Highly divergent and diverse viral community infecting sylvatic mosquitoes from Northeast Brazil

ABSTRACT

Mosquitoes can transmit several pathogenic viruses to humans, but their natural viral community is also composed of a myriad of other viruses such as insect-specific viruses (ISVs) and those that infect symbiotic microorganisms. Besides a growing number of studies investigating the mosquito virome, the majority are focused on few urban species, and relatively little is known about the virome of sylvatic mosquitoes, particularly in high biodiverse biomes such as the Brazilian biomes. Here, we characterized the RNA virome of 10 sylvatic mosquito species from Atlantic forest remains at a sylvatic–urban interface in Northeast Brazil employing a metatranscriptomic approach. A total of 16 viral families were detected. The phylogenetic reconstructions of 14 viral families revealed that the majority of the sequences are putative ISVs. The phylogenetic positioning and, in most cases, the association with a high RNA-dependent RNA polymerase amino acid divergence from other known viruses suggests that the viruses characterized here represent at least 34 new viral species. Therefore, the sylvatic mosquito viral community is predominantly composed of highly divergent viruses highlighting the limited knowledge we still have about the natural virome of mosquitoes in general. Moreover, we found that none of the viruses recovered were shared between the species investigated, and only one showed high identity to a virus detected in a mosquito sampled in Peru, South America. These findings add further in-depth understanding about the interactions and coevolution between mosquitoes and viruses in natural environments.

IMPORTANCE

Mosquitoes are medically important insects as they transmit pathogenic viruses to humans and animals during blood feeding. However, their natural microbiota is also composed of a diverse set of viruses that cause no harm to the insect and other hosts, such as insect-specific viruses. In this study, we characterized the RNA virome of sylvatic mosquitoes from Northeast Brazil using unbiased metatranscriptomic sequencing and in-depth bioinformatic approaches. Our analysis revealed that these mosquitoes species harbor a diverse set of highly divergent viruses, and the majority comprises new viral species. Our findings revealed many new virus lineages characterized for the first time broadening our understanding about the natural interaction between mosquitoes and viruses. Finally, it also provided several complete genomes that warrant further assessment for mosquito and vertebrate host pathogenicity and their potential interference with pathogenic arboviruses.

INTRODUCTION

Mosquitoes transmit several high-impact viral pathogens (arthropod-borne viruses or arboviruses) to humans and animals (1)https://www.zotero.org/google-docs/?U1z3GS. But besides these pathogenic arboviruses, an increasing number of studies reveal that mosquitoes host a much larger diversity of viruses, which are collectively called virome (2–5)https://www.zotero.org/google-docs/?LDfMtq. Mosquito viromes are composed of the most abundant and prevalent insect-specific viruses (ISVs) that only infect arthropods (6, 7)https://www.zotero.org/google-docs/?knRKzR and the less prevalent and human pathogenic arboviruses (8)https://www.zotero.org/google-docs/?kWxU4a. Mounting evidence shows that ISVs may interfere on arbovirus replication either increasing or reducing viral replication in the mosquito body and hence impacting vector competence of specific arboviruses (6, 9, 10)https://www.zotero.org/google-docs/?Rubb9p. Some examples of ISV–arbovirus pairs with known interference phenomenon are as follows: Palm Creek Virus–West Nile virus (11)https://www.zotero.org/google-docs/?QwemBd; Nhumirim virus–Zika virus and Dengue virus 2 (12)https://www.zotero.org/google-docs/?B0IZKV and Eilat virus–Venezuelan equine encephalitis, Eastern equine encephalitis virus, Chikungunya virus and West equine encephalitis virus (13)https://www.zotero.org/google-docs/?jwVPzq.

The discovery and characterization of viruses is historically a laborious process requiring cell isolation and classical virology experimentation and analysis (14–16)https://www.zotero.org/google-docs/?sVKTQm. However, most viruses are not amenable to cell isolation in laboratory conditions limiting our understanding of the natural viral communities to the culturable ones (15)https://www.zotero.org/google-docs/?eKGHko. The nucleic acid sequencing revolution of the last decades has opened new possibilities for more comprehensively characterizing natural mosquito viromes. Among several strategies available, bulk metatranscriptome sequencing stands out as one of the least biased approaches for RNA virus genome sequencing (17)https://www.zotero.org/google-docs/?NJ5Nck. Coupled with large-scale metagenomic sequencing, new bioinformatic tools are accelerating virus discovery and characterization (18, 19)https://www.zotero.org/google-docs/?FlmaJN. However, continued development and integration of new tools is required both because of the increasing data volume and the ever-increasing difficulty to characterize highly divergent virus genomes (virome dark matter), which compose a large fraction of every virome sequenced (14)https://www.zotero.org/google-docs/?XfRMP8.

Currently, the majority of mosquito viruses characterized were identified in species from Culex, Anopheles, and Mansonia genera due to a focus on epidemiological important mosquito species (2, 7, 20)https://www.zotero.org/google-docs/?krVXcA. Hence, the large majority of the mosquito species diversity, which likely transmit viral pathogens in sylvatic transmission cycles, have not been accessed regarding its virome composition. Yet, due to our narrow view of the virosphere and particularly the mosquito viral communities, every new mosquito virome study has revealed many novel viruses. For instance, a recent review of mosquito virome has shown that 14 mosquito genera were positive for viruses, which were assigned into 102 viral families (20)https://www.zotero.org/google-docs/?YzcbCj. In Brazil, there are some scattered studies focusing on characterizing the virome of sylvatic mosquitoes covering different genera, such as Anopheles, Aedes, Culex, Psorophora, Sabethes, Coquillettidia, and Mansonia, sampled at different biomes as the Amazon, Cerrado, Caatinga, Pantanal, and Atlantic forest (21–25)https://www.zotero.org/google-docs/?0rZtwJ. However, some biomes cover extensive territory, and more comprehensive spatiotemporal sampling of mosquitoes is necessary to characterize pathogenic and non-pathogenic viral components of the natural mosquito microbiome.

Here, we performed metatranscriptome sequencing of 10 different mosquito species sampled at a sylvatic–urban interface of the Atlantic forest in Northeast Brazil aiming to characterize its RNA virome composition and potential viral threats to humans. We characterized a total of 16 different viral families. The newly discovered viruses exhibit high divergence from known viruses and clustered with previously identified mosquito-specific viruses, but the low amino acid identity at the most conserved protein [RNA-dependent RNA polymerase (RdRp)] suggests that they belong to several new viral species.

