Human Sulfotransferase SULT2A1 Pharmacogenetics: Genotype-To-Phenotype Studies

Human sulfotransferase SULT2A1 pharmacogenetics: genotype-to-phenotype studies

BA Thomae1 ABSTRACT BW Eckloff2 SULT2A1 catalyzes the sulfate conjugation of dehydroepiandrosterone 1 (DHEA) as well as other steroids. As a step toward pharmacogenetic studies, RR Freimuth we have ‘resequenced’ SULT2A1 using 60 DNA samples from African-Amer- ED Wieben2 ican and 60 samples from Caucasian-American subjects. All exons, splice RM Weinshilboum1 junctions and approximately 370 bp located 5Ј of the site of transcription initiation were sequenced. We observed 15 single nucleotide polymorphisms 1Department of Molecular Pharmacology and (SNPs), including three non-synonymous coding SNPs (cSNPs) that were Experimental Therapeutics, Mayo Medical present only in DNA from African-American subjects. Linkage analysis School-Mayo Clinic-Mayo Foundation, Rochester, MN, USA; 2Department of revealed that two of the nonsynonymous cSNPs were tightly linked. Biochemistry and Molecular Biology, Mayo Expression constructs were created for all nonsynonymous cSNPs observed, Medical School-Mayo Clinic-Mayo Foundation, including a ‘double variant’ construct that included the two linked cSNPs, Rochester, MN, USA and those constructs were expressed in COS-1 cells. SULT2A1 activity was significantly decreased for three of the four variant allozymes. Western blot Correspondence: R Weinshilboum, Department of analysis demonstrated that decreased levels of immunoreactive protein Molecular Pharmacology and appeared to be the major mechanism responsible for decreases in activity, Experimental Therapeutics, Mayo Medical although apparent Km values also varied among the recombinant allozymes. School-Mayo Clinic-Mayo Foundation, In addition, the most common of the nonsynonymous cSNPs disrupted the Rochester, MN 55905, USA Tel: (507) 284 2246 portion of SULT2A1 involved with dimerization, and this variant allozyme Fax: (507) 284 9111 behaved as a monomer rather than a dimer during gel filtration chromato- E-mail: weinshilboum.richardȰmayo.edu graphy. These observations indicate that common genetic polymorphisms for SULT2A1 can result in reductions in levels of both activity and enzyme protein. They also raise the possibility of ethnic-specific pharmacogenetic variation in SULT2A1-catalyzed sulfation of both endogenous and exogenous substrates for this phase II drug-metabolizing enzyme. The Pharmacogenomics Journal (2002) 2, 48–56. DOI: 10.1038/sj/tpj/ 6500089

Keywords: sulfotransferase; SULT2A1; sulfation; pharmacogenetics; genetic polymorphism; dehydroepiandrosterone; SNP

INTRODUCTION Sulfate conjugation is an important pathway in the biotransformation of many drugs, other xenobiotics, neurotransmitters and hormones—especially steroid hormones.1,2 Sulfation or, more accurately, sulfonation of these compounds is catalyzed by cytosolic sulfotransferase (SULT) enzymes that usually exist as homodimers.2 SULT2A1 is a member of this family of enzymes. It catalyzes the sulfate conjugation of dehydroepiandrosterone (DHEA) and other structurally- related steroids (Figure 1)3,4 and has been reported to contribute to the metabolic activation of procarcinogens.5,6 SULT2A1 is highly expressed in the human liver, small intestine and adrenal cortex.7,8 Therefore, orally administered substrates Received: 11 September 2001 would undergo sulfate conjugation catalyzed by SULT2A1. In addition, circulat- Revised: 6 December 2001 ing endogenous levels of DHEA sulfate decrease with increasing age9 and are Accepted: 26 December 2001 inﬂuenced by inheritance.10 In the United States, DHEA is marketed over-the- SULT2A1 pharmacogenetics BA Thomae et al 49

(Figure 2 and Table 1). DNA samples from African-American subjects contained 14 SNPs, and DNA from Caucasian- American subjects contained six SNPs—only one of which was not also present in DNA from African-American subjects. No insertion-deletion events were observed within the area resequenced. Ten of the 14 SNPs in African-American and all six of the SNPs in DNA from Caucasian-American subjects had frequencies of greater than 1% and would, therefore, be considered ‘common’ in these populations. Ј Figure 1 Human SULT2A1-catalyzed DHEA sulfation. PAPS is 3 - SNPs seen in only a single sample were verified by per- phosphoadenosine 5Ј-phosphosulfate, the sulfate donor for the forming a separate, independent amplification, followed by reaction, and PAP is 3Ј-phosphoadenosine 5Ј-phosphate. DNA sequencing. We observed three nonsynonymous cSNPs, all of which were present only in DNA from African- American subjects. The nonsynonymous cSNPs were counter in health food stores as a dietary supplement ‘to G187C, A679G and G781A, which changed amino acids Ala- counter the effects of aging’.11,12 Pharmacogenetic variation 63Pro, Lys227Glu and Ala261Thr, respectively. Only one in SULT2A1 activity could, therefore, influence the SNP was observed in the 5Ј-flanking region at position biotransformation of both orally administered agents and/or (Ϫ122). That SNP had an allele frequency of 0.042 in the endogenous substrates. African-American subjects and was located in a putative AP1 Several years ago we set out to study SULT2A1 pharmaco- binding site, but the polymorphism did not disrupt the AP1 genetics. As a first step, we demonstrated the presence of recognition motif. Nucleotide positions for all SNPs were large individual variations in human liver ‘DHEA sul- numbered with respect to the ‘A’ in the ATG translation fotransferase’ activity—an activity catalyzed primarily by initiation codon. Positions 5Ј to that location were assigned SULT2A1.13 Subsequently, we cloned the cDNA and gene negative, and positions 3Ј were assigned positive numbers. that encode human SULT2A1.7,14 The human gene was A search of publicly accessible DNA sequence databases, approximately 17 kb in length, consisted of six exons and including the EST database (www.ncbi.nlm.nih.gov/dbEST), mapped to chromosome 19q13.3.14–17 In the present experi- dbSNP (www.ncbi.nlm.nih.gov/SNP), public Human Gen- ments, we have used a genotype-to-phenotype strategy to ome Project sequence and all available SNP databases listed test the hypothesis that SULT2A1, like other genes encoding in a recent summary,22 showed that none of the SNPs that human SULT enzymes,18–20 might contain common, func- we had observed were present in public databases. The tionally significant genetic polymorphisms. Specifically, we SULT2A1 SNPs which were listed in those databases were resequenced SULT2A1 using DNA from 60 Caucasian-Amer- located in portions of the gene—eg, introns—that we had ican and 60 African-American subjects and observed 15 polymorphisms, including three novel cSNPs that altered the encoded amino acid—all three of which were present only in DNA from African-American subjects. Expression constructs were created for all of these nonsynonymous cSNPs, including one double variant allele, followed by functional genomic studies of the encoded recombinant allozymes. Three of the four variant allozymes displayed significant decreases in level of activity as compared with the wild-type enzyme and one of them altered the ability of the enzyme to dimerize. These results raise the possibility of ethnic-specific pharmacogenetic variation in sulfate conjugation catalyzed by SULT2A1.

