20 wPip cid genomic islands analyzed in this study. The proposed boundaries of the wPip cid genomic islands are highlighted with a gray background. The (1) positioned after the name of some wPip islands indicates that they belong to the left wPip genomic island of a palindrome, while the (2) indicates the wPip cid genomic islands that belong to the right one. Gene codes from the wPip-Pel reference genome were used to label WO-PC2 genes (WP0275, WP0276, and WP0277). Polymorphic variants in WP0275, WP0276, WP0277, and MutL are presented using different colors and motifs. Dashed black lines indicate the end of a scaffold. b) Occurrence of different alleles in the cid genomic islands and WO-PC2.

Modularity of gene composition among cid genomic islands.

By comparing the gene composition of these islands, we described presence/absence patterns for certain genes, with the exception of the PDDEXK2 transposase, which is the only gene always found associated with cidA–cidB pairs (Fig. 5a and b). Otherwise, the variation in the cid genomic islands gene composition mostly concerned the MutL–rnhA pair that is absent in some islands (Fig. 5a). The group II intron maturase is also occasionally missing (Fig. 5a). These recurrent patterns of gene loss and retention suggest that the cid genomic islands behave as modular and rearrangeable mobile elements, shaped by genomic rearrangement. At the genome scale, all wPip isofemale lines (wPip-I Tunis, wPip-III Slab, and wPip-IV Harash) present different versions of cid genomic islands (Fig. 5a), but different wPip lines can harbor genomic islands with the same gene composition (e.g. Slabp-a, Harashp-b, and Harashp-c) (Fig. 5a). Moreover, both full-length and truncated forms of cidB can be found within cid genomic islands with identical gene composition, as, for instance, in Tunis with Tunisp-a, Tunisp-b, and Tunisp-d(1) (Fig. 5a).

High nucleotide identity of genes composing cid genomic islands.

Contrary to cidA–cidB genes that are highly polymorphic among wPip lineages (Tunis, Slab, and Harash; Fig. 2b) due to block recombination leading to different variants (Bonneau et al. 2018a; Namias et al. 2024a, 2025) (Supplementary Table 5; Supplementary Fig. 6), the other genes of the cid genomic islands are highly conserved (Fig. 4; Supplementary Table 5; Supplementary Fig. 7). Notably, the ubiquitous PDDEXK2 transposase (1,010 bp), always found associated with cid tandems, is monomorphic among all cid genomic islands (Figs. 4 and 5; Supplementary Table 5; Supplementary Fig. 7a). When present, group II intron maturases (1,344 bp) and rnhA (558 bp) are also strictly monomorphic (Figs. 4 and 5; Supplementary Table 5; Supplementary Fig. 7b and c), as well as intergenic regions which share > 97% similarity between cid genomic islands. For MutL, there is 99% identity between all islands for the first 921 bp, also called MutL C-terminal (Figs. 4 and 5; Supplementary Table 5; Supplementary Fig. 7d). The following 1,845 bp of the gene, MutL N-terminal, exhibited polymorphism as previously detected and used for wPip specific MLST (Atyame et al. 2011) (Supplementary Table 5; Supplementary Fig. 7d). Depending on the versions of cid genomic islands, the strictly conserved sequence lengths range from approximately 2,700 bp (e.g. when only the group II intron maturase and PDDEXK2 transposase are present, as seen in Tunisp-c, JHBp-d, and Slabp-b) to around 5,300 bp (when MutL, rnhA, group II intron maturase, and PDDEXK2 transposase are all present). For truncated versions of cidB, the most common cidB versions, 2 palindromic cid genomic islands face each other. Consequently, the strictly conserved sequence of the 2 grouped islands flanking the cidA–cidB–cidB–cidA genes (Namias et al. 2025) ranges from 3,380 bp downstream (for JHBp-d, Tunisp-c, and Slabp-b) to 5,306 bp flanking both sides (for Harasp-b and Slabp-a) (Fig. 4b). Genes from the WO-PC2 region, originally identified in prophages WOPip2, WOPip4, and WOPip5 in wPip-Pel, were consistently found immediately upstream of cid genomic islands. Although an additional ISWpi12 element is occasionally present, WO-PC2 is conserved in gene composition across prophage regions. However, the WO-PC2 genes are highly polymorphic in nucleotide identity (Fig. 5b). Indeed, 2 variants were identified for both WP0277 and WP0276, and 9 variants were found for WP0275 (Fig. 5b). Because of this polymorphism, we decided not to include them in the cid genomic islands composed of highly conserved genes. For genes located in WO prophages but outside the cid genomic islands and WO-PC2, the sequence similarity ranged from 68% to 92% depending on the gene analyzed (Supplementary Table 5). PHI tests for recombination show that some genes exhibit strong recombination rates, whereas others do not (Supplementary Table 5; Supplementary Fig. 5), suggesting that gene rearrangements in WO prophages are frequent, dynamic, and of diverse nature, including both recombination and/or horizontal transfer, within which the cid genomic islands persist with a singular nucleotide conservation.

