Transinfected Wolbachia strains induce a complex of cytoplasmic incompatibility phenotypes: Roles of CI factor genes

Abstract Wolbachia can modulate the reproductive development of their hosts in multiple modes, and cytoplasmic incompatibility (CI) is the most well‐studied phenotype. The whitefly Bemisia tabaci is highly receptive to different Wolbachia strains: wCcep strain from the rice moth Corcyra cephalonica and wMel strain from the fruit fly Drosophila melanogaster could successfully establish and induce CI in transinfected whiteflies. Nevertheless, it is unknown what will happen when these two exogenous Wolbachia strains are co‐transinfected into a new host. Here, we artificially transinferred wCcep and wMel into the whitefly and established double‐ and singly‐transinfected B. tabaci isofemale lines. Reciprocal crossing experiments showed that wCcep and wMel induced a complex of CI phenotypes in the recipient host, including unidirectional and bidirectional CI. We next sequenced the whole genome of wCcep and performed a comparative analysis of the CI factor genes between wCcep and wMel, indicating that their cif genes were phylogenetically and structurally divergent, which can explain the crossing results. The amino acid sequence identity and structural features of Cif proteins may be useful parameters for predicting their function. Structural comparisons between CifA and CifB provide valuable clues for explaining the induction or rescue of CI observed in crossing experiments between transinfected hosts.

Cytoplasmic incompatibility (CI) is the most widely studied phenotype induced by Wolbachia: uninfected females are sterilised by males infected with a Wolbachia strain (unidirectional CI), but the females infected with the same strain will be rescued (Werren et al., 2008).Insects infected with different Wolbachia strains may display unidirectional or bidirectional CI.Embryonic lethality induced by either strain in males cannot be rescued by the other strain in females (bidirectional CI), forming a reproductive barrier between insect hosts (Brucker & Bordenstein, 2012).
The CI factors (Cifs), including cifA and cifB, are involved in CI induction and rescue in host insects (Beckmann et al., 2017;Chen et al., 2019;LePage et al., 2017;Shropshire et al., 2018).Transgenic expression of both cifA and cifB genes, or cifB alone, in the germline of male hosts induced sterility that is highly similar to CI induced by Wolbachia (Adams et al., 2021;Beckmann et al., 2017;LePage et al., 2017;Shropshire & Bordenstein, 2019;Sun et al., 2022).This embryonic lethality could be rescued by crossing transgenic males with Wolbachia-infected females or those with a transgenic cifA gene expressed in the germline (Chen et al., 2019;Shropshire et al., 2018).Recent studies have found that the cif genes from bidirectional CI-inducing Wolbachia pairs were highly divergent, with only 29%-68% amino acid identity (LePage et al., 2017).These results were supported by transgenic research using cif genes of wMel, wRi, and wRec from D. melanogaster, D. simulans, and D. recens (Shropshire et al., 2021).The cifA and cifB genes are co-divergent (Bonneau et al., 2018;LePage et al., 2017;Lindsey et al., 2018), and have been proposed to account at least in part for bidirectional incompatibility probably by modulating CifA-CifB binding (Beckmann et al., 2017;Chen et al., 2019).
Bemisia tabaci (Hemiptera: Aleyrodidae) is a polyphagous agricultural pest feeds on more than 60 plant species and is a devastating pest insect worldwide.The unfertilised eggs of B. tabaci, a haplodiploid species, develop into male offspring, and thus CI can induce a male-biased offspring sex ratio (Hu & Li, 2015).Our previous studies have shown that it is highly receptive to Wolbachia infection: the Wolbachia wCcep strain from the rice moth Corcyra cephalonica and the wMel strain from the fruit fly D. melanogaster could successfully establish and induce CI in B. tabaci, which has implications for biological control of pest insects based on Wolbachia-induced CI.Here, we established a stable double-transinfected (DT) isofemale line of B. tabaci with wCcep and wMel.
Our purpose was to use this model to examine the CI phenotypes induced by co-infecting Wolbachia strains, and further explore the molecular mechanism underlying CI induction.Our results found that DT Wolbachia strains can induce a complex of CI phenotypes, including unidirectional and bidirectional CI, which is consistent with genomic analysis based on cif genes.Furthermore, our data provide valuable clues for explaining the induction or rescue of CI observed in crossing experiments using host insects transinfected with different Wolbachia strains.

