Rodents as potential reservoirs for toroviruses

Abstract

Background

Emerging zoonotic viruses pose a significant threat to global health. The order Nidovirales includes diverse viruses, such as coronaviruses, which are well known for their zoonotic potential. Toroviruses are a less-studied genus within Nidovirales primarily associated with gastrointestinal diseases in ungulates, although some evidence suggests their presence in humans.

Methods

We set out to describe full-length genomes of potentially emerging viruses by collecting feces from dozens of small mammals, mainly rodents, captured in different regions of Spain. Viral reads were obtained by high-throughput Illumina sequencing and analyzed phylogenetically.

Results

In this study, we report the discovery of a novel torovirus from a fecal sample of a dormouse (Eliomys quercinus) in Spain, which we named Dormouse torovirus (DToV). This represents the first complete genome of a rodent-associated torovirus. The 28,555-nucleotide genome encodes the six characteristic torovirus open reading frames, but these exhibit low amino acid sequence identity (44.3–86.3%) compared to other toroviruses, indicating that DToV likely represents a new viral species. Moreover, the basal phylogenetic position of DToV suggests that rodents may represent a reservoir for this viral genus.

Conclusions

Our findings expand the known torovirus host range, underscore their potential for cross-species transmission, and highlight the importance of continued surveillance of wildlife viruses.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12985-025-02933-1.

Background

Viral zoonoses have been responsible for wide-ranging human health impacts [62, 64]. This has led to the strengthening of pathogen surveillance programs to collect information on circulating viruses in human and wildlife populations [64]. To this end, viral metagenomics has become the most widely used tool to characterize circulating viruses in different ecosystems [15]. The search for wildlife viruses has focused on tropical areas, since land-use alterations, high wildlife diversity, and bushmeat consumption, among other factors, are believed to increase disease emergence risk in these regions [3, 28]. However, it is important to extend this search to other regions, as some emerging viral diseases did not originate in tropical areas [23, 26]. In Europe, for instance, animal-to-human viral transmission events include those of tick-borne encephalitis virus, Dobrava virus, and Granada virus, among others [43, 44, 61].

Wildlife reservoirs are responsible for most emerging infectious diseases [15], but certain animal taxa give rise to more zoonoses than others. Although bats seem to harbor a significantly higher proportion of zoonotic viruses than all other mammalian orders [49], rodents are the most speciose group, and some species live in close contact with humans, increasing zoonotic risk [8, 41]. Dwellings and agricultural settings are among the highest-risk interfaces for zoonotic viral transmission, particularly from rodents [35]. For instance, the house mouse has facilitated the worldwide transmission of viruses to sympatric species [41]. In addition, rodents are reservoirs for many human pathogens, including hantaviruses and mammarenaviruses, and probably also embecoviruses [16, 69].

RNA viruses, particularly those with envelopes, are more prone to cross-species transmission and zoonosis than other viruses [22, 60]. A salient example of this group of viruses is the Coronaviridae family (order Nidovirales), which includes viruses that have repeatedly demonstrated their ability to switch hosts [31]. Much less known is the genus Torovirus, historically within the family Coronaviridae, but currently assigned as a unique genus to the subfamily Torovirinae, family Tobaniviridae [63], which includes four species: bovine torovirus (BToV), equine torovirus (EToV), porcine torovirus (PToV), and Bangali virus, as defined by the International Committee on Taxonomy of Viruses (ICTV; https://talk.ictvonline.org/taxonomy). Of note, Bangali virus was identified in a camel-derived tick [70], possibly acting as a transmission vector. Complete torovirus genome sequences have been found in several ungulates and a marsupial carnivore [13, 17, 29], whereas only partial sequences have been detected in humans [18], and there is no strong evidence for the existence of human toroviruses. Indeed, while toroviruses cause gastrointestinal disease in ungulates, pathological associations in humans remain unknown [5, 66, 68].

However, the current lack of knowledge about toroviruses makes it difficult to rule out their potential involvement in future zoonotic events [1]. In this context, given the threat that some members of the Nidovirales group pose to human health, our study focused on identifying viruses classified in this order. To do so, a metagenomic analysis of fecal samples from small terrestrial mammals in Spain was performed. Initially, multiple complete genomes of DNA viruses were identified in these samples [8]. In the present study, the analysis of the RNA fraction allowed the detection of the complete genome of a new species of torovirus, which we called Dormouse torovirus (DToV), identified in a rodent species (Elyomis quercinus).