MATERIALS AND METHODS

Sample collection and species identification

Mosquito sampling and species identification were performed using the same approach described previously (21)https://www.zotero.org/google-docs/?VbTWzw. In brief, we collected mosquitoes using entomological nets during morning and afternoon/evening in two areas of Atlantic forest remains in Northeast Brazil [Fig. S1, Parque Estadual Dois Irmãos (PEDI) (8°00′43.3″S, 34°56′40.7″W) and Jardim Botânico do Recife (JBR) (8°04′33.0″S, 34°57′35.9″W), Recife munici­pality, state of Pernambuco, Brazil]. The collected mosquitoes were transported alive to the Entomology Department of the Aggeu Magalhães Institute (IAM-FIOCRUZ) and stored at −80°C until taxonomic identification. Collected mosquito samples were taxonomically identified using dichotomous keys available in the literature for neotropical Culicidae (https://www.zotero.org/google-docs/?NkleZg26). All specimens were processed in a cooled bench (~0°C) to prevent RNA degradation. After morphological identification, each specimen had its abdomen dissected and processed for DNA extraction using the protocol from Ayres et al. (27)https://www.zotero.org/google-docs/?lc3l0P, while the remaining tissues were stored at −80°C until species-specific pooling and RNA extraction procedures. We amplified the cytochrome c oxidase subunit I (COI) for individual abdomen DNA using the GoTaq (Promega) protocol and carried out sequencing using the Sanger method on an ABI 3500xL (Applied Biosystems), following the manufacturer’s instruction. Electropherograms were analyzed on Geneious Prime 2020.0.5 to extract the consensus fasta sequence. A BLASTn analysis (28)https://www.zotero.org/google-docs/?8MleUU was carried out with these sequences against a custom database composed by National Center for Biotechnology Information (NCBI) and Barcode of Life Data Systems (BOLD) mosquito COI sequences (retrieved on 4 March 2020) and also a database of mitochondrial genomes from local mosquitoes from da Silva et al. (29) to confirm and ensure the mosquito species used in the pooling scheme belong to the same taxon. The only exception was the Aedes albopictus samples that were only identified by morphological characters due the ease in visualization of species-specific morphological characters. The reconstructed phylogenetic tree of the COI sequences corroborated previous database similarity searches (Fig. S2).

RNA extraction and sequencing

Based on previous mosquito identification, the remaining tissues stored at −80°C from the same taxon were pooled in ultrapure water and macerated using a pestle motor tissue grinder. One hundred microliters (µl) of macerated samples was used to extract the total RNA following the Trizol protocol as suggested by the fabricant (Invitrogen, USA). The pellet was eluted in a final volume of 30 μl. The extracted nucleic acids were treated with TURBO™ DNase (2 U/µl—Ambion) following the manufacturer’s instructions. RNA samples were quantified and quality checked through Qubit RNA HS kit and Bioanalyzer, respectively. Total RNA samples were processed for ribosomal RNA depletion with the RiboMinus Eukaryote System v2 kit following manufacturer’s instructions (Invitrogen, USA). The sequencing library was prepared using the TruSeq Stranded Total RNA library kit (Illumina, USA) and sequenced on a NextSeq 500 Illumina platform using a paired-end approach of 75 bp.

Virome characterization and taxonomic classification

The viromes were characterized following the study workflow presented in Fig. S3. Sequenced reads were firstly quality checked using FastQC tool (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) followed by trimming of low-quality reads using Trimmomatic (30)https://www.zotero.org/google-docs/?Df8gCU with the following parameters: LEADING:3 TRAILING:3 SLIDINGWINDOW:4:20 MINLEN:36 and TruSeq2-PE as the adapter file to be removed. The HTML reports from FastQC tool were parsed using the MultiQC tool (31)https://www.zotero.org/google-docs/?L1xkUl that combined the data in a single report used for general sequencing reads statistics recovery. The metatranscriptome assemblies were generated using three different assembly tools: Trinity v.2.11 (32)https://www.zotero.org/google-docs/?25Zkhi, rnaSpades (33),https://www.zotero.org/google-docs/?ADx3nr and metaSpades (34)https://www.zotero.org/google-docs/?79SMK0 in default mode. This approach was used to obtain a more complete assembly following findings by previous studies (35, 36)https://www.zotero.org/google-docs/?LCYcjT. All contigs obtained from the three different assembly tools described above were clustered using the CD-HIT-EST tool (37)https://www.zotero.org/google-docs/?yXLeeO with the following parameters: -c 0.98 G 0 -aS 0.9 g 1 n 9, to exclude redundant sequences having ≥98% of identity and retain the longer contigs. To identify and recover viral sequences from the assembled metatranscriptomes, we first downloaded protein sequences assigned with the viral taxonomy tag (txid10239) from the National Center for Biotechnology Information (NCBI) on 12 December 2020. The sequences were clustered using the CD-HIT tool (37)https://www.zotero.org/google-docs/?7B8vle with the following parameters: -G 0 -aS 0.95 g 1 M 100000 n 5 aiming to remove redundant sequences. Based on the assembled metatranscriptome, we predicted the amino acid sequences encoded from contigs using the Prodigal tool (Hyatt et al., 2010) employing the meta flag. Then, we identified the viral contigs using DIAMOND (38)https://www.zotero.org/google-docs/?zROqoj performing a blastp search of the predicted amino acid sequences against the retrieved viral protein database (txid10239). Four additional searches were performed using specific databases of viral RdRp sequences such as NeoRdRp (39)https://www.zotero.org/google-docs/?BsIGcM, PalmDB (40)https://www.zotero.org/google-docs/?UMCWRv, RVMT (41),https://www.zotero.org/google-docs/?bogd9e and RdRp-Scan (42)https://www.zotero.org/google-docs/?yKBfpc. These databases were used aiming to recover divergent and unclassified viruses that are not present in the NCBI txid10239 database. False-positive viral hits were removed through a second DIAMOND blastp search against the complete non-redundant protein database (NR) from NCBI (retrieved on 12 December 2023). Only contigs where the first five best hits of the predicted amino acid sequences that showed similarities with viral proteins in both DIAMOND analysis were kept for further analysis. All DIAMOND analyses were run using the E-value of 1E⁻³ and the more-sensitive parameter.

To further in-depth characterize more distant related viral contigs that might have been missed from the DIAMOND search, we performed another approach of viral identification based on Hidden Markov model (HMM) profiles using HMMER3 (http://hmmer.org/). The predicted amino acid sequences were used in the hmmsearch analysis comparing them with HMM models available from the RVDB-prot-HMM database and those HMM models available for the previously cited databases (RVMT, RdRp-Scan, and NeoRdRp). Hits showing an E-value of 1E⁻³ were selected for further analysis.

To exclude endogenous viral elements (EVEs) that are actively transcribed from the mosquito genomic loci, we performed a BLASTn search using the previously identified viral contigs (DIAMOND and HMM analysis) as queries against a mosquito genome database ( Tables S2 and S3). First, we downloaded all available mosquito genomes from NCBI (retrieved on 9 September 2023, Table S2) and combined with draft whole mosquito genomes sequenced on Hiseq or Novaseq Illumina technology for the studied species (Table S3). These draft mosquito genomes used here were assembled using the megahit tool (43)https://www.zotero.org/google-docs/?Anvorw. The viral contigs showing hits with genomic regions containing a minimum query coverage of 50% and more than 90% of identity were excluded from further analysis as similarly performed in other studies (44)https://www.zotero.org/google-docs/?uDhiUW.

