Human Arsenic Methyltransferase (AS3MT) Pharmacogenetics

Arsenic contaminates ground water worldwide. Methylation is an important reaction in the biotransformation of arsenic. We set out to study the pharmacogenetics of human arsenic methyltransferase (AS3MT, previously CYT19). After cloning the human AS3MT cDNA, we annotated the human gene and resequenced its 5′-flanking region, exons, and splice junctions using 60 DNA samples from African-American (AA) and 60 samples from Caucasian-American (CA) subjects. We observed 26 single nucleotide polymorphisms (SNPs), including 3 non-synonymous cSNPs, as well as a variable number of tandem repeats in exon 1 within an area encoding the cDNA 5′-untranslated region. The nonsynonymous cSNPs included T860C (M287T) with frequencies of 10.8 and 10% in AA and CA subjects, respectively, as well as C517T (A173W) in one AA and C917T (T306I) in one CA sample. Haplotype analysis showed that Ile306 was linked to Thr287, so this double variant allozyme was also studied functionally. After expression in COS-1 cells and correction for transfection efficiency, the Trp173 allozyme displayed 31%, Thr287 350%, Ile306 4.8%, and Thr287/Ile306 6.2% of the activity of the wild type (WT) allozyme, with 20, 190, 4.4, and 7.9% of the level of WT immunoreactive protein, respectively. Apparent Km values for S-adenosyl-l-methionine were 4.6, 3.1, and 11 μm for WT, Trp173, and Thr287 allozymes, with Km values for sodium arsenite with the same allozymes of 11.8, 8.9, and 4.5μm. The Ile306 and Thr287/Ile306 allozymes expressed too little activity for inclusion in the substrate kinetic studies. Expression of reporter gene constructs for the 5′-flanking region and the variable number of tandem repeats in the 5′-untranslated region demonstrated cell line-dependent variation in reporter gene expression, with shorter repeats associated with increased transcription in HepG2 cells. These results raise the possibility that inherited variation in AS3MT may contribute to variation in arsenic metabolism and, perhaps, arsenic-dependent carcinogenesis in humans.

neurotoxicity, hepatic injury, and carcinogenesis (1)(2)(3). Occupational exposure to arsenic occurs in the smelting industry and during the manufacture of pesticides, herbicides, and other agricultural products (4 -6). Arsenic is also an important environmental carcinogen that contaminates ground water worldwide (7). Finally, As 2 O 3 is used as a therapeutic agent, particularly for the treatment of promyelocytic leukemia (8,9). Methylated and dimethylated arsenic are the major urinary arsenic metabolites in humans (10,11). Although methylation has been regarded as a detoxification reaction, methylated As III is more cytotoxic and genotoxic than are arsenate (the most stable form of arsenic) and arsenite (12)(13)(14). Methylated derivatives are also more potent inhibitors of glutathione reductase (15,16), thioredoxin reductase (17,18) and pyruvate dehydrogenase (19) than is arsenite. Therefore, knowledge of individual variation in the function of the enzyme that catalyzes the formation of methylated arsenicals might be an important step toward increasing our understanding of the biological and pathological consequences of chronic exposure to inorganic arsenic.
An arsenic methyltransferase activity was recently described in the rat that catalyzes the methylation of arsenite, with S-adenosyl-L-methionine (AdoMet) 2 as the methyl donor (20). In rats, that activity is expressed in a variety of organs, and a similar activity is expressed in the human liver, kidney, and brain (21). The cloning of a rat cDNA, encoded by a gene that was originally annotated as "Cyt19," and the subsequent demonstration that the encoded protein was an "arsenic methyltransferase" (20), opened the way for studies of the possible contribution of variation in this gene (subsequently annotated as As3mt) to individual differences in arsenic methylation.
