Metatranscriptomic Analysis Reveals the Virome and Viral Genomic Evolution of Medically Important Mites

Mites are an important group of arthropods that are associated with a variety of human diseases, including scrub typhus and asthma. However, it remains unclear whether or not mites carry viruses that might play a role in human infections or allergic disease.

ABSTRACT

Mites are notorious for being vectors transmitting infectious pathogens and a source of allergens causing allergic conditions in animals and humans. However, despite their huge impact on public health, the virome of mites remains unknown. Here, we characterized the virus diversity and abundance of 14 species of medically important mites based on total RNA sequencing data sets generated in this study, as well as those deposited in the Sequence Read Archive (SRA) database. A total of 47 genetically distinct viruses were identified and classified into 17 virus families or virus supergroups, and the viral sequences accounted for as much as 29.67% of the total non-rRNA transcriptome in one mite library. The most commonly identified viruses are members of Picornavirales, among which we revealed more than one type of viruses that are evolutionarily related to dicistronic viruses but contain a single open reading frame, thus likely representing a recent example of host (i.e., mite)-related parallel evolution from the dicistronic to monocistronic genomic form within the family Dicistroviridae. To our best knowledge, this is the first time a comprehensive and systematic screening of the RNA virome in medically important mites, including house dust mites (HDM), has been performed. Overall, the RNA virome identified here provides not only significant insights into the diversity and evolution of RNA viruses in mites, but also a solid knowledge base for studying their roles in human diseases.

IMPORTANCE Mites are an important group of arthropods that are associated with a variety of human diseases, including scrub typhus and asthma. However, it remains unclear whether or not mites carry viruses that might play a role in human infections or allergic disease. In this study, we used a total transcriptomics approach to characterize and compare the complete RNA virome within mites that are relevant to human health and diseases. Specifically, our data revealed a large diversity, a high abundance, and a flexible genomic evolution for these viruses. Although most of the viruses identified here are not known to be associated with human infectious disease, the abundant presence of viral RNAs may play an immunomodulatory role in the development of allergic reactions, such as asthma, during environmental exposure to mite allergens and therefore may provide important insights into the mite-induced allergy and preparation of mite allergen vaccines.

INTRODUCTION

Mites are a broad group of tiny arachnids from at least three orders, namely, Acariformes, Opilioacariformes, and Parasitiformes (1, 2). Within this group, several can adopt a parasitic life cycle in humans as well as economically important animals and plants, which causes substantial public health concerns and economic impact (3). The medical significance of mites is generally reflected in their roles in (i) causing or contributing to allergic diseases such as asthma and cutaneous dermatitis and (ii) transmission of infectious disease (4, 5).

Despite their importance as potential pathogen vectors, the complete viromes of mites had never been systematically studied. The knowledge base for infectious diseases transmitted by mites is limited to a few classic examples of mite-vectored diseases such as scrub typhus and rickettsial pox, as well as pulmonary acariasis, transmitted by trombiculid mites, gamasid mites, and dust mites, respectively (6–8). Recently, Yu et al. examined roles of the gamasid mites and chigger mites in the transmission of hantavirus and found that these parasitic mites could act as effective vectors for the maintenance and transmission of hantavirus under experimental conditions, which might argue against the current paradigm that rodents are the reservoir of hantaviruses in nature and that humans are infected by inhalation of rodent excreta containing these viruses (9, 10) and raises a much broader question of whether mites could serve as vectors for other human pathogens.

In addition to transmitting infectious diseases, some nonparasitic mites such as house dust mites (HDMs), including Dermatophagoides pteronyssinus, Dermatophagoides farinae, and Euroglyphus maynei, and storage mites, including Blomia tropicalis, Lepidoglyphus destructor, and Tyropahgus putrescentiae, are also the main allergens causing allergic rhino-conjunctivitis, as well as asthma (11). Humans are naturally exposed to mite allergens that are often in complex forms containing both allergic proteins that are responsible for inducing type 2 inflammation and microbial substances such as nucleic acids (RNA/DNA), which stimulate the innate sensing pathways to modulate the progression of type 2 inflammation (12–20). Interestingly, it has been recently reported that double-stranded RNA (dsRNA) derived from HDMs can induce a protective immune response that can attenuate HDM-dependent type 2 inflammatory responses (21). Nevertheless, the identities and sources of those dsRNA species remain undefined. However, it should be pointed out that human respiratory RNA viruses such as rhinovirus and respiratory syncytial virus (RSV) infect and replicate in airway epithelial cells, which often results in asthma exacerbations in children (22). Since replication of RNA viruses also produces dsRNA species, it is therefore important to systematically examine the RNA virome in mites and look for ones that might be associated with allergic reactions.