To obtain the virome profile for the studied mosquito species, we filtered the contigs larger than 600 bp and used the taxonomic annotation from the set of sequences containing RdRp hallmark protein, the most conserved and used marker to viral classification. The best hits from DIAMOND and HMM analyses for each query were used to obtain the viral contig taxonomy information. Since RdRp naming in both NCBI sequences or RVDB-hmm-prot profiles can vary, we performed searches for different terms aiming to filter the RdRp sequences from previous analysis (Table S1). Taxonomic information recovered from DIAMOND was used when sequences showed taxonomic information from both methodologies (DIAMOND and HMM analysis). Sequences showing best hits with unclassified viruses were phylogenetically analyzed and reclassified based on annotation of the closest homologs as long as it clustered in monophyletic clades with high node support.

To obtain the number of reads per viral contig, we mapped quality-trimmed reads against the identified viral contigs using Bowtie2 in default mode, and statistical metrics were collected using coverM tool (https://github.com/wwood/CoverM).

To evaluate viral genome/segment completeness of the identified viral contigs, we used the ViralComplete module from Metaviralspades tool (45)https://www.zotero.org/google-docs/?aAmYJP. Briefly, the ViralComplete analysis predicts the genes and the viral proteins from the contigs and performs a BLAST analysis against the viral RefSeq databases. Moreover, the tool compares if the length of analyzed contig is similar to the viral hit and then classifies them into full-length or partial sequences.

Phylogenetic analysis

Phylogenetic trees were reconstructed for each viral family that showed contigs encoding RdRp sequences larger than 600 bp and amino acid sequences larger than 100 aa. We included in the analysis approximately 50 best hit sequences recovered from DIAMOND blastp against the NR database. The viral proteins were aligned using MAFFT v.7 (46)https://www.zotero.org/google-docs/?QEdBxn with default parameters. The non-aligned blocks were removed from the alignment using the Trimal tool (47)https://www.zotero.org/google-docs/?fBjmjZ. The phylogenetic trees were reconstructed using the IQ-TREE 2.0 (48),https://www.zotero.org/google-docs/?Xc43En and the best evolutionary model for each alignment was selected by the ModelFinder (49)https://www.zotero.org/google-docs/?cI8jeC implemented in the same tool. The node supports were evaluated with 1,000 replicates of ultrafast bootstrapping (50)https://www.zotero.org/google-docs/?42QluA. The consensus trees were visualized and annotated with the ggtree (51)https://www.zotero.org/google-docs/?jgod82 package from R programming language. The genomic maps displayed on trees were built using the gggenomes R package (https://github.com/thackl/gggenomes).

The phylogenies of mitochondrial genomes and COI from mosquitoes were reconstructed using the same approach described above. In brief, draft mitochondrial genomes for each mosquito species were assembled using the metatranscriptome data on MITObim 1.9 (52)https://www.zotero.org/google-docs/?yIHuJw. Then, nucleotide sequences were aligned using MAFFT v.7 and trimmed using the Trimal tool. The phylogenetic tree was obtained and plotted using IQ-TREE 2.0 and Figtree v1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/) respectively.

RESULTS

A total of 194 specimens were processed and identified at species level. The mosquito species were split into 10 pools representing the following species: Mansonia titillans, Mansonia wilsoni, Ae. albopictus, Psorophora ferox, Aedes scapularis, Coquillettidia chrysonotum, Coquillettidia venezuelensis, Limatus durhamii, Coquillettidia hermanoi, and Coquillettidia albicosta (Fig. S1 ;Table S4).

Illumina sequencing generated a total of 88.99 Gbp representing 593.3 million of paired end reads distributed along the 10 mosquito pools (Table 1). The generated reads per pool ranged from 47.1 million of paired reads for Ae. scapularis to 69.1 million of paired reads for Cq. chrysonotum (Table 1). A total of 1,971,030 contigs were generated from the three assembly approaches (Trinity, metaSPAdes, and rnaSPAdes), while 1,285,976 were retained after redundancy removal (Table 1).

TABLE 1.

General sequencing and viral identification statistics

Species name	Library name	Million of raw paired reads	Million of trimmed paired readsa	Total contigsb	Total clustered contigs	Total viral contigsc	Total viral reads	Full-length viral seqsd
Cq. venezuelensis	P2	65.40	61.8	161,589	99,069	14	7,311	4
Ma. wilsoni	P8	50.20	47.1	204,403	129,918	13	89,143	5
Cq. albicosta	P1	61.30	57.1	263,504	178,387	10	21,406	1
Ma. titillans	P10	68.40	64.4	210,612	132,164	9	748,526	3
Li. durhamii	P5	51.40	48.5	139,978	80,031	7	48,365	0
Cq. chrysonotum	P6	69.50	65.6	175,237	115,351	3	18,428	1
Cq. hermanoi	P9	58.80	54.7	289,556	192,588	3	207	0
Ae. albopictus	P7	69.10	64.5	260,363	178,599	2	49,007	2
Ae. scapularis	P3	47.10	44.3	139,261	90,437	2	13,633	2
Ps. ferox	P4	52.10	48.5	126,527	89,432	1	3,173	0
Total	-	593.30	556.5	1,971,030	1,285,976	64	999,304	18

^a Number considering the paired reads plus the unpaired.

^b Considering Trinity, rnaSPAdes and metaSPAdes assemblies.

^c Considering only those viral contigs with RdRp markers.

^d Analysis from ViralComplete.

After performing DIAMOND searches using the viral protein database (NCBI taxid10239), we were able to identify 61 viral contigs (Fig. 1A). The searches based on RdRp databases were able to identify additional 15 viral contigs (Fig. S4A and B). The majority of viral contigs were identified by all four RdRp databases (Fig. S4B). However, in some cases, only two (RDRP-SCAN and RVMT) or three (NEORDRP, RVMT, and RDRP-SCAN) databases have identified specific viral contigs (Fig. S4B). The NEORDRP database was the only one that could identify viral contigs not previously identified by the other databases (Fig. S4B). By comparing the viral contigs identified using RdRp databases together with those identified from NCBI taxid10239 database, we obtained a total of 63 viral contigs (Fig. 1A). The majority of them were identified uniquely by the NCBI taxid10239 database and the minority using the RdRp databases (Fig. 1A). Two viral contigs were only identified by RdRp databases using NEORDRP, RVMT, and RDRP-SCAN (Fig. 1A). The HMM analysis was able to identify a total of 16 viral contigs, and the majority were identified by all four HMM model data sets (Fig. 1B). No viral contig was uniquely identified by one specific HMM model data set. By comparing the viral contigs identified by both approaches, among the 16 identified using HMM analysis, 15 were also identified using DIAMOND analysis, and only one was uniquely identified using HMM models (Fig. S4C).

Fig 1.

General information on viral identification. (A) Venn diagram illustrating the overlap of viral contigs identified using different viral databases on DIAMOND pairwise alignment approach. (B) Venn diagram illustrating the overlap of viral contigs identified using different HMM model search. (C) Sankey diagram showing the viral families identified by the different approaches employed on viral identification.