In the present study, we set out to characterize the human ortholog for the rat Cyt19-As3mt gene to make it possible to resequence the human gene using DNA from subjects of two different ethnic groups. During gene resequencing, we observed a series of genetic polymorphisms and haplotypes in the human AS3MT gene. Functional genomic studies of all variant allozymes encoded by AS3MT alleles that contained nonsynonymous cSNPs, as well as a VNTR located in the 5Ј-UTR of the cDNA, were then performed. These experiments represent a step toward studies of individual variation in arsenic methylation, arsenic pharmacogenetics, and arsenic toxicogenetics in humans.  HD100AA, and PRP00001), were obtained from the Coriell Cell Repository (Camden, NJ). The human DNA samples had been collected and anonymized by the NIGMS, National Institutes of Health. All subjects had provided written informed consent for the use of their DNA for research purposes, and the present studies were reviewed and approved by the Mayo Clinic Institutional Review Board. cDNA Cloning-The human AS3MT cDNA was cloned by PCR amplification based on an assumption of homology between the sequences of the rat As3mt cDNA and its human ortholog, using human liver cDNA as template. 5Ј-and 3Ј-rapid amplification of cDNA ends were performed with Marathon-Ready human liver and prostate cDNA (Clontech, Palo Alto, CA) using nested PCR amplifications with AS3MT-specific primers paired with the AP1 and AP2 primers supplied by the manufacturer. The sequences of primers used to perform these and subsequent amplifications are listed in supplementary materials Table 1.

EXPERIMENTAL PROCEDURES
Gene Resequencing-Annotation of the human AS3MT gene (originally designated CYT19, GenBank accession number NT_008804.8) was performed using the human cDNA sequence obtained by performing cDNA cloning. The PCR was then used to amplify each of the 11 AS3MT exons. We also attempted to amplify one exon using the primate DNA samples obtained from the Coriell Cell Repository as template. Primers were designed to amplify each of the human gene exons, plus portions of the introns flanking that exon. The primers used for the gene resequencing studies included M13 tags at their 5Ј-ends to make it possible to use dye primer sequencing chemistry (see supplementary materials Table 1). Dye primer sequencing was used to enhance our ability to identify heterozygous bases (22). Amplicons from these reactions were sequenced in the Mayo Molecular Biology Sequencing Core Facility with an ABI 3700 DNA sequencer (Applied Biosystems, Foster City, CA) using Big Dye (PerkinElmer Life Sciences) dye primer chemistry. All samples were sequenced on both strands, and those with ambiguous chromatograms, as well as samples that contained SNPs that were observed in only a single sample, were subjected to a second, independent round of amplification, followed by DNA sequencing. Samples resulting from amplifications of the initial exon that contained the VNTR were also subjected to agarose gel electrophoresis to determine VNTR length. Samples homozygous for the VNTR were sequenced directly, and selected heterozygous samples were subcloned into pCR2.1, followed by sequencing of each of the alleles present in that sample.
Transient Expression-The "wild type" (WT) open reading frame (ORF) of the human AS3MT cDNA was amplified, and site-directed mutagenesis was performed to create a synonymous cSNP that removed an EcoRI site. That ORF sequence was then cloned into pCR2.1 (Promega). The insert was excised with EcoRI and was cloned into the eukaryotic expression vector p91023(B), and "circular PCR" was used to perform site-directed mutagenesis to create constructs encoding variant allozymes (see supplementary materials Table 1 for sequences of these primers). Insert sequences after circular PCR were verified by sequencing the entire ORF. These expression constructs were then used to transfect COS-1 cells in serum-free Dulbeccos's modified Eagle's medium (BioWhittaker, Walkersville, MD) using the TransFast reagent (Promega) at a charge ratio of 1:1, with a 1-h transfection time. "Empty" p91023(B), vector that did not contain insert, was used as a control to make it possible to correct for possible endogenous AS3MT in COS-1 cells. However, endogenous activity was found to be negligible, averaging less than 1% of the activity present after transfection with the WT construct. During transfection, 15 g of AS3MT construct DNA was cotransfected with 1 g of pSV-␤-galactosidase DNA (Promega) to make it possible to correct for transfection efficiency. Six separate plates were transfected with each construct studied. The transfected COS-1 cells were incubated at 37°C for 48 h, washed with phosphate-buffered saline, and resuspended in 2 ml of homogenization buffer. They were then lysed with a Polytron homogenizer (Brinkmann Instruments), and the homogenates were centrifuged at 100,000 ϫ g for 1 h at 4°C. The resulting supernatant cytosol preparations were stored at Ϫ80°C prior to assay.
