Discovery of additional genomic segments reveals the fluidity of jingmenvirus genomic organization

Abstract

Jingmenviruses are a distinct group of flavi-like viruses characterized by a genome consisting of four to five segments. Here, we report the discovery of three novel putative jingmenviruses, identified by mining publicly available metagenomics data from mosquito and arachnid samples. Strikingly, these novel jingmenvirus sequences contain up to six genomic segments, with pairs of homologous segments coding for putative structural proteins. Following this discovery, we found an additional homologous segment for two other jingmenvirus genomes, which had gone unnoticed in the initial publications. The presence of a single version of the segments coding for non-structural proteins suggests that we have indeed identified jingmenviruses with infectious units that contain up to six segments. We compared these novel jingmenvirus sequences to published sequences, in particular the segments with multiple open reading frames (ORFs), and we propose that the putative translation initiation mechanisms involved for these segments are ribosomal frameshift resulting in the fusion of ORFs and leaky scanning for overlapping ORFs. These putative mechanisms, conserved for all jingmenvirus sequences analysed, including in homologous segments, require biological confirmation. We also generated structural models of two putative structural proteins in the duplicated segments, and the corresponding alignments enabled us to confirm or identify the homologous relationship between sequences that shared limited nucleotide or amino acid identity. Altogether, these results highlight the fluid nature of jingmenviruses, which is a hallmark of multipartite viruses. Different combinations of segments packaged in different virus particles could facilitate the acquisition or loss of genomic segments and a segment duplication following genomic drift. Our data therefore contribute to the evidence of the multipartite nature of jingmenviruses and the evolutionary role this organization may play.

Introduction

Jingmenviruses are related to flaviviruses but remarkably have a segmented genome, contrary to their close relatives (Colmant et al. 2022). The first jingmenvirus identified was Jingmen tick virus (JMTV) from Rhipicephalus microplus ticks collected in China in 2010 (Qin et al. 2014). Since this initial discovery, dozens of other jingmenvirus genomes have been detected worldwide in a wide range of sample types (Colmant et al. 2022). These sequences phylogenetically group into two main clades: one clade with sequences that are mostly tick- and vertebrate-associated and one clade with mainly insect-associated sequences (Colmant et al. 2022). The best characterized insect-associated jingmenvirus species is Guaico Culex virus (GCXV), isolated from Culex mosquitoes collected in South and Central America between 2008 and 2012 and from Culex mosquitoes collected from Brazil in 2010 (Ladner et al. 2016, Pauvolid-Corrêa et al. 2016). GCXV has been found to have a replication restricted to insects both in vitro and in vivo, and there is evidence that this virus may have a multipartite organization, meaning that viral particles encapsidate fewer fragments than the total number of genomic segments and multiple particles are required to initiate the replication cycle (Ladner et al. 2016). Multipartite viruses are very common in plants and fungi, extremely rare in animals, and have never been identified in bacteria (Michalakis and Blanc 2020). They seem to be more tolerant than their monopartite relatives to fluidity in their genomic organization and to evolutionary processes that facilitate functional gain by duplication or recombination (Michalakis and Blanc 2020).

The multi-segmented genomes of jingmenviruses are generally considered to include four segments except for GCXV, which has been found to have a fifth segment in most isolates but not all (in 4/6 sequenced isolates in Ladner et al. 2016). This fifth segment, named GCXV segment 5 encodes a viral protein called VP7, but no sequence homolog has been identified yet. Segment 5 is not essential for a productive viral infection and does not seem to be providing any fitness advantage in vitro or in vivo (Ladner et al. 2016, Chen et al. 2024).

For most jingmenviruses, segment 1 encodes the non-structural protein NSP1 with RNA-dependant RNA polymerase (RdRp) and methyltransferase functional domains, similar to the non-structural protein NS5 of their close relatives in the Orthoflavivirus genus (Chen et al. 2023, Wang et al. 2024). Segment 2 encodes the putative glycoprotein VP1 or (VP1a and VP1b for some tick-associated jingmenviruses), and has a second open reading frame (ORF) corresponding to the putative small protein nuORF of unknown function(s) for sequences in the tick-associated jingmenvirus clade, or the putative structural protein VP4 for sequences in the insect-associated jingmenvirus clade. The glycoprotein VP1 in tick-associated sequences has recently been described as structurally homologous to the flavivirus envelope protein E, with no identified fusion loop homolog (Mifsud et al. 2024). Segment 3 encodes NSP2, with serine protease and helicase domains, similar to the flavivirus NS3 protein (Gao et al. 2020, Zhang et al. 2021). Finally, segment 4 encodes putative structural proteins VP2 and VP3, with unknown functions.

This nomenclature is different for only two viruses, GCXV and its close relative Mole Culex virus (MoCV) isolated from Culex mosquitoes collected from Ghana in 2016 (Amoa-Bosompem et al. 2020). For these two viruses, segment 1 encodes for NSP1, segment 2 for NSP2, segment 3 for putative structural proteins VP1, VP2, and VP3, segment 4 for putative structural proteins VP4, VP5, and VP6, and segment 5 for VP7 when present (Colmant et al. 2022).

The segments of jingmenvirus genomes are capped at their 5ʹ end and polyadenylated at their 3ʹ end in the case of sequences in the tick-associated clade, making these fragments translation units, in which one to three coding sequences have been identified. The strategies used by jingmenviruses to ensure the translation of each coding sequence in the translation units, such as programmed ribosomal frameshifting or alternative translation initiation, have not been studied in detail to date (Sorokin et al. 2021).

Slippery heptanucleotides have been identified in a few jingmenvirus sequences, which suggests some ORFs are translated by a −1 ribosomal frameshift, the most common programmed frameshifting (GCXV and MoCV between ORFs coding for VP1 and VP3 as well as VP5 and VP6; in JMTV, Alongshan virus and Wuhan cricket virus between ORFs coding for VP2 and VP3; in ALSV between ORFs coding for VP1a and VP1b) (Ladner et al. 2016, Shi et al. 2016, Amoa-Bosompem et al. 2020, Kholodilov et al. 2020). These predictions are yet to be confirmed biologically, with the exception of GCXV frameshift between VP1 and VP3, for which the −1 ribosomal frameshift products have been confirmed by mass spectrometry. Other than these instances of likely ribosomal frameshift, no other mechanism has been suggested for the translation of multiple ORFs in jingmenvirus genomic segments, despite the existence of overlapping ORFs, possibly translated through a cap-dependent deviant translation initiation, a mechanism commonly described in viruses allowing the production of multiple functions from a single RNA (Sorokin et al. 2021).