Establishment of transinfected B. tabaci isofemale lines and trans-generational maintenance
Fifteen (five DT and 10 singly-transinfected [ST]) transinfected B. tabaci isofemale lines were established, which were trans-generationally maintained for 10 generations from G 1 to G 10 .As detected by polymerase chain reaction (PCR) using the ftsZ primers specific to Supergroup A and B Wolbachia, the individuals collected from G 3 to G 10 tested positive for both wMel and wCcep.In addition, a strong signal for wMel and wCcep was detected in the individuals from G 6 and G 10 in a stable manner, although no signal was detected in the individuals from G 1 and G 2 (Figure 1A,B).In ST isolines, wMel or wCcep was detected in the individuals collected from G 3 to G 10 (Figure 1C,D).

Crossing experiments and CI phenotypes
We performed a total of nine groups of crossing experiments.The results showed that there was no significant difference in the number of offspring per female among different crossing groups (Student-Newman-Keuls [SNK], F = 0.686, p = 0.703), while the percentages of males produced among crossing groups were significantly different (SNK, F = 241.295,p < 0.001; Table 1).Specifically, there was a very high percentage of males (97.4%-100%) in wC-ST♀ Â wM-ST♂, wM-ST♀ Â wC-ST♂, wM-ST♀ Â wMC-DT♂, wC-ST♀ Â wMC-DT♂, WT♀ Â wM-ST♂, WT♀ Â wC-ST♂ and WT♀ Â wMC-DT♂, representing a high level of CI.Among them, the two crossing groups wC-ST♀ Â wM-ST♂ and wM-ST♀ Â wC-ST♂ showed typical bidirectional CI phenotype, in which the male and female were ST with different Wolbachia strains and the reproductive anomaly caused by CI induced by one strain cannot be rescued by the other strain.In contrast, the crossing groups WT♀ Â wM-ST♂, WT♀ Â wC-ST♂ and WT♀ Â wMC-DT♂ exhibited unidirectional CI phenotype (Figure 2).

Genome sequence of wCcep
The complete circular genome of wCcep was assembled using long-read PacBio sequencing at 500Â median coverage (GenBank acc.no.CP087954; Figure S1).The total length of wCcep genome was 1,359,904 bp, with an average GC content of 34.0%, within the range of a typical Wolbachia genome (31.7%-38.3%).Annotation of the genome identified 1278 CDSs, 1190 protein-coding genes, 34 tRNA genes that can transfer all 20 amino acids, three rRNA (5S, 16S and 23S rRNA) genes, and one sRNA gene (Table 2).Phylogenetic analysis based on Wolbachia genome sequences confirmed that wCcep belonged to Supergroup B (Figure S2; Table S1).

Identification of cif genes in wCcep and phylogenetic analysis
Genomic analysis identified five cif genes in the wCcep genome, including cifA1 (GenBank acc.no.OP767524), cifB1 (GenBank acc.no.OL539522), cifA2 (GenBank acc.no.OP767525), cifB2 (GenBank acc. no. OP767526) and an incomplete cifA.Phylogenetic analysis using the four Cif proteins of wCcep and reference Cifs showed that CifA1 wCcep and CifA wMel were clustered in a clade belonging to Type I CifA, and CifA2 wCcep was clustered in a clade belonging to CifA Type IV (Martinez et al., 2021; Figure 3A).CifB1 wCcep and CifB wMel were clustered in a clade of Type I CifB, while CifB2 wCcep was grouped in Type IV CifB (Figure 3B).

Structural comparison of CIF proteins between wCcep and wMel
Structural prediction revealed the putative domains of CIF proteins.There was variation only in the length of domains between CifA I wCcep and CifA I wMel , and between CifB I wCcep and CifB I wMel (Figure 4A,C), while substantial structural differentiation was observed between CifA I and CifA IV, and between CifB I and CifB IV (Figure 4B,D).CifA I contained a putative catalase-related (catalase-rel) domain, a domain that shares homology with Puf family RNA-binding proteins and a sterile-like transcription factor (STE) domain (Figure 4A), whereas CifA IV contained a longer Puf domain but no catalase-rel domain (Figure 4B).In contrast, CifB I contained two PD-(D/E)XK superfamily of nucleases domains and a Ulp1 ubiquitin protease module with deubiquitylase (DUB) activity (Figure 4C), but CifB IV contained no Ulp1 domain (Figure 4D).