AS3MT Enzyme Assay-Recombinant allozymes were assayed for AS3MT activity using a modification of the method described by Zakharyan et al. (23). Briefly, 0.10 M Tris-HCl buffer, pH 8.0, 4 mM glutathione, 1 mM MgCl 2 , 12.5 mM sodium arsenite (the methyl acceptor substrate), [methyl-3 H]AdoMet, the methyl donor, (11.8 Ci/mmol, 0.55 mCi/ml, 10 M final concentration), and recombinant enzyme were combined in a final volume of 250 l. Blanks were samples that contained no methyl acceptor substrate. Reaction mixtures were incubated at 37°C for 60 min, and the reaction was terminated by the addition of 750 l of 12 M HCl. Methylated arsenic compounds were isolated using the organic solvent extraction procedure described by Zakharyan et al. (23). Enzyme activity was corrected for transfection efficiency by measuring ␤-galactosidase activity spectrophotometrically with the ␤-Galactosidase Enzyme Assay system (Promega). The same AS3MT activity assay was used to determine apparent K m values for the two cosubstrates for the reaction. Specifically, five concentrations of sodium arsenite that varied from 2.5 to 100 M were tested in the presence of 10 M [methyl-3 H]AdoMet. Activity was also measured in the presence of 12.5 M sodium arsenite with five concentrations of AdoMet that varied from 1.25 to 20 M.
Quantitative Western Blot Analyses-Levels of immunoreactive AS3MT protein were determined for each recombinant allozyme by performing quantitative Western blot analysis. A rabbit polyclonal antibody directed against AS3MT amino acids 341-360 was used to perform these studies. This peptide, linked to keyhole limpit hemocyanin, was used to generate a rabbit polyclonal antibody (Cocalico, Reamstown, PA). The antibody had been tested for specificity by performing Western blot analyses with recombinant AS3MT as well as human liver, kidney, and prostate cytosol preparations. During the Western blot analyses, COS-1 cell cytosol was loaded on 12% SDS mini-gels (Bio-Rad) in quantities that resulted in equal ␤-galactosidase activity to correct for possible variation in transfection efficiency. Electrophoresis was then performed for 1 h at 150 V, and the proteins were transferred to nitrocellulose membranes. The membranes were blocked overnight with 5% milk in Tris-buffered saline with Tween 20 (TBST). The following day, the membranes were incubated for 0.5 h with primary antibody diluted 1:2000 with 5% milk in TBST, followed by three washes. The secondary antibody was a 1:20,000 dilution of goat anti-rabbit horseradish peroxidase (Bio-Rad) that was applied for 1 h in 5% milk in TBST, followed by three washes. Bound antibody was detected by enhanced chemiluminescence performed with the ECL Western blotting system (Amersham Biosciences). The Western blot data were analyzed with the AMBIS radioanalytic Imaging System, Quant Probe version 4.31 (AMBIS, San Diego, CA). Multiple blots were performed for each allozyme, and immunoreactive protein levels were expressed as a percentage of the intensity of the band for the WT construct on the same gel.
Rabbit Reticulocyte Lysate (RRL) Protein Synthesis and Degradation-Transcription and translation of WT AS3MT and allozymes encoded by the variant alleles (Trp 173 , Thr 287 , and Ile 306 ) were performed using the TNT coupled RRL System (Promega) with constructs that had been cloned into pCR3.1 (Promega). Specifically, 25 l of "treated" RRL, plus 2 l of T7 buffer, 1 l of T7 polymerase, 1 l of RNasin, and 2 l of [ 35 S]methionine (1000 Ci/mmol, 10 mCi/ml, 0.4 M final concentration) were used to perform these experiments. With the exception of the RNasin (Promega) and the [ 35 S]methionine (Amersham Biosciences), all reagents were included in the Promega kit. One g of the pCR3.1 expression construct DNA was added to the mixture, the reaction volume was increased to 50 l with nuclease-free water (Promega), and the mixture was incubated at 30°C for 90 min. A 5-l aliquot was then used to perform electrophoresis with a 12% SDS-PAGE gel that was dried and exposed to x-ray film (Kodak).