The flanking regions of the cid genes in wNaev strengthen the hypothesis of cid horizontal transfer between distant Wolbachia

The wNaev genome from the lepidopteran Rhopobota naevana (OX366333.1) was also analyzed, as it was proven to contain multiple cid copies, including 1 cidA–cidB pair suspected to have laterally transferred either from or into a wPip ancestral genome (Namias et al. 2025). Five WO prophage regions were annotated, containing 10 cifA–cifB pairs from different types, including 4 cid pairs (type I) and 1 cin (type IV) (Supplementary Fig. 8). Two of the annotated prophage regions, wNaev region 2 and region 3, harbor large serine recombinases that belong to the sr3WO clade along with wPip prophages (Supplementary Figs. 4a and 5a). Because wNaev harbors a cidA–cidB pair closely related to cidA–cidB pairs from wPip (Supplementary Fig. 6), Namias et al. (2025) hypothesized a lateral transfer between those 2 Wolbachia ancestral genomes. Here, we show that in wNaev, not only region 2 but also region 1 (Supplementary Fig. 8b) contains cid pairs closely related to those of wPip (Supplementary Fig. 6). Strikingly, these 2 cidA–cidB pairs in wNaev are also flanked by 1 to 2 hallmark genes of wPip cid genomic islands: the PDDEXK2 transposase and the group II intron maturase strengthening a horizontal transfer scenario of cid pair along with these genes (Supplementary Fig. 8b). Phylogenetic analysis of these 2 flanking genes showed that the ones from wNaev region 2 perfectly cluster in the same phylogenetic clade, with, respectively, 99% and 100% nucleotide identity, with the PDDEXK2 and group II intron maturase found in wPip cid genomic islands (Supplementary Fig. 7a and b). This result strengthens the hypothesis of recent horizontal transfer of a cid tandem along with putatively other genes of a cid genomic island.

Discussion

Wolbachia wPip genomes contain 2 major types of CI-inducing factors (cif): the cid^wPip (also referred to as type I cif), whose B gene displays a DUB domain, and the cin^wPip (also referred to as type IV cif), whose B gene encodes active nuclease sites. Transgenesis of both cid^wPip and cin^wPip operons in Drosophila can induce strong CI (Beckmann et al. 2017; Chen et al. 2019; Beckmann et al. 2019a; Sun et al. 2022). Since both cinA and cinB are completely monomorphic among wPip, their involvement in the diversity of incompatibility patterns between Culex lineages was previously ruled out (Bonneau et al. 2018a). On the contrary, cid pairs are polymorphic within and between wPip genomes (Bonneau et al. 2018a, 2019; Sicard et al. 2021; Namias et al. 2024a, 2025). Targeted genetic approaches focusing on cid genes, based on Nanopore amplicon sequencing (Namias et al. 2023), have previously shown that each wPip genome harbors a full-length variant (with intact terminal DUB domain) and several truncated forms depleted of their terminal regions (Namias et al. 2025). The combined results from heterologous expression studies (in yeasts, Drosophila cells, and Drosophila insects) (Beckmann et al. 2017; Horard et al. 2022; Wang et al. 2022; Terretaz et al. 2023) and correlative studies in natural situations (i.e. matching cid diversity with Culex crossing experiments) (Bonneau et al. 2018a; Bonneau et al. 2019; Namias et al. 2025) provide an explanation for the diversity of CI in C. pipiens within a toxin-antidote model (Beckmann et al. 2019b). In this model, the variability in full-length CidB variants (the toxins) primarily accounts for the variability in modification (mod) types, while variability in CidA variants (the antidotes) explains the variability in rescue (resc) capacities (Namias et al. 2024a, 2025).

