Genome sequences of primates and rodents 123 now allow genome-wide investigations of evolutionary rates and selective processes. The aforementioned studies, and earlier ones, show that mammalian protein-coding genes vary dramatically in their rates of sequence divergence 13456 and that amino acid sequence divergence is greatest for secreted proteins whose genes are expressed in few tissues and least for intracellular proteins found widely in all tissues 7. These generalizations hold for the great majority (~90%; 1) of genes that are present in single copies in both primate and rodent genomes. The remaining genes have experienced at least one duplication in either one or both of the primate and rodent lineages.

The process of gene duplication provides material for functional diversification. A newly-duplicated gene may subsequently acquire innovative function ("neofunctionalization") or may retain some of its progenitor gene's functional repertoire ("subfunctionalization"), or a combination of both 891011. Such duplicated genes are also rapidly diverging in sequence 13 and they are substantially over-represented in functions relating to chemosensation, reproduction, host defense and immunity, and toxin metabolism. Innovation within these functional categories therefore can occur by sequence divergence and/or gene duplication. Adaptive evolution however can also act by modifying gene expression profiles, for example by restricting expression to few tissues. Among genes conserved as single orthologous copies among primate or mouse species, expression variation appears to evolve neutrally and approximately linearly with time 1213.

We were interested in a related question: How rapidly can expression profiles diverge among duplicated genes? For example, Gu 14 showed an increase in expression diversity in genes involved in Drosophila development. We wish to ask about diversification during other evolutionary process, to wit rapidly diverging duplicated genes involved in the four functions listed above. Because expression patterns appear to diverge substantially over long time periods, such as since the primate-rodent (Euarchontoglires) common ancestor 1516, it is necessary to address this question using many related genes that arose only in very recent times. The most closely-related vertebrate genomes available currently are those of mouse and rat; these lineages diverged approximately 12–24 million years ago (Mya) 1718. Thus, by comparing these two genomes we may be able to identify a gene family that is extensively expanded in one, but not necessarily in the other genome, with which to investigate recent expression divergence.

Indeed, we recently identified two families of paralogous genes, clustered together in a 1–2 Mb region of Mus musculus chromosome 7 that each meet this criterion. We predict that the Euarchontoglires common ancestor, and the rat and mouse common ancestor, each contained a single member of both gene families; these family members were proximally positioned in a tightly linked head-to-head (bidirectional) gene pair 1920. In humans and chimpanzees these genes have accumulated disruptive mutations and are thus likely to be nonfunctional pseudogenes. In the rat lineage, two gene pair duplications have given rise to three gene versions in each family. In the mouse lineage, an extraordinary and rapid burst of duplications has generated over a dozen members of each family, some of which harbor inactivating mutations, whereas others are full-length and apparently functional. Because these mouse genes appear to have all arisen via duplication events since the mouse-rat divergence (12–24 Mya), these are the genes we have chosen to investigate to address the issue of expression divergence.

The two families are termed Abpa and Abpb/Abpg genes (encoding ABPα and ABPβ/ABPγ proteins). They are named according to their proteins that were first described in the literature 2122. Androgen-binding protein α subunit (ABPα) forms covalently linked heterodimers with either β or γ subunits, and these are secreted into the saliva following expression in the submaxillary gland. ABPα, ABPβ and ABPγ proteins are secretoglobins, a family of secreted proteins 23 that bind lipophilic ligands (for a review, see 24) and are present in mammals and birds 2526, but whose roles in cellular and physiological function all remain obscure 2728.

Much of the previous work on salivary ABP has focused on determining its function 22; reviewed briefly in 25. Laboratory tests of female preference for males carrying different genetic variants of salivary ABP have provided evidence that the protein may mediate sexual preference 2930. A role in sexual selection is consistent with the evidence for positive selection in the microevolution of the gene (Abpa) for the α subunit, which has a different allele fixed in each of three subspecies of Mus musculus 3132. These Abpa alleles show significantly reduced polymorphism both in their exons and introns and high ratios of nonsynonymous to synonymous nucleotide substitutions (K_A/K_S) in the coding region 3334. Fixation of different alleles of Abpa in the subspecies of Mus musculus has been proposed to have occurred by selective sweeps 33. Abpb and Abpg also have high K_A/K_S values, suggesting that their microevolution has also been driven by positive selection 192022.