To perform the protein degradation experiments, 60 l of an ATP generating system, 60 l of "untreated" RRL and 12 l of radioac-tively labeled protein were combined. The ATP generating system consisted of 100 l each of 1 M Tris-HCl, pH 7.8, 160 mM MgCl 2 , 120 mM KCl, 100 mM dithiothreitol, 100 mM ATP, 200 mM creatine phosphate, and 2 mg/ml ATP kinase (all from Sigma), plus 300 l of nuclease-free water. This mixture was incubated at 37 or 40°C. Aliquots were removed at various times and were subjected to electrophoresis on a 12% SDS-PAGE gel. A variant allozyme for a genetically polymorphic drug metabolizing enzyme that has been shown to be degraded rapidly by a proteasome-dependent process, thiopurine S-methyltransferase (TPMT)*3A (24), was used as a positive control for these experiments. The gels were dried and exposed to x-ray film, and levels of [ 35 S]methionine radioactively labeled proteins were determined by using the AMBIS system.
Reporter Gene Constructs-Firefly luciferase reporter gene constructs in pGL3-Basic were created for the four different length AS3MT 5Ј-flanking region and 5Ј-UTR VNTR sequences. Inserts in these constructs were then sequenced, and the constructs were used to transfect HepG2 and HEK293 cells using the TransFast reagent at a charge ratio of 1:1, with a 1-h transfection time. During transfection, 2 g of construct DNA was cotransfected with 0.2 g of pRL-TK DNA to make it possible to correct for transfection efficiency. Transfected cells were incubated at 37°C for 48 h, washed with phosphate-buffered saline, and lysed in 1 ml of lysis buffer. Cell lysates were then used to measure firefly luciferase activity with the Dual Luciferase Reporter Assay system (Promega).
Data Analysis-The DNA sequence from the resequencing studies was analyzed using the PolyPhred 4.0 (25) and Consed 8.0 (26) programs. The Wisconsin Genetics Computer Group (GCG) package, version 10, was also used to analyze nucleotide sequence. Human SNP data bases, including dbSNP and the HapMap, build 18, were searched to determine whether the polymorphisms that we observed had been reported previously. The RepeatMasker (University of Washington) program was used to screen for repeat sequences. Values for , , and D values were calculated as described by Tajima (27). Linkage disequilibrium was analyzed by calculating DЈ values for all polymorphism pairs as described by Hartl and Clark (28) and Hendrick (29), and those data were displayed graphically. Haplotype analysis was performed as described by Schaid et al. (30). Apparent K m values were calculated with the method of Wilkinson (31) using a computer program developed by Cleland (32). Differences between mean values were evaluated by use of the Student's t test calculated with the Statview 4.5 program (Abacus Concepts, Berkeley, CA).

RESULTS
Human AS3MT Resequencing-The human AS3MT cDNA sequence, as well as that of the encoded protein, are shown in Fig. 1. The cDNA had a 1125-bp ORF that encoded a 375-amino acid protein. The ORF nucleotide and encoded amino acid sequences were 81 and 76% identical with those of rat As3mt, respectively. The human AS3MT gene mapped to chromosome 10q24, contained 11 exons, and was ϳ32 kb in length (Fig. 2). The location of the human gene was syntenic with the rat ortholog that maps to rat chromosome 1q54, and with the location of the mouse ortholog on chromosome 19D1. It should be pointed out that the sequences of the cDNA and, as a result, the gene that we have shown in Figs. 1 and 2 differ from the current NCBI annotation for the cDNA (NM_020682.2), an annotation that is based on the application of exon prediction software to an early version of the human chromosome 10 sequence. Specifically, ORF nucleotide 1057 in our sequence is missing from the NCBI cDNA annotation, resulting in a frameshift and truncation of the protein within exon 10, with exon 11 converted entirely to 3Ј-UTR. Furthermore, there are 4 variant nucleotides in exon 5 of the present NCBI cDNA annotation, one synonymous and three resulting in the following alterations in encoded amino acids: I132F, Y135N, and G140A (33). Those nucleotides were not polymorphic in any of our resequencing studies of 240 alleles, and the deletion at nucleotide 1057 was also not present in any of our 120 DNA samples. As a result, we concluded that the current NCBI annotation, which is labeled "provisional," is almost certainly in error at those locations. Of note is the fact that the current NCBI genomic sequence for AS3MT matches our WT sequence and that the encoded protein in SwissProt is that which we have shown in Fig. 1.
