Extensive genetic diversity of Rickettsiales bacteria in multiple mosquito species

Rickettsiales are important zoonotic pathogens, causing severe disease in humans globally. Although mosquitoes are an important vector for diverse pathogens, with the exception of members of the genus Wolbachia little is known about their role in the transmission of Rickettsiales. Herein, Rickettsiales were identified by PCR in five species of mosquitoes (Anopheles sinensis, Armigeres subalbatus, Aedes albopictus, Culex quinquefasciatus and Cu. tritaeniorhynchus) collected from three Chinese provinces during 2014–2015. Subsequent phylogenetic analyses of the rrs, groEL and gltA genes revealed the presence of Anaplasma, Ehrlichia, Candidatus Neoehrlichia, and Rickettsia bacteria in mosquitoes, comprising nine documented and five tentative species bacteria, as well as three symbionts/endosybionts. In addition, bacteria were identified in mosquito eggs, larvae, and pupae sampled from aquatic environments. Hence, these data suggest that Rickettsiales circulate widely in mosquitoes in nature. Also of note was that Ehrlichia and Rickettsia bacteria were detected in each life stage of laboratory cultured mosquitoes, suggesting that Rickettsiales may be maintained in mosquitoes through both transstadial and transovarial transmission. In sum, these data indicate that mosquitoes may have played an important role in the transmission and evolution of Rickettsiales in nature.


Phylogenetic analysis of Rickettsiales gene sequences recovered from mosquitoes.
Phylogenetic analysis of the rrs gene sequences revealed that the Chinese mosquitoes collected in this study contained bacteria from the Anaplasma, Ehrlichia, or Candidatus Neoehrlichia genera of the family Anaplasmataceae, and from the genus Rickettsia of the family Rickettsiaceae (Fig. 1). Within the genus Anaplasma, the rrs gene sequences fell into six clusters on the phylogeny (Fig. 1A), corresponding to (i) four documented species: A. bovis, A. marginale, A. phagocytophilum, and A. platys, and (ii) two novel candidate species: Anaplasma sp. Bole  A. phagocytophilum bacteria were also distinct but most closely related to a strain isolated from a R. norvegicus rat 30 . Finally, A. marginale bacteria were closely related to a cattle isolate sampled in Henan province 31 , which adjoins Hubei province, as well as those sampled from cattle in Australia and the USA. The Ehrlichia bacteria from mosquitoes formed three clusters, compatible with the existence of three species (Fig. 1B). One cluster comprised the mosquito rrs gene sequences WHARSA-128 and HHAEAL-113 and Ehrlichia sp. NS101 identified from deer in Japan. In contrast, the mosquito rrs gene sequence WHCUPA-68 was more closely related to those of human E. chaffeensis 32 . The remaining mosquito Ehrlichia bacteria formed a distinct cluster most closely related to the tick bacteria Ehrlichia sp. EHh317. Candidatus Neoehrlichia mikurensis bacteria were also identified in mosquitoes, showing a close relationship with those previously identified in rodents and humans 33 .
We also attempted to amplify other two genes -citrate synthase (gltA) and heat shock protein (groEL) -from all Rickettsiales DNA-positive mosquitoes (see Supplementary Table S1). However, compared to the rrs gene, these gene sequences were difficult to amplify from mosquito samples. Consequently, groEL gene sequences were only recovered from some samples positive for Anaplasma, Ehrlichia, and Rickettsia, while gltA gene sequences were obtained only from some samples positive for Anaplasma and Rickettsia.
In the groEL gene tree ( Fig. 2A   gene sequences recovered from the Ehrlichia positive samples were also closely related each other and formed a distinct cluster. Finally, those mosquito sequences recovered from Rickettsia positive samples were closely related to those from R. belli and R. monacensis (Fig. 2B).
