Bovine proteins containing poly-glutamine repeats are often polymorphic and enriched for components of transcriptional regulatory complexes

In humans there are more than 40 relatively rare progressive neurodegenerative, neurological and neuromuscular diseases linked with repeat instability mutations [1–4]. A distinct subset of this group, which includes for example Huntington disease and spinocerebellar ataxias, is caused by polymorphic variation in trinucleotide repeats, typically CAG repeats, often present in the coding sequences of specific genes [1–3, 5–8]. In this group of 11 diseases the repeats usually encode long poly-glutamine (poly-Q) tracts, although other poly-amino acid tracts also occur but at lower frequencies. These human diseases are caused by extreme expansions of the repeats, which adversely impact protein structure, often causing intracellular protein aggregation and altered protein function [1, 3, 7, 9, 10]. However, trinucleotide repeat tracts which are impure, but maintain the ability to encode poly-Q tracts, have less severe impacts on function in model systems. This suggests that repeat expansions influence both protein and mRNA structure and function [1, 4, 8, 11].

The length of the expanded repeat is typically inversely related to the age of disease onset and directly related to the severity of the disease [6, 11]. Moreover, there is often a tendency for the expanded repeat to further increase in size in subsequent generations thereby resulting in greater disease severity and earlier onset, a phenomenon called genetic anticipation [1]. Trinucleotide coding sequence repeats in many genes are often polymorphic in the normal population but these repeats are typically shorter in length compared with disease-causing alleles. In some instances the distinction between normal and disease states can be minor in terms of repeat length [1, 11].

In a restricted examination of 7,039 human, rat and mouse orthologs 75, 53 and 58 genes, respectively, containing poly-Q encoding CAG repeats (CAG poly-Q genes) were identified [12]. Another analysis of the human genome reported a total of 66 CAG poly-Q genes [11]. These regions were defined as containing five or more CAG repeats encoding poly-Q. Although poly-Q containing proteins are generally more highly conserved in mammals, paradoxically their repeat regions are typically poorly conserved both in sequence and length, and they have been implicated in the rapid functional diversification of these proteins and the continuous morphological evolution of mammals [13–15]. It has also been suggested that normal polymorphic variation in these genes contributes to morphological variation within a species, as many of the affected genes are involved in regulating aspects of development and often encode transcription factors and transcription co-regulators [11, 14]. Indeed, polymorphic trinucleotide repeats in two genes, RUNX2 (runt-related transcription factor 2) and ALX4 (aristaless-like homeobox 4), have been linked with morphological variation in canine skulls and limbs [15, 16]. It is possible therefore that those genes containing polymorphic coding sequence trinucleotide repeats may contribute to variation in a broad range of phenotypes within a species.

What is not clear is whether the propensity for normal polymorphic variation in the CAG repeats of some of these genes also occurs in non-primates. There is significant positive correlation between the length of a pure CAG repeat and its propensity for polymorphic variation in human genes [11]. Interruptions of the CAG repeat, even with the alternate codon for glutamine, CAA, decrease the likelihood of polymorphic variation, as do reading frame shifts maintaining the CAG repeat structure but encoding different amino acids [11]. Moreover, the potential for a gene to cause known human diseases was also associated with greater CAG repeat polymorphic diversity [17]. Thus, it is probable that long and pure CAG tracts encoding poly-Q are predictors of polymorphic variation that is likely to impact on phenotype.

Identification of poly-Q encoding genes in additional mammalian species may identify species-specific differences in both the repertoire of these genes as well as the diversity of polymorphic variants in a subgroup of these genes. The former genes may contribute to an understanding of morphological and functional differences between mammalian species while the latter may contribute to phenotypic variation within a species. The current investigation has identified all bovine poly-Q encoding genes and within this group a number of genes whose orthologs in human and mouse do not contain these repeats. The bovine poly-Q encoding genes were prioritised using multiple criteria for their likelihood of being polymorphic and then the highest ranking group was experimentally tested for polymorphic variation within a cattle diversity panel. Extensive polymorphic variation was identified. The poly-Q encoding genes were strongly enriched for molecular functions relating to transcriptional regulation and their encoded proteins are involved in large transcriptional regulatory complexes in the cell nucleus.