A total set of 64 viral contigs were identified using the two approaches (DIAMOND and HMM analysis), and we were able to identify 16 viral families (Fig. 1C, 2D and E). The majority of viral families (Aliusviridae, Chrysoviridae, Dicistroviridae, Endornaviridae, Flaviviridae, Iflaviridae, Narnaviridae, Orthomyxoviridae, Partitiviridae, Phasmaviridae, Phenuiviridae, Rhabdoviridae, Totiviridae, Virgaviridae, Xinmoviridae) were identified using only DIAMOND and NCBI taxid10239 database, while Solemoviridae was identified by NCBI taxid10239, NEORDRP, RVMT, and RDRP-SCAN databases (Fig. 1C), and Ifaviridae was identified only by NEORDRP, RVMT, and RDRP-SCAN. Only one viral contig was identified using HMM analysis, and it was not able to be classified (Fig. 1C). The majority of the viral contigs were classified into known viral families (Fig. 2A). Three viral genomic architectures were found analyzing all data sets of identified viral contigs: the majority belonged to ssRNA(−), followed by ssRNA(+), and few viral contigs were annotated as dsRNA (Fig. 2C; File S1). Furthermore, we recovered a total of 18 full-length viral sequences from segmented viruses or linear complete genome from eight viral families (Orthomyxoviridae, Partitiviridae, Phenuiviridae, Rhabdoviridae, Solemoviridae, Totiviridae, Virgaviridae, and Xinmoviridae) (Fig. 2B; Table 1), while the remaining sequences were classified as partial. The full-length sequences ranged from 1.75 to 13.08 kb (Fig. S5).

Fig 2.

Virome composition across studied mosquito species. (A) Stacked plot presenting the number of viral contigs assigned or not assigned to known viral families across mosquito viromes. (B) Stacked plot showing the viral genome completeness of identified viral contigs. Number of viral sequences assigned as full-length or partial from this study analyzed by the ViralComplete tool. (C) Stacked plot showing the different viral genome structures of identified viral contigs across the mosquito viromes. Genome structure information for each family was obtained from the NCBI Virus database. (D) Stacked plot displaying the proportion of viral families identified among mosquitoes. (E) Phylogenetic tree of mosquito species with a Sankey diagram of viral families identified across mosquitoes. The phylogenetic tree was reconstructed based on the mitogenomes assembled from sequenced data using MITObim 1.9 and analyzed on IQ-TREE 2.0 performing the ultrafast bootstraping with 1,000 replicates and the evolutionary model GTR + F + G4 selected by the ModelFinder. The root tree was set in an outgroup (Drosophila melanogaster—U37541.1 that was pruned from the tree).

Regarding the mosquito species, Cq. venezuelensis, Ma. wilsoni, and Li. durhamii showed the highest viral contig content, respectively (Table 1), while, Ps. ferox and Ae. scapularis showed the smallest number (Table 1). The viral contig sequences showed a wide length distribution ranging from 602 bp up to 13,088 kb with a median value of 2,832 bp (Fig. S6). We were able to recover five full-length viral sequences for Ma. wilsoni assigned in Xinmoviridae (N = 1), Orthomyxoviridae (N = 2), Virgaviridae (N = 1), and Phenuiviridae (N = 1) families, while only one full-length sequence for each Cq. albicosta and Cq. chrysonotum species, representing viruses from Rhabdoviridae and Solemoviridae, respectively (Table 1; Fig. S4; File S1). Two mosquito species (Ps. ferox and Cq. hermanoi) did not show any full-length sequences.

The virome profile of the mosquito species investigated here have shown a similar pattern of viral families across the mosquito tribes (Fig. 2D and E; Fig. S7). The virome of Limatus durhamii from the Sabethini tribe has shown the most distinct profile (Fig. 2D). In the Aedini tribe (Aedes and Psorophora genera), we identified a limited number of viral contigs that allowed us to have only a narrow view of the virome from these species, while the Coquillettidia and Mansonia species from Mansoniini have shown a similar viral family profile among them (Fig. 2D and E).

In the Li. durhamii species, we identified six viral families: Iflaviridae, Phasmaviridae, Flaviviridae, Narnaviridae, Virgaviridae, and Endornaviridae. Two families were exclusively identified in this mosquito species (Iflaviridae and Enfdornaviridae). Within the Aedini species, a few contigs were classified into the Rhabdoviridae family and identified in both Ae. albopictus and Ae. scapularis. Phenuiviridae was only identified on the Ae. scapularis pool and Chrysoviridae from Ps. ferox (Fig. 2D and E; Fig. S7). Excluding the Chrysoviridae family, which infects predominantly plants and fungi, the viral families identified in Ae. albopictus and Ae. scapularis are known to infect both arthropods and vertebrates (Fig. 2E; Fig. S7).

In the Mansoniini viromes, both species from Mansonia and Coquillettidia genera were infected by viruses belonging to the viral families Xinmoviridae, Virgaviridae, and Solemoviridae (Fig. 2D and E and Fig. S7). However, several other viral families were uniquely identified in specific species within Mansoniini such as Chrysoviridae and Aliusviridae in Cq. venezuelensis; Phasmaviridae in Cq. hermanoi; Narnaviridae and Partitiviridae in Ma. titillans; and Orthomyxoviridae, Flaviviridae, Phenuiviridae, and Dicistroviridae in Ma. wilsoni. In general, Mansoniini mosquitoes showed viral sequences belonging to viral families known to infect arthropods, vertebrates, birds, fungi, and plants (Fig. S7). The Ma. wilsoni species showed the highest number (7) of viral families from Mansonini species studied here: Dicistroviridae, Xinmoviridae, Flaviviridae, Phenuiviridae, Orthomyxoviridae, Solemoviridae, and Virgaviridae, while only Phasmaviridae was identified in Cq. hermanoi.

Phylogenetic reconstruction of virus relationship history

Based on the 16 viral families identified in the mosquito viromes, we reconstructed a total of 15 phylogenetic trees using the RdRp hallmark gene (Fig. S8 to S22) representing 14 viral families (Aliusviridae, Chrysoviridae, Flaviviridae, Iflaviridae, Narnaviridae, Orthomyxoviridae, Partitiviridae, Phasmaviridae, Phenuiviridae, Rhabdoviridae, Solemoviridae, Totiviridae, Virgaviridae, and Xinmoviridae). Besides those, we also reconstructed a phylogenetic tree for one viral sequence with unclassified assignment. From now, we will present the results of viral families known to infect arthropods only (Aliusviridae, Xinmoviridae, Iflaviridae, and Phasmaviridae) and those able to infect two or more hosts, infecting arthropods and vertebrates (Flaviviridae, Phenuiviridae, Orthomyxoviridae, and Rhabdoviridae) (Fig. 3 and 4). Phylogenetic trees of viral families with known plants, fungi, protozoa, and oomycete hosts that were also identified from the mosquito viromes can be found in Supplementary material 1 (Fig. S17 to S22).