RESULTS Figure 2 Human SULT2A1 SNPs. The ﬁgure shows a schematic rep- SULT2A1 Resequencing resentation of the human SULT2A1 gene with the locations of SNPs SULT2A1 was resequenced by performing seven PCR reac- indicated by arrows. Black rectangles represent the open reading tions for each of the 120 DNA samples studied. A total of frame and white rectangles represent portions of exons that over 600000 bp of DNA sequence was analyzed, and 98.7% encode untranslated region (UTR) sequence. ‘AA’ represents data obtained withDNA from African-American subjects and ‘CA’ rep- of that DNA was sequenced on both strands. Dye primer resents results obtained withDNA from Caucasian-American sub- sequencing chemistry was used to enhance the detection of jects. Changes in encoded amino acids resulting from nonsynony- 21 heterozygous nucleotides. These data were compared to mous cSNPs are also indicated. Colors of arrows indicate allele the SULT2A1 sequence deposited in GenBank at the time frequencies in the two populations studied. Red arrows represent that the gene was originally cloned and characterized frequencies of greater than 10%; blue arrows frequencies from 1% (accession numbers U13056–U13061).14 Fifteen SNPs were to 10% and black arrows represent polymorphisms with fre- observed in the 120 DNA samples (240 alleles) studied quencies of less than 1%.

www.nature.com/tpj SULT2A1 pharmacogenetics BA Thomae et al 50

Table 1 Human SULT2A1 genetic polymorphisms

Nucleotide location Location of SNP Nucleotide change Amino acid change Frequency of variant allele in the gene African-American Caucasian-American subjects subjects

5Ј-FR (Ϫ122) G → A 0.042 0.000 Exon 1 6 G → A 0.017 0.000 Exon 1 90 C → T 0.067 0.000 Intron 1 I1(16) C → T 0.033 0.167 Exon 2 187 G → C Ala 63 Pro 0.050 0.000 Exon 2 258 A → C 0.008 0.000 Intron 2 I2(34) G → A 0.100 0.017 Intron 3 I3(57) C → T 0.008 0.000 Intron 3 I3(Ϫ129) G → A 0.050 0.083 Intron 3 I3(Ϫ94) C → G 0.092 0.283 Intron 4 I4(67) C → T 0.008 0.033 Exon 5 679 A → G Lys 227 Glu 0.008 0.000 Exon 6 781 G → A Ala 261 Thr 0.133 0.000 Exon 6 924 T → C 0.000 0.017 Exon 6 935 A → T 0.058 0.000

The table lists the 15 polymorphisms observed during the SULT2A1 resequencing experiments. Polymorphisms within exons have been ‘boxed’. Frequencies for each variant allele in the two population samples studied are listed. ‘5Ј-FR’ represents the 5Ј-flanking region of the gene, and ‘I’ indicates intron. The numbering scheme for nucleotides located in exons and in the 5Ј-flanking region was based on the cDNA sequence, with the ‘A’ in the translation initiation codon designated ϩ1. Positions 5Ј to that location were assigned negative numbers, while those 3Ј to that position were assigned positive numbers. Intron-based primers were numbered on the basis of nucleotide distance from splice junctions, with ϩ1 designated as the first nucleotide at the 5Ј-end of an intron, while (Ϫ1) was the first nucleotide at the 3Ј-end of an intron. not resequenced. Finally, in a recently published study from Therefore, one common haplotype in samples from African- Japan, a series of SULT enzymes were resequenced using American DNA subjects included all three of these SNPs. DNA from 48 healthy Japanese subjects. The SULT2A1 data That fact was of practical importance when expression con- included nine SNPs that were located in introns within areas structs were created to study the functional implications of of the gene that we did not resequence in our experiments, SULT2A1 cSNPs. but none of the 15 SNPs that we observed in African-Amer- ican and Caucasian-American subjects were present in these Recombinant Allozyme Enzyme Activity DNA samples from Japanese subjects.23 Our functional genomic studies focussed on SULT2A1 alleles It has been suggested that haplotype may ultimately prove that contained nonsynonymous cSNPs. Specifically, to be at least as important for understanding function as the expression constructs that encoded each of the three analysis of individual polymorphisms.24 Therefore, we also SULT2A1 variant amino acid sequences, as well as the dou- performed linkage analysis for all SULT2A1 SNPs that were ble variant at codons 63 and 261 (SNPs at ORF nucleotides observed more than once. Specifically, DЈ values were calcu- 187 and 781), were expressed in COS-1 cells. A mammalian lated for all pairwise combinations of SNPs. DЈ values can expression system was used to ensure appropriate post- range from ϩ1.0 when two polymorphisms are maximally translational modification as well as the presence of mam- positively associated, to Ϫ1.0 when two polymorphisms malian protein degradation systems. There have now been never occur together.25,26 Only seven pairs of SNPs had stat- several reports of situations in which the change of a single istically significant Chi-square values (Ͼ3.84) for DЈ, and amino acid has resulted in significant decreases in levels of only the three polymorphisms at nucleotides (Ϫ122), 187 immunoreactive protein—due in some cases to alterations and 781 had absolute DЈ values which were greater than 0.3. in proteosome-mediated protein degradation.20,27–29 Three In all three of those cases, DЈ was 1.0—ie, all three of these of the four recombinant SULT2A1 allozymes had signifi- SNPs were highly linked, even though one was located in cantly decreased enzyme activity, after correction for trans- the 5Ј-flanking region, one was in exon 2 and the final SNP fection efficiency, as compared to that of the ‘wild-type’ was in exon 6. That is, they were spread over a total length sequence (Figure 3a and Table 2). The values shown graphi- of approximately 14.7 kb. These results were of special inter- cally in Figure 3a are averages of those observed during six est because the SNPs at nucleotides 187 and 781 were non- independent transfections. Transfection with the construct synonymous—with frequencies in DNA samples from that contained the Ala63Pro change in sequence, a polymor- African-American subjects of 0.05 and 0.137, respectively. phism with a frequency of 0.05 in African-American sub-