While genetic studies of cid effectors have clarified their central role in the evolution of CI diversity in C. pipiens, the genomic context underlying their amplification and diversification has yet to be explored. In this study, we focused on 3 wPip lineages that induce distinct crossing compatibility patterns and whose cid repertoires have already been fully characterized (Supplementary Table 1) (Atyame et al. 2014; Sicard et al. 2021; Namias et al. 2025). Using long-read sequencing, we failed to reconstruct full, reliable wPip genomes but managed to robustly reconstruct the genomic regions flanking each previously characterized cid variant. Annotation of these contigs enabled us to investigate the genomic drivers underlying cid amplification and diversification. Moreover, such data allowed us to question the difference in polymorphism between the 2 types of cif effectors in wPip, cid^wPip and cin^wPip, which are present in all genomes but exhibit distinct molecular evolution patterns: cinA–B^wPip is found in a single monomorphic copy vs cidA–B^wPip in multiple (up to 6 different) recombinant copies per genome.

Incomplete wPip genome assemblies despite high read coverage

Even though we obtained Wolbachia Nanopore read coverage depths exceeding 40× for 2 of our lines (Harash and Tunis), attempts to assemble complete wPip genomes using standard bioinformatic pipelines were unsuccessful. While these approaches enabled the reconstruction of substantial portions of the genomes, assemblies consistently failed to circularize because they systematically collapsed or broke within prophage regions. These results underscore the exceptional difficulty of assembling wPip genomes. Indeed, wPip genomes harbor prophage regions composed of highly repetitive and strongly conserved sequences that remain challenging to resolve using conventional assembly strategies. In addition, prophages may preferentially accumulate opposite the chromosomal origin of replication, potentially forming a prophage-rich hotspot (Bordenstein and Bordenstein 2022). The presence of multiple adjacent prophages carrying repeated and conserved genes likely contributes to the persistent lack of robustly assembled wPip genomes to date. Future improvements in sequencing technologies and assembly methods, particularly the combination of Nanopore ultra-long reads with highly accurate PacBio long reads, may ultimately enable the complete circularization of wPip genomes. However, because the primary objective of the present study was to investigate the immediate genomic environment of amplified and diversified cif genes within and among wPip genomes, we adopted a manual curation strategy based on extra-long reads comprising cif genes. By focusing on a subset of long reads corrected with Illumina data, we were able to reconstruct several WO prophage contigs containing cif genes for each studied Culex isofemale line. To assess the reliability of these reconstructed contigs, we examined raw-read (Nanopore and Illumina) coverage profiles across assemblies. Coverage patterns were mostly consistent throughout the contigs and did not reveal evidence of major structural misassemblies, supporting their use for downstream analyses.