Comparison of the tertiary structures of CIF proteins between wCcep and wMel
The tertiary structures of CIF proteins were generated by AlphaFold.The confidence of the structures was evaluated by Local Distance Difference Test (lDDT; Figure S3).The results showed that the lDDT scores of all of the six Cif proteins of wCcep and wMel ranged between 68.4 and 84.0, suggesting a confident structural prediction based on the full-length sequences using the AlphaFold programme, except for CifB I wMel containing a DUB domain with a low predicted mass (Table 3; Figure S4).Only two PD-(D/E)XK-nuclease domains of CifB were analysed because they were shared across all types of Cif proteins (Shropshire et al., 2022), and it was difficult to make a direct determination of the structure of CifB wMel because its expression in Escherichia coli was poor (Wang et al., 2022).
To evaluate the similarity between the tertiary structures of CIF proteins, the template modelling (TM) scores were calculated (Xu & Zhang, 2010).The results showed that the TM score between CifA I wCcep and CifA I wMel was 0.86, while they shared an amino acid sequence identity (AA identity) of 67.1%; the TM score between CifB I wCcep and CifB I wMel was 0.63, while shared an AA identity of 57.3%.Although CifA IV wCcep and CifA I wMel shared an AA identity of only 26.3%, the TM score between them was 0.76.The TM score between CifB IV wCcep and CifB I wMel was 0.66, but they shared an AA identity of only 20.3% (Table 4; Figure 5).

CifAI wCcep binds CifBI wCcep through a large distinct interface
A model for the CifA I wCcep -CifB I wCcep 2ND complex was built by using AlphaFold-Multimer (Figure 6A).CifB

I wCcep
2ND comprised residues 1 through 752, which was predicted to contain two PD-(D/E)XK (pseudo) nuclease domains.The obtained end-to-end model was further optimised by molecular dynamics (MD) simulations.The trajectories were performed and achieved equilibrium within 100 ns.The root mean square deviation (RMSD) for one of the simulation trajectories was recorded (Figure S5A).The complex also predicted many hydrogen bonds and salt bridges between CifA I wCcep and CifB I wCcep 2ND at the three interfaces (Figure 6B-E) and only 0.46% of the residues were Ramachandran outliers (Figure S5B), demonstrating that the side chains of the predicted model have been reliably predicted and thus the complex of CifA I wCcep -CifB I wCcep 2ND is reasonable.The interface between CifA I wCcep and CifB I wCcep 2ND was divided into three regions (Figure 6A), and each region mainly involves one pair of structural motifs.In the first region (Interface I), the helices consisting of the residues 38-143 of CifA I wCcep interact with a loop in CifB I wCcep 2ND (residues 402-425) through a network of hydrogen bonds (Figure 6B).At interface I, three residues (ASN144, ARG108 and THR140) from CifA I wCcep interact with the residues from CifB I wCcep 2ND (GLN452, GLY455 and GLU458; Figure 6B).Interface II is located between the helices from CifA I wCcep (residues 38-43) and CifB I wCcep   (residues 399-402), and the residues (GLU37, HIS38 and Lys43) from CifA I wCcep interact with the residues from CifB I wCcep 2ND (ASN409, GLU399 and GLU402) through a network of four hydrogen bonds and two salt bridges (Figure 6C).Interface III involves CifA I wCcep (residues 208, 242-249, 289-297, 337 and 392-398) and a cross-cutting helix (residues 244-254, 377-382) in CifB I wCcep 2ND to stabilise the interaction.Multiple residues interact between CifA I wCcep and CifB I wCcep 2ND and there are 12 hydrogen bonds and six salt bridges at Interface III (Figure 6D,E).