Methods

Study area and sample collection

Box-like traps (Sherman, Ugglan, and Mesh traps) were used to capture small terrestrial mammals from different locations in eight Spanish provinces (Alicante, Cantabria, Huelva, León, Madrid, Salamanca, Valencia, and Zamora; Supplementary Table S1). The sampling period spanned from March to November 2022. A total of 160 individuals were captured, for which species, sex, and age were determined before release, except in a few cases, where samples were taken, without capture, from burrows of previously identified species. Most of the individuals belonged to the order Rodentia, and a small number of cases belonged to orders Eulypotyphla, Soricomorpha, Erinaceomorpha, Carnivora, and Lagomorpha. Fresh fecal samples were collected and kept individually in tubes containing 500 µL of 1X phosphate-buffered saline (PBS) at −20 °C until they were transported to the laboratory and stored at −80 °C for further processing.

Sample processing and nucleic acids extraction

A previous study showed that many of the samples presented an inhibitory agent that prevented sequencing [8]. Consequently, only the seventy-six samples lacking such an inhibitory effect were used here (Supplementary Table S1). These were combined as previously described into a total of thirteen pools [8], each containing between one and twelve samples from the same species (Supplementary Table S1). Processing of each pool was carried out as described before [8, 11]. Briefly, an aliquot of each sample was homogenized in a Precellys Evolution tissue homogenizer (Bertin), and the supernatant obtained after centrifugation was filtered using Minisart cellulose acetate syringe filters with a 1.2 μm pore size (Sartorius). For RNA extraction, 250 µL of the filtrate were cleaned with Trizol LS reagent (Invitrogen) and then 280 µL from this step were used for final extraction using the QIAamp Viral RNA minikit (Qiagen).

Sequencing and virus identification

The preparation of libraries from extracted nucleic acids was carried out using the stranded mRNA preparation kit (Illumina), skipping the mRNA enrichment steps and proceeding straight to the fragmentation step. Paired-end sequencing was performed on a NextSeq 550 device with a read length of 150 bp at each end. Raw reads were processed with fastp v0.23.2 [12], including deduplication, quality filtering, and trimming with a threshold of 20, and those reads below 70 nucleotides in length were removed. De novo sequence assembly was performed using SPAdes v3.15.4 [4] with the meta option and MEGAHIT v1.2.9 [40] using default parameters. Then, assembled contigs were clustered to remove replicates using CD-HIT v4.8.1 [39], and those shorter than 1,000 nt were removed. Taxonomic classification was performed using Kaiju v1.9.0 (Menzel, Ng and Krogh [45]) with the subset of the NCBI nr protein database downloaded on June 6, 2023. Viral contigs were identified using Virsorter2 v2.2.4 [25] and further analyzed with CheckV v1.0.1 [48] for genome quality assessment. Contigs were selected based on size, completeness, and potential vertebrate infectivity, according to the taxonomic classification assigned by Kaiju. Coverage statistics for the viral contigs were determined by remapping the filtered and trimmed reads to their corresponding contigs using Bowtie2 v2.2.5 [37]. The raw sequence reads were deposited in the Sequence Read Archive (SRA) of GenBank under accession numbers SRR32004501-13. The viral contig described in this study was deposited in GenBank under accession number PQ888429.

Genome annotation and phylogenetic analyses

Protein domains were annotated using Interpro 103.0 [6]. Open reading frames (ORFs) were predicted using ORFfinder (https://www.ncbi.nlm.nih.gov/orffinder). The ORFs corresponding to the viral contig described in this study were compared to NCBI databases using BLASTp to obtain identity values and refine annotations. For the phylogenetic analysis of the family Tobaniviridae, ninety-seven complete genome sequences were retrieved from the NCBI database. Genomes lacking either ORF1a or ORF1b were excluded from the analysis. Amino acid sequences corresponding to both ORFs were aligned separately using MAFFT v7.505 [32] with the --genafpair option. The protein domains M^pro, NiRAN, RdRp, ZBD, and HEL, used by ICTV for the taxonomic classification of nidoviruses, were extracted. Each of these domains was then individually realigned using MAFFT v7.505 with the L-INS-i algorithm. The resulting domain alignments were concatenated and redundant sequences were removed using CD-HIT v4.8.1 with a 100% identity threshold. The final dataset consisted of sixty-seven non-identical sequences, which were used to build a maximum likelihood (ML) phylogeny in IQ-TREE v2.3.6 [46] using the built-in ModelFinder function [30] to infer the best substitution model. Branch support was assessed using 1,000 ultra-fast bootstrap replicates [27] and 1,000 bootstrap replicates for the SH-like approximate likelihood ratio test. For the phylogenetic analysis of the genus Torovirus, the forty-eight complete genome sequences currently available in the databases were downloaded. Next, a nucleotide-based multiple alignment was performed using MAFFT v7.505 and then trimmed with trimAl v1.2rev59 (Capella-Gutiérrez, Silla-Martínez and Gabaldón [10]), setting the gap threshold option at 0.2. The ML phylogenetic trees, the best substitution model, and branch support were obtained as described above.