All cif genes are located in WO prophages

The contigs obtained allowed us to cover all variants of cid genes (i.e. the full cid repertoire) as well as the monomorphic pair of cin genes for each different genome (Fig. 2). All these cif genes were found in proximity to core phage genes and/or genes of the EAM, both defined as hallmarks of WO prophages (Bordenstein and Bordenstein 2022). All the prophage regions containing cif genes in wPip belong to the Wovirus genus, more specifically to the sr3WO clade, based on the phylogeny of their large serine recombinases (Bordenstein and Bordenstein 2022) (Supplementary Figs. 4a and 5a). Monomorphic cin genes are embedded in highly syntenic prophage regions in all analyzed wPip genomes, covering over nearly 36,000 bp involving almost all genes constituting WOPip2 described in the reference wPip-Pel genome (Klasson et al. 2008; Bordenstein and Bordenstein 2022). In contrast, most of the cid variants are located in variable WO prophages that differ in gene compositions within and between wPip lineages, which are mosaics of the reference WO prophages described in the wPip-Pel reference genome (Klasson et al. 2008). Altogether, these results indicate that WO prophages are highly dynamic across wPip lineages, as previously suggested by Reveillaud et al. (2019). Their complex repetitive architecture not only poses a major challenge for genome assembly but also makes it impossible for a single reference genome to fully represent their diversity. Our study provides new insights into this difficulty by revealing that WO prophages harboring cid genes exhibit highly variable gene compositions, yet consistently contain multiple repeated core genes (e.g. baseplate, head, and tail) and many copies of the WO-PC2 elements, which likely hinder genome circularization during assembly. The direct examination of Nanopore long reads corrected with Illumina data, combined with a prior, precise identification of the expected cid gene repertoires for each sequenced wPip lineage (Namias et al. 2025), proves to be an effective strategy for circumventing the issue of reconstructing WO prophages. Although some uncertainty regarding the precise assembly cannot be entirely excluded, this manual approach allowed us to delineate the previously unresolved genomic environments of these loci, whereas standard automated pipelines currently fail to resolve it (Supplementary Fig. 1). The cumulative size of WO prophage regions has been identified as a major driver of genome size variation in Wolbachia, as shown by the strong correlation observed across 78 supergroup B genomes (Vancaester and Blaxter 2023). Using this correlation, with a predicted full genome size of ∼1.5 Mb (Klasson et al. 2008), approximately 20% of the wPip genomes are expected to be occupied by WO prophages (Vancaester and Blaxter 2023). In wPip-I Tunis, wPip-III Slab, and wPip-IV Harash, we found that WO prophage regions harboring cif, represent up to 19% of the complete genome. These results suggest that most WO prophages in wPip might harbor cif genes. Importantly, many of these regions lack several canonical structural modules required for the formation of functional phage particles (e.g. complete tail, fiber, or baseplate assemblies). As such, they may not represent intact, mobilizable prophages sensu stricto, but rather defective prophages or horizontal gene transfer islands containing phage remnants.