To date, it is unclear how and if the segmented and putative multipartite nature of jingmenviruses confers them a selective advantage over their monopartite relatives. In order to gain knowledge on these viruses, we have searched for novel jingmenvirus sequences using publicly available next-generation sequencing data. We have also analysed the sequence of the genomic segments of 25 jingmenvirus species to better understand their organization. The data generated lead us to propose multiple new possible genomic organizations for jingmenviruses, with variable numbers of segments.

Material and methods

Discovery of new mosquito-related jingmenvirus sequences

We used the online Serratus platform, specifically the RdRp search functionality, to find novel mosquito-derived jingmenvirus sequences in published RNA sequencing libraries (Edgar et al. 2022). We searched the “Family” entitled Unclassified-899, as we identified that this corresponded to insect-related jingmenviruses in Serratus. The filters used were a percentage identity above 75% and a score above 50. For the 97 matches found, we selected the ones with a mosquito or mosquito-related host, as we were originally looking for novel mosquito-derived jingmenvirus sequences in this study.

Additionally, we screened for divergent hits related to the Flavi-like segmented virus strain US001 NS5 gene (MN811583.1) and manually assembled new sequences from SRA libraries with scores above 15, percentage identity above 50% and more than 25 reads (Vandegrift et al. 2020).

We either downloaded the raw sequencing reads from the NCBI website and assembled the new genomic sequences using Geneious Prime 2024.0.7 or MEGAHIT v1.2.9; or we performed the assemblies using NCBI BLAST with its direct access to the Sequence Read Archive (SRA) database, with the Megablast algorithm optimized for highly similar sequences with maximum target sequences set to the highest setting (5000) for the RdRp or with the blastn algorithm optimized for somewhat similar sequences for all other sequences. In both cases, we used the RdRp contigs assembled by Serratus or published sequences (GCXV KM461666 to KM461670; SAIV7 KR902717 to KR902720; OKIAV332 MW314687) as references for the assemblies.

Sequence analysis

Multiple sequence alignments were performed with MUSCLE using Geneious Prime 2024.0.7, which provided the percentage identity matrices.

The best models to use in order to infer the maximum likelihood phylogenies presented here were selected for each alignment using the online tool Smart Model Selection in PhyML (http://www.atgc-montpellier.fr/sms/; Lefort et al. 2017). The phylogenetic analyses were performed using the PhyML plugin in Geneious Prime 2024.0.7, with an LG substitution model and 100 bootstraps. Phylogenies were mid-point rooted using FigTree v1.4.4 (http://tree.bio.ed.ac.uk/software/figtree/).

Identification of slippery sequences and ribosomal frameshifts

We followed guidelines described by McNair et al. to identify potential heptanucleotide slippery motifs that could result in a ribosomal frameshift in jingmenvirus segments with multiple ORFs (McNair et al. 2024). We used HotKnots 2.0 to compute potential secondary structures in the 100 nucleotides following the identified slippery heptanucleotides and to estimate the minimum free energy of said secondary structures (www.rnasoft.ca) (Andronescu et al. 2003). We then used the PRFECT software developed by McNair et al. and confronted the results obtained to the motifs identified above. When multiple possible motifs were identified, we kept the ones identified using both the manual method and the software results. When multiple motifs remained, we selected the one with the highest probability value generated by the PRFECT software. When this criterion did not help to discriminate between multiple motifs, or when PRFECT had not identified a motif in the sequence, although we had identified several, we performed amino acid sequence alignments with the multiple resulting protein sequences. We selected the frameshift position that resulted in the most conserved amino acid sequence. All predicted amino acid sequences were then translated and aligned to confirm that the resulting frameshifted sequences shared a similar sequence identity with related sequences as the rest of the protein, and that highly conserved residues were translated correctly.

Screening for IRES sequences in jingmenvirus genomic segments with multiple ORFs

We used the tool DeepIRES developed by Zhao et al. to screen the segments with multiple ORFs of jingmenvirus genomes for the presence of internal ribosome entry site (IRES) (Zhao et al. 2024). We then compared the prediction obtained from DeepIRES with the predicted ORFs in these genomic segments to see if the DeepIRES predictions could be associated to a biological relevance.

Protein structure modelling

The structures of VP4 and VP1 of GCXV, and the mosquito-derived jingmenvirus sequences discovered here; as well as both VP2 versions of SAIV7 were predicted using AlphaFold2 Google Colab (Jumper et al. 2021, Mirdita et al. 2022). Model super-impositions were generated using Flexible structure Alignment by Chaining Aligned fragment pairs, allowing Twists (FATCAT 2.0) server (Li et al. 2020) and visualized using UCSF Chimera X (Pettersen et al. 2004). The similarity detected by FATCAT 2.0 is evaluated by a P-value. The smaller the P-value the more statistically significant will be the similarity between the two structures. Pairwise sequence alignments were performed using the Needleman–Wunsch algorithm provided by UCSF Chimera X and analysed with ESPript 3.0 (Robert and Gouet 2014). We searched for VP1 and VP4 distant homologs by analysing the 3D secondary structure of models generated here by AlphaFold2 with Foldseek, a 3D structural alignment tool (van Kempen et al. 2024).

Results

Discovery of new mosquito-related jingmenvirus sequences

Using Serratus to search for novel mosquito-derived jingmenvirus RdRp coding sequences, we identified seven libraries of interest. Five libraries (SRX8323447, SRX8323446, SRX8323444, SRX8323437, SRX8323431) from wild-caught mosquito excreta in Far North Queensland, Australia, in 2018 had a partial match with 76–78% nucleotide identity to the RdRp of GCXV and MoCV (Ramírez et al. 2020). One library (SRX4113301) from a pool of 4400 Cx. tritaeniorhynchus mosquitoes collected from Yunnan, China, in 2016 had a partial match with 79–80% nucleotide identity to the RdRp of GCXV and MoCV (Xiao et al. 2018). The seventh library (SRX833670) from an uncultured mosquito pool collected from Zhejiang, China, in 2013 had a match with 99% identity to the RdRp of Shuangao insect virus 7 (SAIV7). This mosquito pool contained the following species: Aedes albopictus, Armigeres subalbatus, Anopheles paraliae, An. sinensis, Culex pipiens, Cx. sp., Cx. tritaeniorhynchus (Li et al. 2015).

We first focused on the most divergent sequences and assembled the novel jingmenvirus genomic sequences from mosquito excreta and Cx. tritaeniorhynchus pool, then moved on to the SAIV7 positive library. All newly described sequences were deposited in Genbank under accession numbers BK070205-BK070273.