Recently, large-scale metagenomic and metatranscriptomic surveys of the phylum Arthropoda have greatly extended our understanding of the prevalence and abundance of RNA viruses in this most diversified taxonomic group (23–25). These studies have also revealed an unpreceded number of genome architectures that demonstrated great flexibility in the evolution of RNA viruses (26, 27). Furthermore, in addition to metagenomic sequencing, databases containing full transcriptome sequencing data from the arthropod hosts were also proven “gold mines” for virus discovery (28), given that the RNA virome occupies significant proportion of and sometimes dominates the host transcriptome (27, 29).

Currently, the understanding of the virome in mites is limited despite their huge impact on public health and the economy. Recent studies have demonstrated the existence of a diverse and abundant RNA virome in Varroa destructor, one of the economically important mite species infecting the honey bee population (30–32). Nevertheless, to our best knowledge, there have been no systematic investigations of the virome in a wider range of medically relevant mite species. In this study, we revealed the complete virome of several medically related mite groups by directly sequencing the total RNA, as well as by mining the existing transcriptomic data. Specifically, we sought to explore RNA virus diversity as the basis to understand the evolution of viruses in mites and to evaluate their potential roles in transmitting human diseases and the mechanism that induces allergic reactions.

RESULTS

Discovery of RNA viruses in house dust and storage mites by metatranscriptomic analysis.

We performed three transcriptome sequencing (RNA-seq) runs on pools of house dust mites of the species Dermatophagoides farinae and Dermatophagoides pteronyssinus and the storage mite Tyrophagus putrescentiae, respectively, generating a total of 209,826,736 reads (between 67,777,318 and 73,986,168 reads per library), which were assembled de novo into 9,086 to 129,885 contigs per library. Mitochondrial COX1 gene sequences were identified from the assembled contigs to confirm the mite species in each library. A comparison to reference COX1 genes available at the BOLD database (http://www.boldsystems.org/) showed 100, 99.66, and 100% nucleotide identity with existing sequences from D. farinae, D. pteronyssinus, and T. putrescentiae, respectively, which was further confirmed by phylogenetic analyses of relevant species (Fig. 1A).

FIG 1.

RNA viruses identified in house dust and storage mites. (A) Phylogenetic analyses of house dust mites (HDMs) and storage mites based on the mitochondrial cytochrome oxidase subunit 1 (COX1) gene. HDM sequences obtained in this study are shown by purple triangles. The scale bar indicates the number of nucleotide substitutions per site. (B) The diversity of RNA viruses in D. farinae and D. pteronyssinus (Greer Laboratories) and Tyrophagus putrescentiae (Shenzhen University, China). (C) The number and abundance of RNA virus species discovered in the respective HDM species (The threshold was 0.1% of non-rRNA reads). (D) The pie chart shows the percentages of RNA sequencing reads that mapped to the corresponding reference mite genomes (blue) and the viruses (orange), and the remaining reads were designated “others” (gray).

