Zebrafish from the pet trade harbor known and novel microbes that are potentially pathogenic

We sampled outwardly healthy zebrafish from 3 different pet trade sources and a research laboratory housing system to survey microbial populations across environments through sequencing of bulk intestinal RNA. After excluding reads that mapped to the Danio rerio genome, we performed de novo assembly of transcripts and assigned prospective taxonomic classification based on nucleotide or amino acid similarity to known sequences (S1 and S2 Files).

We classified hundreds of zebrafish-associated transcripts as viral and thousands of transcripts as prokaryotic or eukaryotic (S2 File). Many prospective virus sequences were derived from bacteriophages, while others had similarity to viruses that infect diverse eukaryotic hosts. Among the virus families that infect vertebrate hosts, we reconstructed 2 transcripts that constitute a full genome sequence (GenBank accession numbers OR427289 and OR427290) for a novel bisegmented double-stranded RNA virus with similarity to birnaviruses that infect fish (Fig 1A). Phylogenetic analyses confirmed that the birnavirus we discovered, which we named RMBV, is part of a clade of viruses that infect fish and other vertebrates within the family Birnaviridae (Fig 1B). This clade includes infectious pancreatic necrosis virus (IPNV), which can cause severe disease and losses in aquaculture. The closest relative to RMBV in this phylogeny, Wenling jack mackerels birnavirus, shares 62% amino acid identity in the RdRp protein. Thus, RMBV is a newly described virus that appears to naturally infect zebrafish and is related to significant pathogens in aquaculture.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Metatranscriptomic survey of microbes associated with zebrafish from laboratory and pet trade sources.

. (A) Schematic representation of a newly discovered birnavirus associated with zebrafish in this study. Both segments of the linear dsRNA genome are depicted with boxes denoting relative positions of ORFs based on protein domain conservation. Numbers beneath segments indicate nucleotide lengths of each prospective 5′ UTR, coding sequence, and 3′ UTR. (B) Maximum likelihood phylogeny of viruses in the family Birnaviridae. The black dot highlights the novel zebrafish-associated virus schematized in (A), which we name RMBV. (C) Maximum likelihood phylogeny of RdRp protein sequences from viruses in the order Picornavirales. Black dots highlight the zebrafish-associated viruses that were detected (ZfPV) or newly discovered in this study. Trees in (A and C) are midpoint rooted. Scales are in amino acid substitutions per site. Branches are colored based on the known or inferred host group. (D) Taxonomic classification and quantification of host and microbe transcripts identified in intestinal tissues of zebrafish from laboratory and pet trade sources. Columns show estimated abundances in tissue samples from independent individuals grouped by source. RdRp, RNA-dependent RNA polymerase; RMBV, Rocky Mountain birnavirus; ZfPV, zebrafish picornavirus.

https://doi.org/10.1371/journal.pbio.3002606.g001

Among the other viruses with known or possible vertebrate hosts, we detected an endemic zebrafish picornavirus (ZfPV) that we and others recently discovered [14,15] and dozens of other sequences with similarity to members of the order Picornavirales. Eight of the novel sequences contained full-length RNA-dependent RNA polymerase (RdRp) sequences that we used along with representative RdRp sequences across the order Picornavirales to perform phylogenetic analyses. A maximum likelihood phylogeny placed all novel picornavirus-like sequences in a clade with viruses isolated from non-vertebrate hosts (Fig 1C), which suggests that they infect eukaryotes associated with zebrafish as opposed to zebrafish themselves.

We next estimated host and microbe abundances by mapping all reads to an integrated set of zebrafish and microbial transcripts (Fig 1D and S3 File). ZfPV was broadly detected in zebrafish from both laboratory and pet trade sources, extending its known presence beyond the laboratory to other environments [14,15]. In contrast, RMBV was only detected in zebrafish from a single pet trade source. Zebrafish from this source also expressed significantly higher levels of interferon-stimulated genes than control zebrafish from the laboratory, while zebrafish from the other pet trade sources did not exhibit signatures of antiviral gene expression (S1 Fig and S4 File). These comparisons reveal a canonical antiviral defense response associated with the presence of RMBV in the absence of visible disease phenotypes in pet trade animals.