New jingmenvirus sequences from mosquito excreta

The assemblies performed on the mosquito excreta libraries identified with Serratus yielded the genomic sequence for a putative novel jingmenvirus, most closely related to GCXV according to NCBI BLASTx search results. The genome was first assembled using SRX8323431 and all segments were subsequently detected in the other four libraries identified with Serratus. Strikingly, we identified six putative viral segments instead of five, as is the case for GCXV or four, as is the case for all other described jingmenviruses.

The full nucleotide sequences of segments 1 (NSP1) and 2 (NSP2) shared only 71% and 50% identity, respectively, with GCXV segments 1 and 2. Among the other four segments, two putative versions of both segments 3 (34.9% and 50.4% nucleotide identity with GCXV) and 4 (49.7% and 48.2% nucleotide identity with GCXV) were assembled from the data (Fig. 1), while no sequence related to GCXV segment 5 was identified. We searched for evidence of co-infection or multiple infections in the libraries by searching for additional segments 1 and 2. We did not find any in the data, even when assemblies were performed with less stringent mapping criteria, using GCXV and the new sequences as references, suggesting the six segments participate in a single infection unit. Moreover, the initial screening tool, Serratus, only provided one GCXV RdRp-related contig per library, even with less stringent amino acid percentage identity criteria.

Figure 1.

Genomic organization of two new mosquito-derived jingmenviruses, FNQJV1 and CTJV1 compared to the published sequence of GCXV. ORFs are represented by blocks of colour, and translation resulting from a ribosomal frameshift is represented by overlapping blocks in pastel colours, with the position of the frameshift and slippery heptanucleotide sequence specified underneath.

In the two segments designated as segment 3, three ORFs coding for VP1 to VP3 were identified in the first one (hereafter designated segment 3-1), similarly to GCXV. Only two ORFs, coding for VP1 and VP3 were identified in the second one (hereafter designated segment 3-2).

In segment 3-1, most of the genetic distance was attributed to the second half of the sequence (Fig. 2). Indeed, VP1 and VP2 coding sequences were identified (50% nucleotide identity with GCXV) whereas the VP3 coding sequence only shared 24% nucleotide identity with GCXV VP3, and was almost 600 nucleotides shorter than the GCXV VP3 ORF, while the 3ʹUTR of this segment (163 nucleotides) had a length similar to that of GCXV segment 3 3ʹUTR (199 nucleotides), indicating the segment is indeed shorter due to the VP3 ORF length.

Figure 2.

Percentage identity derived from multiple sequence alignment of GCXV, MoCV, FNQJV1, and CTJV1 genomic sequences. The bottom left of each table represents nucleotide identity over the whole genomic segment. The top right of each table corresponds to amino acid percentage identity over the corresponding ORF. Highlighted in blue are the sequences discovered and assembled in this study (Genbank accession numbers starting with BK).

The other version of segment 3 (segment 3-2) shared 50% nucleotide identity overall with GCXV and 33% nucleotide identity with the first version in segment 3-1. While ORFs for VP1 and VP3 were identified, no obvious additional ORF with START and STOP codon or possible frameshifting with VP1 and VP3 was identified, thus VP2 coding sequence is not present in this version.

We chose the designations segment 3-1 and 3-2 based on the similarity of genomic organization compared to GCXV, with segment 3-1 being the version with the genomic organization closest to GCXV.

The two versions of segment 4 shared 41% nucleotide identity with each other and 48–50% nucleotide identity with GCXV. ORFs coding for VP4 and VP5-6 were detected in both versions. Considering that the genomic organization was the same for GCXV and these two segments, we chose to assign numbers to the versions based on the percentage nucleotide identity with GCXV over the whole segment. Segment 4-1 is slightly more similar to GCXV than segment 4-2 with 49.7% nucleotide identity, versus 48.2% (Fig. 2).

We searched the 52 SRA libraries from this BioProject (PRJNA631724) for each of the six identified segments using NCBI BLAST. We found highly similar reads mapping to these jingmenvirus-associated sequences in 11 libraries, all from Far North Queensland, none from South East Queensland. We tentatively named this group of six sequences Far North Queensland jingmenvirus 1 (FNQJV1). The six segments were always found together, except in two libraries where the number of identified reads were below 50 for the most detected segments, therefore, which might not have had sufficient levels of viral RNA for the whole genome to be detected (Table 1). Of note, segments 3-2 and 4-2 were present in higher numbers of reads than 3-1 and 4-1, respectively in the majority of libraries.

Table 1.

Number of reads mapping to FNQJV1 six genomic segment sequences (BK070205–BK070210) in libraries from BioProject PRJNA631724 and to CTJV1 (BK070211–BK070216) in libraries from BioProject PRJNA472635, divided by the length of the respective segment and multiplied by 1000 (average number of reads per kilobase) The assemblies were performed using BLAST NCBI with the megablast algorithm, optimised for highly similar sequences, with a maximum target sequence set to the highest setting (5000; >LD = higher than limit of detection, >5000 reads). This method did not detect reads similar to FNQJV1 sequences in any of the other 41 libraries from PRJNA631724, nor did it detect reads similar to CTJV1 in the other two libraries from PRJNA472635. In bold are the libraries that had been identified using Serratus, and underlined is the library used to assemble FNQJV1 and CTJV1 segment sequences. seg: Segment.

Library	Reference	seg1	seg2	seg3-1	seg3-2	seg4-1	seg4-2
SRX8323428	FNQJV1	51	59	102	139	50	191
SRX8323431	FNQJV1	>LD	>LD	>LD	>LD	>LD	>LD
SRX8323435	FNQJV1	21	65	129	189	134	268
SRX8323436	FNQJV1	30	92	30	96	45	123
SRX8323437	FNQJV1	226	502	807	901	554	>LD
SRX8323439	FNQJV1	0,0	15	1,4	10	4,6	7,5
SRX8323443	FNQJV1	0,3	1,1	1,9	2,3	0,8	0,0
SRX8323444	FNQJV1	251	516	281	642	258	1029
SRX8323446	FNQJV1	321	439	432	757	573	1270
SRX8323447	FNQJV1	832	1282	1613	>LD	1620	>LD
SRX8323477	FNQJV1	12	12	29	77	8,87	419
SRX4113301	CTJV1	121	49	126	93	64	1602

New jingmenvirus sequence from Culex tritaeniorhynchus

Another library of interest identified with Serratus was from a pool of Cx. tritaeniorhynchus mosquitoes collected from Yunnan, China in 2016 (SRX4113301). From this library, we obtained the genomic sequence for another novel jingmenvirus, Culex tritaeniorhynchus jingmenvirus 1 (CTJV1), and despite a lower coverage for the assemblies than for FNQJV1, we identified that CTJV1 and FNQJV1 share a common genomic organization. Indeed, two versions of segments 3 and 4 were assembled from the data, closely related to the corresponding versions of FNQJV1 (one version of segment 3 has three putative ORFs for VP1 to VP3 and the second version has no VP2 ORF; both versions of segment 4 have three putative ORFs), while only one version of segments 1 and 2 was assembled for each (Fig. 1, Fig. 2).