The Pharmacogenomics Journal SULT2A1 pharmacogenetics BA Thomae et al 51

the African-American subjects studied, enzyme activity was decreased to 42% of the wild-type level (Figure 3a and Table 2). However, activity of the Ala261Thr variant alone—an allozyme encoded by approximately 8% of all alleles in African-American subjects—did not differ signiﬁcantly from that of the wild-type enzyme. The Lys227Glu polymorphism, with a frequency of 0.008 in DNA samples from African-Americans, resulted in the largest decrease in activity. Lysates from COS-1 cells transfected with that construct had only 15% of the wild-type activity (Figure 3a and Table 2). The next step in this series of experiments involved studies of mechanisms responsible for differences in levels of activity among these recombinant SULT2A1 allozymes.

Recombinant Allozyme Substrate Kinetics One way in which a change in amino acid sequence can affect enzyme activity is by an alteration of kinetic proper-

ties. Therefore, we determined apparent Km values for both DHEA and the sulfate donor cosubstrate, 3Ј-phosphoadenosine 5Ј-phosphosulfate (PAPS), for all of the recombinant SULT2A1 allozymes. Because SULTs, including SULT2A1, can show profound substrate inhibition,1 experiments involving the sulfate acceptor cosubstrate were conducted in two stages. Initially, a wide range of concentrations was

tested—followed by a determination of Km values within a narrower range of concentrations. Although there were stat- istically signiﬁcant variations among the allozymes in

apparent Km values for both DHEA and PAPS, those vari- Figure 3 Recombinant human SULT2A1 allozyme activity and ations were not large quantitatively and—in the case of Western blot analysis. (a) Average levels of enzyme activity are PAPS—the apparent K values for three of the variant allo- shown for each of the SULT2A1 recombinant allozymes with DHEA m zymes were significantly lower than that of the wild-type as the sulfate acceptor substrate. All values have been corrected for transfection efficiency. Eachbar represents theaverage of six enzyme (Table 2). As a result, decreases in level of enzyme independent transfections (mean Ϯ SEM). * Indicates P Ͻ 0.01 activity observed for three of the four variant allozymes when compared with the wild-type construct. (b) SULT2A1 recom- could not be attributed entirely to alterations in substrate binant allozyme Western blot analysis. The gel was loaded on the affinity. Therefore, our next experiments involved an basis of ␤-galactosidase activity to correct for variations in transfec- attempt to determine whether decreases in activity might be tion efficiency. See text for details. due to alterations in levels of SULT2A1 immunoreactive protein.

jects, resulted in only 57% of the average activity present Recombinant Allozyme Western Blot Analysis after transfection with the wild-type sequence. When the Quantitative Western blot analysis showed that levels of Ala63Pro polymorphism was combined with Ala261Thr, immunoreactive SULT2A1, corrected for transfection thus creating the double variant that occurred commonly in efﬁciency, paralleled relative levels of enzyme activity

Table 2 Human SULT2A1 functional genomic studies

␮ Construct Enzyme activity, Immunoreactive Ratio of enzyme DHEA, Km, M PAPS, Km,nM % protein, % activity to immunoreactive protein

Wild-type 100 100 1.0 0.247 Ϯ 0.071 21.8 Ϯ 0.86 Ala63Pro 57 Ϯ 7* 27 Ϯ 9* 2.1 0.210 Ϯ 0.010 8.2 Ϯ 1.32* Lys227Glu 15 Ϯ 3* 2 Ϯ 1* 7.5 0.905 Ϯ 0.053* 10.2 Ϯ 0.96* Ala261Thr 93 Ϯ 679Ϯ 7 1.2 0.245 Ϯ 0.012 22.7 Ϯ 0.62 Ala63Pro/Ala261Thr 42 Ϯ 2* 26 Ϯ 14* 1.6 0.290 Ϯ 0.012 11.4 Ϯ 0.30*

The table lists average levels of activity for recombinant SULT2A1 allozymes after six independent transfections, as well as average levels of immunoreactive protein = = ± determined by Western blot analysis (n 3), ratios of these two values and apparent Km values (n 3). All values are mean SEM. * Values that differ signiﬁcantly from that for the wild-type allozyme (P Ͻ 0.05).