By comparing assembled contigs with the nonredundant protein (nr) sequence database (27), we discovered a total of 15 novel viruses belonging to 7 families or virus supergroups, namely, Sobemo-like virus, Narnaviridae, Tombus-like virus, Rhabdoviridae, Reoviridae, Dicistroviridae, Iflaviridae, and Picorna-like viruses (Fig. 1B). Notably, these viruses are highly divergent to previously described arthropod viruses (27.2 to 54.5% amino acid diversity in the RNA-dependent RNA polymerase [RdRP] domain [see Table S1 in the supplemental material]). Specifically, nine new viruses were discovered from T. putrescentiae, three from D. farinae, and four from D. pteronyssinus (Fig. 1B). Among the newly identified viruses, eight exhibited relatively high abundance (≥0.1% of total non-rRNA), including all in D. farinae and D. pteronyssinus, as well as one in T. putrescentiae, namely, Shenzhen reo-like virus 2 (Fig. 1C). Furthermore, we assessed the relative quantities of viral RNA reads in each library after removing host and microbial rRNA reads (see Materials and Methods). It turned out that viral RNA comprised 0.56 to ∼29.67% of the total RNA sequences (rRNA excluded) within each library, in comparison to 19.06 to ∼89.34% of the mites’ own whole-genomic RNA (Fig. 1D). The viruses found in this study were named mainly based on the location of sample sources and according to related viral clades and are sometimes followed by a number: for example, Shenzhen Rhabdo-like virus 2 (Table S1).

RNA virome identified in medically important mites.

In addition to the three sequencing runs performed in this study, we also explored the total RNA virome of medically important mites by analyzing RNA sequencing data sets obtained from the Sequence Read Archive (SRA) database. As of 30 June 2020, there were a total of 1,588 data sets under the taxonomy Acari (mites and ticks). Since this study primarily focused on mites that are relevant for human disease and health, we aimed at those families or orders (i.e., gamasid, sarcoptic, scab, chigger, dust, and oribatid mites) that may be associated with human disorders, such as allergy, dermatitis, and pulmonary acariasis. After the filtration according to the criteria described in Materials and Methods, the scope of data mining was therefore reduced to 40 data sets in total, and the corresponding mites belonged to 14 species belonging to 12 families, including the ones described in the previous section: i.e., D. farinae, D. pteronyssinus, and T. putrescentiae (Table S1). The host (mite) information for these data sets was originally obtained from the database and was confirmed or corrected by analyzing the host’s COX1 gene (Fig. 2; Table S1). Of note, some of SRA data sets in our study were prepared with oligo(dT) priming due to the original use for transcriptome analysis, and thus we expected a general loss of sensitivity in detecting viral sequences without poly(A) tails.

FIG 2.

Phylogenetic analyses of mites based on the COX1 gene. Samples of HDMs collected and sequenced are marked as purple circles, data sets filtered from the SRA database are marked as solid green circles, and the relative host names are shown on the right. The scale bar indicates the number of nucleotide substitutions per site.

Overall, from the 43 libraries, a total of 47 genetically distinct viruses with complete or nearly complete genomes were identified (Fig. 3A). They belonged to 17 virus families or unclassified virus groups that included 11 single-stranded positive-sense RNA viruses [ss(+)RNA viruses], three single-stranded negative-sense RNA viruses [ss(−)RNA viruses], and three double-stranded RNA viruses (dsRNA viruses), respectively (Fig. 3B). All of these viruses were not previously described and thus belong to new viruses. Members from the order Picornavirales (i.e., Dicistroviridae, Iflaviridae, Seco-like viruses, and other Picorna-like viruses) had the highest number of viruses (n = 14), followed by Reoviridae (n = 6) and Narnaviridae (n = 5). Sequence comparisons revealed that the newly identified viruses shared 24.4 to 62.2% amino acid identity in the RdRp domain with those described previously (Table S1), and therefore all merit the definition of new viruses. The distribution of virus families in each library is summarized in Table S1.

FIG 3.

RNA viruses identified in medically important mites in this study. (A) The number of data sets involved in medically important mite families, the host species included in each family, and the number of RNA viruses discovered in this study. (B) The number of virus species distributed in each virus families. (C) The co-occurrence graph displaying the congruent relationships between mite hosts and their associated viral families for each virus identified in this study. The length of the pie edge represents the number of the viruses in mite species and in each viral family.

