Glutathione S-transferases (GSTs) are a widely-expressed family of enzymes that play an important role in detoxification by catalyzing the conjugation of many hydrophobic and electrophilic compounds with the reduced form of the tripeptide glutathione (GSH). Substrates for these enzymes include an extensive array of xenobiotics, including drugs, toxins and environmental carcinogens [1]. The conjugation of xenobiotics with GSH is catalyzed by GSTs in the cytosol, endoplasmic reticulum and mitochondria. Based on their biochemical, immunologic, and structural properties, the cytosolic or soluble GSTs are categorized into 4 main classes: alpha (A), mu (M), pi (P), and theta (T). Each class can be subcategorized into various related isozymes.

GST pi 1 (GST-P1) is expressed in extrahepatic tissues such as placenta, lung, gut and erythrocytes [2, 3], where it is involved in the biotransformation of carcinogens such as acrolein [4], and benzo[a]pyrene diol epoxide (BPDE) [5], endocrine-disrupting pesticides such as atrazine [6], and numerous drugs, including cisplatin [7] and chlorambucil [8]. While humans and rats express one subunit belonging to the P1 class of GSTs, GSTP1, mice express two pi subunits, GSTP1 and GSTP2. The situation may be analogous in dogs, who appear to have one annotated GSTP1 gene, located on chromosome 18 (chr18:49,905,161–49,908,182 on the CanFam3 dog genome assembly), and another uncharacterized “GSTP-like’ gene (chr18:49,950,873–49,953,818) adjacent to it.

Polymorphisms in the GSTP1 gene can lead to impaired GST function. These DNA variants influence susceptibility to carcinogens and the toxic effects of some pesticides, affect the response to chemotherapy used in the treatment of metastatic colon cancer and multiple myeloma, and modify the risk of several forms of cancer. For example, a single nucleotide polymorphism (SNP) in the coding region of human GSTP1 (c.313 A > G, denoted as rs1695), present in 35% of the population, leads to decreased enzyme activity [9, 10]. This polymorphism results in a Ile105Val change in the GSTP1 enzyme that is, for example, less able to conjugate carcinogens such as BPDE [11]. In fact, this allelic variant has been associated with increased susceptibility to acute myeloid leukemia, prostate, lung, bladder, gastric, oral and breast cancer [12–18]. Genetic and epigenetic variability in the promoter region of GSTP1 also affects gene expression and influences susceptibility to several types of cancer [19–22].

In dogs (Canis lupus familiaris), several studies suggest a possible causal relationship between exposure to certain environmental chemicals and canine cancers. An increased risk of canine malignant lymphoma has been associated with owner application or commercial treatment of lawns with the commonly used herbicide 2,4-dichlorophenoxyacetic acid [23, 24], as well as heterogeneous pesticide mixtures [25]. Living in industrial areas, or in proximity to polluted sites, incinerators, and/or radioactive waste, are also significant risk factors [26, 27]. The expression of GSTs in dogs can vary widely. GST-T activities have been characterized in livers from research beagles, and 12% of these dogs had virtually no GST-T protein expression or activities [28]. In addition, GSTT1 polymorphisms have been associated with lymphoma in dogs [29]. It is not known if the expression of GSTP1 shows a similarly wide variability between individual dogs, or whether GSTP1-null allelic variants exist in the population. A preliminary study we have conducted (unpublished data) indicated that, in contrast to the coding region, the canine GSTP1 promoter sequence contained multiple polymorphisms. By affecting gene expression, these polymorphisms could potentially impact detoxification, and consequently modulate the risk of cancer and other diseases in dogs.

Therefore, the objective of this study was to investigate the variation in the GSTP1 gene promoter in a diverse population of domestic dogs, compare the incidence of these variants between dog breeds, and predict the potential functional effects of these genetic variants. In order to determine which polymorphisms may have been introduced by domestication or breed formation, we also studied the same genetic region in other species representing the Canidae family, the gray wolf (Canis lupus), coyote (Canis latrans), red fox (Vulpes vulpes) and gray fox (Urocyon cinereoargenteus).

Fourteen polymorphic loci, including twelve SNPs and two microsatellites were found in the proximal GSTP1 promoter region (chr18: 49,908,620–49,908,119) of domestic dogs (Fig. 1). The five variant alleles between positions c.-21 and c.-46 were located within the microsatellite sequence on the 5′ untranslated region (5′UTR) in exon 1.