www.nature.com/tpj SULT2A1 pharmacogenetics BA Thomae et al 52

among the variant allozymes (Table 2). Western blot analysis data were obtained by assaying cytosol for each allozyme from three independent transfections performed on differ- ent days to ensure that results were reproducible. The data listed in Table 2 are averages of those three transfections. The table also lists the ratio of enzymatic activity to immunoreactive protein for each allozyme. A representative West- ern blot gel is shown in Figure 3b. It should be noted that none of the alterations in encoded amino acid for the variant allozymes studied were located within the sequence of the peptide used to generate the rabbit polyclonal antibody utilized to perform the Western blots.30 These studies showed that alterations in the amino acid sequence of SULT2A1 as a result of genetic polymorphisms, as has been reported for other proteins,20,27,28 significantly altered levels of enzyme protein. Figure 4 Recombinant human SULT2A1 gel filtration chromato- SULT2A1 Crystal Structure and Dimerization graphy. The figure shows the results of SULT2A1 assays performed The x-ray crystal structure of SULT2A1 has been solved at a withSephadexG-100 column fractions using DHEA as a substrate resolution of 2.3 Å.31 As a result, we were able to determine for recombinant ‘wild-type’ and Thr261 variant allozymes. The the locations of polymorphic SULT2A1 amino acids within arrows show the elution from the column of standards: (1) Dextran that structure. The polymorphism that resulted in the Blue 2000; (2) bovine serum albumin (67 kDa); (3) ovalbumin Ala63Pro change in amino acid was located 14 residues (45 kDa); (4) chymotrypsinogen A (25 kDa); and (5) ribonuclease A downstream from the conserved ‘Region I’ sequence motif (13.7 kDa). Vo is the void volume. that is present in all known human SULTs and which is thought to be involved in PAPS binding.1,32,33 That change in amino acid resulted in a 43% decrease in enzyme activity it can participate in the metabolic activation of procarcino- as well as a decrease in immunoreactive protein (Table 2). gens.3,5–7 Sulfation is the major metabolic pathway for the This amino acid was located on the ‘surface’ of the protein. metabolism of DHEA in humans,9 just as it is for many other Lys227Glu was located four amino acids upstream of the endogenous and exogenous steroids.35 Circulating levels of putative SULT2A1 substrate binding site.31 Like Ala63Pro, DHEA sulfate vary widely among individuals, decrease with this residue was also on the surface of the protein. This latter age and are known to be influenced by inheritance.9,10 All alteration in sequence resulted in a striking decrease in of those observations made SULT2A1 an attractive candidate enzyme activity (to 15% of the activity of the wild-type for pharmacogenetic studies. sequence) as well as a very low level of immunoreactive pro- The present experiments were initiated to determine tein (Table 2). whether common SULT2A1 genetic polymorphisms might Of interest was the fact that the Ala261Thr polymor- exist, whether they might vary among ethnic groups and, if phism, present in 13.3% of the African-American popu- so, to determine whether they might be of functional sig- lation sample studied, was located within the dimerization nificance. We applied a genotype-to-phenotype strategy that motif found at the C-terminus of the conserved ‘Region IV’ began with the resequencing of SULT2A1, including all SULT signature sequence.34 Therefore, the possibility arose exons, portions of introns flanking exons and the 5Ј-flank- that this polymorphism might alter the formation of the ing region of the gene—using DNA samples from African- SULT2A1 homodimer. To test that hypothesis, we subjected American and Caucasian-American subjects. Fifteen SNPs, recombinant wild-type and codon Thr261 variant allozyme including three non-synonymous cSNPs, were observed in preparations to gel filtration chromatography through the 120 DNA samples (240 alleles) resequenced. The DNA Sephadex G-100. Those experiments showed that the major samples from African-American subjects had more than peak for wild-type enzyme was compatible with the pres- twice as many SNPs within the area resequenced as did ence of SULT2A1 dimer, while that for the variant allozyme samples from Caucasian-American subjects, and all three of was compatible with monomer (Figure 4). As a result, this the non-synonymous cSNPs were observed only in DNA appears to be the first demonstration of a common SULT from African-American subjects. These observations are genetic polymorphism that disrupts dimerization—with similar to those obtained during other resequencing efforts functional consequences that are presently unknown. which have demonstrated greater DNA sequence variation in samples from African subjects than in those from Euro- DISCUSSION pean-Caucasian subjects—even within areas of DNA that Sulfate conjugation is an important pathway in the contained no genes and, thus, were thought to be less likely biotransformation of both exogenous and endogenous com- to be influenced by natural selection.36 The majority of the pounds.1,2 Human SULT2A1 catalyzes the sulfate conju- SULT2A1 SNPs, both in samples from Caucasian-American gation of DHEA and other structurally-related steroids, and and African-American subjects, were present at frequencies

The Pharmacogenomics Journal SULT2A1 pharmacogenetics BA Thomae et al 53

greater than 0.01 (1%), and none were available in public niﬁcantly decreased activity for three of the four constructs DNA sequence databases. However, it was puzzling that we (Figure 3a and Table 2). Those decreases were due both to failed to observe either of the two cSNPs reported during a decreased levels of immunoreactive SULT2A1 protein and previous study of DNA from Caucasian subjects30 that was alterations in substrate kinetics (Table 2). The next step in performed prior to a time when automated sequencing was this series of genotype-to-phenotype studies will involve a widely available and prior to the development of sophisti- determination of possible in vivo alterations in the pharma- cated DNA sequence analysis software such as PolyPhred cokinetics and/or the effects of orally administered agents and Consed. metabolized by this phase II drug-metabolizing enzyme as a We also studied recombinant allozymes for each of the result of these common SULT2A1 genetic polymorphisms. SULT2A1 SNPs that altered amino acid sequence as well as the common double variant allozyme that occurs in African- MATERIALS AND METHODS American subjects. After expression in COS-1 cells, three of DNA Samples the four variant sequences resulted in both decreased DNA samples from 60 Caucasian-American and 60 African- enzyme activity and decreased levels of immunoreactive American subjects were obtained from the 100-item sample protein, although the extent of those decreases differed sets HD100CAU and HD100AA held by the Coriell Cell among the polymorphisms studied (Table 2 and Figure 3). Repository (Camden, NJ, USA). These samples had been