The host groups for these viruses included 10 species belonging to 9 families of medically important mites (Fig. 3C). Notably, the number of viruses identified varied substantially in different host species. While no viruses have been identified from the species Ornithonyssus bacoti, Psoroptes cuniculi, Hypochthonius rufulus, and Platynothrus peltifer, which belong to the families Macronyssidae, Psoroptidae, Hypochthoniidae, and Crotoniidae, respectively, other mite species can be associated with 1 to ∼18 viruses. The largest virus diversity (18 viruses) was found in a single species, T. putrescentiae of the family Acaridae, although it contained most of the data sets (22 of the 43 SRA data sets examined in this study). Besides the diversity, the viral abundance level also showed a great variation. The highest abundance level (29.67% of total non-rRNA) was observed in our sequenced library of D. pteronyssinus, followed by SRA data set Df494 from D. farinae (17.39% [Fig. 4A]), and both of which were sequenced by the total RNA approach, suggesting that the library generated with the oligo(dT) primer may have left out a large portion of viral RNA sequences. Indeed, all the other viruses in these two libraries showed high abundance after read normalization (Fig. 4B).

FIG 4.

The distribution and abundance of RNA viruses identified in medically important mites. (A) The frequency and diversity of RNA viruses in each mite library. The short name of each data set is shown on the abscissa. Red lines represent the virus frequency of each library, including the sequenced HDMs and SRA data sets. The numbers of high-abundance (≥1% non-rRNA reads) and low-abundance (<1% non-rRNA reads) viruses in each data set are shown as orange and blue bars, respectively. (B) Heat map showing the presence and abundance of mite viruses across different host species and geographic regions. The relative abundance of the viruses in each sample was estimated as RPM.

Evolutionary features of the mite RNA viruses.

Based on the phylogenetic analyses of the RdRp domains, all the newly identified viruses were assigned to corresponding families and orders (Fig. 5 and 6). Although these viruses had a highly diverse genetic backgrounds and were found in 17 families/clades, most of the mite-related viruses within each virus family tend to form monophyletic clusters with each other and are closely related to those found in related hosts, such as ticks and spiders, as well as other arthropod host groups (Fig. 5 and Fig. 7 to 11). Nevertheless, there were also a few cases where the viruses discovered in mites were more closely related to Totiviruses (Fig. 8) and Narnaviruses (Fig. 10), infecting fungi and basal eukaryotes, suggesting alternative hosts for the mites, although this needs to be confirmed with more data. Importantly, none of the viruses discovered in this study fell into arbovirus groups (i.e., genera Flavivirus, Alphavirus, Phlebovirus, and Orthobunyavirus, among others), suggesting that viruses identified in this study are unlikely to directly cause infectious disease in vertebrates.

FIG 5.

Phylogenetic diversity of RNA viruses discovered in mites. Phylogenetic analyses of virus families or virus supergroups were carried out based on virus RNA-dependent RNA polymerase alignment with the ML algorithm. Names of genera or arbitrary virus groups are marked on the right side. Positions of novel species identified in this study are marked as solid red circles in each tree. The scale bar indicates the number of amino acid substitutions per site.

FIG 6.

Phylogenetic analysis and genomic features of the family Dicistroviridae. Maximum likelihood phylogenetic trees show the diversity of dicistroviruses and the positions of dicistroviruses identified in this study (marked as solid red circles) and previous species belonging to the genera Aparavirus, Cripavirus, and Triatovirus (marked as solid blue circles). The virus genome structures are shown next to the phylogenetic trees. Blue lines marked in the phylogenetic tree represent the dicistro-like virus clades, whereas the red lines are dicistrovirus-related monocistrovirus. For viruses discovered in this study, those with two-ORF genomes (ancestral forms) are marked with a blue star, whereas those with single-ORF genomes are marked with a red star.

FIG 7.

Maximum likelihood phylogenies of the negative-sense single-stranded RNA viruses. The phylogeny analysis is based on the viral RdRp domain. The viruses discovered this study are marked as solid red circles. The identities of viruses in the same clade are shown on the right. The genome structures of these newly discovered viruses are shown next to their corresponding phylogenies. Predicted ORFs of these genomes are labeled with information on the potential protein or protein domain that they encode.

FIG 8.

Maximum likelihood phylogenies of double-strand RNA viruses. The figure legend follows the legend to Fig. 7.

FIG 9.

Maximum likelihood phylogenies of positive-sense single-stranded RNA viruses of Negev-like, Sobemo-like, and Toga-like viruses. The figure legend follows the legend to Fig. 7.