Using NCBI BLAST, we searched the three SRA libraries from this BioProject (PRJNA472635) for each of the six identified CTJV1 segments and found reads mapping to CTJV1 only in the original Cx. tritaeniorhynchus library identified with Serratus (Table 1).

Interestingly, the two newly identified mosquito-derived sequences (FNQJV1 and CTJV1) are more closely related to each other than to any other published sequence, for all segments, despite the geographical distance of the samples that yielded the positive libraries (Fig. 2). Indeed, FNQJV1 and CTJV1 share 89% amino acid identity over NSP1, 80% over NSP2, 53% over VP1/VP3 and 69% over VP2, for segment 3-1, 64% over VP1/VP3 for segment 3-2, 88% and 75% over VP4 for segment 4-1 and 4-2, respectively, 75% and 61% over VP5-6 for segment 4-1 and 4-2, respectively (Fig. 2). These values are much higher than the percentage identity they shared with GCXV or MoCV, their closest published relative. Moreover, in addition to their shared genomic organisation, FNQJV1 segment 3-1 is much more closely related to its homolog CTJV1 segment 3-1 than it is to FNQJV1 segment 3-2. This relationship is true for all homologous segments described here and can be observed in the phylogenetic analyses presented in Fig. 3. It should be noted that the lowest percentage identity values were observed for segment 3 and in particular for VP1-3. These low values are heavily influenced by the divergence of the VP3 sequence in segments 3-1.

Figure 3.

Maximum likelihood analysis of homologous jingmenviruses amino acid sequences: NSP1, NSP2, VP1 (mosquito-derived sequences) or VP2 (insect-derived sequences), and VP4 (mosquito- and insect-derived). The phylogenetic trees were constructed using the LG model, gamma distributed with 100 bootstraps (branch labels) and midpoint rooted using FigTree. The scale bar represents the number of amino acid substitution per site. The sequences highlighted in red were discovered and described in this study with 5 or 6 segments and the sequences highlighted in blue are part of viruses for which we identified additional segments (5 segments). Guaico Culex virus is the only other described jingmenvirus with more than 4 segments (5 segments).

Identification of an additional genomic segment for Shuangao insect virus 7

The last library of interest identified with Serratus was an uncultured mosquito pool collected from Zhejiang, China, in 2013 (SRX833670). The assemblies performed on these data yielded the genomic sequence (segments 1, 2, 3, 4) for a new strain of SAIV7. These nucleotide sequences share 99.3–99.5% identity to the SAIV7 reference sequences (KR902717 to KR902720), obtained from another uncultured insect pool (SRX833685) from the same BioProject (PRJNA271540) (Shi et al. 2016).

When comparing these four sequences to published sequences using BLAST, we noticed that segments 1 (coding for NSP1), 3 (NSP2) and 4 (VP2 and VP3) shared 98% identity with Dipteran jingmen-related virus (OKIAV332) (MW314686, MW314688, MW314689), while segment 2 (coding for VP1 and VP4, thereafter designated segment 2-1) shared only 43% nucleotide identity with OKIAV332 segment 2 (MW314687; thereafter designated segment 2-2) (Fig. 4) (Paraskevopoulou et al. 2021). Moreover, SAIV7 segment 2-1 shared 99% identity with Sichuan mosquito virus 1 (SCMV1) segment 2, this segment being the only published sequence for this virus (MZ556307) (Zhao et al. 2022). In order to determine whether these libraries were proof of segment 2 reassortment between viruses or an indication that this virus genomic sequence actually includes five segments, we searched for the five sequences (segments 1, 2-1, 2-2, 3 and 4) using NCBI BLAST/SRA in these records as well as in the library identified with Serratus (SAIV7: SRX833685; OKIAV332: SRX798056; SCMV1: SRX10979868; Serratus: SRX833670). We detected all five segments in the four libraries (Table 2), with VP1 and VP4 ORFs in both segment 2-1 and 2-2 (Fig. 4).

Figure 4.

Genomic organization of Shuangao insect virus 7 (SAIV7) and Jingmenvirus Cameroon (JVC). ORFs are represented by blocks of colour, and translation resulting from a ribosomal frameshift is represented by overlapping blocks in pastel colours, with the position of the frameshift and slippery heptanucleotide sequence specified underneath.

Table 2.

Average number of reads per kilobase mapping to the five genomic segment sequences of SAIV7 in the libraries of the three studies with published sequences highly similar to SAIV7, and in the libraries identified with Serratus The assemblies were performed using BLAST NCBI with the megablast algorithm, optimized for highly similar sequences, with a maximum target sequence set to the highest setting (5000; >LD = higher than limit of detection, >5000 reads). The references used for the BLAST search were KR902717 (segment 1), KR902718 (segment 2-1), MW314687 (segment 2-2), KR902719 (segment 3), and KR902720 (segment 4). seg: Segment.

Library	seg1	seg2-1	seg2-2	seg3	seg4	Sample	Country	BioProject
SRX833685 (SAIV7)	>LD	>LD	>LD	>LD	>LD	Chrysopidae sp., Psychoda alternata, Diptera sp.	China	PRJNA271540
SRX798056 (OKIAV332)	283	458	479	247	252	Clogmia albipunctata	USA	PRJNA267928
SRX10979868 (SCMV1)	7	49	4	22	10	Mosquito	China	PRJNA680461
SRX833670 (Serratus)	182	1525	237	323	210	Mosquito	China	PRJNA271540
SRX5887211	235	683	409	304	388	Anser indicus	India	PRJNA526291
SRX13680292	326	>LD	859	614	564	Hystrix brachyuran	China	PRJNA795267
SRX13667241	>LD	>LD	>LD	>LD	>LD	Hystrix brachyuran	China	PRJNA795267
SRX13680277	373	>LD	2361	854	1447	Paguma larvata	China	PRJNA793740
SRX14333807	824	2534	631	1485	>LD	Wastewater	USA	PRJNA771693
SRX14333808	763	1353	406	1001	1777	Wastewater	USA	PRJNA771693
SRX14333809	508	1578	279	779	984	Wastewater	USA	PRJNA771693
SRX3035916	117	1421	316	248	217	Wastewater	Brazil	PRJNA395784