DISCUSSION
We established stable double-and ST B. tabaci isofemale lines with wMel and wCcep.These isolines were used for reciprocal crossing experiments, showing that a complex of CI phenotypes was induced, including unidirectional and bidirectional CI.B. tabaci is a haplodiploid species, and it is capable of laying haploid eggs that develop into males (arrhenotokous parthenogenesis) when CI is induced (Hu & Li, 2015;Zhou & Li, 2016).A high male-biased offspring sex ratio indicates a high CI level.Here, the percentages of males produced in the crossing groups wC-ST♀ Â wM-ST♂, wM-ST♀ Â wC-ST♂, wM-ST♀ Â wMC-DT♂, wC-ST♀ Â wMC-DT♂, WT♀ Â wM-ST♂, WT♀ Â wC-ST♂ and WT♀ Â wMC-DT♂ were 100% or nearly 100%, indicating that a complete or nearly complete CI had been induced in these crossings.As expected, no CI phenotype was observed in wMC-DT♀ Â wMC-DT♂, which might have been restored through what is usually called "the rescue (resc) function" in the eggs infected with the same Wolbachia strains.In wC-ST♀ Â wM-ST♂ and wM-ST♀ Â wC-ST♂, the male and female were ST with different Wolbachia strains and the reproductive anomaly caused by CI induced by one strain cannot be rescued by the other strain, which represents a typical bidirectional CI phenotype.In contrast, the crossing groups WT♀ Â wM-ST♂, WT♀ Â wC-ST♂ and WT♀ Â wMC-DT♂ exhibited unidirectional CI phenotype, in which transinfected males induced CI in wildtype females.
It has been shown that the Cifs (CifA and CifB) were determinants for CI induction and rescue (Adams et al., 2021;Shropshire & Bordenstein, 2019;Sun et al., 2022), and the expression of CifA in female was sufficient for rescuing CI (Shropshire et al., 2018).We, therefore, sequenced the whole genome of wCcep, identifying two cifA and two cifB genes.In contrast, the genome of wMel contained only one cifA and one cifB (Wu et al., 2004).Crossing experiments showed that both wCcep and wMel are CI-inducing Wolbachia strains, and the strength of CI induced by wCcep (WT Â wC-ST) was not significantly different from that by wMel (WT Â wM-ST; Table 1).It seems that the higher copy number of cif genes in wCcep cannot significantly differentiate it from wMel in CI induction.Similarly, the strength of CI induced by DT males is not significantly different from that by ST males, although the CI phenotype induced by double transinfection is more stable than single transinfections.
Previous studies showed that the amino acid similarity between CifA proteins influenced the rescue of CI phenotype.The rescue happened only when the cif genes were closely related between different Wolbachia strains (Bonneau et al., 2018;LePage et al., 2017;Shropshire et al., 2018).For instance, wRi could rescue wMel-induced CI in a homozygous cross, because the amino acid sequences of their CifA proteins were nearly identical (99% AA identity) (Shropshire et al., 2021).In contrast, wHa could not rescue CI induced by wMel, and vice versa, because the AA identity between their CifA proteins was only 67% (LePage et al., 2017).In our study, the AA identities between CifA I wCcep and CifA I wMel and between CifB I wCcep and CifB I wMel were 67.1% and 57.3%, respectively.The divergence of Cif proteins between wCcep and wMel can explain the bidirectional CI observed in wC-ST♀ Â wM-ST♂ and wM-ST♀ Â wC-ST♂, in which CI induced by one strain cannot be rescued by the other.It is noteworthy that the AA identities between CifA IV wCcep and CifA I wMel and between CifB IV wCcep and CifB I wMel were only 26.3% and 20.3%, respectively.The substantial divergence between these Cif proteins may enhance the incompatibility between the two Wolbachia strains.