Here we investigate the expression of 17 predicted Abpa, Abpb and Abpg genes in 23 mouse glands and other organs. We chose organs and glands to obtain as wide a representation of gene expression in the mouse as possible. We specifically included any tissues where expression of other secretoglobins has been identified (Harderian gland, lacrimal gland, lung, kidney, uterus, skin/sebaceous glands, salivary glands and prostate; 25). In the head we surveyed the brain and (separately) olfactory lobes, parotid glands, sublingual glands, submaxillary glands, lacrimal glands, major olfactory epithelium (MOE), vomeronasal organ (VNO) and Harderian glands; in the body, we tested skin, adrenal gland, heart, spleen, kidney, testis, lung, liver, pancreas, small intestine, bladder, prostate, ovary and uterus. To test for expression similarity or divergence among paralogous genes we sought to identify transcripts in spatially-distinct tissues and from embryonic or adult individuals of different ages and/or sexes. Our results show how adaptive evolution among paralogous genes has led to expression divergence within a relatively short (12–24 My) period of time. We use these expression patterns to develop a model of paralogue neofunctionalization.

Our goal was to survey the glands and other organs of male and female, embryonic, juvenile and adult mice for expression of nine Abpa and eight Abpbg paralogues previously predicted from the mouse genome to be full-length and apparently functional. Hereinafter Abpa and Abpbg paralogous genes will simply be identified as "a" and "bg" with numerical suffixes as described previously 19. We developed primer sets specific for each putative Abp gene (Fig. 1), excluding those paralogues that are predicted to be pseudogenes 19. Forward and reverse primers of each set fall in exons 2 and 3, respectively (see 20, for the structures of the ABPα, β and γ subunit genes) in order that a much larger product (ca. 1–1.2 kb) is obtained when genomic DNA is the template, compared to that obtained from cDNA templates (ca. 200–300 bp). This design allowed us to test the primer sets, using genomic DNA as templates. Each primer pair amplified a product of the expected size from genomic DNA (not shown). In preliminary experiments the RNA extracted from C3H/HeJ tissues and subsequently used to make cDNA was contaminated with genomic DNA (Fig. 2, Panel A). In subsequent experiments we removed the contaminating genomic DNA by RNase-free DNase treatment, as shown in Fig. 2 (Panel B). In all cases where transcripts were detected, they were of the expected sizes. We verified a representative of each paralogue with DNA sequencing.

The quality of the cDNA made from RNA extracts was tested using an internal control, a primer set designed to amplify the housekeeping gene Gapdh 35. Only cDNA that showed the clear presence of a band of the expected size with the Gapdh primers was subsequently tested for Abpa/Abpbg transcripts (see Fig. 2, Panel B for an example of Gapdh expression). Because RNA preparations were treated with RNase-free DNase prior to their use as templates in PCR reactions, the product obtained with the Gapdh primers could confidently be ascribed to amplification of that transcript, rather than to the amplification of pseudogenes of corresponding size 36.

Although it was not our purpose in this study to precisely quantitate levels of expression of the individual paralogues, we did note reproducible variation in intensity of RT-PCR products. A series of dilutions of a typical template indicated that we could detect expression over a broad range. Most expressing tissues gave a negative result between 10⁴ and 10⁶ dilution.

The recent sequencing of numerous genomes has made possible comparative studies aimed at enhancing our understanding of gene evolution. It is clear from studies of mammalian genomes that the vast majority of genes have strongly conserved their coding sequences, and generally occur only in single copies 123. Genes involved in adaptation and functional innovation, on the other hand, often show the footprints of positive selection in elevated ratios of nonsynonymous to synonymous nucleotide substitutions rates (K_A/K_S; 39) in their coding regions. In addition, they are subject to frequent duplication, deletion and pseudogene formation 40. Prevalent amongst rapidly evolving genes are those involved in immunity, reproduction, chemosensation and toxin metabolism 40.

Mouse salivary androgen-binding protein genes Abpa and Abpbg have evolved under strong positive selection 1931333441 and they have been duplicated extensively, mostly in pairs of Abpa/Abpbg genes, across a 1–2 Mb region of mouse chromosome 7 19. This has been a recent expansion peculiar to the mouse, since the rat genome contains only three pairs of Abpa/Abpbg paralogues, and the human and chimpanzee genomes contain only a single pair whose members are both pseudogenes. These and other observations led to the hypothesis 19 that the common ancestor of mice and rats possessed only a single Abpa/Abpbg gene pair.