Ethics approval

According to the European directive regulating the protection of animals used for scientific purposes (2010/62/EU, Article 1), subsequently transposed into Spanish legislation (Royal decree 53/213, 1 February, Article 2), procedures used in this study (i.e. capture, non-invasive handling and in situ release of wild animals) are not subject to the condition of animal experimentation and, therefore, an IACUC approval document is not required, but specifically a permit from the competent regional authority.

Results

From metagenomics analysis, only one viral genome, identified in a fecal sample from a dormouse (Elyomis quercinus), was assigned to the order Nidovirales and classified as a torovirus by Kaiju, hereafter designated as Dormouse ToV (DToV). According to the ICTV, the members of the genus Torovirus present genomes of about twenty-eight kb in length. Congruently, DToV showed a genome size of 28,555 nt and a total of 8,085 reads from its corresponding library were remapped to this contig (average coverage = 39.12 ± 13.13). The assignment of DToV to the genus Torovirus was further confirmed by a phylogenetic analysis. From the complete genomes of all representatives of the Tobaniviridae family, the amino acid sequences of the most conserved regions were aligned (see Methods section) and a ML tree was constructed. This tree showed that DToV was included in the Torovirinae subfamily, but was not closely related to any of the four species described to date (Fig. 1).

Fig. 1.

ML tree of the Tobaniviridae family based on amino acid sequences of five conserved protein domains (M^pro, NiRAN, RdRp, ZBD, and HEL). Nodes were collapsed at the subfamily level, except for the Torovirinae subfamily, where only a single node comprising BToV sequences was collapsed. The colour legend represents the four accepted species of torovirus and two unclassified sequences from goat and Tasmanian devil (Torovirus sp.). The sequence identified in our study (DToV) is labelled in red. Substitution model LG + F + I + G4 was used. SH-aLRT and ultrafast bootstrap values higher than 80 and 95, respectively, are indicated in red circles. The tree is midpoint-rooted. Scale bar indicates the evolutionary distance in amino acid substitutions per site

For a more detailed study of phylogenetic relationships at the genus level, an analysis was performed using all currently available complete torovirus genomes. This phylogenetic analysis showed that DToV was clearly distant from the four currently accepted torovirus species and from other unclassified toroviruses identified in different hosts (Fig. 2A). From the data presented here, DToV emerged as the first complete genome of a rodent-associated torovirus. However, a metagenomic analysis recently identified torovirus sequences in fecal samples from the rodent Myodes glareolus species [51]. A complete genome could not be assembled in that study, but reanalysis of the raw sequencing data identified eight contigs associated with toroviruses. From these contigs, the complete sequence of the HE and N genes was recovered, and phylogenetic analysis involving these genes showed that DToV remained in a basal position, while the other rodent-associated torovirus was much closer to viruses identified in other hosts (Fig. 2B). This suggests that virus-host specificity in toroviruses may not be as high as proposed [21]. The main incongruence between the two phylogenetic analyses shown in Fig. 2 corresponds to the position of the antelope ToV, which is explained by its involvement in an HE gene recombination event [17].

Fig. 2.