Furthermore, using Serratus looking specifically for SAIV7-related sequences, we found 12 libraries that had RdRp sequences with >95% identity to SAIV7 (SRX5887211, SRX13680292, SRX13667241, SRX833670, SRX833685, SRX13680277, SRX14333774, SRX14333807, SRX14333808, SRX14333809, SRX16702048, SRX3035916). We had already screened two of these libraries (SRX833670 and SRX833685) for the presence of the five segments. Two other libraries (SRX14333774 and SRX16702048) were not available for assembly with NCBI BLAST/SRA and were not included in the further search. We found that the five genomic segments (1, 2-1, 2-2, 3, and 4) were represented in the remaining eight libraries. All SAIV7 sequences assembled here are available on Genbank under accession numbers BK070217 to BK070267. The positive libraries originated from samples as diverse as an oral swab of bar-headed goose Anser indicus from India, a nasal swab and faecal sample from Malayan porcupine Hystrix brachyura from China, a faecal sample from masked palm civet Paguma larvata from China and wastewater from the USA and Brazil (Table 2).

In this study, we therefore found five SAIV7 genomic segment sequences in a wide distribution of hosts: a sample with pooled Chrysopidae sp., Psychoda alternata, Diptera sp.; a sample with pooled Ae. albopictus, Ar. subalbatus, An. paraliae, An. sinensis, Cx. pipiens, Cx. sp, Cx. tritaeniorhynchus; and samples containing each Clogmia albipunctata, Hystrix brachyura swabs, Paguma larvata swabs, Anser indicus swabs and waste water; in samples collected from China, India, Brazil, and the USA between 2012 and 2021 (Li et al. 2015, Paraskevopoulou et al. 2021, He et al. 2022, Zhao et al. 2022).

Identification of an additional genomic segment for Jingmenvirus Cameroon

As we were comparing the SAIV7 genomic sequences to published sequences using NCBI BLAST, we noticed that their most closely related sequences are the genomic segments of a jingmenvirus sequence detected from human blood samples from Cameroon (70–83% nucleotide identity for segments 1, 2-1, 3, and 4) (Fig. 5) (Orf et al. 2023). The authors reported four genomic segments for this virus they named Jingmenvirus sp. strain Cameroon/U172471/2017 (hereafter described as Jingmenvirus Cameroon or JVC). Upon our request, the raw sequencing reads for this metagenomics study were added to the NCBI SRA public database (SRR31402545, SRR31402546, SRR31402547) and we were able to assemble a fifth genomic segment from the raw data, highly similar to SAIV7 segment 2-2 (93–99% nucleotide identity to SAIV7 segment 2-2, Genbank accession number BK070268) (Fig. 5).

Figure 5.

Percentage identity derived from multiple sequence alignment of SAIV7 strains SKC, OKIAV332, SCMV1, and JVC genomic sequences. The bottom left of each table represents nucleotide identity over the whole genomic segment. The top right of each table corresponds to amino acid percentage identity over the corresponding ORF. Highlighted in blue are the sequences discovered and assembled in this study (Genbank accession numbers starting with BK).

These nucleotide sequences (SAIV7 segment 2-2 and JVC segment 2-2) have only one significant match using BLASTn, the published sequence of OKIAV332 seg2-2 (MW314687), but are similar to other jingmenvirus segment 2 sequences using BLASTx (nucleotide query compared to protein databases).

Genomic fluidity is also found in the tick-associated jingmenvirus clade

Following these discoveries in the insect-associated clade of the jingmenvirus phylogeny, we found a novel jingmenvirus genomic sequence with five segments in library SRX3777434 from a single adult female Dysdera bandamae spider body (no legs or pedipalps) from the Canary Islands, Spain in 2015 (Fig. 6) (BK070269-BK070273) (Vizueta et al. 2019). This virus was tentatively named D. bandamae jingmenvirus 1 (DBJV1) and its most closely related sequences are the genomic segments of Hainan jingmen-like virus (HJLV), detected in Chinese forest soil metagenome collected in 2018, one of the most divergent sequence in the tick-associated clade of the jingmenvirus phylogeny (Fig. 3) (Chen et al. 2022). DBJV1 genome has two versions of segment 2, one coding for two ORFs (nuORF and VP1; hereafter designated segment 2-1) while the other version does not seem to code for nuORF (hereafter designated segment 2-2). The segment 2-1 nuORF sequence has no match on BLASTx; however, we have identified an ORF in the closely related HJLV segment 2 (128-526) that shares 39.6% amino acid identity with DBJV1 nuORF and that was not annotated upon publication.

Figure 6.

Genomic organization of Dysdera bandamae jingmenvirus 1 (DBJV1). ORFs are represented by blocks of colour, and translation resulting from a ribosomal frameshift is represented by overlapping blocks in pastel colours, with the position of the frameshift and slippery heptanucleotide sequence specified underneath.

Interestingly, the UTRs are more conserved than the coding regions between segment 2-1 (with nuORF and VP1) and segment 2-2 (with VP1 only) as they share 52.3% nucleotide identity (5ʹ UTR) and 57.1% nucleotide identity (3ʹ UTR) while the coding regions only share 44.3% nucleotide identity. This was also found for FNQJV1 segment 3-1 (VP1-3 with non-conserved VP3; and VP2) and 3-2 (VP1-3 only) which share 46.3% and 43.4% nucleotide identity over the 5ʹ and 3ʹ UTRs respectively while they only share 37.3% nucleotide identity over the coding sequences. This discrepancy was not observed for FNQJV1 and CTJV1 segments 4-1 and 4-2 or for SAIV7 segments 2-1 and 2-2. We were not able to assemble the full UTR sequences of CTJV1 segments 3-1 and 3-2 so the comparison cannot be made for these two segments.

We searched for DBJV1 reads in all other samples of the BioProject containing SRX3777434 and found that none had evidence of DBJV1 sequence, including the libraries containing the metagenome of the legs and pedipalp of the same spider (#2 spider ID CRBA2178). In library SRX3777434, the average number of reads per kilobase mapping to the five genomic segment sequences of DBJV1 were as follows: 144 reads for segment 1; 721 for segment 2-1; 465 for segment 2-2; 312 for segment 3 and 1213 for segment 4.