Structural comparisons of Cif proteins between wCcep and wMel revealed that CifA I and CifB I proteins of wCcep and wMel are structurally similar, containing common domains of CIF proteins.However, substantial structural variation was observed in CifA IV and CifB IV of wCcep, with some missing domains (Figure 4).It is still unknown whether they are functionally incompetent or not before further functional studies are carried out.Interestingly, a comparison of the tertiary structures of Cif proteins between wCcep and wMel resulted in a high TM score (0.86) between CifA I wCcep and CifA I wMel , whereas a moderate TM score (0.63) was obtained between CifB I wCcep and CifB I wMel , even lower than the TM scores between CifA IV wCcep and CifA IV wMel (0.76) and between CifB IV wCcep and CifB IV wMel (0.66).Crossing experiments using transgenic models are needed for functional verification of these cif genes.
Previous studies suggested that the interfacial residues of the CifA-CifB complex that simulated CifA-CifB binding in the female were essential for the rescue of CI in transgenic D. melanogaster (Wang et al., 2022;Xiao et al., 2021).In this study, CifA I wCcep -CifB I wC- cep 2ND complex (except for the DUB domain) built by homology modelling is very similar to the structure of CidA wMel -CidB wMel ND1-ND2 complex in which CidA binding does not block the DUB domain of CidB catalytic activity (Beckmann et al., 2017).Both of two complexes form three binding sites in similar positions.Interestingly, the residues involved in the interactions at the three interfaces are very different between the both complexes (Wang et al., 2022;Figure 2C-E), which helps explain their cognate-specific binding.
In summary, we established stable transinfected B. tabaci isofemale lines.Successful establishment of DT isolines demonstrated that wCcep and wMel can co-exist in the same recipient host, confirming that B. tabaci is highly receptive to Wolbachia infection.This research model is useful for deciphering the molecular mechanisms underlying CI induction by Wolbachia.We performed a series of crossing experiments using these transinfected isolines, and a complex of CI phenotypes were induced.We therefore sequenced the whole genome of wCcep and identified four Cif genes in its genome.Comparative analysis revealed that the CI phenotypes induced by wCcep and wMel can be explained by the divergence between their cif genes.Our study confirmed that the amino acid sequence identity is an important parameter for defining CI phenotypes.Nonetheless, structural comparisons in terms of the interfacial residues in the binding regions of the CifA-CifB complex can provide valuable clues for explaining the induction or rescue of CI observed in crossing experiments using host insects transinfected with different Wolbachia strains.
Transinfection and establishment of double-and ST isofemale lines of B. tabaci with wMel and wCcep Wolbachia strains A volume of 46 nL bacterial suspensions of wCcep and wMel (1:1) in SPG buffer (220 mM sucrose, 4 mM KH-2PO4, 9 mM Na-2HPO4, 5 mM L-glutamate, pH 7.4) were injected separately into the fourth-instar nymph (pseudopupa) of B. tabaci placed in a petri dish (Φ 9 cm) covered with 10% agar on the platform of Nanoliter 2000 (World Precision Instruments, Sarasota, Florida, USA).The pupa was then placed in a climate incubator until adult emergence after injection (25 C, 65% ± 1 RH and 14L:10D).The newly emerged adults were separately maintained on potted cotton plants in pairs (♀/♂; G 0 ), and the offspring (G 1 ) from the pairs tested positive for Wolbachia was kept for establishing isofemale lines.For establishing ST isofemale lines, the wCcep strain or wMel strain was used for transinfection following the same procedures.