ML phylogenetic trees based on the nucleotide sequence of ToV representatives for the whole genome (A) and the region including HE and N genes (B). The partial sequence corresponding to a recently described rodent-associated ToV [51] was also included in B. Trees were rooted using as outgroup a snake virus sequence (NC_046963) from another genus of the family Tobaniviridae. Substitution models used were GTR + F + G4 and TIM2 + F + I + G4, respectively. SH-aLRT and ultrafast bootstrap values higher than 80 and 95, respectively, are indicated in red circles. The tree is rooted at the midpoint. Scale bars indicate the evolutionary distance in nucleotide substitutions per site. Some branches have been extended by dotted lines to align taxon names

The DToV genome showed a G + C content of 38.7%, slightly higher than in other genomes, such as bovine, antelope, and porcine toroviruses (38, 37, and 35%, respectively) [17]. This genome included 5’ and 3’ non-translated regions (NTR, 877 and 199 nt in length, respectively), six characteristic ORFs, and also the two commonly deduced CUG-initiated ORFs encoding U1 and U2 putative proteins within the 5’-NTR and ORF1a, respectively [56] (Fig. 3; Supplementary Table S2). The replicase polyprotein (pp1ab), jointly encoded by ORF1a and ORF1b, included eleven predicted domains: ADP-ribose 1-phosphatase (ADRP), papain-like protease (PLP), 3 C-like main protease (M^pro), cyclic phosphodiesterase (CPD), nidovirus RdRp-associated nucleotidyltransferase (NiRAN), RNA-dependent RNA polymerase (RdRp), Zn-binding domain (ZBD), Helicase (Hel), 3’-to 5’ exoribonuclease domain (ExoN), nidoviral uridylate-specific endoribonuclease (NendoU), and ribose 2’-O-methyltransferase (MT). The CPD domain (IPR039573), detected at the 3’-end of ORF1a, is related to NS2, identified only in a lineage A betacoronavirus but at a different genome location [24].

Fig. 3.

Schematic representation of the DToV genome showing ORF distribution. Conserved protein motifs/domains are highlighted in coloured boxes as follows: A represents the ADP-ribose 1-phosphatase (ADRP); P represents the papain-like protease (PLP); M^pro represents the 3 C-like main protease; C represents the cyclic phosphodiesterase (CPD); Ni represents the nidovirus RdRp-associated nucleotidyltransferase (NiRAN); Rd represents the RNA-dependent RNA polymerase (RdRp); Z represents the Zn-binding domain (ZBD); Hel represents the Helicase; Ex represents the 3’-to 5’ exoribonuclease domain (ExoN); N represents the nidoviral uridylate-specific endoribonuclease (NendoU); MT represents the ribose 2’-O-methyltransferase. Furin predicted cleavage sites are indicated by blue arrows. In addition, pp1ab included multiple peptidase 3 C cleavage sites (not shown). The coordinates of the genomic annotation are shown in Supplementary Table S2

The last third of the DToV genome contained genes coding for spike (S), membrane (M), haemagglutinin-esterase (HE), and nucleocapsid (N) proteins (Fig. 3; Supplementary Table S2). ORF1b and S were slightly overlapping, while short intergenic regions were present between the four 3’-terminal structural genes. The S protein showed the two typical furin cleavage sites found in toroviruses [58]. In addition, this protein lacked the cysteine-rich region between the transmembrane domain and the cytoplasmic tail at the C-terminal end, which is present in coronaviruses but not in toroviruses [50]. Upstream of the M, HE, and N genes, a short intergenic motif (5′-ACN₃₋₄CUUUAGA-3′), previously described as a trans-regulatory sequence (TRS), is always detected in toroviruses [55]. This motif showed a small sequence variation upstream of the N gene (5′-ACN₃-CUUUAGT-3′), which may have implications for the complex transcription mechanism of this group of viruses [58].

To compare genome-wide characteristics between DToV and other toroviruses, a pairwise alignment comparison analysis was performed against different toroviruses, including a representative of the four currently accepted species and other complete or near-complete genomes identified in different hosts. Specifically, sequences from antelope, goat, and Tasmanian devil (Torovirus sp.) toroviruses were used. From pairwise comparisons against DToV, nucleotide and amino acid identities were estimated for the six major ORFs, as well as the coverage of each comparison, defined as the percentage of gap-free positions in either of the two sequences (Table 1). For each ORF, identities were very similar across the different toroviruses, except Bangali virus, which had lower values for ORFs 1b, M and N. This finding was consistent with the associated coverage values, which were also lower for these three ORFs in Bangali virus. Overall, nucleotide and amino acid identities were considerably low, around or below 60% for ORFs 1a, HE and N, below 70% for S, and around 80% for 1b and M. These results confirmed the phylogenetic analyses above, showing that DToV was different from previously described torovirus sequences and that these differences strongly varied across the genome.