Most of the other prospective eukaryote-infecting viruses we discovered were only detected in pet trade sources. The closest match to a novel virus in phylum Negarnaviricota discovered here is Amsterdam virus (61% amino acid identity), which is associated with parasitic nematodes that infect rodents [16]. Together with a pronounced abundance of nematode transcripts in the pet trade zebrafish, these observations suggest that the newly described negative-sense RNA virus infected nematodes that infected zebrafish. Parasitic worm transcripts were detected in animals from both environments but were generally more abundant in zebrafish from the pet trade, while microsporidia transcripts were abundant in both laboratory and pet trade zebrafish. Clustering samples based on the abundance of bacteria families in our data did not distinguish between their sources (S5 File), highlighting that similar bacterial communities were associated with zebrafish in laboratory and pet trade environments. Thus, we did not observe identifiable bacterial infections specific to pet trade zebrafish. The results from this metatranscriptomic survey underscore the use of zebrafish from the pet trade as potential sources of viruses and eukaryotic parasites that are not commonly found in laboratory zebrafish.

Transmission of RMBV from pet trade to laboratory zebrafish is associated with necrotic disease

To determine if microbes found in pet trade zebrafish could be naturally transmitted to laboratory zebrafish, we devised a cohousing experiment using closed circulation aquaria containing zebrafish from both environments (Fig 2A). A control tank was seeded with only zebrafish from the laboratory, and a cohousing tank was seeded with zebrafish from the laboratory along with zebrafish from pet trade source 3 (Fig 1D). All zebrafish in the control tank remained healthy throughout the experiment (Fig 2B). In contrast, we observed hemorrhaging in 3 laboratory zebrafish approximately 1 month after cohousing them with pet trade zebrafish (Fig 2C). At this point, we collected intestines, kidneys, and spleens from control and cohoused laboratory zebrafish for RNA isolation and sequencing. The metatranscriptomes of these samples revealed pronounced levels of RMBV in the 3 diseased laboratory zebrafish that were cohoused with pet trade zebrafish (Fig 2D). We detected RMBV in all tissues that were sampled, revealing that infections were systemic. No other viruses that were unique to pet trade zebrafish in our survey above (Fig 1D) were detected in cohoused laboratory zebrafish. RNA was not collected from the 7 cohoused zebrafish that did not exhibit disease, and the infection status of these animals is unknown. These experiments demonstrate that viruses circulating in seemingly healthy zebrafish from the pet trade can be naturally transmitted to laboratory zebrafish and cause disease.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Overt disease observed in laboratory zebrafish cohoused with zebrafish from the pet trade is associated with RMBV infection.

(A) Diagram of cohousing experiment. Tanks were seeded with adult zebrafish from the laboratory (control) or zebrafish from the laboratory and the pet trade (cohoused). (B) A representative individual from the control tank. (C) A representative laboratory zebrafish 1 month after cohousing with zebrafish from the pet trade. Red arrows highlight necrotic lesions. (D) Transmission of microbes from pet trade to cohoused laboratory zebrafish inferred from host and microbe transcript abundances. Three laboratory fish were sampled from each tank. Samples are grouped by tank and tissue. Columns show estimated transcript abundances in tissue samples from individual laboratory zebrafish. RMBV, Rocky Mountain birnavirus.

https://doi.org/10.1371/journal.pbio.3002606.g002

In a second cohousing experiment, another tank was seeded with zebrafish from the laboratory along with additional zebrafish from pet trade source 3. Zebrafish were cohoused for the same length of time as in the experiment described above but no hemorrhaging was observed in laboratory or pet trade zebrafish. We collected intestines from pet trade and laboratory zebrafish after 1 month of cohousing for RNA isolation and sequencing, which revealed RMBV transcripts in one of 2 pet trade zebrafish but no evidence of RMBV transcripts in either of the 2 cohoused laboratory zebrafish (S2 Fig and S6 File). These results demonstrate that in some contexts, zebrafish can carry RMBV without transmitting infection to potential recipients in a shared environment. Although we detected no evidence for the transmission of RMBV in this experiment, we observed exceptionally high levels of parasitic nematode transcripts in all cohoused zebrafish. Furthermore, we detected transcripts from the novel virus in phylum Negarnaviricota described above in cohoused pet trade and laboratory zebrafish, lending additional support to the proposition that nematodes serve as hosts. These experiments highlight examples of virus and parasite transmission from pet trade to cohoused laboratory zebrafish and reveal variability in the frequency of transmission.