Fig 3.

Phylogenetic trees of the Iflaviridae, Xinmoviridae, and Phasmaviridae families. The phylogenetic trees were reconstructed based on aligned sequences representing the RdRp identified for each viral family and analyzed on IQ-TREE2.0 performing the ultrafast bootstraping with 1,000 replicates. Full phylogenetic trees can be seen in Supplementary material 1—Fig. S8 to S11. The trees were set as the midpoint root. The tip colors represent the different mosquito host species from this study, while silhouettes represent the host species recovered from the NCBI. Gray boxes in phylogenies indicate the subgenus of viruses according to information from NCBI and ICTV. Black features on genomic maps indicate the portion corresponding to the reference genome included in the analysis. Red features represent the regions used from our contigs. Gray links represent the region that matches with the reference.

Fig 4.

Phylogenetic trees of Phenuiviridae, Orthomyxoviridae, and Rhabdoviridae families. The phylogenetic trees were reconstructed based on aligned sequences representing or the RdRp or PB2 for Orthomyxoviridae family and analyzed on IQ-TREE2.0 performing the ultrafast bootstrapping with 1,000 replicates. Full phylogenetic trees can be seen in Supplementary material 1—Fig. S13 to S15. The trees were set as the midpoint root. The tip colors represent the different mosquito host species from this study, while the silhouettes represent the host species recovered from the NCBI. Gray boxes in phylogenies indicate the subgenus of viruses according to information from NCBI and ICTV. Black features on genomic maps indicate the portion corresponding to the reference genome included in the analysis. Red features represent the regions used from our contigs. Gray links represent the region that matches with the reference.

Arthropod exclusive viruses

Iflaviridae

For the Iflaviridae family, we identified one sequence derived from a virus that we designated as Limatus durhamii Iflaviridae virus 1 in Li. durhamii species. Our phylogenetic reconstruction revealed this virus grouping together with other viruses from the Iflavirus genus (Fig. 3A; Fig. S8). This virus was placed as a basal branch together with other viruses identified mainly in Aedes and Culex spp. However, Limatus durhamii Iflaviridae virus 1 has shown to be highly divergent from viruses recovered from the literature. The RdRp amino acid identity values among Limatus durhamii Iflaviridae virus 1 and those viruses inside this clade ranged from 55% to 59% (Fig. S24; File S2).

Xinmoviridae family

For the Xinmoviridae family, we reconstructed a tree including seven sequences identified in Cq. albicosta (N = 4), Cq. chrysonotum (N = 1), and Ma. wilsoni (N = 2) (Fig. 3B; Fig. S9B and S10). In general, our sequences were placed into two distinct clades (Fig. 3B; Fig. S9B and S10). The clade I grouped partial and complete sequences from Cq. albicosta (Coquillettidia albicosta Xinmoviridae virus 2) and viral sequences from Ma. wilsoni (Mansonia wilsoni Xinmoviridae virus 1), while the basal sequence from this clade was Malby virus (UYE93836.1), a virus identified in Aedes communis from Sweden (Fig. S10). A distinct separation of viruses from Mansonia and Coquillettidia was evident. Specifically, Mansonia wilsoni Xinmoviridae virus 1 and Mansonia titillans Xinmoviridae virus 1 that clustered together in a subclade, while Coquillettidia albicosta Xinmoviridae virus 2 and Coquillettidia albicosta Xinmoviridae virus 4 were grouped together into a separated subclade (Fig. S10). The identity values of our sequences with Malby virus ranged from 47% to 59% suggesting a high divergence of those sequences in relation with sequences from the literature (Fig. S24; File S2). In general, both major clades were represented by sequences derived from mosquito hosts.

We also analyzed partial Xinmoviridae sequences from Cq. albicosta (Coquillettidia albicosta Xinmoviridae virus three and Coquillettidia albicosta Xinmoviridae virus 1) and Cq. chrysonotum (Coquillettidia chrysonotum Xinmoviridae virus 1), which encoded truncated small open reading frames (ORFs) (Fig. S9B). These sequences were grouped in clade II together with Gordis virus (QRW42745.1) and Culex mononega-like virus 1 (QGA70931.1) (Fig. S10). The sequence from Coquillettidia albicosta Xinmoviridae virus 1 showed to be more divergent being placed as a sister branch of the remaining sequences of Cq. albicosta and Cq. chrysonotum (Fig. S10). However, the positioning inside the clade, was not supported due the low UFboot value for most of the nodes.

Phasmaviridae family

For the Phasmaviridae family, we positioned three sequences from Cq. hermanoi (N = 2) and Li. durhamii (N = 1) (Fig. 3C; Fig. S11). Our sequences were assigned into Orthophasmavirus genus and clustered in distinct clades. The sequence from Limatus durhamii Phasmaviridae virus one was placed together with Miglotas virus (QRW41773.1), identified in Culex erythrothorax from the United States. Coquillettidia hermanoi Phasmaviridae virus 1 is more divergent and was placed as a sister branch of other Phasmaviridae viruses identified in Culex, Aedes, and Mansonia species, besides the Coquillettidia hermanoi Phasmaviridae virus 2 (Fig. S11). The RdRp identity among Coquillettidia hermanoi Phasmaviridae virus one and other viruses from the clade ranged from 31% to 63%.

Aliusviridae family

Two viral sequences encoded from a contig (Coquillettidia venezuelensis Aliusviridae virus 1) identified in Cq. venezuelensis were analyzed (Fig. S9A and S12). All sequences analyzed from Coquillettidia venezuelensis Aliusviridae virus 1 grouped into a clade together with Atrato Chu-like virus 5 (YP_010798469.1), a virus originally identified in Ps. albipes. These viral sequences showed an RdRp identity between 56% and 62% (Fig. S24; File S2). The Coquillettidia venezuelensis Aliusviridae virus 1 was classified as members of Obscurusvirus according to phylogenetic positioning and grouped in a major clade with other mosquito-derived sequences.

Dual host viruses infecting arthropod and vertebrates

Phenuiviridae family

For the Phenuiviridae family, we positioned the two sequences representing complete segments identified in Ae. scapularis and Ma. wilsoni (Fig. 4A; Fig. S13). Our Phenuiviridae viral sequences were grouped in two distinct clades. Mansonia wilsoni Phenuiviridae virus 1 grouped together with the Narangue virus (YP_010839995.1) with an RdRp identity of 65%. This virus was characterized from Ma. titillans species sampled in Colombia, while the Phenuiviridae sequence from Ae. scapularis was placed together with Salarivirus Mos8CM0 (API61884.1) showing an RdRp amino acid identity of 65%. This virus was identified firstly in non-identified mosquito samples from the United States.