Table 1.

Nucleotide and amino acid (between brackets) identities (%) of six open reading frames between DToV and representative torovirus sequences (accessions are provided between brackets)

ORFs	BToV (ON376728)	PToV (MT684462)	EToV (MG996765)	Bangali virus (NC_076940)	Antelope ToV (MZ438674)	Goat ToV (NC_034976)	Torovirus sp. (MK521914)
1a	62.0 (53.5)	62.4 (53.6)	62.7 (54.9)	60.3 (52.6)	61.6 (53.9)	62.0 (54.0)	62.2 (54.2)
1a-cov	80.7 (94.3)	80.9 (95.4)	81.5 (94.8)	78.1 (93.2)	80.3 (94.2)	80.7 (94.2)	81.0 (95.1)
1b	75.9 (80.8)	76.0 (80.7)	77.2 (82.4)	71.0 (74.3)	76.0 (79.9)	76.1 (82.0)	76.1 (81.8)
1b-cov	96.6 (100)	95.9 (100)	96.6 (100)	91.4 (96.1)	96.0 (100)	96.4 (100)	96.4 (100)
S	66.9 (68.5)	65.9 (65.6)	68.2 (68.2)	67.0 (67.0)	66.3 (66.9)	66.9 (68.6)	66.0 (68.6)
S-cov	87.6 (98.9)	85.2 (98.7)	88.1 (98.3)	86.3 (98.6)	86.8 (99.1)	86.6 (98.9)	85.6 (98.9)
M	78.1 (86.3)	75.9 (84.6)	77.0 (83.8)	69.4 (68.2)	73.6 (84.2)	78.1 (85.0)	78.1 (86.3)
M-cov	97.5 (100)	94.2 (100)	97.2 (100)	89.1 (91.4)	93.1 (100)	96.9 (100)	98.3 (100)
HE	61.7 (50.7)	57.4 (50.2)	-	57.9 (47.4)	59.1 (46.4)	61.6 (51.3)	60.7 (51.8)
HE-cov	79.7 (95.1)	74.7 (93.4)	-	76.1 (91.3)	77.3 (93.3)	80.1 (97.4)	79.6 (97.4)
N	56.7 (54.9)	59.1 (56.9)	62.2 (55.5)	49.8 (44.3)	56.4 (56.7)	61.1 (55.8)	61.0 (55.8)
N-cov	73.8 (85.6)	76.1 (92.2)	78.65 (93.9)	64.5 (68.5)	73.6 (87.1)	79.9 (89.0)	78.6 (89.0)

For each ORF, the row below shows the corresponding coverage (cov, percentage of gap-free positions in either of the two sequences) for pairwise nucleotide and amino acid (in parentheses) comparisons. S, Spike glycoprotein; M, membrane protein; HE, haemagglutinin-esterase; N, nucleocapsid protein

Discussion

The diversity of toroviruses is probably much greater than currently considered by the ICTV. Only four species have been identified, all of them in ungulate hosts (BToV, EToV, PToV, and Bangali virus), although toroviruses appear to infect a wide range of animal species. This is supported by indirect evidence, such as the detection of antibodies against toroviruses in the sera of other animals, including goats, sheep, rabbits, and mice [7, 67]. Moreover, electron microscopy studies have detected torovirus-like particles in cat and human faeces [5, 47]. However, direct characterisation requires the identification of viral sequences, ideally complete genomes. The majority of torovirus complete genome sequences are of porcine and bovine origin, potentially as a result of sampling bias. As new sequences are added to the databases, it is noted that toroviruses are not exclusive to ungulates, since they have already been observed in other taxonomic groups, such as in a marsupial carnivore [13] and the one detected in this work in rodents. This will probably lead to the definition of new species and probably also new subgenera. For instance, Bangali virus has been recently included into a new subgenus (Bantovirus) within the genus Torovirus according to the Virus Taxonomy 2023 Release, while the other toroviruses have been assigned to the subgenus Renitovirus. In this context, therefore, it is likely that DToV should be considered not only a new species, but at least a new subgenus. In addition, the remarkable divergence between DToV and the other torovirus sequences makes it tempting to speculate on the possibility that rodents are the ancestral host of toroviruses.