The recently released dog genome contains a single full-length Abpa/Abpbg pair, a finding which supports our proposal that this is the ancestral state in eutherian mammals 19. The cat genome also contains at least one Abpa/Abpbg pair and these are expressed as the subunits of the Fel dI dimer, the major allergen in the cat 42. We argue, based on the evolutionary distances (K_S values) of the genes, that the alpha and beta/gamma-like genes of cat Fel dI are likely to be the orthologues of the Abpa and Abpbg genes. The median K_S values between rodent Abpa or Abpbg genes and cat Fel dI are 1.25 and 0.87, respectively which are consistent with the reported substitution rate between rodent and cat orthologous sequence of 0.60 43 given that neutral rates typically vary within the genome by up to 3 fold. Since the cat orthologue, like a number of Abpabg orthologues, is expressed mainly in salivary glands 2942 it is possible that the original function of the ancestral Abpa/Abpbg genes involves secretion from salivary glands and subsequent application to the pelt of the animal.

The large cluster of mouse Abp paralogues qualifies in three respects as genes that are rapidly evolving under strong positive selection, namely elevated K_A/K_S and a fast duplication rate, and now data showing rapid diversification of gene expression following duplication. The question is which of the four most likely adaptive functions best fits our current understanding of ABP: immunity, reproduction, chemosensation or toxin metabolism? Karn and Dlouhy 32 showed that salivary ABP is ubiquitous in Old World and New World rodents, regardless of their diets, and noted that ABP binds androgens specifically and with relatively high affinity (see also 44 and 41). They speculated that the function of ABP must be one general to rodents, such as mate recognition, rather than a diet-specific one, such as toxin metabolism. The expression patterns we report here also augur against a role in toxin metabolism because we could not find expression of any of the Abp paralogues in mouse liver or spleen, where detoxification of toxic metabolites might be expected to occur.

The developmental sequence of expression of these genes provides clues that help further narrow the choices of functional role for ABP. Abpa/Abpbg expression begins in the head of the embryo as early as 13 days of gestation, comprised of a relatively simple pattern of a11/bg11 and bg10. By the first day of life, this has changed little, but by six days many other paralogues are expressed in the submaxillary and lacrimal glands. Expression proceeds to 15 days of life without yet revealing the sexually dimorphic patterns seen in several glands in adult mice. The transition to sexually dimorphic expression following puberty appears incompatible with roles for these in host immune defense.

It is fascinating that the MOE and VNO are the last tissues to achieve an adult expression pattern. Unlike the lacrimal gland and the salivary glands, these are not secretory tissues. It may be that early expression in the pups' secretory tissues provides signals by which the female parent recognizes her offspring. The pups would not need MOE and VNO expression themselves at this early stage. This distinction hints at a role for ABP in chemosensation, perhaps by participating in detection and recognition of ABP subtypes. Paralogues expressed in the MOE and VNO could be involved in phenotype matching and/or in a system whereby a bound ligand is transferred between paralogues to augment the signal of a proteinaceous pheromone.

Recently, Grus et al. 45 studied V1R genes expressed in the VNOs of the dog, cow and opossum, reporting 8, 32 and 49 intact V1R genes, respectively, in these three species. The numbers of V1R genes in the genomes of mice and rats had been previously reported to be 187 and 102, respectively 46 while the human genome contains about 200 V1R genes, all but 4 of which appear to have been pseudogenized 4748. Grus et al. 45 showed a concordance between V1R repertoire size and the complexity of VNO morphology and suggested that these characteristics of the VNO are indicative of the sophistication of pheromone communications within species. These findings are particularly interesting in the context of Abp evolution. We previously reported finding a single, pseudogenized Abp gene pair in both the human and the chimpanzee genomes, while the genomes of the mouse and rat had approximately 14 and 3 pairs of Abp genes, respectively, most of which had full-length ORFs 19. A single, full-length Abp gene pair in each of the dog and cat genomes would be consistent with a pattern of increasing importance of Abp genes in conspecific communication, proceeding from potentially no importance in primates to low importance in carnivores to relatively high importance in murid rodents, with the mouse utilizing the most number and diversity of Abp gene pairs.

In our recent paper 19, we demonstrated extensive sequence diversification in the numerous Abpa/Abpbg paralogues we predicted. In this report, we have shown that Abp genes have evolved multiple, often distinct, expression patterns. Once these patterns are mapped to these genes' proposed phylogenetic tree a surprising lack of congruency is observed (Fig. 5): the evolution of gene expression appears not to recapitulate the evolution of coding sequence. Thus it appears that neofunctionalization, rather than neutral drift, has driven the evolution of these genes' expression patterns over relatively short time periods.