AS3MT was resequenced using DNA samples from 60 AA and 60 CA subjects. Eleven PCR amplifications were performed for each sample and, as a result, a total of over 1,300,000 bp of DNA sequence was analyzed on both strands. Those experiments resulted in the identification of 26 SNPs, 22 in DNA samples from AA and 21 in DNA from CA subjects. The locations of the polymorphisms that we observed are listed in Table 1 and are depicted graphically in Fig. 2. All polymorphisms were in Hardy-Weinberg equilibrium. Three of the SNPs were nonsynonymous and altered encoded amino acids in the following codons: R173W, M287T, and T306I. The frequency of the polymorphism resulting in the M287T alteration in sequence was 10% or greater in both populations studied. This polymorphism was located in exon 9 of human AS3MT. Therefore, exon 9 was also amplified in the 10 primate DNA samples that we had obtained (rhesus monkey, pigtailed macaque, bonobo, gorilla, chimpanzee, Sumatian orangutan, redchested mustached tamarin, black-handed spider monkey, common woolly monkey, and ring-tailed lemur). This codon encoded Thr in all 10 of these primate DNA samples, Ile in the rat, mouse, and dog orthologs, and Glu in the chicken gene. Twenty-one of the 26 human AS3MT SNPs had frequencies of greater than 1% in at least one of the Amino acids altered as a result of nonsynonymous cSNPs are also indicated. "VNTR" refers to a variable number of tandem repeats located in the 5Ј-UTR (see Fig. 3).
two ethnic groups and, as a result, would be considered "common" in these populations. An additional polymorphism involved a VNTR in exon 1 within an area encoding the 5Ј-UTR. This VNTR consisted of repeat elements that were either 35 or 36 bp in length, differing only in the loss of the 3Ј-terminal "a," with 2 to 4 repeats in each allele (Fig. 3). The repeat consisted of 1 to 3 copies of the 36-bp "A" sequence, plus one copy of the 35-bp "B" sequence, with the B repeat at the 3Ј-terminus of the VNTR (Fig. 3). Four repeat elements (three A and one B sequence) were observed only in AA subjects. Allele and genotype frequencies for the VNTR in both populations are listed in Table 2. When public data bases, including dbSNP, were searched, 7 of the AS3MT polymorphisms that we observed were already in data bases. The HapMap, build 18, included only 3 of the polymorphisms that we observed. Our AS3MT resequencing data have been deposited in the NIH data base PharmGKB, with accession number PA128747780.
We calculated "nucleotide diversity," a quantitative measure of genetic variation, adjusted for the number of alleles studied, for these data. Two commonly used measures of nucleotide diversity are , average heterozygosity per site, and , a population mutation measure that is theoretically equal to the neutral mutation parameter (34). In the samples that we studied, was 9.16 Ϯ 4.96 ϫ 10 Ϫ4 for AA and 8.24 Ϯ 4.52 ϫ 10 Ϫ4 for CA subjects, whereas was 7.82 Ϯ 2.40 ϫ 10 Ϫ4 for AA and 7.46 Ϯ 2.32 ϫ 10 Ϫ4 for CA subjects. Therefore, the D values of Tajima (27) for AA and CA subjects were 0.51 and 0.30, respectively. These values were similar to those reported by Stephens et al. (35) for 292 human autosomal genes (average ϭ 5.8 ϫ 10 Ϫ4 and average ϭ 9.6 ϫ 10 Ϫ4 ).