Identification of translation initiation mechanisms

Slippery heptanucleotides have previously been proposed as a mechanism for coding several proteins in a single segment for GCXV, MoCV, JMTV, and Alongshan virus. As we analysed the sequences and genomic organization of the newly discovered viruses, we noticed that most jingmenvirus segments with multiple ORFs contained slippery heptanucleotide motifs that could facilitate a ribosomal frameshift. We therefore systematically analysed genomic sequences of 25 jingmenvirus species for the presence of such motifs.

In the genomes of all mosquito-derived jingmenviruses (GCXV, MoCV, FNQJV1, CTJV1), we identified strictly conserved heptanucleotide motifs in segments 3 (VP1-3 GGAUUUU) and 4 (VP5-6 AAAAAAC) (Table S1). We also identified motifs in all sequences of the more distantly related insect-derived jingmenvirus we analysed (no strict consensus, segment 2 VP4-1; GGUUUUU in most cases, segment 4 VP2-3) and in tick-associated jingmenvirus sequences (segment 2 VP1a-b AAAAAAC; segment 4 VP2-3 GGUUUUU or AAAUUUU) (Table S1). Therefore, ribosomal frameshift is the putative mechanism that initiates translation of multiple ORFs on a single segment and is conserved among all jingmenviruses: tick-associated and insect-associated (including mosquito-derived), and in all versions of homologous segments.

Some remaining ORFs that require alternative translation initiation have no evidence of ribosomal frameshift: VP4/VP5 and VP1/VP2 in mosquito-derived jingmenviruses and nuORF/VP1 in tick-associated jingmenviruses (Figs 1 and 6 for reference). For these ORFs, the mechanism allowing translation initiation also seems conserved and is most likely to be leaky scanning. Indeed, in almost all instances, the first AUG in the segment corresponds to the methionine START of the first ORF to be translated (VP4 or VP1 for mosquito-derived jingmenviruses; nuORF for tick-associated jingmenviruses), while the second AUG corresponds to the methionine START of the second ORF to be translated (VP5 or VP1 for mosquito-derived jingmenviruses; VP1 or VP1a for tick-associated jingmenviruses) (Table S2). For some viruses, the methionine START of the second ORF is the second or third instance of AUG after the initial one (Table S2).

Finally, we used the tool DeepIRES to screen for the presence of IRES in the jingmenvirus genomic segments with multiple ORFs and found no instances where the predictions from DeepIRES matched a predicted ORF, suggesting this mechanism for alternative translation initiation might not be relevant to jingmenvirus biology.

The predicted mechanisms we propose here (ribosomal frameshift and leaky scanning) are conserved for all jingmenvirus sequences analysed but need to be confirmed with biological samples, e.g. using mass spectrometry as it was done to confirm ribosomal frameshifting in GCXV segment 3 (Ladner et al. 2016).

Structural analysis

We chose which segment would be designated x-1 or x-2 based either on which sequence displayed similar segment organization to the closest relative (similar ORFs: segment x-1; ORFs missing: segment x-2) or in case the segment organization was the same, based on which sequence had a higher nucleotide percentage identity to the closest relative (highest similarity: segment x-1; lowest similarity: segment x-2). However, these rules based on sequence comparisons might not be applicable to more divergent sequences with low to no homology with other published jingmenvirus sequences. In general, the percentage identities shared between different jingmenvirus species or even between homologous segments coding for putative structural proteins of one virus species can be around 30% or lower, similar to the percentage obtained by aligning two unrelated sequences (as shown in Fig. 2 and Fig. 5). Given this important genetic distance, we generated structural models for the putative structural proteins in the FNQJV1, CTJV1, SAIV7, and GCXV sequences using AlphaFold2 and performed structural alignments with FATCAT and Chimera X to determine whether the homologous segments we identified indeed code for homologous proteins. Using this method, we also attempted to confirm which segments encode similar proteins between the mosquito-derived sequences (GCXV, CTJV1, FNQJV1) and more distantly related SAIV7, as its genomic organization follows a different nomenclature compared to mosquito-derived jingmenviruses, due to low sequence identity in the segments that are not coding for non-structural proteins.

We first generated the structural models of VP1 the only protein in common for GCXV segment 3, FNQJV1 segments 3-1 and 3-2, and CTJV1 segments 3-1 and 3-2. The models were superimposed in order to generate a structural alignment. The structures showed statistically significant similarity (Table S3) and conserved motifs were identified (Fig. 7). VP1 is organized in three domains, a N-terminal domain composed of three α-helices, a central domain containing three antiparallel β-strands, and a C-terminal domain organized in six β-strands surrounded by five α-helices (Fig. 7). We then performed pairwise comparisons between GCXV VP1 and SAIV7 putative structural proteins and found that GCXV VP1 shares statistically significant structure similarities with SAIV7 VP2 (P = 1.38 × 10⁻⁴, Table S3) but is not statistically significantly similar to SAIV7 VP4 (P = .459 and .271 for segment 2-1 and 2-2, respectively). We were therefore able to include SAIV7 VP2 in the multiple structure alignment presented here (Fig. 7, Table S3) and identify the three domain-organization, and conserved residues and motifs shared between SAIV7 VP2 and all mosquito-derived jingmenvirus VP1 sequences, including four cysteines involved in two putative disulphide bonds contributing to the structural organization of the N- and C-terminal domains. For reference, a pairwise amino acid alignment between GCXV VP1 and SAIV7 VP2 results in 18.6% amino acid identity.

Figure 7.

Structural comparison of representative structures of VP1 and bioinformatics analysis. (a) GCXV_VP1 (grey), CTJV1_seg3-1_VP1 (beige), CTJV1_seg3-2_VP1 (blue), FNQJV1_seg3-1_VP1 (green), FNQJV1_seg3-2_VP1 (orange), and SAIV7_VP2 (pink) structures were generated using AlphaFold2 (Jumper et al. 2021). Structures were superimposed using flexible FATCAT (Li et al. 2020) and visualized in Chimera X (Pettersen et al. 2004). (b) A structural alignment was generated by Chimera X and analysed with ESPript. The numbers on top of the alignment indicate the amino acid positions in the GCXV_VP1 sequence. Spirals and arrows above the alignment indicate the position of α-helices and β-strands, based on the AlphaFold2 models. Cysteines that may be involved in disulfide bridges are indicated by filled triangle.