FIG 10.

Maximum likelihood phylogenies of positive-sense single-stranded RNA viruses of Narnaviridae and Nodaviridae. The figure legend follows the legend to Fig. 7.

FIG 11.

Maximum likelihood phylogenies of positive-sense single-stranded RNA viruses of Picornavirales. The figure legend follows the legend to Fig. 7.

We also examined the evolutionary relationship between mites and their associated viruses. Based on the host distribution on virus phylogenies, the clustering of mite-associated viruses as well as the observation of closer relationship between viruses found in mites and ticks suggested a potentially long-term association between viruses and hosts. Furthermore, the comparison of virome compositions among different mite species revealed a lack of host-switch events in general, with a Shenzhen dicistrovirus-related monocistrovirus (belonging to the family Dicistroviridae) as the only case of a single virus shared by different host species (i.e., D. pteronyssinus from China and the United States and T. putrescentiae from China) (Fig. 4B). On the other hand, the same virus species tend to have wide geographic distributions: for example, Skokie dicistro-like virus and Skokie reo-like virus were found in both the United States and China, and Shenzhen reo-like virus 4 and Shenzhen reo-like virus 5 were from both the Czech Republic and China (Fig. 4B).

Viruses identified in HDMs possess monocistronic genomes but phylogenetically belong to dicistronic Dicistroviridae.

The majority of viruses of the family Dicistroviridae discovered so far, including the genera Aparavirus, Cripavirus, and Triatovirus, are known to contain two nonoverlapping open reading frames (ORF 1 and ORF 2, encoding nonstructural and structural proteins, respectively), which are separated by an intergenic untranslated region (IGR) of about 170 to 530 nucleotides in length and with an internal ribosome entry site (IRES) element (33) (Fig. 6). Surprisingly, we found that some of the mite-associated dicistroviruses contained only a single ORF (Fig. 6). No IGRs and IRESs were detected in these new viral genomes, although they shared the same gene order—i.e., nonstructural genes (helicase-protease-RdRp) followed by structural genes (VP2-VP4-VP3-VP1)—as their two-ORF counterparts of the family Dicistroviridae (Fig. 6). On the phylogenetic tree, viruses with the single-ORF genome formation were deeply nested within those with the two-ORF formation, which suggested that the single-ORF genome formation was more recently derived. Importantly, the merge of two ORFs occurred independently for two times in the evolutionary history of the family Dicistroviridae, each within the diversity of mite-associated virus groups (Fig. 6 and 11).

Identification of Jingmenviruses in mites.

In the Histiostoma sp. mite, we discovered a four-segmented virus related to Jingmenviruses, which are Flavi-like viruses infecting a wide range of arthropod hosts (34, 35) and potentially humans (36, 37). We tentatively named the newly discovered virus as Histiostoma Jingmenvirus (HJMV). Phylogenetic analyses of its four segments revealed that despite high sequence divergence (22.04 to 33.41% amino acid identity), HJMV formed a monophyletic cluster with Toxocara canis larva agent (TCLA) and Jingmen tick viruses (JMTVs) (Fig. 12 and Table 1), a group of multisegmented and enveloped Jingmenviruses. Although longer, the genome of HJMV resembled that of JMTV with regard to the distributions of transmembrane helices, as well as the presence of signal peptides (Fig. 12). Furthermore, for HJMV, the abundance levels were similar across the four genome segments (Table 1), suggesting that it is more likely to be an enveloped multisegment virion-like virus (similar to Jingmen tick virus) rather than a multipartite virion-like virus (like Guaico Culex virus) (38). The latter tends to have significantly different abundance levels for different genomic segments (35).

FIG 12.

Evolutionary history and genomic features of HJMV. (A) Phylogenetic tree of the multipartite Jingmenviruses and members of the multisegmented Jingmenvirus group based on the NS3 and NS5 alignments. The trees are midpoint rooted. The virus discovered in the present study is marked with a solid red circle. The trees shown here were inferred using an ML method. Branches are drawn to the scale of substitutions per site. (B) Comparison of polyprotein structures among members of the Jingmenvirus group, which includes one newly discovered virus and three prototypical members of the Jingmenviruses described previously. A unified length scale is used for all polyproteins. Predicted or known transmembrane domains (black vertical bars) and cleavage sites for the host signalase (red arrow) are indicated.