Linkage Disequilibrium and Haplotype Analysis-Haplotype information may be more useful than individual polymorphisms for application in association studies (36,37). Therefore, we also performed linkage disequilibrium and haplotype analyses for the 26 SNPs and the VNTR that we had observed in AS3MT. For the linkage disequilibrium analysis, ͉DЈ͉ values were calculated for all pairwise combinations of SNPs. ͉DЈ͉ values can range from 1.0 when two polymorphisms are maximally Locations and frequencies of the 26 SNPs and one VNTR observed in the human AS3MT gene are listed. Nucleotide locations, except those in introns, have been numbered on the basis of the A in the translation initiation codon, with that nucleotide designated (ϩ1), and with positive numbers 3Ј and negative numbers 5Ј to that location. Intron (IVS) nucleotides have been numbered relative to splice junctions, with the initial 5Ј nucleotide in an intron designated (ϩ1) and the final 3Ј nucleotide designated (Ϫ1). Polymorphisms in exons have been "boxed." The numbers at the right are those used to designate polymorphisms for the linkage disequilibrium analysis. Asterisk, polymorphism present in dbSNP. associated, to zero when they are randomly associated (28,29). Ninety pairs of SNPs for DNA samples from AA subjects and 56 pairs for CA subjects had ͉DЈ͉ values Ն0.8, with p values Ͻ 0.05. Those linkage disequilibrium data are displayed graphically in Fig. 4.
Haplotypes can be determined unequivocally only if no more than one polymorphism in an allele is heterozygous, but it is also possible to infer haplotypes computationally (30). There were 6 unequivocal AS3MT haplotypes for AA subjects and 4 for CA subjects on the basis of our resequencing data, as well as 29 inferred haplotypes for AA subjects and 26 for CA subjects. We "named" haplotypes based on the amino acid sequence of the encoded allozyme, with the WT sequence designated as *1. We then designated the relatively common Thr 287 variant allozymes as *2, the Thr 287 /Ile 306 variant as *3 (Ile 306 was only observed in the presence of Thr 287 ), and the Trp 173 variant allozyme as *4, based on allele frequency. Letter designations were then added to the numerical designations on the basis of descending allele frequencies, starting with the AA data. For example, *1A was more frequent than *1B and both were more frequent than *1C. All haplotypes, both unequivocal and inferred, with frequencies of greater than 1% in either populations are listed in Table 3.
AS3MT Allozyme Activity and Protein Levels-Functional genomic studies were performed after transient expression of the WT AS3MT sequence and the four variant allozymes (three with only a single variant amino acid and one, Thr 287 /Ile 306 , with two) in COS-1 cells. A mammalian cell line was used to perform these experiments to ensure that appropriate mammalian post-translational modification and protein degradation systems would be present. There have now been many reports that have demonstrated that the alteration of only one or two amino acids as a result of common genetic polymorphisms can be associated with significant changes in the level of protein, often as a result of accelerated variant allozyme degradation through a proteasome-mediated process (24, 38 -41). After correction for transfection efficiency, the Trp 173 variant displayed 31 Ϯ 2.6% (mean Ϯ S.E., n ϭ 6), the Thr 287 variant 350 Ϯ 89%, the Ile 306 variant 4.8 Ϯ 2.1%, and the Thr 287 /Ile 306 variant 6.2 Ϯ 1.9% of the activity of WT AS3MT with sodium arsenite as a substrate (Fig. 5A and Table 4). One way in which alterations in the encoded amino acid might result in decreased enzyme activity would involve changes in substrate kinetics. Therefore, we also determined apparent K m values for the WT and variant AS3MT allozymes (Table 4). It was not possible to calculate apparent K m values for the Ile 306 or Thr 287 /Ile 306 variant allozymes because they displayed such low levels of activity. The only K m values that differed significantly from that for the WT were those for the Thr 287 allozyme, with an apparent K m for sodium arsenite that was double the WT value and an apparent K m for AdoMet that was half of the WT value. Although statistically significant, these differences may not have major biological consequences. Therefore, we also measured the level of immunoreactive protein for WT and the four variant allozymes.