The current study identified 178 bovine poly-Q encoding genes and demonstrated that a substantial number of the ranked genes were polymorphic in their repeat containing regions. Assessment of larger populations of cattle will be required to define the full repertoire of polymorphic poly-Q encoding genes as lower frequency alleles were probably excluded in the current analysis. The lack of conservation of the repeat regions associated with polymorphic variation in poly-Q encoding genes indicate that these are rapidly evolving gene regions [13]. Indeed, it has been estimated that the insertion-deletion mutation rate of microsatellites, such as CAG repeats in mammals, is approximately 100,000 fold greater than single nucleotide substitutions [14]. As alterations in the number of trinucleotide repeats do not cause frameshifts in coding sequences, variation in these regions both between and within species is probably much better tolerated than other types of indels. Longer CAG repeats with little interruption by the alternative glutamine encoding codon CAA, are more likely to be associated with higher allele diversity (Figure 3) [11, 35–39]. Polymerase slippage, unequal crossing-over during replication, and repair associated mechanisms, are the most likely explanations for this increased genetic diversity [4, 40, 41]. The overall balance between the mutational mechanisms that promote repeat expansions and the accumulation of point mutations that decrease repeat purity and therefore decrease the probability of repeat expansion is likely to dictate the nature of the poly-Q tract within a gene and its propensity for allelic variation.

Although there are a number of human diseases that are caused by instability in the repeat regions of poly-Q encoding genes, particularly expansion in subsequent generations (anticipation) and during aging, there was no evidence for meiotic instability in the 16 polymorphic poly-Q encoding bovine genes that were tested. A similar result was also demonstrated for porcine HTT[42]. One likely explanation is that the identified bovine polymorphic gene variants represented normal alleles of relatively short lengths and not the extreme expansions characteristic of some poly-Q encoding genes that cause human diseases [1, 6, 11].

It has been argued that proteins containing repeats simply reflect the propensity for variation and that these regions are not subject to selection and have no functional impact [43, 44]. These conclusions are at odds with substantial evidence demonstrating that amino acid repeat size in a wide spectrum of human genes impacts on molecular function and that these regions are indeed subject to strong selection [4, 9, 10, 13, 24, 30, 45–51]. In particular, Q-tracts have been shown to be autonomous length-dependent activators of DNA binding activities in transcription factors [46, 47, 49]. It is noted that there was strong enrichment for transcription factors in the bovine poly-Q encoding genes and thus it is likely that intra-species and inter-species poly-Q tract expansions (or contractions) directly impact the transcriptional activities of these proteins. This is an appealing concept as it links propensity for increased genetic diversity with subtle functional differences in a class of proteins that can amplify these genetic influences because they impact the transcriptional regulation of a much broader group of genes. Moreover, based on evidence linking normal polymorphic variations in two poly-Q encoding genes, RUNX2 and ALX4, with morphological variation in dogs, it has been proposed that poly-Q encoded genes contribute raw genetic material for rapid evolutionary change in a species [10, 14, 15, 52].

Those genes containing conserved repeats, which often tend toward shorter and less pure repeat sequences, may reflect strong functional constraints on the encoded proteins in these regions and therefore selective pressure against variation despite increased mutational opportunity. Consistent with this concept, these genes are rarely polymorphic in their repeat regions within a species population [11]. This information highlights local regions which may have important functions. The absence of polymorphic variation in the poly-Q tract of the bovine androgen receptor (AR), a co-activator for androgen dependent gene transcription, is interesting in view of the links between poly-Q variation in the human protein and androgen-dependent diseases, many of which influence reproductive success [53, 54]. As the AR gene is carried on the X chromosome some of these diseases result in somatic mosaicism [35]. The long history of cattle domestication involving intensive sire-based selection, primarily focussed on growth and reproductive traits, may have selected against variation in the AR gene in cattle.