The extensive diversification in expression of this family of proteins provides two lines of evidence for a pheromonal role for ABP: 1) different patterns of Abpa/Abpbg expression in different glands; and 2) sexual dimorphism in the expression of the paralogues in a subset of those glands. It is clear from our observations that expression patterns of many of the Abp paralogues differ dramatically among various glands that are located almost exclusively in the head and neck, where the sensory organs are located. Since mice are nocturnal, it is expected that they will make extensive use of olfactory as opposed to visual cues. The glands expressing Abp paralogues produce secretions (lacrimal and salivary) or detect odors (MOE and VNO) and thus it appears highly likely that ABP proteins play a role in olfactory communication. This supports the earlier suggestion, based on behavioral testing, that salivary ABP (the Abpa gene, now Abpa11) mediates mate preference 2930. In the studies of Laukaitis et al., 29, females showed a stronger preference in experiments where the choice was between males tethered to the ends of the test chamber, than they did for the males' territories when the males were absent. Sniffing the face of the male would present the female with much more information (Fig. 4) than the male might leave in his territory in the form of shed skin flakes or hair which had been coated with saliva during grooming 29.

If the variety of proteins produced by expression of the mouse Abp paralogues act as components of a complex pheromone system, then the secretions of the lacrimal gland and salivary glands provide a constellation of olfactory information, such as an animal's age and sex. Indeed it has been proposed previously that polymorphism in salivary ABP communicates the species/subspecies of the animal 2930313249. Thus we suggest that the strong evidence for involvement of Abp paralogous genes in adaptive evolution is consistent with a pheromonal role for their protein products.

Laukaitis et al. 20 proposed that the region between Abpa and Abpb (a11 and bg11 in 19 and in this study) could contain a sequence that coordinately regulates their expression. Emes et al. 19 suggested that this was possibly true of the bidirectional 5'-5' paired Abpa and Abpbg paralogue sets in general. The expression data we report here for a11 and bg11 is consistent with this idea since we have not seen either expressed independently of the other. However, it does not appear to hold true generally, since we observed independent expression of the members of a number of other pairs (e.g., bg10 in submaxillary and sublingual glands; a10 in VNO and MOE, etc.). Thus, regulation of the expressions of Abpa and Abpbg members of pairs must be more complex than simple coordinate up- or down-regulation of pairs. The lack of correlation between the evolutionary relationships of the Abpa/Abpbg paralogues 19 and their expression patterns (summarized in Fig. 5) support this notion, suggesting that sequences regulating expression of these genes have evolved in a manner that did not parallel the evolution of the structural genes. This implies rapid evolution of expression diversity in the Abpa/Abpbg paralogues.

Huminiecki and Wolfe 50 studied divergence of transcription profiles of genes in homologous tissues between human and mouse. They focused on loci where recent species-specific gene duplication occurred within one or the other species, allowing them to compare the transcription profiles of young paralogues in one species with a single orthologue in the other. They found that the presence of species-specific gene duplication accelerates the rate of expression divergence and that the recent duplicates are subject to reduced constraints on their protein sequences. In most cases, they observed that multiple changes in the spatial expression profile have occurred and they concluded that the expression in a new tissue suggests neofunctionalization. These results are consistent with those we reported earlier 19 and those that we report here for the recently and extensively duplicated Abpa/Abpbg genes in mice. This is perhaps an even more striking example because the single orthologous pair of Abpa/Abpbg genes in primates both degenerated into pseudogenes while expression in rodents has expanded into at least 9 tissues from the 2 tissues reported for the cat 42.

The multiple changes in the spatial expression profile of these genes resulting in various combinations of expression in glands and other organs in the head and face of the mouse strongly suggest that neofunctionalization of these genes, driven by adaptive evolution, has occurred following duplication. Primates represent an interesting contrast insofar as their single Abpa/Abpbg gene pair was pseudogenized, suggesting that its function has become dispensable. Many genes involved in olfactory and pheromonal cues have become pseudogenes in primates 40. Evolutionary and expression information thus both indicate a role for ABP and its paralogues in pheromonal communication among mice and, perhaps in other animals, one that is likely to be less important in the primate line, at least in the great apes.

This work arose from a comparative genomic analysis in which RDE, RCK, CML and CPP agreed to divide the bioinformatics analyses (published in 2004) and the expression analyses (this paper) between the two laboratories, Ponting's at Oxford University and Karn's at Butler University. SRD directed and participated in the harvesting of the many mouse tissues required for the study and he contributed to the design of the expression detection experiments. CML and RCK carried out the PCR expression analyses. RDE and CPP created the annotated gene trees. All authors made substantive intellectual contributions to this work and all authors read and approved the final manuscript.