Orthomyxoviridae family

For the Orthomyxoviridae family, we reconstructed a tree comprising one sequence of PB2 subunit that compose the RdRp of Orthomyxoviridae viruses (Fig. 4B; Fig. S14). The sequence derived from Mansonia wilsoni Orthomyxoviridae virus 1 grouped into a clade of sequences classified as Quaranjavirus and was close to Astopletus virus (QRW42566.1) showing high divergence with a PB2 amino acid identity of 46% (Fig. 4B; Fig. S14).

Rhabdoviridae family

For the Rhabdoviridae family, we included five sequences representing complete genomes detected in Ae. scapularis (N = 1), Ae. albopictus (N = 2), Ma. titillans (N = 1), and Cq. albicosta (N = 1) (Fig. 4D and E; Fig. S15). In general, our sequences were clustered into three distinct clades and classified as members of Alphahymhavirus (Aedes scapularis Rhabdoviridae virus 1), Almendravirus (Aedes albopictus Rhabdoviridae virus 1 and Aedes albopictus Rhabdoviridae virus 2), and Merhavirus (Mansonia titillans Rhabdoviridae virus 1 and Coquillettidia albicosta Rhabdoviridae virus 1) genera, respectively. The Aedes scapularis Rhabdoviridae virus 1 grouped into a clade together with other viruses identified in Aedes and Culex genera such as Atrato Rhabdo-like virus 3 (QHA33680.1), San Gabriel mononegavirus (DAZ85658.1), Primus virus (QIS62334.1), and XiangYun mono-chu-like virus 4 (UUG74104.1). The Aedes scapularis Rhabdoviridae virus 1 was positioned as a basal branch in relation to San Gabriel mononegavirus, Primus virus, and XiangYun mono-chu-like virus 4 (Fig. 4C - Clade I; Fig. S15) and showed an RdRp amino acid identity of 52% with Atrato Rhabdo-like virus 3. The two viruses identified in Ae. albopictus (Aedes albopictus Rhabdoviridae virus 1 and Aedes albopictus Rhabdoviridae virus 2) were positioned in two distinct branches within the Almendravirus clade (Fig. 4D - Clade II). Specifically, Aedes albopictus Rhabdoviridae virus 1 clustered closely with Puerto Almendras virus (YP_009094394.1), a virus identified in Aedes fulvus from Peru. Our analysis revealed that these two viruses are probably the same virus given the remarkably high amino acid identity of the RdRp—98%. The other Almendravirus identified in Ae. albopictus was Aedes albopictus Rhabdoviridae virus 2 that clustered together other Almendraviruses identified in Aedini mosquitoes from Psorophora and Armigeres genera (Fig. 4D - Clade II; Fig. S15). The last clade of Rhabdoviridae sequences identified in Mansoniini species from this study, clustered Mansonia titillans Rhabdoviridae virus 1 and Coquillettidia albicosta Rhabdoviridae virus 1 together with Merida virus (AWJ96718.1), a virus identified in Cx. quinquefasciatus from the United States. The identity of RdRp from those viruses with Merida virus was approximately 49% (Fig. 4E - Clade III; Fig. S15).

Flaviviridae family

For the Flaviviridae, we analyzed three partial sequences that encoded truncated small ORFs (Fig. S9C and S16). These sequences were placed into Flavivirus genus and we defined as Mansonia wilsoni Flaviviridae virus 1, Mansonia wilsoni Flaviviridae virus 2 and Limatus durhamii Flaviviridae virus 1 (Fig. S9C and S16). The sequences from Mansonia and Limatus were placed into distinct clades, where the viruses derived from Ma. wilsoni clustered together with Mansonia flavivirus (BCI56826.1) that was characterized from mosquitoes of the Mansonia genus in Bolívia. The identities of analyzed sequences ranged from 65% to 72% that suggest a high divergence of the Flaviviridae viruses identified here in relation to those available in the literature. Limatus durhamii Flaviviridae virus 1 was placed into another clade together with other viruses identified in Sabethes, Culex, and Culiseta mosquito genera. However, their positioning inside the Flaviviridae tree was not solved due the low UFBoot node support values.

DISCUSSION

Mosquitoes carry an abundant and diverse viral community. A recent review has shown that viruses were investigated and detected in at least 128 mosquito species and 14 genera around the globe (20)https://www.zotero.org/google-docs/?dbJ6hb. Most of these viruses have been identified in species of Culex, Aedes, and Anopheles genera that have been the focus of extensive research efforts, once there are medically important pathogen vectors within these genera (7, 20, 53, 54)https://www.zotero.org/google-docs/?us1Xq9. However, it is crucial to study the virome composition from other mosquito genera and species including species inhabiting sylvatic and urban–sylvatic interfaces once the majority of arthropod-borne viruses have emerged and are maintained in the sylvatic environment and may be transmitted to humans through the so-called bridge vectors (55)https://www.zotero.org/google-docs/?LsdxN9. Studying the virome from sylvatic mosquito could not only bring information from novel viruses that are maintained in nature but also pinpoint potentially human pathogenic viruses as well as reveal non-pathogenic viruses that have the potential to reduce replication from other arboviruses.

While metatranscriptome sequencing offers an unbiased approach to characterizing a diverse number of viruses, some challenges for accurate viral identification still exists. Previous studies have highlighted that the choice of reads assembler can impact virome characterization (35, 36)https://www.zotero.org/google-docs/?NY1hWv. Additionally, the use of specific databases on similarity sequence methods can also impact viral identification. With the aim of reducing the knowledge gap in the virome of sylvatic mosquitoes from Brazil, we conducted metatranscriptome sequencing of 10 distinct species and evaluated different viral identification approaches coupled with different viral databases. The databases included in this study encompassed sequences available on the NCBI to specialized RdRp databases that included divergent and unclassified viral sequences that are not present in most popular databases. Our analysis indicates that conventional databases containing viral proteins are generally adequate for detecting the majority of viruses including more divergent ones. However, some specific viral contigs were exclusively identified using specialized databases or using HMM models. Therefore, coupling of different approaches and databases on viral identification offer a comprehensive and complementary understanding of virome composition.

In Brazil, there have been few studies focusing on the virome of mosquitoes, and the ones published covered a limited number of mosquito species from different biomes such as Cerrado, Caatinga, Pantanal, Amazon and Atlantic forest from Southeast and Northeast Brazil (21–25, 56, 57)https://www.zotero.org/google-docs/?j7L6TJ. Our analysis identified a total of 64 viral sequences bearing RdRp domain representing 16 viral families from mosquito species sampled in the Atlantic forest, Northeast Brazil. We were able to reassess the virome composition of two mosquito species from a previous study (Ma. wilsoni and Cq. hermanoi from da Silva et al., 2021), and we expanded the knowledge including another eight sylvatic species from five genera, some of those were screened for viruses for the first time. Our findings revealed a diverse set of viruses that are highly divergent from known viruses from databases and are likely insect-specific viruses of the virome of studied mosquitoes, with the exception of two sequences that showed high amino acid RdRp identity (>90%) with viral sequences available in the literature, probably representing the same virus lineage. The majority of RdRp sequences were distinct across mosquito species, genera, and from those already available in the databases (Fig. S24; Supplementary Material 2).