cid genes are part of a modular and mobile genomic island

The presence of WO core genes along with WO-PC2 in the close genomic environments of all wPip cid pairs proved that they belong to prophagic regions. However, the gene compositions of these WO prophages are highly variable, suggesting important structural genomic rearrangement in these regions (Fig. 3). The phylogenetic analysis of the wPip WO prophage core genes showed that each of them has its own evolutionary history, suggesting recurring horizontal transfer and/or recombination events (Supplementary Table 5; Supplementary Figs. 4 and 5). Unexpectedly, in these highly dynamic prophagic environments, we show that cid genes are tightly associated with 4 modular genes that exhibit high sequence conservation. We propose that these genes form, along with the cid genes, genomic islands that move between different WO prophage regions. Among the 4 non-cid genes forming the “cid genomic islands,” 2 are mobile elements, and the other 2 appear to be implicated in recombination and DNA repair. The first mobile element, a PDDEXK2 transposase (WP0281) consistently found just upstream of cidA in wPip genomes, was already described as broadly associated with cif genes in Wolbachia (Tan et al. 2024; Amoros et al. 2025). The PD-(D/E)XK domain from the PDDEXK2 transposase shares the same name as nucleases found in all the cifB types (Beckmann et al. 2017; LePage et al. 2017; Martinez et al. 2021), although showing no alignable sequence similarity. Transposases such as Insertion Sequences (IS) had already been proposed as potential drivers of horizontal transfers of cif genes in other Wolbachia and bacterial endosymbionts (Cooper et al. 2019; Madhav et al. 2020; Amoros et al. 2025; Oswalt et al. 2025). These transfers, involving cif genes flanked by IS elements, can occur within distinct WO prophage regions, as observed in the Wolbachia endosymbiont of the horn fly Haematobia irritans irritans, wIrr (Madhav et al. 2020). The second mobile element found associated with cid genes in wPip is the Group II Intron maturase (intII), a type of transposable element with a reverse transcriptase activity previously reported in several Wolbachia strains (Leclercq et al. 2011), and sometimes also found in association with cif genes in Wolbachia and other bacterial endosymbionts (Bordenstein and Bordenstein 2022; Tan et al. 2024; Amoros et al. 2025). Apart from mobile elements, 2 other genes are strongly associated with cid genes in their genomic islands: (i) the rnha gene, which encodes an RNase H enzyme that may play a role in RNA regulation, and (ii) MutL, a gene involved in the DNA mismatch repair system. MutL was punctually found near the cif genes in other genomes, such as wAlbB (inside the WO-like island WOAlbB3) or in the Candidatus Mesenet longicola, a bacterium phylogenetically close to Wolbachia (Takano et al. 2021; Bordenstein and Bordenstein 2022; Amoros et al. 2025). While primarily known for its role in DNA mismatch repair, the recurrent presence of MutL in wPip cid genomic islands suggests a role in the maintenance, regulation, or even remodeling of these islands. Its activity could contribute to preserving the integrity of horizontally transferred genes, modulating local mutation rates that could otherwise disrupt the integrity of the islands or facilitating recombination events, thereby supporting both the stability and adaptive diversification of cid loci within and across Wolbachia genomes. The MutL found in the cid genomic islands has been proposed to have a phage-related origin, as it is divergent from chromosomally encoded MutL orthologs (Fallon 2021). The rnhA gene remains also functionally uncharacterized in Wolbachia. However, its known role in resolving RNA–DNA hybrids (Schroeder et al. 2023) suggests a potential involvement in transcriptional regulation, maintenance of genome stability, or modulation of recombination within the dynamic genomic environment of cid genomic islands. Although the PDDXK2 transposase, intII, and MutL have been previously identified separately as frequently associated with cif genes, they have never been reported as being jointly associated within a single conserved unit alongside cif genes. Altogether, this configuration forms a mobile, identifiable unit that is both stable and recombinogenic, making the wPip cid genomic island a unique case for cif genes. Interestingly, the wPip cid genomic islands share architectural similarities with another genomic island, the Octomom module known for its strong instability in wMelpop (Chrostek et al. 2013; Chrostek and Teixeira 2015, 2018; Bénard et al. 2021; Duarte et al. 2021; Monnin et al. 2021). The Octomom module is a genetic element whose copy number can evolve rapidly (increase or decrease over a few generations), and which is directly correlated with Wolbachia density within the host (Chrostek et al. 2013; Chrostek and Teixeira 2015, 2018; Bénard et al. 2021; Duarte et al. 2021; Monnin et al. 2021). Similar to the cid genomic islands in wPip, the Octomom module in wMelCS and wMelpop is flanked by a PDDEXK2 transposase, 2 reverse transcriptases (showing homology with intII), and a MutL gene (Bordenstein and Bordenstein 2022). This shared genetic architecture between cid genomic islands in wPip and Octomom modules in wMel suggests that both may rely on a common set of mobility and regulatory genes to facilitate their amplification, structural rearrangement, and/or stability.