Several polymorphisms appear to be unique or much more frequent in dogs, when compared to their closest relatives, wolves and coyotes. For example, the c.-185 delT variant was found in 8% of dogs but not in the other Canis species (Table 1). MIDAS analysis also indicated that this base deletion appears to be the most recently evolved polymorphism in the dog GSTP1 promoter region. The c.-228 C > A polymorphism, present in almost half the dogs genotyped, was found in only 11% of wolves. At c.-46, the least frequent variant in dogs (C) is the major allele in wolves and coyotes. Similarly, the c.-305 GCCCTC repeat genotype differed between the three groups; in wolves and coyotes, the minor allele is the (GCCCTC)₂ repeat as opposed to the (GCCCTC)₃ variant in dogs. A total of 4 variants were only observed in wolves and coyotes.

The 5′UTR tandem repeat was predicted to form a hairpin RNA structure at the 5′ end of the GSTP1 transcript. The stability of this secondary structure, as indicated by the Gibbs’ energy (dG), decreases with smaller numbers of tandem repeats. Comparison of alleles of equal length but differing purity (for example 12*1 vs 12*2, and 16*1 vs 16*2) showed that the compound repeat alleles were expected to be half as stable as their simple repeat counterparts, perhaps because the 5′ hairpin loop is missing or is structurally altered (Additional file 2a). This decreased stability appears to extend to the entire GSTP1 transcript (Additional file 2b). The VARScores indicate that the highest mutation rates, after consideration of the number of GCC repeat units, unit length and repeat purity should occur in the commonly occurring 16*1 and 17 alleles in dogs and the 18 allele in wolves (Table 2). Conversely, the repeats least likely to experience GCC deletions or insertions are the smallest-length 10 repeat and the compound 16*2 and 22*1 repeats.

The gray fox GSTP1 promoter sequence was more similar to dog (98.2%), than to red fox (96.1%), which in turn showed more similarity to gray fox (96.1%) than dog (95.9%). Alignment of the promoter sequence of red and gray fox with dog showed differences at 16 and 12 loci respectively, with four of these loci shared between the two fox species (Additional file 3). With one exception (the 5′ UTR microsatellite), none of these were located at the sites of the genetic polymorphisms observed in the Canis species. In the grey fox, there were two repeating GCCCTC units at the locus which corresponds to c.-305 in the dog; this repeat number was only observed in three wolves and two coyotes but in none of the domestic dogs. In the red fox, two of the four repeating units were changed to GCCCCCGCCCCC. In the 7 red foxes sequenced, no polymorphisms were observed within the entire promoter region, including the tandem repeat. The red foxes were all homozygous for a 12-repeat exon 1 microsatellite (GCC(GCC)₃ACC(GCC)₃(ACCGCC)₂) that was distinct from the equal-length alleles observed in wolves (12*1) and coyotes (12*3) (Table 2). The single gray fox sequenced had a unique 8-repeat/16-repeat microsatellite genotype (GCT(GCC)₇ACCGCC and GCT(GCC)₃GCTGCCACTGCTGGCT(GCC)₅ACCGCC), with a single base (underlined) interrupting the triplet repeat sequence.

Alignment of the ‘wild-type’ GSTP1 transcript (which includes the 17-repeat allele) with other mammalian transcripts showed evidence of the microsatellite in a number of other species, with the highest number of repeats (13) observed in the white-tailed deer and Bactrian camel. The other species all showed repeat numbers lower than 10, with the two rodent species not showing any GCC repeat units in their 5′UTR (Additional file 4). No information is available on whether these sequences are polymorphic, as has been observed in the three canid species described in this study.

Some uncommon haplotypes, such as *1C (10), were detected in one or two breeds only (3 alleles in two Chow Chows; 1 allele in a Tibetan terrier). Interestingly, this haplotype was also observed in 6 coyotes. The otherwise rare *6A(17) haplotype was seen 5 times in the three New Guinea Singing Dogs investigated, with the other Singing Dog having the commoner *6A(16) haplotype.