Virome of Sabethini tribe

Current knowledge about Sabethini viromes in Brazil is limited to few studies that have been conducted only on species of the Sabethes and Wyeomyia genera (23, 24, 56)https://www.zotero.org/google-docs/?ED87aA. These studies revealed several viral families, such Flaviviridae, Chuviridae, Reoviridae, Phenuiviridae, Virgaviridae, Lispiviridae, Partitiviridae, Parvoviridae, Partitiviridae, and Xinmoviridae in the virome of Sabethes mosquitoes (23, 56)https://www.zotero.org/google-docs/?Ph3vfr. Despite these findings, two studies investigating the virome of the Wyeomyia genus have been unable to detect viral sequences (23, 24)https://www.zotero.org/google-docs/?TCx0kc. The current study represents the first analysis of a virome of a mosquito from the Limatus genus in Brazil revealing the presence of six viral families (Iflaviridae, Flaviviridae, Narnaviridae, Virgaviridae, Phasmaviridae, and Endornaviridae) in which Endornaviridae, Virgaviridae, and Phasmaviridae were the most abundant (Fig. S7). Although Flaviviridae sequences were identified in this mosquito species (58)https://www.zotero.org/google-docs/?cmumsU, the phylogenetically relationship of sequences from this viral family grouped with other ISVs (Fig. S16). Furthermore, the contigs encoding these Flaviviridae sequences exhibited multiple split ORFs; hence, additional analyses are warranted to determine if these contigs belong to bona fide viruses or EVEs that were not removed by our filters.

Virome of Aedini tribe

Studies on the virome of Aedini mosquitoes have focused on Ae. albopictus and Ae. aegypti (20, 59)https://www.zotero.org/google-docs/?sU0te6, as well as Psorophora (60)https://www.zotero.org/google-docs/?N8tW5z, Ochlerotatus (61),https://www.zotero.org/google-docs/?diJs8T and Armigeres species (62)https://www.zotero.org/google-docs/?gZYw7g. In our analysis, based on three Aedini species, we observed a low diversity of viral families within sampled pools (Fig. 2). Ae. scapularis showed the highest diversity within the tribe, but with only two viral families, while both Ae. albopictus and Ps. ferox showed only one viral family each. The Aedini species showed the lowest diversity of viral families in relation to other genera also assessed in the current study such as Limatus, Mansonia, and Coquillettidia mosquitoes (Fig. 2D and E). Another study has also shown the virome of Ae. albopictus with a lower number of viral sequences in relation to Ae. aegypti mosquitoes in Colombia (63)https://www.zotero.org/google-docs/?PWctau. On the other hand, a study on the virome of Ae. albopictus from China revealed a higher number of identified viral families, with at least 50 families specifically associated with vertebrates, invertebrates, plants, fungi, bacteria, and protozoa hosts (59)https://www.zotero.org/google-docs/?Ln7PZC. In Brazil, there have been few studies investigating the virome in Aedini, and the results are similar with our findings, where the species exhibited a lower number of viral families or, in certain cases, no viral sequences (23, 24, 56, 64, 65)https://www.zotero.org/google-docs/?ib8ELd. Each of these studies were conducted using distinct mosquito sampling, sequencing, and bioinformatics approaches hindering direct comparison with the results presents in this study. More studies focusing on the virome of Aedini species in different geographical regions are needed to uncover the main differences in viral diversity among Aedini species around the world.

Virome of Mansoniini tribe

Our group has previously assessed the virome of Ma. wilsoni and Cq. hermanoi species from the Mansoniini tribe (21)https://www.zotero.org/google-docs/?vxEwrs. In the current study, we expanded the knowledge of viral families interacting with mosquitoes from this tribe. Characterizing the virome of different samples of the same species allowed us to detect five viral families already identified before (Orthomyxoviridae, Xinmoviridae, Phenuiviridae, Flaviviridae, and Virgaviridae). Moreover, we identified another three that were reported for the first time: Phasmaviridae for Cq. hermanoi and Dicistroviridae and Solemoviridae for Ma. wilsoni. These viral families are known to infect arthropods and plants, respectively. However, we have not identified Rhabdoviridae, Chuviridae, and Partitiviridae viral sequences in those species as previously identified (21)https://www.zotero.org/google-docs/?Yg4ujQ. The only study from another research group assessing viruses from Ma. wilsoni has identified a few sequences classified in Totiviridae, Partitiviridae, and Chuviridae (24)https://www.zotero.org/google-docs/?XlFdij. But these authors used a different mosquito tissue (salivary glands) and sequencing approach, which may explain the differences found (24)https://www.zotero.org/google-docs/?Z6mLp4. Our analysis revealed the presence of a sequence from the Aliusviridae family that belongs to the Jingchuvirales order in Cq. venezuelensis. Recently, the Jingchuvirales order originally consisted only of the Chuviridae family that was split into five viral families, including the Aliusviridae family (66)https://www.zotero.org/google-docs/?UyQKdA. Further exploration of the Jingchuvirales within the Mansoniini tribe could elucidate whether the previous Chuviridae have been reclassified or if these mosquito species harbor different viral families within this order. The Aliusviridae sequences also exhibited split ORFs, such as the flavivirus sequences detected in this study, requiring additional investigation to determine their exogenous or endogenous origin. In general, our results for Mansoniini were similar as observed in a previous study (61)https://www.zotero.org/google-docs/?Ik27vB wherein the majority of sequences from Ma. uniformis were classified into the Xinmoviridae family with the exception of Cq. hermanoi where the majority of sequences belonged to Phasmaviridae.

Arthropods and vertebrates viruses

We identified some viral contigs belonging to families (Phenuiviridae, Orthomyxoviridae, Rhabdoviridae, Flaviviridae) known to harbor viruses that infect arthropods and vertebrates (Fig. 1E). The Phenuiviridae family encompasses arboviruses, such as the Rift Valley Virus, which is transmitted mainly by Aedes and Culex mosquitoes (67, 68)https://www.zotero.org/google-docs/?1NiuzN. In our study, we positioned two Phenuiviridae sequences, one from Ae. scapularis that is clustered with Salarivirus Mos8CM0 (API61884.1) and one from Ma. wilsoni grouping with Narangue virus (QHA33858.1) (Fig. S13). Due to the limited knowledge about these viruses, further studies are required to evaluate their effect in different cell lines and their infection capacity in invertebrate and vertebrate cells.

The Orthomyxoviridae also encompass known arboviruses that can infect both arthropods and vertebrates such as Quaranjarviruses (Quaranfil virus) and Thogotovirus (Thogotovirus Thogoto) (https://www.zotero.org/google-docs/?kykLxJ69, 70)https://www.zotero.org/google-docs/?kykLxJ. Here, we recovered the Orthomyxovirus previously detected in Ma. wilsoni (21)https://www.zotero.org/google-docs/?oxk7Iy. The Ma. wilsoni Othomyxoviridae virus 1 was assigned into the quaranjavirus genus. There is evidence of vertebrate and humans cell infection for viruses of this genus highlighting its medical significance (71, 72)https://www.zotero.org/google-docs/?W0tftZ.