High nucleotide identity of the genomic islands increases recombination probabilities during DNA replication

Although the gene composition of the cid genomic islands is modulable (Figs. 4 and 5), the 4 genes show high sequence similarities (>99%) between islands. This high nucleotide conservation is especially striking considering that the module is located within highly polymorphic, recombination-prone WO prophage regions (Supplementary Table 5). Indeed, upstream of each cidA–cidB pair, there are between 2,700 bp and 5,300 bp that are strictly identical between 2 islands within the same genome or between genomes of different wPip lineages. Some palindromic structures harboring truncated cid pairs are flanked by identical sequences totaling up to 10,600 bp. The presence of highly conserved regions near cid pairs (especially in a palindromic manner) increases the probability of recombination (homologous or gene conversion) between genomic regions flanked by these repetitions (Rocha 2003; Vos 2009; Paulsson et al. 2017). High recombination frequency may explain: (i) why the number of cid pairs increases or decreases between wPip lines and over the evolution of a single isofemale lineage (Namias et al. 2024a); and (ii) why there is a high probability of recombination between cid genes, which has been shown to be a major driver of their diversification (Namias et al. 2025). As components of genomic islands, the copy number of cif genes can evolve rapidly within a single isofemale line, leading to the coexistence of distinct clones within the same Culex line, as previously reported (Namias et al. 2024a). It is therefore possible that some of the prophages we annotated originate from different coexisting wPip clones within the same isofemale Culex line. This may be the case in the Tunis isofemale line, where Tunisp-d could represent an alternative version of Tunisp-c. Indeed, both contigs display highly similar syntenic organization while harboring distinct cidA and cidB variants. Moreover, Tunisp-d shows markedly uneven read coverage, suggesting that it may be present in only a subset of coexisting clones.

Among the available genomes, another Wolbachia strain, infecting the lepidopteran Rhopobota naevana, so-called wNaev, also exhibits the pattern of cid amplification with all 5 prophagic regions harboring 1 cif tandem (Supplementary Fig. 8). It has been proposed that the formation of palindromic structures involving truncated cidB genes in wPip may have resulted from a horizontal transfer event between ancestral wNaev-like and wPip-like strains (Namias et al. 2025), although the directionality of this event remains uncertain. Our genomic data show that even if prophage regions from wNaev and wPip both belong to srWO3 clade based on large serine recombinase, other WO prophage markers are phylogenetically discordant ruling out the hypothesis of a full WO phage exchange between the ancestral genomes of these 2 strains (Supplementary Fig. 4). However, in the prophagic regions in wNaev, the cid tandems are also associated with genes found in wPip cid genomic islands including the PDDEXK2 transposase and the intron group II maturase, some of which exhibit > 99% similarities with those of wPip (Supplementary Table 5; Supplementary Fig. 7). These results suggest that cid genomic islands could have been a main driver of horizontal transfer between these 2 ancestral strains.

Genomic islands as drivers of cid amplification and diversification in wPip

Wolbachia represent a distinctive group of endosymbiotic bacteria in that they are heavily infected by bacteriophages, unlike their closest relatives among the Rickettsiales (Darby et al. 2007). The high prevalence of cif genes in Wolbachia compared to other Rickettsiales may be linked to WO prophages, which constitute the predominant genomic environment of cif and are thought to facilitate horizontal transfer (Martinez et al. 2021; Vancaester and Blaxter 2023; Amoros et al. 2025). Previous studies have proposed that these bacteriophages, by alternating between prophagic and particle forms, could mediate the horizontal transmission of key genes between distantly related Wolbachia strains co-infecting the same arthropod cells (Bordenstein and Wernegreen 2004; Bordenstein and Reznikoff 2005; Bordenstein and Bordenstein 2016, 2022; Perlmutter et al. 2019). However, sequencing of WO phage particles remains limited, and none have been shown to encapsidate cif genes to date (Bordenstein and Bordenstein 2016; Kupritz et al. 2021). In wPip, particle-like structures resembling WO phages have been observed (Sanogo and Dobson 2006), yet no genomic evidence has confirmed that these particles are indeed WO phages that might carry cid genes. This absence of direct evidence suggests that transduction may not be the principal mechanism underlying the amplification and diversification of cid genes, and that alternative processes are likely involved. In this sense, the modularity and rearrangeability nature of the cid genomic islands could promote recombination during DNA replication, providing a robust mechanism for their rapid diversification. Across all analyzed wPip genomes, we consistently found a maximum of 5 to 6 cid genomic islands, suggesting an upper limit for this modularity, which could be consistent with a “cut-and-paste” mechanism of transposition for the cid genomic islands. Furthermore, previous studies have reported that some isofemale lines exhibit variation in specific cid variants between individuals, suggesting that rapid genomic remodeling can occur (Namias et al. 2025). Outside Wolbachia, genomic islands play a crucial role in the evolution of diverse bacterial species by spreading accessory genes, including those related to antibiotic resistance, defense systems, virulence, and catabolic functions, ultimately enabling the emergence of new metabolic pathways (Juhas et al. 2009; Li and Wang 2021; Krishnakant Kushwaha et al. 2024; Wang and Dagan 2024). By rapidly supplying new adaptive traits, they also favor the local and fast adaptation of their bacterial hosts to changing environments (Juhas et al. 2009; Bardaji et al. 2017; Li and Wang 2021; Krishnakant Kushwaha et al. 2024). Here, we provide evidence that such genomic islands may also promote the diversification of CI genes in Wolbachia.