TABLE 1.

Information on the top blastx hit and protein identity of coding proteins, as well as the abundance of HJMV viral segments from Histiostoma mite species^a

Segment	Top blastx hit	BLAST accession no.	Coding protein(s)	% of identity	Length (nt)	No. of mapped reads	Expression level	Total reads (%)
1	JMTV (NS5)	QHW06950	NS5	33.41	3,687	29,105	1,487.0000	0.08
2	TCLA (ANT-30)	ACF19854	Glycoprotein	33.33	3,333	34,632	2,215.2976	0.10
3	TCLA (ANT-34)	ACF19855	NS3	32.63	3,374	33,972	2,505.0000	0.09
4	JMTV (capsid protein)	QCW07570	Capsid, membrane	22.04	3,325	45,754	2,573.0663	0.13
Total					13,701	143,463		0.4

^aThe genome coverage was 10.47-fold.

DISCUSSION

In this study, we performed, for the first time, a comprehensive and systematic screening of the RNA virome in medically important mites, including HDMs by using a metatranscriptomic approach. Our data show that mites-associated viruses are highly prevalent and diverse, indicating that similar to many other arthropods, mites can serve as a natural reservoir for RNA viruses (27, 39, 40). Previous studies have suggested a relatively long-term relationship between hosts and their virome in many vertebrate RNA virus groups (41) and insect-specific virus groups (28). Such a trend is also observed in this study because the cross-species virus transmission among mites was not frequently found from our current analysis. However, more sampling and experiments would be needed in order to strengthen such a statement. Moreover, the high abundance of RNA viruses found in the transcriptome of D. farinae or D. pteronyssinus raises an interesting question on how the HDM deals with viral infections. Animals such as Drosophila and Caenorhabditis elegans usually use small interfering RNAs (siRNAs) and Piwi-interacting RNAs (piRNAs) to defend against transposable elements (TEs) and viruses. However, the HDM is unique because it lacks a functional piRNA pathway. Instead, it has been recently reported that the HDM carries a plant-like Dicer-dependent siRNA machinery to defend against transposable elements and RNA viruses (42).

Another important finding is the remarkable genomic diversity and flexibility of the viruses found in different species of mites. Specifically, in medically important mites, we demonstrate a monocistronic genomic structure in dicistroviruses, which is derived from the dicistronic ancestries; dicistronic formation is the genomic signature of the family Dicistroviridae. Importantly, this transformation occurred twice and independently in the evolutionary history of dicistroviruses, which is in consistent with previous findings on the occurrence of multiple monocistronic genome forms in the family, which is defined based on “dicistronic” genome formation (27). Together, these results suggest more flexible genome structures in the “Dicistroviridae” and call for a new name for the family. Importantly, the missing intergenic region contained an IGR-IRES element (43), which can recruit ribosome to independently regulate the translation of structural polyproteins during viral infection (44). It is believed that the IGR-IRES allows the virus to produce a much larger amount of structural proteins than nonstructural proteins (45, 46), although such regulations may not be evolutionary advantages for viruses infecting a host that leads a largely parasitic life cycle. Lack of the IGR-IRES element in the dicistronic mite viruses found in this study indicates a different genome expression and regulation mechanism from that of prototypical dicistronic viruses.

Our study has successfully revealed the complete genome (4 segments) of HMJV. Three segments of the virus were revealed by protein sequence homology analysis. One segment (segment 4), showing no homology to any of the existing proteins, was identified based on its high abundance level (1,281.5027 reads per million [RPM]), which is close to the other three segments (814.6564 to 969.9917 RPM). The candidate segment was later confirmed by the presence of conserved regulatory sequences at both ends of the genome, the ORF arrangement, the presence of a signal peptide in the first ORF, and the secondary structure of transmembrane helicase domains in the second ORF, all of which were similar to segment 4 of related Jingmenviruses. Nevertheless, not all segmented viruses had as high an abundance level as that of HMJV. Under the circumstances of low abundance level, it is often difficult to identify potential nonhomologous segments because there are numerous candidate contigs with similar abundance levels. Such is the case for Ganwon-do orthomyxo-like virus and Skokie reo-like virus. For reo-like viruses, only at most 3 segments were found when 9 to 12 segments were expected. While the rest of the segments were mostly sequenced and assembled, it was unlikely, based on current data, to confirm that they were from the same reo-like viruses.