In the gltA gene tree, all Anaplasma sequences formed three clusters (Fig. 3A). Within the Candidatus A. boleense cluster, the bacteria formed a cluster clearly distinct from that previously documented in ticks. Strikingly, although the mosquito Anaplasma bacteria (ZJARSA-8, WHANSL-27-1, JXARSA-29, WHAEAL-17-2, JXANSA-19, WHARSL-30, WHAEAP-26) were most closely related to A. platys bacteria in both the rrs and groEL trees, in the gltA tree they grouped with strain Anaplasma sp. clone SY124 previously identified in ticks from Shenyang of China 36 . Further studies are needed to determine whether these bacteria belong to A. platys or represent co-infection (and/or recombinant event). Within the genus Rickettsia, the gltA gene sequences from mosquitoes were closely related to those from R. bellii and R. monacensis, respectively, while sequences WHCUTA-121 and WHCUTA-130 formed a distinct lineage (Fig. 3B). As no gltA sequences from the Rickettsia symbiont of Nephotettix cincticeps were available, we could not determine their phylogenetic relationship.

Rickettsiales in different life stages of mosquitoes sampled in nature. The detection rate of
Rickettsiales in adult mosquitoes from our three sampling sites was 19.78% in An. sinensis, 23.90% in Ar. subalbatus, 3.33% in Ae. albopictus, 4.86% in Cu. quinquefasciatus, and 6.25% in Cu. tritaeniorhynchus. Overall, Anaplasma bacteria showed the highest prevalence (10.2%), exhibiting species, geographic and annual variation ( Table 1, Supplementary Table S2). In contrast, the detection rates of bacteria from the Ehrlichia, Candidatus Neoehrlichia, and Rickettsia genera were lower in adult mosquitoes (0.72%, 0.51%, and 1.13%, respectively), possible caused by bias in the PCR assay. Finally, among all the Rickettsiales identified in mosquitoes, A. marginale had the highest detection rate (4.43%), followed by A. phagocytophilum (2.16%).
To better understand the mosquito circulation of Rickettsiales, eggs, larvae, and pupae were collected from aquatic environments in Wuhan. Although only A. marginale and A. phagocytophilum were identified in two pools of An. sinensis eggs, more Rickettsiales were found in these five mosquito species at the larvae and pupae stages ( Table 2, see Supplementary Table S3). In similar pattern to that observed in adult mosquitoes, Anaplasma bacteria were at higher prevalence in larvae and pupae than Ehrlichia, Candidatus Neoehrlichia, and Rickettsia. Notably, A. phagocytophilum and A. marginale were highly prevalent in larvae. As with the adults, three species of Ehrlichia (E. chaffeensis, Candidatus Ehrlichia sp. EHh317, and Candidatus Ehrlichia sp. NS101) were detected in the larvae and pupae stages. With the exception of Ar. subalbatus, Candidatus N. mikurensis were identified in four other species of mosquitoes. Finally, a novel candidate species (or symbiont/endosybionts) (Candidatus Rickettsia sp. Culex tritaeniorhynchus) was identified in Cu. tritaeniorhynchus larvae. In juvenile mosquitoes, the prevalence of these bacteria was high in Ar. subalbatus (52.08% larvae and 12.50% in pupae), An. sinensis (27.08% in larvae and 4.55% in pupae), and Ae. albopictus (18.75% in larvae and 4.86% in pupae), but relatively low in Cu. quinquefasciatus (7.64% in larvae and 2.08% in pupae) and Cu. tritaeniorhynchus (2.78% in larvae and 2.31% in pupae). In sum, these data suggest that Rickettsiales may be transmitted through the transstadial and/or transovarian transmission routes in mosquitoes, although the number of observations in eggs was small (95% CI 0-24.8%).
More than one species of Anaplasma bacteria was detected in some individuals of both adult and juvenile mosquitoes ( Table 4), indicative of co-infection. In adult mosquitoes, the most common combinations were A. marginale and A. bovis, and A. marginale with A. platys. The co-infection of three bacteria (A. bovis, A. marginale, and A. platys) was detected in one An. sinensis individual. Interestingly, co-infection with A. phagocytophilum and Ehrlichia sp. EHh317 was also detected in Ar. subalbatus. Notably, co-infection was relatively more common in An. sinensis mosquitoes, but absent in Cu. quinquefasciatus and Cu. tritaeniorhynchus mosquitoes. For juvenile  (Table 4). The most common combination was also A. marginale or A. phagocytophilum with other bacteria. Finally, co-infection with bacteria from the Ehrlichia and Anaplasma genera was also detected in juvenile mosquitoes.

Rickettsiales in different life stages of laboratory reared mosquitoes.