Naturally transmitted RMBV infections elicit inflammatory immune responses in laboratory zebrafish

To identify host transcriptional responses to RMBV, we tested for differential gene expression in tissues from infected laboratory zebrafish compared to control laboratory zebrafish and identified more than 800 genes at an adjusted p < 0.05 that were induced by at least 2-fold (S7 File). RMBV infection was associated with elevated levels of genes related to interferon signaling, inflammation, innate immunity, and adaptive immunity (Fig 3A). We identified an overlapping set of differentially expressed genes in all 3 organs, but the spleen had more virus-induced genes than infected intestine or kidney tissues (Fig 3B and S7 File). We next compared the RMBV-induced genes to those induced in another virus infection context. Spring viremia of carp virus (SVCV) is an RNA virus naturally found in a range of fish species that can also cause hemorrhagic disease when experimentally administered to zebrafish [17]. A recent study measured gene expression in zebrafish intestine, kidney, and spleen tissues 12 h after intraperitoneal injection with SVCV [18]. Despite several differences in experimental context, more than a third of the genes that were highly induced by RMBV were also induced during infection with SVCV (Fig 3C). We observed enrichment for antiviral defense processes among both virus-induced gene sets, enrichment for B cell activation among RMBV-induced genes, and enrichment for lymphocyte chemotaxis and other processes among SVCV-induced genes (Fig 3D). We speculate that the divergent enrichment for lymphocyte activation and chemotaxis between RMBV- and SVCV-induced gene sets may reflect differences in the stage of infection progression, but defining the full extent of overlap between the gene sets induced by either virus will require further experimentation. These observations highlight genes associated with inflammatory disease observed during natural transmission of birnavirus infections in zebrafish.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Naturally transmitted RMBV infections induce systemic inflammatory antiviral responses in laboratory zebrafish.

(A) Differential gene expression in laboratory zebrafish infected by RMBV compared to uninfected laboratory zebrafish. Average log2 fold change expression is plotted for each gene in a comparison of 3 RMBV-infected and 3 uninfected samples for each tissue. All comparisons had adjusted p-values of <0.05 unless denoted with a black circle. (B) Comparison of RMBV-induced gene sets by tissue. Euler plot includes all genes with log2 fold changes >1 and adjusted p-values <0.05. (C) Comparison of virus-induced gene sets in intestine, spleen, or kidney during RMBV or SVCV infections. Euler plot includes all genes with log2 fold changes >2 and adjusted p-values <0.05. SVCV RNAseq data are from (PMID: 34908463). (D) Enrichment of Gene Ontology biological processes among the genes that are induced with log2 fold changes >2 during RMBV or SVCV infections. All biological processes with false discovery rates of <0.05 for at least 1 condition are plotted. RMBV, Rocky Mountain birnavirus; SVCV, spring viremia of carp virus.

https://doi.org/10.1371/journal.pbio.3002606.g003

Pet trade survey and cohousing assay

This study was conducted under the approval of the Office of Institutional Animal Care and Use Committee (IACUC # 00001439) of the University of Utah’s animal care and use program. Nine zebrafish from 3 local pet trade sources in the Salt Lake Valley were purchased. Fish were euthanized and intestinal tissue harvested and processed individually for RNA extraction. Cohousing assay was set up with closed circulation aquaria using 3 identical 10-gallon tanks. Tanks were inoculated with ceramic filter rings (Aquapapa), these filters had been housed in a zebrafish research facility for 1 month previously to build up a nitrifying microbial community. Identical filtration systems were set up in each tank with activated charcoal and sponge filters (AquaClear 50). Tanks were cycled for 6 weeks using pure ammonia (Great Value). The cohousing experiments were initiated when ammonia and nitrite levels were undetectable. Each tank housed 15 fish. Fish were fed identical diets (Zeigler, Adult Zebrafish Diet). Control tanks contained 15 laboratory-reared Tübingen zebrafish. The laboratory samples presented in Fig 1 were collected from the control tank of zebrafish. Each cohousing tank contained 10 laboratory-reared Tübingen fish and 5 pet trade fish. Pet trade animals were obtained from the same source for both cohousing experiments and their genetic backgrounds were unknown. In the first experiment, the pet trade fish were all females and the laboratory fish were all males (Fig 2), and in the second experiment, the pet trade fish were all males and the laboratory fish were all females (S2 Fig). No fighting or evidence of aggression was observed throughout the cohousing experiments. None of the control or cohoused fish exhibited abnormal swimming behaviors throughout the experiments. Approximately 1 month into the first cohousing experiment several laboratory-reared fish began exhibiting hemorrhaging. These were euthanized and processed for RNA extraction. All additional fish used in the study were subsequently euthanized and processed within 1 week.