Crossing experiments
Two-day-old adult male and female whiteflies collected from the 6th-8th generation (G 6 -G 8 ) were used for crossing experiments.A total of nine groups were set up for reciprocal crossing: DT wMC-DT ♀ Â wMC-DT♂; ST wC-ST♀ Â wM-ST♂; wM-ST♀ Â wC-ST♀; wM-ST♀ Â wMC-DT♂; wC-ST♀ Â wMC-DT♂; wild-type WT♀ Â wM-ST♀; WT♀ Â wC-ST♂, and WT♀ Â WT♂ (control group).For each group, the mating pairs were confined in a leaf-clip cage on a cotton plant for 5 days, and the eggs laid on the plant were then placed into a climate incubator for further development till adult emergence (L14:D10 and 65 ± 1% RH at 28 C).The progenies were collected, and the number of off-spring per female and the percentage of males were calculated.CI level was assessed as the proportion of male offspring.Three biological replicates were performed for each group.

Genome sequencing and identification of Cif genes in wCcep
The genome of Wolbachia wCcep strain was sequenced by Illumina and PacBio technologies (Ellegaard et al., 2013).For isolation and purification of Wolbachia wCcep, the infected rice moth was allowed to oviposit, and the eggs were collected.The eggs were dechorionated in bleach, rinsed with sterile distilled water, and then homogenised in phosphate-buffered saline (PBS) buffer with a sterile micropestle.The homogenate was centrifuged at 400 Â g for 5 min, and the supernatant was transferred to a new tube and centrifuged at 5400 Â g for 5 min.The pellet was resuspended in PBS and then centrifuged at 400 Â g for 5 min to remove remaining debris.The supernatant was slowly pushed through a 5 μm pore size filter (Millipore, Bedford, MA) with a syringe, followed by a 2.7 μm pore size filter (Whatman, USA).The filtrate was centrifuged at 6900 Â g for 15 min.After decanting the supernatant, a bacterial pellet (Wolbachia cells) was obtained.The purified Wolbachia was confirmed by PCR using the primers 81F/522R targeting wsp of Supergroup B Wolbachia (Zhou et al., 1998), and then quickly ground and transferred to a preheated 50-mL centrifuge tube containing 15 mL CTAB solution, which was mixed and placed in a constant-temperature water bath at 65 C, lysed for 60 min, and mixed once every 15 min.The sample was cooled to room temperature before centrifuged at 5000 rpm for 10 min at room temperature.The supernatant was added to an equal volume of phenol/chloroform/isoamyl alcohol and centrifuged.The upper liquid was extracted one more time, and then transferred to a new tube with addition of 2/3 volume of isopropanol and 50 μL 3 M sodium acetate, and placed in a refrigerator (À20 C) overnight for precipitation.The mixture was added with 750 μL ethanol (75%) and centrifuged at 5000 rpm for 5 min.This step was repeated once.The dried pellet was finally dissolved in Tris-EDTA buffer (10 mM Tris, 1 mM EDTA, pH 8.0) and frozen at À20 C for genome sequencing on a PacBio RS II and Illumina HiSeq 4000 platform (Illumina Inc., San Diego, CA, USA) by Beijing Genomics Institute (Beijing, China).
The draft genome of wCcep was assembled using the Celera Assembler against a high-quality corrected circular consensus sequence subreads set.The assembled genome was annotated using Prokka version 1.14 (Seemann, 2014) with default parameters.The Cif genes in the genome of wCcep were searched with TBlastN (https://blast.ncbi.nlm.nih.gov/Blast.cgi)using all cif gene sequences available to date.

Phylogenetic analysis of Cif proteins in wCcep and wMel
The Cif proteins encoded by the cif genes identified in the wCcep genome sequenced in this study and retrieved from the wMel genome (GenBank acc.no.GCF_000008025.1) were aligned with reference homologous sequences (Table S2).Phylogenetic analysis was performed using the maximum likelihood algorithm on MEGA7.1.The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) are shown at the node.The evolutionary distances were computed using the Poisson correction method.All positions containing gaps and missing data are eliminated.

Comparison of the tertiary structures of Cif proteins between wCcep and wMel
The tertiary structures of CifA I wCcep , CifB I wCcep , CifA IV wCcep , CifB IV wCcep , CifA I wMel and CifB I wMel proteins were generated by AlphaFold2.2.3 as a local server and based on full AlphaFold database (Jumper et al., 2021).The max_template_date was set to ensure that all proteins are generated off the same set of starting templates (last accessed: 21 September 2020).The similarity between tertiary structures was calculated using the Zhanglab TM-score Server (http:// zhanggroup.org/TM-align/;Xu & Zhang, 2010).The structural difference between the CIF proteins of wCcep and wMel was evaluated based on their similarities.AlphaFold-Multimer (Evans et al., 2021) was used to predict the binding complex of CifA I wCcep -CifB I wC- cep full sequences with multiple sequence alignments set as the all genetics database used at CASP14.The prediction of complexes was run fifth with different random seeds and 25 models were obtained.Finally, the complex with the highest quality score (p lDDT = 0.726) was selected.Among the complexes, the tail of CifB I wCcep that contained a DUB domain (residue 753-1107) was removed due to being far away from the core and having a low predicted mass, which was consistent with previous studies (Wang et al., 2022).The remainder contained two PD-(D/E)XK (pseudo) nuclease domains (residue 1-752) for further optimization with subsequent MD simulations.

