Comparative evolutionary genomics of androgen-binding protein genes

Abstract

Allelic variation within the mouse androgen-binding protein (ABP) alpha subunit gene (Abpa) has been suggested to promote assortative mating and thus prezygotic isolation. This is consistent with the elevated evolutionary rates observed for the Abpa gene, and the Abpb and Abpg genes whose products (ABPbeta and ABPgamma) form heterodimers with ABPalpha. We have investigated the mouse sequence that contains the three Abpa/b/g genes, and orthologous regions in rat, human, and chimpanzee genomes. Our studies reveal extensive "remodeling" of this region: Duplication rates of Abpa-like and Abpbg-like genes in mouse are >2 orders of magnitude higher than the average rate for all mouse genes; synonymous nucleotide substitution rates are twofold higher; and the Abpabg genomic region has expanded nearly threefold since divergence of the rodents. During this time, one in six amino acid sites in ABPbetagamma-like proteins appear to have been subject to positive selection; these may constitute a site of interaction with receptors or ligands. Greater adaptive variation among Abpbg-like sequences than among Abpa-like sequences suggests that assortative mating preferences are more influenced by variation in Abpbg-like genes. We propose a role for ABPalpha/beta/gamma proteins as pheromones, or in modulating odorant detection. This would account for the extraordinary adaptive evolution of these genes, and surrounding genomic regions, in murid rodents.

Figure 1

A graphical representation of the relative position and transcriptional orientation of the Abpabg-like genes and pseudogenes located on Rattus norvegicus Chromosome 1, Mus musculus Chromosome 7, and Homo sapiens Chromosome 19. Coordinates are taken from genome releases rn3 (Baylor RGSC v3.1), mm3 (NCBI build 30), and hg16 (NCBI build 34), respectively. The 5′-to-3′ orientations of the genes are shown by the direction of the arrowheads. Scn1b and Uble1b genes, which lie in orthologous genomic regions in all three species, are numbered 1 and 2, respectively. Abpa-like genes are shown in blue, Abpbg-like genes in red, and primate SCGB4A1–4(P) genes in green. Filled arrowheads represent predicted functional genes whereas open arrowheads denote predicted pseudogenes. The sequence of duplication events among rodent genes inferred from phylogenetic trees (see text) is implied from the dendrogram, shown in black. Blue and red lines in this dendrogram represent branches that are not supported by the phylogenetic trees shown in Figure 3 (see text). The scale bar shows approximate genomic distance in megabases. Gaps (>5 kb) in the genomic assembly of each species are shown as black boxes.

Figure 2

Amino acid sequence multiple alignment of rodent and primate ABPα and ABPβγ homologs, and primate secretoglobin homologs. Conceptual translations of both genes and pseudogenes (denoted by Ψ) are shown, with stop codons replaced by “X” symbols in white on black; codons containing a frame shift are also shown in white on black. Pseudo-exon sequences that could not be identified in genomic sequence are replaced by “.” characters; “-” represents a gap position. Multiple sequence alignments were produced by CLUSTAL W, manually adjusted, and colored using Chroma (Goodstadt and Ponting 2001) and an 80% consensus; gap positions were ignored in the calculation of a consensus sequence. Exons are shown as horizontal bars, and intron and exon boundaries were determined from alignments of the gene predictions and genomic DNA sequence using the UCSC genome browser. Protein secondary structure predicted by PHDSec (Rost and Sander 1993; PHD) shows four α-helices (H) in α-like chains, as has been demonstrated for cat Fel dI, rabbit uteroglobin, and human uteroglobin, and five α-helices in βγ-like chains. Codons predicted to be subject to positive selection by Codeml with a posterior probability of p > 0.9 in one model, and of at least p > 0.5 in one other model, are termed ω⁺ sites and are marked with an asterisk above the alignment. Positions of rabbit uteroglobin (UTG) residues predicted to interact with a bound ligand are each indicated by “X” above the aligned rabbit UTG sequence. (Mm) Mus musculus; (Hs) Homo sapiens; (Rn) Rattus norvegicus; (Pt) Pan troglodytes; (As) Apodemus sylvaticus; (Af) Apodemus flavicollis; (Aa) Apodemus agarius; and (Oc) Oryctolagus cuniculus.

Figure 3

(A) Schematic representation of gene structures and repeat elements between the Abpa11 (Abpa) and Abpbg11 (Abpb) gene pair. The position and size of genes and repeat elements are shown to scale. Coordinates were obtained from the genome browser at UCSC (Kent 2002). The upper portion of the figure shows the genomic architecture of Abpa11 and Abpbg11 genes in greater detail. Highlighted regulatory elements correspond to those described previously (Laukaitis et al. 2003). The dashed lines represent the upstream regions used to generate the 5′ phylogenetic trees shown in B. (B) 5′ trees: phylogenetic relationships of rodent Abpa-like and Abpbg-like genes. Repeat-masked genomic DNA sequences 5′ upstream of Abpa-like genes and Abpbg-like genes were aligned (see Methods). For Abpa-like and Abpbg-like genes, 300 bp and 1 kb, respectively, 5′ to the translational start site were used for generation of the trees. Trees were generated using the neighbor joining method. The lineages containing the proposed roots of the trees are shown by black dots. Bootstrap values >50% are shown.

Figure 5

Site-specific K_A/K_S analysis of Abpa-like and Abpbg-like genes mapped to the tertiary structure of cat Fel dI. ω⁺ codons that are predicted to be under positive selection for mouse paralogs are mapped to a ribbon representation of the tertiary structure of feline major allergen Fel dI (PDB 1POU; Kaiser et al. 2003). (A) Ribbon diagram of the Fel dI dimer. Side chains of mapped predicted ω⁺ sites are highlighted in red. A single ligand molecule (2-methyl-2,4-pentanediol) present in the crystal structure is colored orange. (B) A representation of the molecular surface of the Fel dI chain 1 and 2 heterodimer. Mapped ABPα ω⁺ sites are shown in blue, and ABPβγ ω⁺ sites are shown in red. (C) Mapped ω⁺ sites predicted for the Abpa-like mouse paralogs. (D) Mapped ω⁺ sites predicted for the Abpbg-like mouse paralogs. (E) Prediction of the ABPβγ terminal extension and the mapped location of identified ω⁺ sites. The ABPβγ teminal extension is predicted to form a helical structure of ∼24 residues and is shown to scale. Swiss-PDBviewer (http://www.expasy.org/spdbv/; Guex et al. 1999) was used for all structural manipulations, and POVRAY (http://www.povray.org) was used to generate images.

Figure 4

Multiple nucleotide sequence alignment of mouse and rat Abpbg-like exons 3 and surrounding genomic DNA. Genomic DNA corresponding to exon 3 (98 positions) and 100 nucleotide positions of both flanking intronic and 3′-UTR sequence was aligned with HMMER, and manually adjusted. We found that 81.3%, 50.5%, and 92.6% of the sites in the intron, exon, and 3′-UTR, respectively, exhibited ≥70% consensus. In these calculations, positions with fewer than 50% gaps were considered. The 14 codons of exon 3 corresponding to predicted ω⁺ sites are shown by horizontal bars.