Viruses from the Rhabdoviridae are generalist and were identified into a wide range of hosts (73)https://www.zotero.org/google-docs/?30Xa4J. The sequences generated in this study clustered with other Rhabdoviridae viruses found in mosquitoes. Interestingly, one clade that clustered the sequences identified in Ae. albopictus revealed to be a basal clade encompassing viruses known to infect a wide range of vertebrates and arthropods (Fig. S15). The sequences of this basal clade have been assigned in the Almendravirus genus, which includes the mosquito Rhabdoviruses. One sequence identified in Ae. albopictus (Ae. albopitus Rhabdoviridae virus 1) showed a high amino acid identity with Puerto Almendras virus that was previously identified in Aedes fulvus in Peru (74)https://www.zotero.org/google-docs/?fCyGlD. Although this virus induced no cytopathic effect in vertebrate cells, it could be detected in the cell culture supernatant suggesting active replication (74)https://www.zotero.org/google-docs/?E2NFrW.

Our phylogenetic analysis revealed that the sequences from Flaviviridae, Rhabdoviridae, Phenuiviridae, and Orthomyxoviridae clustered with other ISVs that have been previously characterized although with substantial amino acid divergence. Based on known hosts of these clades, our findings suggest we characterized several mosquito-specific viruses (MSV) (Fig. 3 and 4). Our results are consistent with previous studies that have also found a low proportion of pathogenic arboviruses among the metatranscriptomic uncovered viruses (4, 7, 20, 75)https://www.zotero.org/google-docs/?nJmPdH. The MSV capacity to infect vertebrate hosts remains unclear, and further research is needed to assess the infection potential of these novel viruses.

Arthropod viruses

Several viral contigs identified in this study were classified within known viral families bearing members that interfere with arboviruses replication such as Xinmoviridae (Anphevirus genus), Phasmaviridae, and Phenuiviridae families. ISVs from those families, such as Phasi Charoen virus (PCV) from the Phenuiviridae family, together with a cell-fusing agent virus from the Flaviviridae, have been shown to interfere with replication of arboviruses, such as DENV, ZIKV, and La Crosse Virus, from Flaviviridae and Peribunyaviridae respectively (76)https://www.zotero.org/google-docs/?iezHt9. Furthermore Olmo et al.’s study showed a wide distribution of Phasi Charoen-Like Virus on Aedes mosquitoes in different locations and its influence in DENV and ZIKV transmission in vivo (77)https://www.zotero.org/google-docs/?UjKCFX. No viral contig with similarity with this virus was found in our data set. One study has described the capacity of Aedes anphevirus from the Xinmoviridae family to interfere with Dengue virus replication in cell lines (78)https://www.zotero.org/google-docs/?7OT0gf. We found several contigs from both Mansonia and Coquillettidia genera that were clustered in a major clade with Aedes anphevirus.

In summary, we uncovered a diverse virome of sylvatic mosquitoes from Northeast Brazil largely composed of highly divergent ISVs or MSVs.

Limitation of the study

The analysis of virome composition of eukaryotic species, including mosquitoes, has revealed a substantial dynamism through time linked with some factors that affect viral richness and abundance (3, 5, 75, 79)https://www.zotero.org/google-docs/?oTxFWN. Factors such as sampling collection in different seasons (59, 80)https://www.zotero.org/google-docs/?sqDCVR or host feeding pattern (81)https://www.zotero.org/google-docs/?Wwn7kD have shown to influence the mosquito virome. In this study, we did not evaluate the virome composition through time and limited our sampling to a narrow species range distribution; therefore, our data are a snapshot of the viral diversity carried by sylvatic mosquitoes. Moreover, due to the different mosquito tissues, viral enrichment, sequencing, and bioinformatic analysis performed in other studies from Brazil and the world, we could not directly compare the virome abundance and diversity with our findings. Anyhow, the data presented here is in line with a broad picture found in other studies in the sense that mosquito viromes are mainly composed of highly divergent ISVs. More intense sampling and virome characterization using less biased wet and dry lab protocols are needed to fully uncover sylvatic mosquitoes viromes in Brazil.

Virome findings may be biased by the presence of EVEs, viral genome remnants endogenized into host genomes that can be transcribed (7, 82)https://www.zotero.org/google-docs/?A5SIkq. Insects, including mosquitoes, have been shown to harbor several RNA virus EVEs into their genomes (83)https://www.zotero.org/google-docs/?rCATO7. To differentiate EVEs from bonafide circulating viruses, one could compare the identified viruses' genomes against the host genomes, as performed in previous studies (21, 84)https://www.zotero.org/google-docs/?5nGE7j. Due to the lack of host genomic sequences and small RNA data sets for the majority of mosquitoes investigated, here, we compared the identified viral sequences against all genomes available for mosquito species and included draft genomes for some species aiming to reduce the interference of EVEs-derived RNA in our virome output. However, further assessment of viral sequences found here is warranted to validate the findings particularly for the viral contigs from Aliusviridae and Flaviviridae families. Although some of the viral contigs identified warrant further in-depth analysis, we identified several complete viral sequences that likely represent true bona fide viruses once the large majority of EVEs are composed of short fragments of viral genomes (85, 86)https://www.zotero.org/google-docs/?C7r2EE.

It is also important to point out that we focused our analysis on the most conserved RdRp gene/protein once it is the best marker for detecting and performing evolutionary studies with distantly related viruses, but we did not perform detailed analysis of other viral genes; this can particularly bias the viral completeness results regarding segmented viruses. Hence, future analysis is warranted to more in-depth characterization of segmented virus genomes. Moreover, the metatranscriptomic approach used is better suited to characterize viruses with RNA genome, but it can also detect actively transcribed DNA viruses. Therefore, we have analyzed more specifically the sylvatic mosquito RNA virome and the viral community composed of DNA genomes should be further accessed.

Conclusion

In this study, we conducted a virome profiling of 10 mosquito species from Northeast Brazil and identified viral sequences from 16 different families, including a total of 18 full-length sequences of viral segments or complete linear genomes. Our phylogenetic analysis revealed that the majority of the newly characterized viruses represent new viral species based on RdRp amino acid identities. Our study provides important basic knowledge about the virome composition of mosquitoes in Northeast Brazil and highlights several viral families that require further investigation to evaluate their potential to infect vertebrates or impact arbovirus replication. These findings enhanced our understanding of how these viruses are maintained in nature, their interactions with the sylvatic mosquito fauna, and the natural viral community that infect those species.

DATA AVAILABILITY

Raw reads generated in this study and the identified viral sequences are available through the ENA under project number PRJEB63303 and run accessions ERR12411109 to ERR12411118. Supplementary material 1 and Files S1 and S2 are available in Figshare (https://doi.org/10.6084/m9.figshare.23584827.v5).