Apparent Km values for PAPS and DHEA, the two cosub- anonymized prior to their deposit in the Coriell Cell Reposi- strates for the reaction, also differed significantly when vari- tory. All subjects had provided written consent for the use ant recombinant allozymes were compared with the wild- of their DNA for experimental purposes, and the present type enzyme. However, those differences were smaller than studies were reviewed and approved by the Mayo Clinic were variations in protein quantity (Table 2). These obser- Institutional Review Board. vations indicate that—for SULT2A1—mechanisms by which changes in encoded amino acids decrease enzyme activity DNA Sequencing involve both alterations in quantity of enzyme protein and The PCR was used to amplify each of the six SULT2A1 exons, alterations in substrate kinetic parameters. There is increas- as well as approximately 370 bp of the 5Ј-flanking region of ing evidence that changes of single amino acids as a result the gene. An M13 ‘tag’ was added to the 5Ј-terminus of each of genetic polymorphisms often result in decreased levels of primer to make it possible to use dye primer sequencing protein.20,27,28 Those decreases could result from a series of chemistry. Dye primer chemistry enhanced our ability to mechanisms—including variations in transcription, RNA identify heterozygous sequence.21 Primer locations were splicing, mRNA stability, mRNA translation or protein chosen to be SULT2A1-specific and to avoid repetitive degradation. In one case which has been studied in detail, sequence. Sequences for the primers used in these and sub- that involving the genetically polymorphic drug-metaboliz- sequent experiments are listed in Table 3. PCR amplifi- ing enzyme thiopurine S-methyltransferase, the mechanism cations were performed using 2.5 units of AmpliTaq Gold has been shown to involve an increased rate of proteosome- DNA polymerase (Perkin Elmer, Foster City, CA, USA), 0.5 ␮l mediated degradation.29 A similar mechanism for decreased of DNA as supplied by the Coriell Cell Repository, 12.5 pmol protein levels has been described for the most common cys- of each primer, 0.25 mM dNTPs (Boehringer Mannheim, tic fibrosis transmembrane conductance regulator (CFTR) Indianapolis, IN, USA), and 5 ␮lof10ϫ buffer (Perkin Elmer) ⌬ 37 ␮ mutation, CFTR F508. All of these mechanisms should be that contained 15 mM MgCl2 in a final 50 l reaction vol- explored systematically for the SULT2A1 genetic polymor- ume. Reactions were initiated with a 12 min ‘hot start’ at phisms in the course of future experiments—experiments 94°C, followed by 35 cycles at 94°C for 30 s, a primer-spe- that should also address the possible functional implications cific annealing temperature (see Table 3) for 30 s and 45 s at of the 5Ј-flanking region SNP located at nucleotide (Ϫ122). 72°C, with a final 10 min extension at 72°C. Reactions were Included among those functional studies should be a deter- performed in a Perkin Elmer model 9700 thermal cycler. mination of the functional consequences—if any—of the Amplification products were sequenced in the Mayo Mol- disruption of the formation of the SULT2A1 dimer by the ecular Biology Core Facility with an ABI 377 DNA sequencer common codon 261 polymorphism (Figure 4). Our results (Applied Biosystems) using Big Dye (Perkin Elmer) dye indicated that this common SNP disrupted the ability of the primer chemistry. Samples were sequenced on both strands, variant allozyme to dimerize. The full functional impli- and those with ambiguous sequencing chromatograms, as cations of that observation remain to be explored in the well as samples with SNPs that were observed in only a sin- course of future experiments. gle sample, were subjected to a second, independent ampli- In summary, we set out to determine whether an fication, followed by DNA sequencing. important human sulfotransferase gene, SULT2A1, might DNA sequence was analyzed using the PolyPhred 3.038 display common genetic polymorphisms. Resequencing and Consed 8.039 programs. These programs align sequence SULT2A1 in 120 DNA samples revealed 15 SNPs, many of chromatograms and identify areas in which polymorphisms which showed striking differences in frequency between the might be present. Each chromatogram was then evaluated two ethnic groups studied (Table 1). Functional genomic individually to confirm ‘flagged’ sequence. The Wisconsin studies performed by transfecting COS-1 cells with con- Genetics Computer Group (GCG) package, version 10, was structs that encoded SULT2A1 variant allozymes showed sig- also used to analyze nucleotide sequence. The SULT2A1

www.nature.com/tpj SULT2A1 pharmacogenetics BA Thomae et al 54

Table 3 Human SULT2A1 DNA resequencing and site-directed mutagenesis primers

Primer name Primer location Primer sequence Annealing temperature, °C

Resequencing UF(Ϫ370) 5Ј-Flanking Region TGTAAAACGACGGCCAGTGCTTTGTTCTTGTTTTTAAGTTTGCAC 62 UR(Ϫ44) 5Ј-Flanking Region CAGGAAACAGCTATGACCCACCGCTGGAGGCTGTGGCAGCTACAG 62 UF(Ϫ118) 5Ј-Flanking Region TGTAAAACGACGGCCAGTGGGACAAGGTTAAAGATCGTTTTATC 62 I1R103 Intron 1 CAGGAAACAGCTATGACCAAGCATTGTACACTGTCTGAC 62 I1F(Ϫ50) Intron 1 TGTAAAACGACGGCCAGTATGTCCGGCTGAGATGGTACA 62 I2R112 Intron 2 CAGGAAACAGCTATGACCAACACAGACCTGTTGAAGGAG 62 I2F(Ϫ112) Intron 2 TGTAAAACGACGGCCAGTATTGATATTATTAGGCATTATCCA 60 I3R129 Intron 3 CAGGAAACAGCTATGACCAGGTTTGGCCTCCAGGGGTGG 60 I3F(Ϫ176) Intron 3 TGTAAAACGACGGCCAGTACATCTGTTATATACTGTAAA 57 I4R94 Intron 4 CAGGAAACAGCTATGACCGGTGGAATGAAGACACAG 57 I4F(Ϫ76) Intron 4 TGTAAAACGACGGCCAGTGGGATTACACGGTGAACCACC 57 I5R98 Intron 5 CAGGAAACAGCTATGACCAAGCTCTCTTTCATCTCAACT 57 I5F(Ϫ95) Intron 5 TGTAAAACGACGGCCAGTCTGCTATGTTAGCTACAATGTCCA 62 R1011 Exon 6 CAGGAAACAGCTATGACCTTTTAACAAGGAAGGGATCAG 62