To confirm the transstadial and transovarian transmission of Rickettsiales, adult (and/or larvae for Cu. tritaeniorhynchus) mosquitoes collected in the field from Wuhan were reared in the laboratory for an entire life cycle (adult-egg-larva-pupae-adult) and tested for the presence of bacteria. From the adult (parent) mosquitoes collected in field, Ehrlichia and Rickettsia bacteria were identified; Ehrlichia sp. EHh317 in Ae. albopictus, R. japonica in all five species of mosquitoes, and R. monacensis and R. sibirica in Cu. tritaeniorhynchus (Table 3). Accordingly, these bacteria were also identified in their offspring. Notably, R. japonica was detected in all life stages of Ar. subalbatus and Cu. quinquefasciatus, and in three stages of the other three mosquito species. Additionally, both R. japonica and R. monacensis were successfully identified at each life stage in Cu. tritaeniorhynchus, again supporting the transstadial and transovarial transmission of Rickettsiales. As Anaplasma and Candidatus N. mikurensis were not identified in adult mosquitos sampled in the field for this experiment, they were similarly not found in laboratory reared eggs, larvae, pupae and adult mosquitoes.

Discussion
Rickettsiales are associated with a wide range of animals including diverse arthropods, mammals and birds 3,6,14-20 . High levels of genetic diversity in Rickettsiales have been identified in both ticks and vertebrates 6,14,36 , including multiple infection by distinct bacteria in a single tick species 18,37 . Our phylogenetic analysis revealed the co-circulation of nine documented and five tentative species bacteria, as well as three symbionts/endosybionts in five species of mosquitoes. Of particular note in this context was that two distinct Rickettsia species (or symbionts/endosybionts) were identified in mosquitoes and that a single mosquito can harbor two or more species from the genera Anaplasma and Ehrlichia. As more than 3,500 species of mosquitoes are distributed worldwide, it is likely that additional (and/or novel) mosquito-associated Rickettsiales (or symbiont/endosybionts) will be discovered in the future.
Ticks are considered the primary vectors for Rickettsiales, especially in the case of Anaplasma and Ehrlichia 14,18,19 . Although mosquitoes are both diverse and abundant, with the exception of the genus Wolbachia 24,26 there is little evidence for mosquitoes serving as competent vectors or hosts of Rickettsiales bacteria. Previous studies showed that mosquito cell lines (i.e. Ae. albopictus and An. gambiae cells) could be used to propagate A. marginale, R. felis, R. montanensis, and R. peacockii bacteria 38,39 . Lindh and colleagues also reported the detection of the rrs gene of A. platys and A. ovis in the Anopheles midgut, but could not exclude that the Anaplasma bacteria were derived from ingested blood 27 . Recently, Dieme and colleagues reported the transmission potential of R. felis infection by An. gambiae mosquitoes 28 . Herein, we document the presence of diverse bacteria of the genera Anaplasma, Ehrlichia, Candidatus Neoehrlichia and Rickettsia in five species of adult and juvenile mosquitoes sampled from three Chinese provinces. In addition, Rickettsiales bacteria were identified in each mosquito life stage (egg, larvae, pupae, and adult). Hence, these data clearly show that diverse Rickettsiales are present in mosquitoes in nature, such that mosquitoes may have played an important role in the transmission, and likely evolution, of Rickettsiales bacteria. In addition, it was noteworthy that some of the Rickettsiales sequences (for example, WHCUTL-65, WHCUTA-121, WHCUTA-130 and WHANSA-146) recovered from mosquitoes were closely related to those of symbionts or endosymbionts. Finally, as only PCR was used here, which may result in some bias, it is possible that the diversity and the prevalence of Rickettsiales bacteria in mosquitoes may be higher than we report. Intracellular parasites are transmitted by either transstadial and/or transovarial mechanisms (i.e. vertical transmission) in arthropods or by horizontal transmission via infected vertebrates. As horizontal transmission is dependent on the density of susceptible hosts and their intervals of patent infection, the transovarial and/or  transstadial pathways may be more reliable routes to transmit intracellular parasites 40,41 . Although transovarial and/or transstadial transmission has been documented in members of the Rickettsiaceae 15,17,42,43 , this process is not thought to occur to a significant degree in those Anaplasma and Ehrlichia