An important goal of this study is to identify parts of the virome that have the potential to cause human or animal diseases. However, among the 47 novel viruses we identified from medically important mites, none fell into the genera or families of “vector-borne” viruses that are known to infect humans or other mammalian hosts. While it is possible that the mites examined in this study do not typically carry vector-borne viruses, it should also be noted that vector-borne viruses usually constituted only a very small proportion of the total virome, even within disease vectors such as mosquitoes and ticks (25, 47, 48). In addition, in comparison to other arthropod-specific viruses, vector-borne viruses had a much smaller diversity, lower prevalence, lower abundance, and higher restrictions in terms of geographical and temporal distributions (25, 27). Given that the current data set is still limited in scope, future studies should be performed with a much broader survey using a bigger sample size collected from wider geographic regions in order to identify potential “vector-borne” viruses in mites. Since RNA viruses occupied a significant proportion of the mite transcriptome, it is also important to examine their potential effect on allergy. However, unlike rhinoviruses or other respiratory viruses that directly infect and replicate in human cells, the mite-associated virome may play an immunomodulatory role via activating innate immune responses during the process of environmental exposure to the mite-allergen complexes. In this regard, the findings described in this study may provide important insights into the pathogenic mechanisms of mite-induced allergies and allergy immunotherapies using mite allergen extracts.

MATERIALS AND METHODS

Samples.

The storage mite T. putrescentiae was obtained from Shenzhen University, China. The house dust mite (D. farinae and D. pteronyssinus) samples involved in this study were obtained from Greer Laboratories, Lenoir, NC. T. putrescentiae mites were collected alive and stored at −80°C for RNA extraction, whereas D. farinae and D. pteronyssinus (no. B81 and B82, respectively) samples were reconstituted from the frozen samples purchased from Greer Laboratories (21). All three samples were collected from home environments and continuously cultured for many generations.

Sample processing and RNA sequencing.

Prior to homogenization, approximately 30 mg of pooled mite samples was first washed three times with 500 to 700 μl standard sterile RNA- and DNA-free 1× phosphate-buffered saline (PBS) solution (GIBCO). The samples were then homogenized in liquid nitrogen. Subsequently, total RNA was extracted from homogenates using TRIzol LS reagent (Invitrogen) according to the manufacturer’s instructions, followed by purification using an E.Z.N.A total RNA kit (Omega). The quality of the extracted RNA was evaluated by an Agilent 2100 bioanalyzer (Agilent Technologies) before library construction and sequencing. Then, total RNA sequencing library preparation was carried out following an rRNA removal step using the Ribo-Zero Gold (Epidemiology) rRNA removal kit (Illumina) (35). Libraries were constructed by using a TruSeq total RNA library preparation kit (Illumina). Next-generation sequencing was performed using Illumina HiSeq platform, and 150-bp paired-end reads were generated for T. putrescentiae libraries and 100-bp paired-end reads for D. farinae and D. pteronyssinus libraries. All library preparation and sequencing were performed by BGI Tech (Shenzhen, China).

Mite RNA sequencing data mining from the SRA database.

In order to discover more viruses, we screened 1,588 mite transcriptomes available in the SRA database under the taxonomy ID txid6933. Among them, 40 were selected for further virus discovery analysis under the following criteria: (i) a data set size of >1,000 Mb was the first condition, (ii) medically important or medically related mite species were filtered under the taxonomies Acariformes, Opilioacariformes, and Parasitiformes, and (iii) the data sets had never been used for viruses explored previously. The associated samples can be found in Table S1.

RNA virus discovery.