Finally, we aligned the structures of VP4 sequences from GCXV segment 4, FNQJV1 segment 4-1 and 4-2, CTJV1 segment 4-1 and 4-2 and found significant similarity for all (Table S3, Fig. 8). We also included both versions of VP4 from SAIV7 in the multiple structure alignment and found significant similarity between the seven VP4 structures (Fig. 8), all organized in five α-helices. For reference, a multiple amino acid alignment of these sequences results in 18–26% amino acid identity between the mosquito-derived VP4 sequences and SAIV7 VP4. As shown in Fig. 5, the VP4 homologs of SAIV7 share only 30–32% amino acid identity.

Figure 8.

Structural comparison of representative structures of VP4 and bioinformatics analysis. (a) GCXV_VP4 (grey), CTJV1_seg4-1_VP4 (dark blue), CTJV1_seg4-2_VP4 (light blue), FNQJV1_seg4-1_VP4 (dark orange), FNQJV1_seg4-2_VP4 (light orange), SAIV7_seg2-1_VP4 (dark green) and SAIV7_seg2-2_VP4 (light green) structures were generated using AlphaFold2 (Jumper et al. 2021). Structures were superimposed using flexible FATCAT (Li et al. 2020) and visualized in Chimera X (Pettersen et al. 2004). (b) A structural alignment was generated by Chimera X and analysed with ESPript. The numbers on top of the alignment indicate the amino acid positions in the GCXV_VP4 sequence. Spirals above the alignment indicate the position of α-helices, based on the AlphaFold2 models.

The structural models of mosquito-derived VP2, VP5-6, and VP7 (GCXV only) generated with AlphaFold2 were not obtained with sufficient confidence to be included in this study (Table S3, Figure S1).

No structural homologue was retrieved for either VP1 or VP4 models generated by AlphaFold2 using Foldseek (van Kempen et al. 2024).

Discussion

In this study, we report the discovery of three novel jingmenvirus genomes, which have prompted us to revisit our knowledge of the genomic organization of jingmenviruses. Until now, jingmenviruses were described as multisegmented viruses with positive single-stranded RNA genome containing four segments, with the exception of GCXV, with a genome containing five segments. Using a screening approach based on similarity to known jingmenvirus RdRp sequences in public sequencing raw data, we identified that more than four jingmenvirus genomic segments co-occur in different libraries, and that this observation was valid for both the tick-associated and insect-associated clades of the jingmenvirus phylogeny. In all instances, all versions of each segment were always found together in sequencing libraries, including in a library derived from a single spider (SRX3777434; DBJV1). In addition to this, we found a single version of the segments coding for the two non-structural proteins NSP1 and NSP2 in each library, and homologous versions of other segments. Taken together, these data suggest that the jingmenvirus sequences we found in one library are part of a single virus species, and that jingmenviruses could have up to six segments.

In addition to discovering new jingmenvirus genomic sequences with more than four segments, we uncovered the existence of additional segments in two published jingmenvirus species (SAIV7 and JVC). These newly uncovered segments had remained unidentified in the studies describing these virus species most likely because only four segments were expected to be found in jingmenvirus genomes. It would seem that this is not an isolated occurrence: while this manuscript was in preparation, Tang et al. published a new tool to reconstruct segmented virus genomes from metatranscriptomics data: SegVir, and Liu et al. published a pre-print article on a tool to identify RNA virus genome segments: SegFinder (Liu et al. 2024, Tang et al. 2024). Liu et al. conclude that they have identified multiple occurrences of additional genomic segments in virus species (unrelated to jingmenviruses) with multisegmented and multipartite genomes, compared to the number of segments classically described. This observation is reinforced by the recent identification of jingmen-related sequences in plant-derived samples using the conserved untranslated regions as markers (Zhang et al. 2024). Indeed, Ailanthus jingmen-related virus 1 was identified in the metagenome of the Ailanthus altissima tree and was found to have six genomic segments, two related to the jingmenvirus NSP1 and NSP2, respectively, and four sequences with low identity to jingmenvirus sequences, including two segments with ORFs similar to each other (Zhang et al. 2024). These data closely match our findings, and are in agreement with the existence of an evolutionary continuum shared by insect and plant viruses. These publications, taken together with our study, highlight that with the existing knowledge on multipartite viruses, assigning their genomic segments is not straightforward, especially when using metagenomics data rather than sequencing an isolate. While we cannot prove without a doubt with the data currently available that the duplicated segments identified here complete the genome of a single virus species, the arguments developed above are in favour of this hypothesis. In order to confirm the number of genomic segments within a single viral genome, and identify the role of each in viral replication, it would be valuable to obtain an isolate of one of these viruses, other than GCXV which does not seem to contain any multicopy segment (raw sequencing data were unavailable publicly, but the virus isolate reported have four or segments segments), and to develop associated molecular and infections tools.

Indeed, it remains to be seen whether the homologous versions of the segments are necessary for infection and/or replication or if they confer some sort of advantage for replication fitness, e.g. by conferring entry into different cell types. First, it is notable that GCXV VP7 homologs were not found, in any data set. The function of this protein/segment and the reason for its presence or absence in some isolates remains a mystery, especially considering that no fitness advantage to the presence of this segment was identified in vitro or in vivo to date (Ladner et al. 2016, Chen et al. 2024). Then, in the sequences identified, some homologous segments had ORFs that were absent (nuORF in DBJV1 segment 2-2; VP2 in FNQJV1 and CTJV1 segment 3-2), truncated or poorly conserved protein sequences (VP3 in FNQJV1 and CTJV1 segment 3-1). In other instances, we did not identify organizational differences between the ORFs of the two homologous versions of segment (SAIV7 segments 2-1 and 2-2; FNQJV1 and CTJV1 segments 4-1 and 4-2). This discrepancy could suggest different evolutionary processes that drove jingmenvirus diversification (Fig. 9). It should be noted that for the ORFs that were conserved in all versions, we were able to find statistically significant structural homology between the proteins they encoded (mosquito-derived VP1 and VP4 e.g.), despite the sequence divergence observed. It is likely that this structural similarity implies a function similarity, but as we have not performed functional studies here, the biological relevance of this structural redundancy is not clear at this stage. Indeed, as structural conservation is not necessary for functional conservation as exemplified by the Rossmann fold, a prevalent structural motif that can accommodate a wide range of functions (Medvedev et al. 2021).

Figure 9.

Fluidity of the jingmenvirus genomic organization. Jingmenviruses contain four to six homologous or non-homologous genomic segments per infectious unit. These homologous versions may share the same organization (ORFs, size) or may have complementary putative functions with ORFs lacking in one version (3-1 vs 3-2) but not the other (4-1 vs 4-2), for example. Different ORFs are represented in different colours, with grey representing missing or non-conserved ORFs.