bacteria documented to date, which may in part be due to lack of the aldolase/adducing domain protein 44 . Hence, the known Anaplasma and Ehrlichia bacteria (excluding endosymbionts such as Midichloria) transmitted by ticks require feeding on an infected vertebrate 14 . Notably, we identified Anaplasma, Ehrlichia, Candidatus Neoehrlichia, and Rickettsia bacteria in eggs, larvae, and pupae, as well as adult mosquitoes collected in the field. Additionally, Ehrlichia and Rickettsia bacteria were identified in eggs, larvae, pupae, and adults reared in the laboratory. Together, these data suggest that Rickettsiales, including Anaplasma and Ehrlichia spp., may be transmitted transovarially and transstadially in mosquitoes in nature. However, as the detection rates of Ehrlichia and Rickettsia bacteria in laboratory reared mosquitoes were relatively low, further studies are needed to determine the efficiency of transovarial transmission. The prevalence of Anaplasma, Ehrlichia, and Rickettsia bacteria in nature varies substantially with respect to vectors, hosts and geographic regions 4,14,45 . Similarly, ticks collected at different time points in the same locality can display different infection rates 44 . Such variation can be attributed to several factors, including host susceptibility and competence, the availability of different reservoir hosts, and geo-ecologic factors 6,14,36,[45][46][47] . In addition, adult ticks have an additional blood meal compared to nymphs, such that the infection rates of A. phagocytophilum in adult ticks are higher than in nymphs 36 , and because of a lack of transovarial transmission, tick larvae are considered free of Anaplasmataceae bacteria 45 . Our analysis of mosquitoes also revealed variation in infection rate according to geographic location, mosquito species and their life stage, and sampling times. For example, infection rates were higher in adult mosquitoes of An. sinensis (20.50%) and Ar. subalbatus (23.90%), but lower in adult Ae. Albopictus (3.33%), Cu. quinquefasciatus (4.86%), and Cu. tritaeniorhynchus (6.25%), perhaps reflecting further differences in susceptibility. Furthermore, compared with Ehrlichia, Candidatus N. mikurensis, and Rickettsia, Anaplasma bacteria showed the highest infection rates in adult and juvenile mosquitoes collected in the field.
In all the gene trees inferred here, A. marginale from mosquitoes were closely related to strains sampled from cattle 31 . For Candidatus A. rodmosense, a close relationship between mosquito-and rodent-associated bacteria is observed in the rrs gene tree. Hence, these data are compatible with the inter-species transmission of bacteria among mosquitoes and mammals, such that mosquitoes may act as transmission vectors. However, for other Anaplasma bacteria (Candidatus A. boleense, A. bovis, and A. platys), the strains identified here were phylogenetically distinct in all three gene trees, such that additional studies are needed to determine whether they are specifically adapted to mosquitoes.  Bovine anaplasmosis caused by A. marginale is widely distributed in tropical and subtropical regions globally, and responsible for substantial economic losses 4,48 . Indeed, cattle can develop persistent infections and serve as reservoirs of A. marginale 4 . It was therefore notable that the A. marginale bacteria identified here were more common than other bacteria in An. sinensis and Ar. subalbatus mosquitoes, and that these strains are closely related to widely distributed ruminant strains 31,49 . Additionally, An. sinensis and Ar. subalbatus mosquitoes prefer to feed on large animals including cattle, and the geographic distribution of a variety of mosquito species (An. sinensis, Ar. subalbatus, Ae. albopictus, and Cu. quinquefasciatus) overlaps with the distribution of bovine anaplasmosis in China. Consequently, it is possible that mosquitoes are involved in the spread of bovine anaplasmosis, beyond serving as mechanical fomites.
The identification of extensive genetic diversity of Rickettsiales (especially Anaplasma) in adult and juvenile mosquitoes indicates that mosquitoes may have played an important role in the transmission of Rickettsiales bacteria, and that their role as vectors needs to be investigated further. These data are also compatible with the notion that Rickettsiales can be maintained in mosquitoes through transstadial and transovarial transmission. Due to the global distribution of these mosquitoes, greater efforts are clearly needed to determine their role in the evolution and natural transmission of Rickettsiales.