Conclusion

By exploring the genomic environments of cif^wPip genes, our study unveils new insights into the microevolution of highly related Wolbachia genomes. Our investigation of the genomic environments of cif types in wPip reveals a strong link between their genetic diversity, CI pattern diversity, and the genomic architecture of the prophage regions in which they are embedded. The cin^wPip genes, which are monomorphic and not involved in the diversity of CI patterns in Culex, are located within a conserved prophage region across all available genomic data. In contrast, cid^wPip genes are embedded within conserved genomic islands, but in highly variable WO prophages both within and between wPip genomes. These genomic islands appear to promote the dispersion and diversification of cid genes across genomes. This genomic dynamic likely underlies the extensive diversity of CI patterns observed in C. pipiens, which can evolve rapidly over just a few generations, as it alters the repertoires of toxins (CidB) and antidotes (CidA) (Namias et al. 2024a). Therefore, the evolution of CI patterns in C. pipiens would be mainly driven by mobile elements within WO prophage regions. Identification of the drivers of Wolbachia genomic evolution is of primordial importance to develop safe and sustainable Wolbachia-based antivectorial methods. The ability of WO prophage regions to rapidly diversify CI genes suggests that Wolbachia strains introduced into mosquito populations might change over evolutionary timescales, potentially affecting the stability and efficacy of vector control.

Data availability

The raw Nanopore and Illumina sequencing reads were deposited in the Sequence Read Archive (SRA) at the National Center for Biotechnology Information (NCBI) under BioProjects PRJNA1426725 and PRJNA629472 (BioSamples SAMN55890339, SAMN55890340, and SAMN14776761). The accession numbers are SRR37316918 (Tunis Illumina), SRR37316919 (Tunis Nanopore), SRR36099207 (Slab Illumina), SRR11651065 to SRR11650728 (Slab Nanopore), SRR37316916 (Harash Illumina), and SRR37316917 (Harash Nanopore). The WO prophage contigs reconstructed in this study for wPip-I Tunis, wPip-III Slab, and wPip-IV Harash are available under the accession numbers PX571960 to PX571970. The subset of Illumina-corrected Nanopore sequences used for manual contig assemblage is available on https://doi.org/10.5281/zenodo.17245713. The codes and resources used in this study are available on GitHub (https://github.com/julien01A/wPip_cid_island).

Supplementary material is available at GENETICS online.

Funding

This work was supported by the I-SITE Montpellier Université d'Excellence (MUSE) (ANR-16-IDEX-0006) and the French National Research Program for Environmental and Occupational Health of ANSES (ANSES-25-EST-254).

Author contributions

Conceptualization: J.A, A.N., M.W., M.S.; Funding acquisition: M.W., M.S, H.A.; Investigation: J.A, A.N, T.D, H.A., J.M., M.S.; Data curation: J.A, A.N, T.D, H.A., J.M., M.S.; Formal analysis: J.A.; Methodology: J.A., M.S; Supervision: M.S.; Visualization: J.A.; Writing—original draft: J.A., M.S.; Writing—review & editing: All authors; Validation: M.S.