The mechanisms by which these polymorphisms could affect GSTP1 expression via effects on transcription factor binding were investigated using the web-based program LASAGNA-Search. The wolf and coyote −539 C > A SNP may delete the binding sites for ARE86 (TRANSFAC Accession number M00415), Pax-2 (M00486), and C/EBP (M00159) and introduce novel sites for interactions with CHOP:C/EBP alpha (M00249) and STAT4 (M000498), while the 408G variant deletes an ARP-1 (M00155) site and creates a new site for Brachyury protein (M00150) binding. The common 350 C > A and −228 C > A SNPs result in new sites for the binding of NRF-2 (M00108) and GATA (M00076, M00462, M00075) respectively. The rare -127C allele is predicted to result in a new GC box (M00255) and E2 binding site (M00107, M00181). The wolf and coyote −82 C > T, located just inside the 5′UTR, may generate a new PPARalpha:RXR-alpha (M00242) binding site, while the other 5′UTR SNP, −68C > T, also unique to these wild canids, creates new sites for the binding of HNF-4alpha1 (M00411) and COUP-TF (M00158). The Wilms Tumor zinc finger (WT1-KTS) protein (T01839) was predicted to bind to the multiple EGR1 GGCG core sequences (on the + strand) within the microsatellite. Binding sites for this transcription factor were decreased for alleles with a smaller number of repeat units, and in repeats with decreased purity. For example, while there were 25 potential binding sites on the 16*1-repeat allele, there were only 11 sites on the 11-repeat allele, and just 5 sites on the similar size but compound 16*2-repeat allele (Additional file 6). On the other hand, alleles with the GCTGCCACTGCTACC motif (12*2, 16*2, 20, 22-repeats) were predicted to have a new binding site for KLF12 (M00468). In the case of the *22-repeat alleles, the unique GCA triplet within the repeat introduces a binding site for MyoD (M00184) that binds to the consensus TGCAACTGCT.

The promoter region of a gene is essential for its proper function, containing enhancers and repressors that control its rate of expression. Certain promoter features, such a TATA box, cis mutations, and the presence of unstable tandem repeats, increase expression divergence, or the capacity of a gene to evolve in expression [37]. Genes containing these type of promoters preferentially diverge when the effects of natural selection are minimized [38], a good example of which is canine domestication. The DNA resequencing performed in this study demonstrated that the canine GSTP1 promoter region (with a TAATAA box located at c.-102) contains a number of genetic polymorphisms of varying frequency and a highly polymorphic tandem repeat sequence located in the 5′ untranslated region. It is possible that the promoter architecture found in the canine GSTP1 gene has arisen both to allow flexibility in the biotransformation of existing and new xenobiotics as well as a direct consequence of inbreeding practices.

Molecular phylogeny studies have shown that the dog family, or Canidae, can be classified into four clades, including the wolf-like canids (dog, gray wolf, coyote), the fox-like canids (red fox), the South American canids, and the smallest, most primitive clade represented by the grey fox and the island fox [39]. The dog GSTP1 promoter sequence showed the best similarity with the other Canis species, gray wolf and coyote, followed by the gray fox, and red fox. Although sample sizes for the wild canids used in our study were too small for definite conclusions, we can tentatively conclude that our data supports the aforementioned phylogeny studies.

A total of 16 polymorphic loci were identified in the canid GSTP1 promoter, of which 13 were polymorphic in the dog population, 11 were polymorphic in the wolf, and 12 were polymorphic in the coyote. Almost half the dogs sequenced had the wild-type *1A(17) and the *6A(16) haplotype. Three loci (c.-82, c.-179, c.-539) were seen to have the allele fixed in dogs but polymorphic in both wolves and coyotes, which likely reflects the bottlenecks associated with domestication or breed formation. The c.-46C > T allele was observed in 50% of dogs but only in 14% of wolves and 36% of coyotes, suggesting that this polymorphism may have arisen from domestication, and that the ancestral allele for this SNP (rs22633673) should therefore be ‘C’. Nine different microsatellite repeat lengths were observed in dogs, while six and five repeat lengths were seen in wolves and coyotes respectively. The *3A(11) dog haplotype is characterized by the 11-repeat microsatellite, which is in strong linkage disequilibrium with the deletion variant at c.-185. This haplotype was not observed in wolves and coyotes and appears to have arisen as a consequence of dog domestication. The *3A(11) haplotype may thus prove useful in DNA marker studies seeking to distinguish between domestic dogs and their wild relatives. Another potential marker is the 1H(n) haplotype, which is found at high frequency in wolves but is absent in both dogs and coyotes. Haplotypes which are unique to and found in almost half of the coyotes studied were the *1C (10) and the *21A (12) haplotypes. Interestingly, the *1C (10) was also found in 2 Chow Chows and 1 Tibetan Terrier, both of which are ancient breeds [40] and would conceivably retain more ancestral alleles than the more recent breed groups.