Material and methods
Mosquito collection and identification. During the summer (June to August) in 2014 and 2015, 971 adult mosquitoes were collected by ultraviolet light traps from sheep folds, cattle pens, ponds, creeks, and indoors at night in three regions of China: (i) Wuhan city in Hubei province, (ii) Yudu county in Jiangxi province, and (iii) Cixi city in Zhejiang province (see Supplementary Fig. S1). Similarly, mosquito eggs, larvae and pupae were collected from aquatic environments in Wuhan. All mosquitoes were identified to the species level and life stage based on morphologic criteria 50 and further by molecular differentiation as described previously 51 . The main morphological characters distinguishing mosquito pupae from larvae are that the latter are comma-shaped in their lateral aspect and the head and thorax are merged into a cephalothorax. All samples collected were stored at − 80 °C until DNA extraction.
Mosquito culture. For mosquito culture, adult mosquitoes (fed or unfed, Table 3), including Ae. albopictus, An. sinensis, Ar. subalbatus and Cu. tritaeniorhynchus collected from the natural aquatic environments in Wuhan were reared for a complete life cycle (adult-egg-larvae-pupae-adult) as described previously 50,52 . Larvae and adult Cu. quinquefasciatus were also collected for a complete life circle (adult-egg-larvae-pupae-adult or larvae-pupae-adult-egg-larvae-pupae-adult). Samples were collected from adult/larvae mosquitoes caught in the field and at each life stage of cultured mosquitoes. DNA extraction, PCR amplification, and sequencing. After washing twice with PBS, adult mosquitoes, larvae and pupae were individually homogenized with a mortar and pestle. Approximately 450 An. sinensis and 500 Cu. tritaeniorhynchus eggs were pooled (a total of 19 pools) and homogenized. After homogenization, the suspension was incubated at 4 °C for 1 h and centrifuged at 2,500 g for 5 min, and the upper fraction collected. DNA was extracted from individual mosquitoes or mosquito pools with the QIAamp DNA Mini Kit (Qiagen GmbH, Hilden, Germany) according to the manufacturer's instructions, and then subjected to PCR for amplification of both bacterial gene sequences (rrs, groEL and gltA genes) and mosquito 18S rRNA genes 51 .
Rickettsial DNA was detected using nested PCR targeting a conserved sequence of the Rickettsiales rrs gene using the primers Eh-out1/Eh-out2 (outer primers) and Eh-gs1/Eh-gs2 (inner primers) 53 . Primers designed in-house were also used to amplify complete rrs gene sequences and partial groEL and gltA gene sequences by nested PCR. All primer sequemnces are described in Supplementary Table S4. DNA samples of E. chaffeensis strain Arkansas were used as positive controls, with distilled water used as a negative control. Amplified DNA was purified by electrophoresis in low-melting point agarose and ligated into the cloning vector pMD19-T. Subsequently, each vector was transformed into E. coli and plated onto agarose culture dishes. Twenty clones were picked from each 20 th dish and sent to the Sangon Biotechnology Company (Shanghai, China) for sequencing. To prevent contamination, the pre-PCR mix was prepared in a separate room and template DNA was added using dedicated pipets and tips.
Sequence data and genetic analyses. DNA sequences of the three bacterial genes obtained here were aligned with existing reference sequences in GenBank using ClustalW (default parameters) as implemented in the MEGA program, version 5.2 54 . Nucleotide and amino acid sequence identities were calculated using DNAStar (DNASTAR, Inc., Madison, WI). Data sets of the following sizes were then used in an evolutionary analysis: (i) a 1425 bp rrs alignment (N = 131 sequences); (ii) a 1656 bp groEL gene alignment (N = 82); and (iii) a 1227 bp gltA gene alignment (N = 59). The sequences recovered in this study were named according to their relatedness to known bacteria, geographic origins, and sample numbers. All sequences obtained here have been submitted to GenBank and assigned accession numbers KU585921-KU586334.
Phylogenetic analyses. The best-fit evolutionary model for all sequence alignments was determined using jModelTest 55 and found to be the General Time Reversible (GTR) nucleotide substitution model with a gamma (Γ )-distribution model of among-site rate variation and a proportion of invariable sites (i.e. GTR + Γ + I).
Phylogenetic trees using this model were then estimated using the Maximum Likelihood (ML) method implemented in PhyML (version 3) 56 . All trees were mid-point rooted for clarity only.