Although coronaviruses and toroviruses are relatively close groups [58], the latter have been the subject of fewer studies, probably because they are not currently associated with human diseases. Nevertheless, BToVs cause economic losses due to calf diarrhoea [19, 68], while PToVs seem to cause less severe symptoms, although these may be worsened by co-infections [7, 52, 53]. In any case, new evidence supports the zoonotic potential of toroviruses. First, the detection of a torovirus in an insect suggests the possibility of vector-borne transmission [21]. Second, numerous cases of recombination have been observed, not only between different toroviruses but also with viruses of other genera, which can either enhance their pathogenicity or facilitate unexpected host adaptation [14, 29, 38, 54]. Third, wild ungulates have been identified as potential hosts of undiscovered zoonotic viruses [49]. Indeed, the detection of an antelope torovirus provides evidence of cross-species transmission between wild and domestic ungulates [17]. Finally, the identification of new toroviruses in wild animals, especially rodents, further underscores their cross-species transmissibility and suggests some degree of zoonotic potential, as the proximity of rodents to humans and farm animals may facilitate host-jumping events [21]. It must be acknowledged that the above is a body of indirect evidence pointing to some zoonotic potential, although the genetic divergence between the different toroviruses could be primarily explained by virus-host coevolution. Coevolution analyses are particularly useful to quantify the magnitude of possible cross-species transmission events. In papillomaviruses, for example, their identification in multiple mammalian species has enabled to study the extent of co-speciation in their evolution [9, 33]. In toroviruses, despite their apparently wide host range in mammals, viral sequences have only been identified in a relatively small number of species. This is a major limitation for coevolutionary analyses and, therefore, it will be necessary to characterize new torovirus sequences in different mammalian hosts to gain further insight into their evolutionary dynamics.

Apart from their apparently low pathogenicity, little is known about toroviruses because they have not been cultured, with a few exceptions. In the mid-1980s, EToV (Berne virus) was the first torovirus to be propagated in cell cultures [65]. Since then, no other EToVs have been isolated, suggesting that Berne virus may be a mutant adapted to growth in cultures [34]. More recently, several BToVs have been propagated in cell cultures [2, 29, 36], which showed, for example, that HE protein is not essential for viral growth under these conditions. BToV is also the only torovirus for which a reverse genetics system is available [59], which is likely to become an essential tool for studying important viral aspects such as pathogenesis and vaccine development. In this scenario, studying the S protein, which determines cellular and tissue tropism and host range [57], is crucial to characterise the zoonotic potential of toroviruses. The use of viral pseudotypes has recently allowed us to characterise the spike protein of dozens of viruses from multiple RNA virus families [20]. Unfortunately, our attempt to obtain a VSV pseudotype to study the cell tropism of a torovirus was unsuccessful. In the future, other approaches, such as those based on obtaining recombinant viruses [42], may help to shed light on the zoonotic potential of toroviruses.

Conclusions

To date, most of the known variants of torovirus have been identified in ungulates [17]. However, the complete genome reported here was recovered from a faecal sample of a Spanish dormouse, suggesting that the host range of toroviruses is broader than currently described. The basal phylogenetic position of the newly described torovirus led us to speculate that rodents may serve as a reservoir for this viral genus. Although the zoonotic potential of toroviruses may be low, some of their characteristics, such as their tendency to recombine and possible vector-borne transmission, point to the need to include them in surveillance programs. Further efforts to understand the infectivity determinants of this group of viruses will be necessary to better characterize their potential for host-jumping events.

Author contributions

JB obtained viral contigs, analyzed data and drafted the manuscript. AV obtained rodent feces samples. RMR prepared samples for sequencing. JSM supervised work and obtained rodent feces samples. RS obtained funding, supervised work, and co-wrote the manuscript. JMC obtained funding, supervised work, and co-wrote the manuscript. All authors read and approved the final manuscript.

Funding

This work was supported by the Spanish Ministerio de Ciencia e Innovación (MICINN) [PID2020-118602RB-I00] to R.S and J.-M.C.; the Conselleria de Educación, Universidades y Empleo (Generalitat Valenciana) [CIAICO/2022/110] to R.S., and ERC Advanced Grant [101019724-EVADER] to R.S.

Data availability

All data generated or analysed during this study are included in this published article [and its supplementary information files].

Declarations

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.