The fox species were not polymorphic at any of the loci common to the Canis species, and exhibited sequence dissimilarities at 21 different positions. The lack of variation between the 7 foxes sequenced, including those polymorphisms within the microsatellite region, was surprising in view of the fact that 4 red foxes originated from different geographical areas in the US and the other 3 were Russian domesticated foxes. It would be interesting to further investigate microsatellite variation in a greater number of coyotes, wolves and foxes, as well as other Canis and Canidae species.

Since most dog breeds represent closed breeding populations that receive limited genetic variation [41], we expected to find significant differences between dog breeds, as well as a lack of variation within these breeds. The *3A (11) haplotype predominated in Afghans, Basenjis and Greyhounds, while the commonest haplotypes in Standard Poodles were *2A(16) and *8A(19), the latter haplotype occurring only in this breed. A lack of genetic diversity in the GSTP1 promoter was observed in Boxers and Siberian Huskies. Boxers are known to have low levels of heterozygosity, this being one of the reasons why a female dog from this breed was selected to have its genome sequenced at the Broad Institute [42]. Despite the fact that the 6 Siberian Huskies in the study were not related and came from different states within the US, they had identical *1A(17) haplotypes. This lack of genetic diversity is surprising in view of the fact that Siberian Huskies have ancient origins and display high heterozygosity in microsatellite data compared with other breeds [40].

The canine GSTP1 promoter containsa microsatellite mainly consisting of GCC triplet units tandemly repeated anywhere from 10 to 23 times. This repeat exhibits a very diverse distribution across species, even across closely related species, indicating rapid evolutionary change. Significant expansion of this microsatellite may have occurred in species belonging to the genus Canis, as it appears that the corresponding region in other mammals is much shorter. The red fox sequences were not polymorphic, perhaps indicating that the longer repeats occurred with the appearance of extant canids around 10 million years ago. Similarly, a GCC dodecamer repeat found in the coding region of the canine EPM2 gene has been shown to be polymorphic across the Canidae but not to other close relatives of the dog [43].

The microsatellite in the canine GSTP1 promoter has a very high GC content, which means that there is a greater tendency to form hairpin-loop and slipped-strand DNA structures during replication, recombination, and repair. These structures result in genetic instability and increased polymorphisms [44]. Two mechanisms are probably the reason for the microsatellite polymorphisms observed. One is replication slippage, which occurs when the daughter strand slips back one or more repeating unit in pairing with the template strand, leading to a trinucleotide expansion or contraction during DNA replication. This probably accounts for alleles which only differ by one repeat unit, as in the 14, 15 and 16*1 repeats seen in the *2A and *6A haplotypes, and possibly even those alleles that differ by two GCC units (for example, 14 and 16*1). The other mechanism is unequal recombination or crossing-over, which is a type of gene duplication or deletion event that deletes a sequence in one strand and replaces it with a duplication from its sister chromatid in mitosis or from its homologous chromosome during meiosis. Based on the microsatellite alleles observed, several types of crossover events can be postulated to occur via intra-repeat and/or inter-repeat recombination mechanisms. The simplest and most common event may involve the exchange of one GCC repeat at multiple sites, generating, for example, the 15 and 17 alleles from two parent 16*1 alleles (Fig. 2a). Identical-length alleles may also recombine only at specific sites on the tandem repeat, as may occur when two 16*2 alleles exchange 4 repeats to form the compound 12*2 and 20 alleles, as was observed in coyotes (Fig. 2b). Unequal exchanges of repeats can occur between alleles with different lengths, as between the 17 and 16*1 alleles to form daughter alleles with 11 and 22 repeats (Fig. 2c), or alleles with 15 and 18 repeats. Whilst the longer allele was observed in only two dogs, the truncated 11 allele was the third most frequent allele observed, probably because the 16- and 17-repeat unit alleles are the most common in the dog. The 19-repeat unit allele seen in standard poodles could be another example of unequal crossing-over between 2 different alleles, in this case the 17 and 18 repeat alleles, which also generates the common 16*1 allele (Fig. 2d). Support for this mechanism comes from the fact that all the standard poodles that were heterozygous for the 19 allele had the 16*1 allele. Moreover, the 16*1 and 19 alleles represent the major forms of the microsatellite in the standard poodles sequenced, with allele frequencies of 0.59 and 0.32 respectively. Finally, an ancestral recombination event could account for the origin of the common dog 16*2 and 17 alleles from crossing-over the 13- and 20-repeat variants that appear to be found only in wolves.