For each library, sequencing reads were first adaptor and quality trimmed using the Trimmomatic program with the following options: SLIDINGWINDOW:4:5, LEADING:5, TRAILING:5, and MINLEN:25 (41). The remaining reads were assembled de novo using the Trinity program (version 2.1.1) (49) and MEGAHIT program (version 1.2.9) (50) with default parameter settings. To identify viral contigs, the assembled contigs were first compared against the nonredundant nucleotide (nt) and nonredundant protein (nr) sequence databases using the blastn program and diamond blastx program (27), respectively. Virus hits were identified by searching keywords “virus” within the taxonomy column of the BLAST hit result. The contigs were also compared against the database comprising reference RNA virus proteins downloaded from GenBank, and the parameter settings were similar to those described above. The E value cutoff for both comparisons was set at 1 × 10⁻⁵. Viral contigs with unassembled overlaps or from the same scaffold were merged to form complete or partial virus genomes using the SeqMan program implemented in the Lasergene software package (version 7.1, DNAstar).

To rule out the possibility that the discovered contigs were endogenous virus elements, we first compared all potential viral contigs against corresponding host genome available in the Whole Genome Shotgun (WGS) database, and those showing a high percentage of similarity to genome fragments were excluded from further analyses. Finally, reads were mapped back to the full-length genome with Bowtie2 (51) and inspected using the Integrated Genomics Viewer (52). The composition of each HDM transcriptome and the proportion of the remaining sequence data that mapped to viral RNA were analyzed after all rRNA reads of each library had been removed, including those from host species and microbes, by comparing them against the entire rRNA database. The proportion of reads mapped to the host genome were estimated by mapping against host whole-genome shotgun sequences available in GenBank (Dfa_Genome_UMICH_USM_1.1, ASM190122v2, and UMMZ_TP_P3.0F6 for D. farinae, D. pteronyssinus, and T. putrescentiae, respectively, after the removal of rRNA reads). The abundance of each viral genome within the library was estimated using the percentage of non-rRNA reads that were mapped to the target genome and normalized as reads per million (RPM).

Virus genome characterization.

The predication of the potential open reading frames (ORFs) was based on the ORF Finder program available at the NCBI website. Subsequent functional annotation of each ORF was carried out by BLAST searching against the conserved domain database (available at https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi) or by comparing each ORF against the nr database. Secondary structures of proteins were predicted using PredictProtein (https://www.predictprotein.org/), Pfam (http://pfam.xfam.org/), SOSUI (http://harrier.nagahama-i-bio.ac.jp/sosui/sosui_submit.html), and SignalP (http://www.cbs.dtu.dk/services/SignalP/).

For those segmental viruses, we followed previously reported strategies to search for genome segments other than the RdRp (27). In brief, we first identified the viral segments based on their homology to the proteins of related viruses. If such a strategy failed, we then retrieved unrelated contigs that had similar abundance to other known segments and searched for conserved genome termini or protein structures.

Phylogenetic analyses.

Phylogenetic analyses were carried out based on virus RNA-dependent RNA polymerase (RdRp) alignment (27). Viral RdRp sequences obtained from in this study were compared against reference proteins from relevant taxonomy groups downloaded from the GenBank. All sequences were then aligned using the E-INS-i algorithm implemented in MAFFT (version 7.4) (53). Subsequently, TrimAl (version 1.2) and the MEGA6.0 program were used to remove ambiguously aligned regions so that only RdRp domain and its neighboring conserved domains were kept (27). Finally, the PhyML program (version 3.3) was employed to estimate the phylogenetic relationships by using the maximum likelihood (ML) method, with the LG amino acid substitution model and Subtree Pruning and Regrafting (SPR) branch-swapping algorithm. In addition, an approximate likelihood ratio test (aLRT) with the Shimodaira-Hasegawa-like procedure was used to evaluate support for each internal node on the tree (54).

Accession number(s).

All sequence reads generated in this study have been submitted to the NCBI SRA database under BioProject accession no. PRJNA641992. All RNA virus genome sequences obtained in this study have been deposited in GenBank under accession no. MT747990 to MT748000, MT757474 to MT757511, MT721839 to MT721847, and MW177748 to MW177763.