Site-directed mutagenesis Mutated nucleotide F(Ϫ36) cDNA TGAAACCCTCACACCACGCAGGAAGA N/A R895 cDNA ATGTCATTCTCCATATAAGATCCAGA N/A F187p Exon 2 CCAAGGGGGATCCCAAGTGGATC G187C R187p Exon 2 GATCCACTTGGGATCCCCCTTGG G187C F679p Exon 5 TGAAAGAAAACGAGATGTCCAAT A679G R679p Exon 5 ATTGGACATCTCGTTTTCTTTCA A679G F781p Exon 6 ACTTCACAGTGACCCAAGCTGAA G781A R781p Exon 6 TTCAGCTTGGGTCACTGTGAAGT G781A

The underlined portions of primers used for the resequencing experiments are M13 tags that were added at the 5Ј-terminus to make it possible to use dye primer DNA sequencing chemistry. The underlined nucleotide in the sequences of primers used to perform site-directed mutagenesis is the ‘mutated’ base. ‘F’ represents forward; ‘R’, reserve; ‘U’, upstream; and ‘I’, intron. ‘N/A’ indicates ‘not applicable’. The numbering scheme for nucleotide locations is described in the legend for Table 1. sequences against which the resequencing data were com- pended in 2 ml of 5 mM potassium phosphate buffer, pH pared were those with GenBank accession numbers 6.5. The cells were then lysed with a Polytron homogenizer U13056–U13061. (Brinkmann Instruments, Westbury, NY, USA), and homo- genates were centrifuged at 100000 ϫ g for 1 h. The SULT2A1 Expression Studies resulting cytosol preparations were stored at Ϫ80°C prior Expression constructs for the wild-type SULT2A1 cDNA to assay. sequence construct, as well as constructs with the three nonsynonymous cSNPs that we observed and a construct con- Enzyme Assays taining two linked cSNPs, were created by site-directed Recombinant allozymes were assayed for SULT2A1 activity mutagenesis using overlap extension.40 The sequences of the as described by Hernańdez et al4 and Wood et al30 using a primers which were used to create these constructs are listed modification of the method of Foldes and Meek.41 Specifi- in Table 3. Inserts with the wild-type and mutated sequences cally, 5 ␮M DHEA served as the sulfate acceptor cosubstrate were cloned into the expression vector pCR3.1 (Invitrogen, and 0.4 ␮M 35S-PAPS was the sulfate donor. Blanks were Carlsbad, CA, USA), and those expression constructs were samples that contained no sulfate acceptor substrate. Levels used to transfect COS-1 cells using TransFast reagent of enzyme activity were corrected for transfection efficiency (Promega, Madison, WI, USA) at a charge ratio of 1:1 in by measuring ␤-galactosidase activity. Activity in cells trans- serum free Dulbecco’s Modified Eagles Medium (DMEM) fected with empty vector was also assayed to determine (Bio Whittaker, Walkersville, MD, USA) with a 2-h transfec- endogenous SULT2A1-like activity and those values—which tion time. Transfection was also performed with ‘empty’ never exceeded 1% of the activity observed after transfec- vector, ie, pCR3.1 that lacked an insert, to make it possible tion with the ‘wild-type’ construct—were subtracted from to correct for endogenous COS-1 cell SULT activity. Seven levels of activity in cells transfected with the expression con- ␮g of construct DNA was co-transfected with 7 ␮g of pSV-␤- structs.

Galactosidase DNA (Promega) to provide a control to make This same assay was used to determine apparent Km values it possible to correct for transfection efﬁciency. After trans- for each of the two cosubstrates with each recombinant allo- fection, the COS-1 cells were incubated at 37°C for 48 h, zyme. Because SULTs display profound substrate inhibition,1 washed with phosphate buffered saline (PBS) and resus- Km values for DHEA were determined in a two-step process.