In humans, it has been estimated that a third of promoter variants may alter gene expression to a functionally significant extent [45]. Several studies have demonstrated that tandem GGC or CCG repeats in gene promoters have significant effects on gene function. For example, expansion of a CGG repeat in the 5′UTR of the fragile site mental retardation 1 (FMR1) gene causes epigenetic changes that reduce transcription factor binding and are the cause of Fragile X Syndrome [46, 47]. Two-repeat differences in GCC repeat length in a microsatellite located in the thiopurine S-methyltransferase (TPMT) 5′UTR may be responsible for ultra-high enzyme activity and consequent resistance to treatment by thiopurine drugs [48]. Tandem repeats can modulate GSTP1 expression by changing the copy number of binding sites for transcription factors, as has been observed with the binding of Sp1 to the human EGF gene [49], the interactions of Sp1 and AP2 to the human RFC gene [50], and the binding of p53 to PIG3 [51]. The GSTP1 microsatellite GCC repeat polymorphisms were predicted to affect the number of potential binding sites for the Wilms Tumor zinc finger (WT1-KTS) protein. The Wilms tumor gene 1 (WT1) is a transcription factor that acts as an oncogene in acute myeloid leukemia [52], and a recent Chip-Seq investigation has revealed that binding of the WT1-KTS isoform in leukemic cells is correlated to active transcription of numerous genes [53]. The 16*2 allele, which was found in 12% of the dogs in the study and over-represented in breeds such as Pembroke Welsh Corgi and Golden Retrievers, has very few binding sites for the WT1 activator compared to the other major repeat alleles (11, 16*1 and 17). The CAGTGGc motif present in the 16*2 allele provides a binding site for the WT-1 homolog transcriptional silencer KLF12 (Ap-2rep). It has been hypothesized that KLF-12 silences WT-1-dependent gene expression [54]. The decreased binding of WT1 and novel binding of KLF12 to the repeat could decrease GSTP1 expression. Although there is as yet no experimental evidence of the binding of WT1 and/or KLF12 to the canine GSTP1 promoter, it would be interesting to establish whether these transcription factors do indeed bind to this repeat sequence and whether the polymorphisms observed affect the number of binding sites and, ultimately, gene expression.

The tandem repeat is located in the 5′UTR of the canine GSTP1 gene and therefore forms part of the mature mRNA. The microsatellite is also located in proximity to the core promoter (20–30 base pairs downstream of the predicted transcription start site and the TATA box respectively), so variations in the tandem repeats could interfere with the binding of RNA polymerase by changing the spacing between functional elements in promoters [55]. The microsatellite may also affect stability of the secondary structure of the GSTP1 transcript. In the dog, two types of microsatellites consisting of 16 triplet repeats were observed. The 16*1 allele was predicted to form a more stable RNA hairpin-loop secondary structure than the 16*2 allele. The 16*2 allele alters the structure and position of the hairpin loop, potentially decreasing the stability of the entire GSTP1 transcript. These hairpin RNA structures may play a role in disease pathogenesis by affecting the processing of the primary transcript, resulting in deficit of the corresponding protein, or interacting with RNA-binding proteins, altering their normal activity [56].

In summary, this study has demonstrated the existence of multiple polymorphisms in the proximal promoter of the canid GSTP1 gene. Distinct haplotype differences exist between various dog breeds and between dogs, wolves, coyotes and foxes. The high degree of polymorphism in a 3′UTR microsatellite may be due to replication slippage and unequal allelic recombination events. Some of these variants may affect the binding of transcription factors and/or decrease mRNA stability, altering the detoxification potential of glutathione conjugation catalyzed by GSTP1. In order to investigate any effects that these promoter polymorphisms may have on the glutathione conjugation of xenobiotics, reporter gene assays are underway to determine the effect of the various microsatellite alleles on canine GSTP1 expression, together with the determination of GSTP1 mRNA levels and GSTP1 activity in the blood of genotyped dogs.