Comparison of allele frequencies in Bos taurus indicus and Bos taurus taurus breeds showed significant differences for 15 of 16 polymorphic poly-Q encoding genes. Since the repeat regions in these genes are subject to relatively high rates of mutation, then these data could suggest differences in the microsatellite mutational rates within the two breeds. This possibility seems unlikely as there were no systematic effects on the repeat sizes or the number of alleles for each gene. An alternative explanation is that the allele frequency differences reflect variation that has independently arisen since divergence of these two cattle breeds from an ancestral population. The frequencies of alleles for four genes, CACNA1A, FXC1, HTT and TBP were significantly different for Bos taurus taurus cattle specialised for milk or meat production. All of these genes have ubiquitous expression patterns (GEO database accessions; GDS182 (mouse) and GDS1402 (human)). TBP (TATA box binding protein) and HTT are transcriptional co-regulators located in the nucleus, CACNA1A is a subunit of a plasma membrane voltage dependent calcium channel and, FXC1 is a component of a hetero-oligomeric translocase complex present in the mitochondrial inner membrane where it plays a role in the import of proteins into the inner mitochondrial membrane. Thus, structural variations that impact on the functions of these proteins have potential to modify broad transcriptional activities and metabolism. These influences may be subject to different selective pressures in beef and dairy cattle populations. Notably, extreme polymorphic repeats in three of four of these genes (CACNA1A, HTT and TBP) are associated with a spectrum of human diseases [53].

FXC1 was the most polymorphic gene with 18 detected alleles. Although the evolutionary conserved FXC1 encoded protein has not been directly associated with any human diseases, a loss of function mutation in a related component of the mitochondrial hetero-oligomeric translocase complex, TIMM8A, has been linked with the neurodegenerative disorder, Mohr-Tranebjaerg Syndrome [53]. The extensive protein-protein interactions occurring in the inner mitochondrial membrane translocase complex suggest that variation in FXC1 protein sequence caused by polymorphisms in its poly-Q region may have potential for impacting mitochondrial function [55]. Alternatively, the positioning of the poly-Q tract near the N-terminus of this small protein may have insulated its primary functions from the influence of polymorphic variation.

The number of poly-Q encoding genes in the cow (178) was similar to human (201) especially taking into account that the former number was determined using a draft genome assembly which is missing representation of all genomic sequence (~90% coverage) and does not have completed gene models for all genes [19]. Surprisingly, there was only 61% (109/178) gene overlap between cow and human indicating substantial differences between the species, although the functional enrichments in both groups were similar [11]. The figure of 201 human poly-Q encoding genes (Q ≥ 5) is consistent with an independent figure of 158 human poly-Q genes (Q ≥ 4) in a dataset of 6,477 vertebrate orthologs [37]. However, the subset of 123 bovine CAG poly-Q genes (≥ (CAG)₅) was substantially larger than that estimated for humans (66-75) and rodents (53-58) [2, 11, 12]. This observation may partially reflect the impact of the large expansion of CAG repeats in cow genomic DNA compared with other mammalian species (~10 fold increase compared with humans) [19]. As the scale of the repeat expansions in bovine coding sequences is much less than the genome-wide figure, it is likely that coding sequence CAG repeats are subject to different selective pressures compared with non-coding intergenic CAG repeats, as has been previously reported [13, 48].

A total of 26 bovine poly-Q encoding genes did not have murine and human orthologs with poly-Q tracts. This group of genes potentially highlights evolutionary adaptations that contribute to the unique biology of ruminants. While some of these encoded proteins contained unique poly-Q tracts, others were identified because there were no orthologs in either the mouse or human genomes or the protein models did not contain sequence corresponding to the orthologous region of the bovine poly-Q tract proteins. The latter case may reflect species specific differences in alternative splicing. More trivial explanations include potentially incomplete gene and protein models or that the Q-tracts in the human and murine orthologs did not reach the threshold size of five residues used to define a Q tract. Notably, these 26 genes were not different from all bovine poly-Q encoding genes in relation to encoded protein size or function.

Comparison of bovine and human poly-Q proteins revealed that the lengths of ortholog Q-tracts were generally similar (Figure 1). This result is different from the finding that rodent Q-tracts were generally shorter than human Q-tracts [2]. As the maximum length of the bovine CAG Q-tract was directly related to allelic diversity (Figure 3) then it may be concluded that the overall extent of polymorphic variation in the poly-Q encoding genes of cattle is likely to be similar to human populations.