The Pharmacogenomics Journal SULT2A1 pharmacogenetics BA Thomae et al 55

During the first step, DHEA concentrations that varied in culated with the method of Wilkinson43 using a computer 44 10-fold increments over a range from 100 nM to 1.0 mM program written by Cleland. Apparent Km values were were tested. After we determined that peak activity for all compared by the use of unpaired Student’s t-test. Linkage allozymes occurred near 1 ␮M, assays were repeated in the analysis involved testing all possible pairwise combinations presence of 0.4 ␮M 35S-PAPS using eight concentrations of of SNPs by calculating DЈ values.25,26 DЈ is a method of link- ␮ DHEA that varied from 7.8 nM to 1.0 M. Apparent Km age analysis that is independent of polymorphism fre- values for PAPS were determined in the presence of 5 ␮M quency. DHEA using six PAPS concentrations that varied from 3.13 nM to 100 nM. ␤-Galactosidase activity was measured ACKNOWLEDGEMENTS spectrophotometrically with the ␤-Galactosidase Enzyme The authors thank Dr Edward Carlini for assistance with the preparation Assay System (Promega) as suggested by the manufacturer. of the expression constructs, Dr Araba Adjei and Mr Thomas Wood for All results were corrected for transfection efficiency. their advice and Ms Luanne Wussow for her assistance with the preparation of this manuscript. Supported in part by NIH grants RO1 GM28157 (RMW), RO1 Western Blot Analysis GM35720 (RMW) and UO1 GM61388 (RMW and EDW). Levels of immunoreactive SULT2A1 were determined for The DNA resequencing data described in this manuscript have been each recombinant allozyme by performing quantitative deposited in the NIH-sponsored Pharmacogenetics Research Network Western blot analysis. Rabbit polyclonal antibody directed database PharmGKB with accession number PKBFK00000190. against SULT2A1 amino acids 268–285, with an additional amino terminus cysteine (described in detail by Wood et DUALITY OF INTEREST al30), at a dilution of 1:2000 was used to perform these stud- None declared. ies. Specifically, recombinant protein was loaded on a 12% SDS mini-gel (BioRad, Hercules, CA, USA) in quantities that REFERENCES resulted in equal ␤-galactosidase activity, ie, loading was cor- 1 Weinshilboum R, Otterness D. Sulfotransferase enzymes. In: Kauffman FC (ed). Conjugation-Deconjugation Reactions in Drug Metabolism and rected for transfection efficiency. Electrophoresis was perfor- Toxicity,ch2,‘Handbook of Experimental Pharmacology’ series, vol med for 1 h at 150 volts, and proteins were transferred to 112. Springer-Verlag: Berlin Heidelberg, 1994, pp 45–78. nitrocellulose membranes. The membranes were then 2 Falany CN. Enzymology of human cytosolic sulfotransferases. FASEB J ‘blocked’ overnight at 4°C with 5% dried milk in Tris buff- 1997; 11: 206–216. 3 Falany CN, Vazquez ME, Kalb JM. Purification and characterization of ered saline with Tween-20 (TBST). Subsequently, they were human liver dehydroepiandrosterone sulphotransferase. Biochem J incubated with primary antibody for 1 h, followed by three 1989; 260: 641–646. TBST washes. The secondary antibody was a 1:10000 4 Hernańdez JS, Watson RWG, Wood TC, Weinshilboum RM. Sulfation ␤ dilution of goat anti-rabbit horseradish peroxidase (BioRad) of estrone and 17 -estradiol in human liver: catalysis by thermostable phenol sulfotransferase and by dehydroepiandrosterone sulfotransfer- that was applied for 1 h in 5% milk in TBST, followed by ase. Drug Met Dispos 1992; 20: 413–422. multiple washes. Bound antibody was detected by enhanced 5 Glatt H, Pauly K, Czich A, Falany JL, Falany CN. Activation of benzylic chemiluminescence performed with the ECL Western Blot- alcohols to mutagens by rat and human sulfotransferases expressed in ting System (Amersham Pharmacia, Piscataway, NJ, USA) as Escherichia coli. Eur J Pharmacol 1995; 293: 173–181. 42 6 Glatt H. Sulfation and sulfotransferases 4: bioactivation of mutagents described previously. Multiple blots were assayed for each via sulfation. FASEB J 1997; 11: 314–321. allozyme, and immunoreactive protein levels were 7 Otterness DM, Wieben ED, Wood TC, Watson RWG, Madden BJ, expressed as a percentage of the intensity of the wild-type McCormick DJ et al. Human liver dehydroepiandrosterone sul- construct on the same gel. Blots were analyzed with the fotransferase: molecular cloning and expression of cDNA. Mol Pharma- col 1992; 41: 865–872. AMBIS Radioanalytic Imaging System, Quant Probe v. 4.31 8 Otterness DM, Weinshilboum R. Human dehydroepiandrosterone sul- (AMBIS, San Diego, CA, USA). fotransferase: molecular cloning of cDNA and genomic DNA. Chem- Biol Interact 1994; 92: 145–159. Gel Filtration Chromatography 9 Orentreich N, Brind JL, Vogelman JH, Andres R, Baldwin H. Long-term longitudinal measurements of plasma dehydroepiandrosterone sulfate Gel filtration chromatography of recombinant human in man. J Clin Endocrinol Met 1992; 75: 1002–1004. SULT2A1 was performed at 4°C with a 2 ϫ 25 cm column 10 Rotter JI, Wong FL, Lifrak ET, Parker LN. A genetic component to the of Sephadex G-100 superfine. The mobile phase was 5 mM variation of dehydroepiandrosterone sulfate. Metabolism 1985; 34: K phosphate buffer, pH 7.5, and 3 ml fractions were col- 731–736. 11 Lane MA, Ingram DK, Ball SS, Roth GS. Dehydroepiandrosterone sulf- lected at a flow rate of approximately 1.8 ml per h. The frac- ate: a biomarker of primate aging slowed by calorie restriction. J Clin tions were then assayed for SULT2A1 activity. Standards for Endocrinol Met 1997; 82: 2093–2096. column calibration were eluted separately and included 12 Villareal DT, Holloszy JO, Kohrt WM. Effects of DHEA replacement on Dextran Blue 2000, bovine serum albumin (67 kDa), ovalbu- bone mineral density and body composition in elderly women and min (45 kDa), chymotrypsinogen A (20 kDa) and ribonu- men. Clin Endocrinol 2000; 53: 561–568. 13 Aksoy IA, Sochorova´ V, Weinshilboum R. Human liver dehydroepiand- clease A (13.7 kDa). rosterone sulfotransferase: nature and extent of individual variation. Clin Pharmacol Ther 1993; 54: 498–506. Data Analysis 14 Otterness DM, Her C, Aksoy S, Kimura S, Wieben ED, Weinshilboum Average levels of enzyme activity were compared by ANOVA RM. Human dehydroepiandrosterone sulfotransferase gene: molecular cloning and structural characterization. DNA Cell Biol 1995; 14: 331– performed with the StatView program, version 4.5 (Abacus 341. Concepts, Berkeley, CA, USA). Apparent Km values were cal- 15 Otterness DM, Mohrenweiser HW, Brandriff BF, Weinshilboum RM.