Quantitative Western analysis, corrected for transfection efficiency, showed that differences in levels of immunoreactive protein paralleled the variations that we observed in levels of AS3MT enzyme activity (Table 4 and Fig. 5, B and C). The antibody used to detect AS3MT in these experiments was directed against amino acids 341-360 (see Fig.  1), an area that did not include any of the amino acids that were altered in the variant allozymes. This series of experiments demonstrated the following consequences of these naturally occurring alterations in AS3MT amino acid sequence: 1) a marked increase in level of immunoreactive protein for the Thr 287 allozyme and 2) marked decreases in levels of immunoreactive protein for the Trp 173 , Ile 306 , and Thr 287 /Ile 306 allozymes. Decreased levels of encoded protein have been observed for other genetically polymorphic enzymes with alteration in only one or two encoded amino acids, and those decreases often result from accelerated degradation of the variant allozyme (24, 38 -40). Therefore, in the next set of experiments, we attempted to determine whether rapid

TABLE 3 Human AS3MT haplotypes
All haplotypes with frequencies of 1% or greater in at least one population are listed. Frequencies for both populations are listed. Frequencies that are in italics, bold type with shaded background indicate that the haplotype was "observed," while all others were "inferred." The "wild type" for each position is indicated in bold type with shaded background. White letters on a black background indicate that an amino acid change resulted from that polymorphism. I, "intron" (IVS), E, exon. degradation might contribute to the decreased levels of Thr 173 , Ile 306 , and the Thr 287 /Ile 306 AS3MT allozymes.
AS3MT Degradation-The degradation of WT AS3MT and the three variant allozymes showing decreased levels of protein (Trp 173 , Ile 306 , and Thr 287 /Ile 306 ) were studied in the RRL, an experimental system that is often employed to perform this type of experiment (24, 38 -42). As a first step, each allozyme, plus the human TPMT*3A allozyme, was translated using the TNT RRL system (Promega). The TPMT*3A allozyme is degraded very rapidly in the RRL (24) and was included as a "positive control" for rapid degradation. The [ 35 S]methionine radioactively labeled proteins generated during in vitro translation were added to an untreated RRL that included an ATP generating system. These mixtures were incubated at 37°C for 24 h, and aliquots were removed at regular intervals to perform SDS-PAGE. However, during this time period, no significant differences were observed among the AS3MT allozymes (data not shown). A second set of samples was then incubated at a slightly higher temperature, 40°C, to determine whether the WT, Trp 173 , Ile 306 , or Thr 287 /Ile 306 allozymes might degrade more rapidly than did the Thr 287 protein, the allozyme with the highest level of immunoreactive protein. Those experiments were each performed three separate times but, once again, no significant differences were observed among allozymes in rates of degradation. In all cases, the control human TPMT*3A allozyme displayed the anticipated rapid degradation, indicating that the assay system was functioning properly. Therefore, we concluded that large differences in proteasomemediated degradation did not appear to be responsible for the striking differences in levels of AS3MT immunoreactive protein observed after the transient transfection of COS-1 cells.
AS3MT Reporter Gene Studies-AS3MT includes a VNTR that begins 5Ј of exon 1 and encompasses that entire exon (Fig. 2). To determine whether this VNTR might influence transcription, luciferase reporter gene constructs were generated for the AS3MT 5Ј-flanking region and 5Ј-UTR in an area that included the VNTR. The VNTR contained A and B sequences, and we observed alleles with one, two, or three copies of the A sequence, always joined to a single copy of B (Fig.  2). In addition, the *V2 sequence (one A and one B repeat) was in linkage disequilibrium with a SNP at position (Ϫ114), so constructs were created that did and did not include this SNP. All of these constructs were expressed in both HepG2 and HEK293 cells, cell lines that expressed AS3MT on the basis of Western blots (data not shown). The results demonstrated cell line-dependent variation in transcription (Fig. 6). In HepG2 cells, shorter repeat length was associated with enhanced reporter gene transcription (Fig. 6). The common SNP at nucleotide (Ϫ114) that was linked with the two repeat allele did not appear to significantly alter transcription.