MD simulations
MD simulations were performed by using Desmond programme of Schrödinger 2021-3 (Bowers et al., 2006) and the OPLS4 (Lu et al., 2021) protein force field.The binding complex of CifA I wCcep -CifB I wCcep 2ND obtained in the last step was explicitly solvated with TIP3P (Jorgensen et al., 1983) water molecules under cubic periodic boundary conditions for a 15 Å buffer region.The overlapping water molecules were removed and 0.15 M KCl was added, and the systems were neutralised by adding K + as counter ions.The electrostatic interactions were calculated under elastic simulations by Verlet and cg algorithms and particle-mesh Ewald method, and energy minimization was performed using the steepest descent method for the maximum number of steps (50,000 steps).The Coulomb force cut-off distance and the van der Waals radius cut-off distance were both 1.4 nm.The system was equilibrated using a regular system (NVT) and an isothermal isobaric system (NPT), followed by 100 ns MD simulations at constant temperature and pressure.The V-rescale temperature coupling method was used to control the simulation temperature to 300 K. RMSD was used to observe the local site metastability of the system during the simulation and calculated based on C-alpha atoms.The plot was presented using PyMOL 2.4.1.The Ramachandran plot of the eventual model was generated with Schrödinger 2021-3.

Data analysis
Statistical differences were analysed using One-way analysis of variance followed by SNK test and Tukey's post hoc tests ( p < 0.05) on SPSS v.27.0 software (SPSS Inc., Chicago, IL, USA).

FF
I G U R E 2 Crossing experiments between double-transinfected (DT), singly-transinfected (ST) and wild-type B. tabaci whiteflies.A total of nine groups of crossings are performed.WT, wild-type; wMC-DT, DT with wMel and wCcep; wC-ST, ST with wCcep; wM-ST, singly-transinfected with wMel.The means are indicated by vertical lines.T A B L E 2 Genome statistics of Wolbachia wCcep strain (host: C. cephalonica).I G U R E 3 Phylogenetic analysis of CIF proteins using maximum likelihood method on MEGA 7.1.The CIF proteins used in the analysis are listed in Table S2.(A) CifA.(B) CifB.The CIF proteins of wCcep are indicated by solid squares, and those of wMel by solid stars.

2NDF
I G U R E 4 Comparisons of the organisations and domains of CIF proteins between wMel and wCcep.(A) CifA I wCcep and CifA I wMel ; (B) CifA IV wCcep and CifA I wMel ; (C) CifB I wCcep and CifB I wMel ; (D) CifB IV wCcep and CifB I wMel .The domains include a putative catalase-related (catalase-rel) domain (gold), a domain sharing homology with Puf family RNA-binding proteins (green), a sterile-like transcription factor (STE) domain (brown), a Ulp1 deubiquitinase module (indigo blue), and a PDDEXK nuclease domain (dark brown).The numbers indicate the start and stop positions of each domain.T A B L E 3 Alphafold structural confidence of CIF proteins.
AA identities are shown above the diagonal.TM-scores are shown below the diagonal.TM-scores are calculated by the Zhanglab Server and range from 0 to 1: TM > 0.5, a very similar fold; 0.5 > TM > 0.17, significant similarity and TM < 0.17, random structural similarity.F I G U R E 5 Overlapping CIF proteins of wMel and wCcep generated by Alphafold programme.(A) Overlapped CifA I wCcep and CifA I wMel ; (B) Overlapped CifA IV wCcep and CifA I wMel ; (C) Overlapped CifB I wCcep and CifB I wMel ; (D) Overlapped CifB IV wCcep and CifB I wMel .The domains include a putative catalase-related (catalase-rel) domain (grey), a domain sharing homology with Puf family RNA-binding proteins (green), a sterile-like transcription factor (STE) domain (brown), a Ulp1 deubiquitinase module (gold), and a PDDEXK nuclease domain (light grey).

F
I G U R E 6 Interaction between CifA and CifB proteins through a large conserved tripartite interface.(A) A structural model of CifA I wCcep in complex with CifB I wCcep 2ND generated by AlphaFold-Multimer.CifA I wCcep binds to CifB I wCcep 2ND through three regions shown in green, pink and cyan for Interface I, II and III, respectively.(B-E) Each interface of the CifA I wCcep -CifB I wCcep 2ND complex involves a pair of structural motifs.Representative residues directly involved in the interaction are labelled and tabulated.2ND, two PD-(D/E)XK nuclease domains.