Poly-Q encoding genes can potentially generate considerable transcriptional diversity through a variety of mechanisms. The tissue-specific mRNA splicing of NFYA alters the length of the poly-Q encoded region in the absence of polymorphic variation. The poly-Q region is part of the activation domain of this transcription factor and modifications to the region have marked effects on function [33]. Thus, the tissue specific splicing may be a means of generating NFYA functional variants which could impact the transcription of the larger repertoire of genes that are regulated by NFYA. Although no poly-Q tract polymorphic variants in NFYA were demonstrated in the current analysis, if they existed, such allelic variants could exert additional tissue specific functional influences through this splicing mechanism. Polymorphic repeat regions in other poly-Q encoding genes traverse splice sites (e.g. ABCF1) or are closely juxtapositioned with splice sites (e.g. MED12). Moreover, a considerable number of poly-Q encoding genes contained repeats immediately adjacent to or at splice sites. In the case of MED12 this was associated with considerable transcriptional variation likely caused by the influence of coding sequence CAG repeats on splicing. There are similarities between CAG repeats and splice donors (consensus sequence (A/C)AG at the 3' end of the exon) and splice acceptors (consensus sequence (C)AG at the 3' end of the intron). This similarity could cause additional mRNA splicing variation in poly-Q encoding genes. Notably the protein MBNL1 (muscleblind-like) is involved in pre-mRNA splicing site choice and it also binds strongly to transcribed CAG repeats [56]. This dual function reinforces the view that splice site choice can be influenced by coding sequence CAG repeats. The presence of poly-Q encoding tracts contained within small alternatively spliced exons for a considerable number of poly-Q encoding genes suggests that this is another mechanism which generates considerable transcriptional and functional diversity. Indeed, the latter is accentuated by the large size of many poly-Q encoding genes, which often encode multiple autonomous functional domains each encoded by discrete exons. Alternative splicing of these exons is frequent and can generate many combinatorial mRNAs for each of these genes, possibly producing a range of protein products with subtle functional variations and a diversity of tissue specific variants.

The bovine poly-Q proteins were strongly enriched for large multi-domain transcriptional regulators (Additional file 4). Moreover, many of these proteins participate in large and common physical interaction networks as well as an extensive network associated with their transcriptional regulation (Figure 4). The central importance of CREBBP in this nuclear network is emphasised by its ability to act as a transcriptional co-activator in conjunction with EP300 to help assemble large regulatory protein complexes at sites of active transcription. Both of these proteins have intrinsic histone acetylase activity, which is involved in modifying chromatin structure in nucleosomes and thereby regulating gene expression [53]. The broad tissue expression patterns of the proteins in these nuclear complexes indicate that they are co-expressed in cells and hence these regulatory networks are likely to be biologically operational, with their components acting cooperatively in large transcriptional regulatory complexes. There could be even greater functional complexity inherent in these complexes caused by polymorphic variation in some of these proteins. Indeed, there is potential for amplification of the phenotypic impact of poly-Q polymorphic variations of multiple components in this regulatory complex - an epistatic effect in genetics terms. Even in the absence of polymorphic variation, alternative splicing influenced by CAG tracts may generate poly-Q protein functional diversity that could impact the activities of large transcriptional regulatory complexes in the nucleus. This process could generate additional functional complexity tailored to the needs of specific tissues, developmental pathways or tissue responses to external stimuli.

A total of 178 poly-Q encoding genes have been identified in the bovine genome. By using a hierarchical prioritization process we established that at least 16 of 32 top ranked genes were polymorphic in regions encoding poly-Q tracts. There was significant correlation between the extent of allelic diversity and the length of the poly-Q tract. This information, in conjunction with the purity of the repeat region encoding the poly-Q tract and the lack of conservation of repeat length, provide indicators for the propensity for polymorphic variation in the full repertoire of poly-Q encoding genes. Unlike some human poly-Q encoding genes there was no evidence for repeat instability, which may be attributable to the lack of large repeat expansions in the normal cattle population. This may be an inherent biological characteristic of cattle or a reflection of strong selective pressures in breeding programs that exclude these variants. Polymorphic variations in bovine poly-Q proteins have strong potential to generate epistatic biological effects due to the involvement of these proteins in transcriptional regulatory complexes and protein-gene regulatory interactions. Thus, the bovine poly-Q encoding genes may be contributing to phenotypic variation in cattle populations both directly and indirectly through epistatic interactions. CAG repeat tracts in poly-Q encoding genes can also influence splice site choice and thereby increase the functional diversity of transcripts from these genes.