Given the structural and organizational (conservation of slippery sequences, order, and number of ORFs) conservation of the segments and related viral proteins, it would be tempting to classify these homologous segments as paralogs, resulting from a duplication of the parental segment, followed by a drift of their sequence. However, multipartite viruses seem to be more tolerant than their monopartite counterparts to fluidity in their genomic organisation and to evolutionary processes that facilitate functional gain by different mechanisms, including duplication recombination, reassortment or gene acquisition (Michalakis and Blanc 2020). As mentioned in the introduction, multipartite virus particles encapsidate fewer fragments than the total number of genomic segments, multiple particles are required to initiate the replication cycle and they are extremely rare in animals (Michalakis and Blanc 2020). The genome plasticity observed here for jingmenviruses would be in line with a multipartite organization. Indeed, the fluid genomic organization could be explained by packaging of different combinations of segments leading to a reshuffling of genomic diversity and organization, with gain or loss of genomic material and gain, loss or overlap of function (Fig. 9). The segment duplication observed here could result from different mechanisms, e.g. a virus with an initial infectious unit of 4foursegments acquiring an extra version of a gene from another virus species during co-infection, therefore becoming a virus with an infectious unit of five segments—as opposed to segments reassortment, which would not result in a change in the number of segments in the infectious unit. Another possible mechanism for gene duplication in multipartite jingmenviruses is dynamic evolution by several rounds of adaptive pressure selection on multiple quasi-species at once in one virus species, once again increasing the number of segments in the infectious unit by one, by adding a homolog of an existing gene to the infectious unit. These mechanisms would be facilitated by the putative multipartite organization, with packaging of different combinations of segments in different virus particles as mentioned above, and therefore requiring multiple particles to obtain an infectious unit.

Moreover, for all sequences identified here, we observed variations in the sequencing depth between segments, which suggests a variation in the quantity of RNA of each segment in the different biological samples. This could be linked to the notion of genomic formula for multipartite viruses, as it has been observed that different segments of the multipartite genome might vary in orders of magnitude with relative frequencies that seem to be host-specific (Lucía-Sanz and Manrubia 2017, Michalakis and Blanc 2020). While this characteristic may increase the threshold of number of particles needed for an effective infection, it may also play a role in facilitating differential gene expression in different hosts (Michalakis and Blanc 2020). With these observations and the conclusions from Ladner et al. on GCXV following a virus dilution and low multiplicity of infection experiment (Ladner et al. 2016) raises the hypothesis that the presence of homologous segments could provide a selective advantage by increasing the host spectrum of jingmenviruses.

This study has highlighted the need for a simpler nomenclature for jingmenvirus genomic organization. The discrepancies between segment numbers and ORF designations have and will lead to confusion when comparing tick-associated (e.g. JMTV, DBJV1), insect-derived (e.g. SAIV7) and mosquito-derived (GCXV, MoCV, FNQJV1, CTJV1) jingmenvirus genomes. Here we confirm a structural homology between proteins from mosquito-derived jingmenvirus genomes and insect-derived jingmenvirus genomes which have a different nomenclature: VP1 from mosquito-derived jingmenvirus segment 3 and VP2 from insect-derived jingmenvirus segment 4; and VP4 from mosquito-derived jingmenvirus segment 4 and VP4 from insect-derived jingmenvirus segment 2. The structural comparisons yielded results with a much higher confidence than those obtained by sequence comparisons, as the percentage identities obtained by multiple sequence alignment were similar to those that can be obtained when comparing two unrelated sequences (≤30%). Continuing to confirm homology between ORFs in different clades of the jingmenvirus phylogeny would help simplify the nomenclature.

Considering the fluid nature of the jingmenvirus genome highlighted in this study, and the high sequence divergence observed between sequences in different phylogenetic clades, the simplest nomenclature would be to have segment 1 carrying the RdRp and methyltransferase functional domains; segment 2 carrying the serine protease and helicase domains; segment 3 carrying the putative envelope; segments 4, 5, 6, etc., for additional segments with unknown or untested functions, ranked according to size, as has historically been the case; a numbered suffix for any obvious homolog (3-1, 3-2 e.g.). Search of distant homology by protein structures alignments in addition to nucleotide or amino acid sequences alignment to attribute the segment number would contribute to rationalizing the nomenclature. ORFs with ribosomal frameshifts could be named ORF1a and ORF1ab for the version coding for the shorter or longer protein, respectively, coding for VP1 and VP1-3, rather than being named VP1 and VP3 (e.g.). Once the function of these ORFs encoding putative structural proteins is confirmed, the product of these ORFs could be renamed with that function (envelope e.g.) rather than “VP” followed by a number, which might differ between jingmenviruses from different clades. Implementing these changes could result in challenging circumstances if already published records were not corrected.

In addition to these proposed modifications in nomenclature, special care should be taken when annotating the ORFs in a new jingmenvirus genome. Indeed, based on sequence analyses, we propose different mechanisms for the alternative translation initiation in jingmenvirus segments with multiple ORFs: ribosomal frameshifts and leaky scanning, depending on the segment and ORF (Sorokin et al. 2021). In particular, the ribosomal frameshifts lead to updated ORF annotations for a number of segments in a number of jingmenvirus species (e.g. VP2-3 rather than VP2 and a separate VP3). Again, biological verification would be needed to confirm the relevance of these proposed molecular mechanisms, e.g. the same way that GCXV ribosomal frameshifts were confirmed: with peptides dependent on the frameshift detected in virus samples by mass spectrometry (Ladner et al. 2016).

Overall, particular care should be taken when publishing and depositing new jingmenvirus sequences, for all the reasons discussed above: considering the presence of homologous segments when metagenomic data are analysed will help to find “hidden” segments and ORFs on segments with multiple ORFs should be able to be translated following a molecular mechanism to be biologically relevant.

With the discovery of novel jingmenvirus-related sequences, we have been able to put forward new evidence of the multipartite nature of jingmenviruses, and of the evolutionary role this organization may play. With these new considerations, a harmonized nomenclature should be put in place whenever possible, aiming at clarifying, correcting, and simplifying the formal genomic organization of jingmenviruses.

Supplementary data

Supplementary data is available at VEVOLU online.

Conflict of interest:

None declared.

Funding

This work was supported by the European Virus Archive GLOBAL (EVA-GLOBAL) project that has received funding from the European Union’s Horizon 2020 Research and Innovation Program under grant agreement no. 871029.

Data availability

All newly described sequences were deposited in Genbank under accession numbers BK070205-BK070273.