DISCUSSION
Methylation of inorganic arsenic to form methylarsonic acid and dimethylarsinic acid is an important reaction in arsenic biotransformation (7,10). Because of the importance of this metabolic reaction in humans, we set out to determine whether AS3MT, like many other human methyltransferase genes (24,41,(43)(44)(45), might include common, functionally significant genetic polymorphisms, polymorphisms that might contribute to risk for arsenic-dependent carcinogenesis or for individual variation in response to treatment with arsenic-containing therapeutic agents (8,9). A genotype-to-phenotype experimental strategy was used to test that hypothesis. The first step included annotation of the AS3MT gene, resulting in several differences from the current "provisional" NCBI annotation. We then resequenced AS3MT in 60 DNA samples from AA and 60 samples from CA subjects. The resequencing studies confirmed that the current provisional NCBI annotation is almost certainly incorrect. In addition, 27 polymorphisms, including three nonsynonymous cSNPs and a VNTR, were identified during the gene resequencing experiments. A mammalian expression system was then used to express the WT and variant AS3MT allozymes, including a "double variant" allozyme. There were striking differences in levels of enzyme activity and immunoreactive protein among these variant allozymes (Table 4 and Fig. 5). For the two "infrequent" allozymes (Trp 173 and Ile 306 ), levels of enzyme activity and immunoreactive protein were strikingly decreased when compared with the WT allozyme. However, the Thr 287 variant, encoded by an allele with a frequency of ϳ10% in both the AA and CA DNA (Table 1), demonstrated increased levels of both enzyme activity and immunoreactive protein. Furthermore, the human WT amino acid at codon 287, Met, was not encoded by any of the other 10 primate DNA samples studied, all of which encoded the human variant amino acid, Thr, implying that the mutation resulting in the sequence encoding the human WT amino acid at this position had occurred very recently in primate evolution. The possible role of natural selection in this alteration in sequence for a major arsenic-metabolizing enzyme in humans remains speculative, but should be pursued in the future.
Many proteins encoded by genes containing common nonsynonymous cSNPs that alter one or two amino acids display patterns of alteration in levels of enzyme activity and protein similar to those observed here for AS3MT (24,38,39,41,(43)(44)(45)(46)(47). Furthermore, in all of those situations in which the experiments have been performed, results obtained after transient expression in COS-1 cells directly parallel relative levels of expression in human tissue biopsy samples (44 -46, 48). In addition, in those cases that have been studied in detail, accelerated degradation of the variant protein has been the most frequent explanation for striking reductions in protein level such as those that we observed for Ile 306 (26, 38 -41, 47). Therefore, we tested the possibility that the Trp 173 , Ile 306 , and Thr 287 /Ile 306 variant AS3MT allozymes might be degraded more rapidly than the WT or Thr 287 allozymes. Those experiments were performed using the RRL, an experimental system that has been used frequently to perform protein degradation studies (24,38,39,41,42,47). However, we failed to observe significant alterations in degradation for any of the variant allozymes. We also studied the possibility that the AS3MT VNTR might influence transcription and observed cell-line dependent variation in transcription, with the association of shorter repeat length (*V2) with increased transcription in HepG2 cells. Obviously these preliminary experiments will require replication and pursuit in the course of future experiments.
Finally, one of the striking observations made in the course of these experiments involved the increased levels of both enzyme activity and protein for the common Thr 287 allozyme when compared with WT AS3MT. Because the frequency of this polymorphism is ϳ10% in both AA and CA subjects, ϳ1% of subjects in both populations would be expected to be homozygous for this allele, might express elevated levels of AS3MT, and perhaps, display increased arsenic methylation. Those subjects might also be at increased risk for the cytoxic and genotoxic effects of arsenic exposure if, as appears to be the case, methylation enhances arsenic toxicity (12)(13)(14). In summary, these genomic and  functional genomic results will make it possible to test the hypothesis that sequence variation in AS3MT might contribute to individual variations in risk for arsenic toxicity, arsenic-dependent carcinogenesis, or response to therapy with arsenic-containing drugs.