RNA extraction and sequencing

Fish were euthanized and intestines, whole kidney marrows, and spleens were harvested for processing. All tissues were homogenized using mechanical lysis. RNA was extracted using the directzol kit (Zymo), and 1 ml of TRIzol was used per intestine and 0.5 ml of TRIzol used for individual spleens and kidneys. Following RNA extraction, samples were DNAse treated (Zymo). RNA ScreenTape (Agilent) was used to assess quality of RNA samples. Only samples with RIN scores >8 were used for analysis. RNA libraries were prepared by the High Throughput Genomics Shared Resource at the University of Utah with the Illumina TruSeq Stranded Total RNA Library Prep Ribo-Zero Gold and sequenced on a Novaseq with using a 150 × 150 bp sequencing kit to a depth of 25 million reads per sample. Raw sequencing reads were deposited at NCBI under BioProject PRJNA1005695.

Metatranscriptomic classification and quantification of RNA-sequencing data

Reads were mapped to the GRCz11 Danio rerio genome assembly (Ensembl release 100) with STAR v2.7.3a [40]. Unmapped reads from all samples were concatenated and assembled de novo using Trinity v2.11.0 [41]. Prospective taxonomic lineages were assigned to each assembled Trinity “gene” by extracting the best hit (highest bit score) per query from BLASTn [42] or DIAMOND v2.0.14 [43] searches of the nt or nr NCBI sequence collections (downloaded in May 2022), respectively. All contigs prospectively assigned to microbial lineages were then aligned against all available zebrafish nucleotide sequences with BLASTn. Any contig that aligned a zebrafish sequence with a bitscore of 100 or greater was removed from the collection of contigs. Estimated zebrafish and microbe transcript counts were obtained by mapping reads from each sample to an index of zebrafish transcripts and all newly assembled sequences using Salmon v1.3.0 [44] with the validateMappings flag and 20 Gibbs samples. Counts were summed at the gene or taxonomic family level, and normalized counts were generated using DESeq2 v.28.1 [45].

Phylogenetic analyses

Phylogenies were estimated for viruses in the order Picornavirales using amino acid sequences from prospective 3C protease and 3D RNA-dependent RNA polymerase genes. Sequences from at least 1 species of all known genera in family Picornaviridae were included, as well as all available sequences from picornaviruses that infect fish. Several representative viruses from other clades were also included to span the currently known breadth of Picornavirales. Sequences were aligned with MAFFT version 7 [46] and manually trimmed to remove sequences at ends with mostly empty columns, resulting in an alignment of 753 amino acid sites across 95 taxa. Evolutionary relationships were inferred using maximum likelihood phylogenetic analyses carried out in RAxML-NG version 1.0.2 [47]. The LG+I+G+F substitution model was selected as the best-fit model for the alignment under the Akaike information criterion as determined by ProtTest version 3.4.2 [48]. Bootstrapping converged after 650 replicates. Birnavirus phylogenies were estimated using amino acid sequences from VP1 (RdRp) genes. Representative viruses were included from vertebrate and invertebrate hosts. Sequences were aligned using MAFFT-L-INS-i with DASH structures, resulting in an alignment of 1,093 amino acid sites across 12 taxa. Evolutionary relationships were inferred using maximum likelihood phylogenetic analyses carried out in RAxML-NG [47]. The LG+G+I+F substitution model was selected as the best-fit model for the alignment under the Akaike information criterion as determined by ProtTest [48]. Bootstrapping converged after 300 replicates.

Comparisons to Spring viremia of carp virus

Raw fastq reads in BioProject PRJNA690234 [18] were downloaded from the Sequence Read Archive at NCBI with the SRA Toolkit. Estimated zebrafish and microbe transcript counts were obtained by mapping reads from each sample to the same index of zebrafish transcripts and all newly assembled sequences using Salmon v1.3.0 [44] as described above. Counts were summed at the gene or taxonomic family level, and normalized counts were generated using DESeq2 v.28.1 [45]. All genes with log2-fold changes greater than 2 and adjusted p-values less than 0.05 were included in comparisons of birnavirus- and SVCV-induced gene sets. Euler plots were generated in R with the eulerr package [49]. Gene Ontology enrichment was determined with the PANTHER overrepresentation test (released February 2022) [50].