An expanded clade of rodent Trim5 genes

Abstract

Trim5alpha from primates (including humans), cows, and rabbits has been shown to be an active antiviral host gene that acts against a range of retroviruses. Although this suggests that Trim5alpha may be a common antiviral restriction factor among mammals, the status of Trim5 genes in rodents has been unclear. Using genomic and phylogenetic analyses, we describe an expanded paralogous cluster of at least eight Trim5-like genes in mice (including the previously described Trim12 and Trim30 genes), and three Trim5-like genes in rats. Our characterization of the rodent Trim5 locus, and comparison to the Trim5 locus in humans, cows, and rabbits, indicates that Trim5 has undergone independent evolutionary expansions within species. Evolutionary analysis shows that rodent Trim5 genes have evolved under positive selection, suggesting evolutionary conflicts consistent with important antiviral function. Sampling six rodent Trim5 genes failed to reveal antiviral activities against a set of eight retroviral challenges, although we predict that such activities exist.

Figure 1. A paralogous expansion of rodent Trim5 genes

(A) The mouse and rat Trim5 loci are compared to the human locus. All three loci sit in an olfactory receptor gene expansion, here shown with only two of the genes on each side (white arrows). Distances between genes are drawn roughly to scale, with numbers on the top (for mouse) and bottom (for rat) indicating distance between genes, based on BAC sequences. The Trim5 homologs are shown in green arrows, with dark green color for a human Trim5-like pseudogene (P1). The asterisk (*) denotes partial Trim5-like sequences found in the mouse genome. Trim34 homologs are shown in orange, Trim6 homologs in blue, and Trim22 is shown in magenta. Mouse Trim12-X is not included because it is on the X chromosome. 30-L1 and 30-L2 refer to rat Trim30-like1 and Trim30-like2, respectively. (B) The DNA sequence for RBCC domains were used to construct a maximum likelihood tree with 100 replicates for estimating bootstrap support. Mouse Trim12-X and Trim30-4 are not included because they do not contain the RBCC domains. Color profiles match those used in (A). Scale for branch length is shown below the tree. (C) Phylogeny of the species used in (B), based on reference (Hedges, 2002).

Figure 2. Domain organization and expression of rodent Trim5 loci genes

Domains of rodent Trim5, Trim6 and Trim34 genes are compared to those of human Trim5. Five of the mouse and three of the rat Trim5 paralogs contain all four of the canonical domains of Trim5α. Previously existing (if available) RefSeq or Genbank accession numbers are shown below each gene name and are gathered from the NCBI and UCSC genome browser online servers. The expression files for mouse Trim5 genes containing the four canonical domains of Trim5α are shown for NIH3T3 cells. Expressed indicates that we were able to amplify cDNA from NIH3T3 RNA extracts, using reversetranscriptase-PCR. Not expressed indicates that we were not able to amplify cDNA from NIH3T3 RNA extracts, even though were designed primers according to predicted sequences. Not applicable indicates that the NIH3T3 cell line was not suitable to amplify that gene.

Figure 3. Recombination among rodent Trim5 genes

(A) The online GARD tool identifies a breakpoint upstream of the PRYSPRY (exon8) domain for the full-length canonical Trim5α sequences from mice, rats and rabbits. The second-order Akaike Information Criterion (c-AIC) value is shown on the y axis as a measure to show the goodness of fit of this location of the breakpoint compared to other locations identified by GARD (not shown). The model average support value (numbers on the y axis) by itself has no meaning; they are used by GARD to identify the best single breakpoint among a series of possible breakpoints. Nucleotide position is indicated on the x axis. DNA sequences of the (B) RBCC domains (excluding exon8), and of the (C) PRYSPRY (exon8) domain for rodent Trim5 homologs were used to construct a maximum likelihood tree. Rabbit Trim5α was used as an outgroup. Sequences that showed evidence of recombination are shown in bold text. Mmusc stands for Mus musculus. Mmusc Trim12 was excluded because it lacks a PRYSPRY domain (exon8). Scale for branch length is shown below the tree.

Figure 4. Rodent Trim5 homologs are rapidly evolving

Maximum likelihood trees for DNA sequences of (A) RBCC domains (excluding exon8) and (B) PRYSPRY domain (exon8) are shown for rodent Trim5 homologs, with bootstrap values at each node out of 100 replicates. Rabbit, rhesus and human Trim5α are used as out-groups. Branch dN/dS values calculated by PAML are shown above each branch. The number of non-synonymous changes (N) and synonymous changes (S) are shown below each branch in parentheses as (N,S). Branch dN/dS values significantly greater than 1 are highlighted as a bold line. Lineages with the highest dN/dS on each tree are highlighted in bold text (Mmusc Trim12 for RBCC, Mdunn Trim30 for PRYSPRY). MMusc stands for Mus musculus, and Musdunn for Mus dunni. Mmusc Trim12 was excluded from the PRYSPRY tree (B) because it lacks a PRYSPRY domain.

Figure 5. Codons of mouse Trim30 orthologs under positive selection are locate in the variable loops of the PRYSPRY domain

(A) Phylogeny of Mus species (Tucker 2006) used in our study are shown for Trim30 orthologs, with an out-group to rat Trim30 paralogs. Branch dN/dS values calculated by PAML are shown above each branch. The number of non-synonymous changes (N) and synonymous changes (S) are shown below each branch in parentheses as (N,S). Branch dN/dS values significantly greater than 1 are highlighted as a bold line. The black circle indicates the node for the ancestral Trim30 sequence for Mus dunni. (B) Protein sequence alignment of the PRYSPRY (exon8) domain for Mus dunni and its ancestral Trim30 is shown. Differences are highlighted with a gray box. The position of the beta-sheets and variable loops are based on structural work from reference (James et al., 2007). Codons under positive selection identified by PAML based on Bayes-Empirical-Bayes analysis (with posterior probability greater than 95%) are shown in *. White arrows and gray rectangles indicate beta-sheets and variable loops in the tertiary structure, respectively.

Figure 6. Retroviruses used in our assays were not restricted by mouse Trim5 homologs

CRFK cell lines were transduced to stably express Trim5 homologs containing all four of the canonical domains of Trim5α, cloned from NIH3T3 and MDTF cells, representing Mus musculus and Mus dunni respectively. Stable cell lines for Trim34-1 from NIH3T3 cells and Rhesus Trim5α were also generated. Each Trim gene was cloned with an N-terminal HA tag for detection via immunoblotting. (A and B) 50 ug of lysates per well, immunoblotted with anti-HA to check for expression of each Trim5 homolog. Anti-actin was used as a loading control. A second attempt at stable cell line generation was needed to boost the levels of MDTF Trim30 and Trim30-3 (B). We were unable to generate CRFK cells stably expressing Trim30-2 from either NIH3T3 or MDTF cells. (A and B) Each stable cell line expressing one of the listed Trim genes was challenged with the list of retroviruses at sufficient amounts to give non-saturating levels of infection. Each virus is generated to express GFP upon infection. These retroviruses were used because they represent a broad range of viruses restricted by antiviral Trim5 orthologs from primates, cows and rabbits. HIV-1 (human immunodeficiency virus-1), SIVmac (simian immunodeficiency virus macaque strain), MPMV (mason-pfizer monkey virus), N-MLV (N-tropic murine leukemia virus), B-MLV (B-tropic murine leukemia virus), NB-MLV (N- and B-tropic murine leukemia virus), FIV (feline immunodeficiency virus), EIAV (equine infectious anemia virus). Percentage of cells that are infected are plotted on the y-axis as a log-scale.