www.nature.com/tpj SULT2A1 pharmacogenetics BA Thomae et al 56

Dehydroepiandrosterone sulfotransferase gene (STD): localization to et al. Enhanced proteasomal degradation of mutant human thiopurine human chromosome 19q13.3. Cytogenet Cell Genet 1995; 70:45–47. S-methyltransferase (TPMT) in mammalian cells: mechanism for TPMT 16 Luu-The V, Dufort I, Paquet N, Reimnitz G, Labrie F. Structural charac- protein deficiency inherited by TPMT*2, TPMT*3A, TPMT*3B or TPMT* terization and expression of the human dehydroepiandrosterone sul- 3C. Pharmacogenetics 1999; 9: 641–650. fotransferase gene. DNA Cell Biol 1995; 14: 511–518. 30 Wood TC, Her C, Aksoy IA, Otterness DM, Weinshilboum RM. Human 17 Durocher F, Morissette J, Dufort I, Simard J, Luu-The V. Genetic linkage dehydroepiandrosterone sulfotransferase pharmacogenetics: quanti- mapping of the dehydroepiandrosterone sulfotransferase (STD) gene tative western analysis and gene sequence polymorphisms. J Steroid on the chromosome 19q13.3 region. Genomics 1995; 29: 781–783. Biochem Mol Biol 1996; 59: 467–478. 18 Raftogianis RB, Wood TC, Otterness DM, Van Loon JA, Weinshilboum 31 Pedersen LC, Petrotchenko EV, Negishi M. Crystal structure of RM. Phenol sulfotransferase pharmacogenetics in humans: association SULT2A3, human hydroxysteroid sulfotransferase. FEBS Lett 2000; 475: of common SULT1A1 alleles with TS PST phenotype. Biochem Biophys 61–64. Res Commun 1997; 239: 298–304. 32 Varin L, Marsolais F, Brisson N. Chimeric flavonol sulfotransferase 19 Raftogianis RB, Wood TC, Weinshilboum RM. Human phenol sul- define a domain responsible for substrate and position specificities. J fotransferases SULT1A2 and SULT1A1: genetic polymorphisms, allo- Biol Chem 1995; 270: 12498–12502. zyme properties and human liver genotype-phenotype correlations. 33 Weinshilboum RM, Otterness DM, Aksoy IA, Wood TC, Her C, Raftog- Biochem Pharmacol 1999; 58: 605–610. ianis RB. Sulfotransferase molecular biology: cDNAs and genes. FASEB 20 Freimuth RR, Eckloff B, Wieben ED, Weinshilboum RM. Human sul- J 1997; 11:3–14. fotransferase SULT1C1 pharmacogenetics: gene resequencing and 34 Petrotchenko EV, Pedersen LC, Borchers CH, Tomer KB, Negishi M. functional genomic studies. Pharmacogenetics 2001; (in press). The dimerization motif of cytosolic sulfotransferases. FEBS Lett 2001; 290 21 Chadwick RB, Conrad MP, McGinnis MD, Johnston-Dow L, Spurgeon :39–43. 35 Hobkirk R. Steroid sulfotransferases and steroid sulfatases: characteriz- SL, Kronick MN. Heterozygote and mutation detection by direct auto- ation and biological roles. Can J Biochem Cell Biol 1985; 63: 1127– mated fluorescent DNA sequencing using a mutant Taq DNA poly- 1144. merase. Bio Techniques 1996; 20: 676–683. 36 Kasessman H, Heissig F, von Haeseler A, Paä¨bo S. DNA sequence vari- 22 Boguslavsky J. Find SNPs online. Drug Discovery Develop 2001; June: ation in non-coding region of low recombination on the human X 67–68. chromosome. Nature Genet 1999; 22:78–81. 23 Iida A, Sekine A, Saito S, Kitamura Y, Kitamoto T, Osawa S et al. Catalog 37 Meacham G, Patterson C, Zhang W, Younger JM, Cyr DM. The Hsc of 320 single nucleotide polymorphisms (SNPs) in 20 quinone oxidore- 70 co-chaperone CHIP targets immature CFTR for proteasomal degra- ductase and sulfotransferase genes. J Hum Genet 2001; 46: 225–240. dation. Nature Cell Biol 2001; 3: 100–105. 24 Drysdale CM, McGraw DW, Stack CB, Stephens JC, Judson RS, Nanda- 38 Nickerson DA, Tobe VO, Taylor SL. PolyPhred: automating the detec- ␤ balan K et al. Complex promoter and coding region 2-adrenergic tion and genotyping of single nucleotide substitutions using fluor- receptor haplotypes alter receptor expression and predict in vivo escence-based resequencing. Nucl Acids Res 1997; 25: 2745–2751. responsiveness. Proc Natl Acad Sci USA 2000; 97: 10483–10488. 39 Gordon D, Abajian C, Green P. Consed: a graphical tool for sequence 25 Hartl DL, Clark AG. Principles of Population Genetics, 3rd edn. Sinauer finishing. Genome Res 1998; 8: 195–202. Associates: Sunderland, MA, 1997, pp 95–107. 40 Ho SN, Hunt HD, Horton RM, Pullen JK, Pease LR. Site-directed 26 Hedrick PW. Genetics of Populations, 3rd edn. Jones and Bartlett Publ: mutagenesis by overlap extension using the polymerase chain reac- Sudbury, MA, 2000, pp 396–405. tion. Gene 1989; 77:51–59. 27 Szumlanski C, Otterness D, Her C, Lee D, Brandriff B, Kelsell D et al. 41 Foldes A, Meek JL. Rat brain phenolsulfotransferase—partial purifi- Thiopurine methyltransferase pharmacogenetics: human gene cloning cation and some properties. Biochim Biophys Acta 1973; 327: 365–374. and characterization of a common polymorphism. DNA Cell Biol 1996; 42 Aksoy S, Klener J, Weinshilboum RM. Catechol O-methyltransferase 15:17–30. pharmacogenetics: photoaffinity labeling and western blot analysis of 28 Preuss CV, Wood TC, Szumlanski CL, Raftogianis RB, Otterness DM, human liver samples. Pharmacogenetics 1993; 3: 116–122. Girard B. Human histamine N-methyltransferase pharmacogenetics: 43 Wilkinson GN. Statistical estimations in enzyme kinetics. Biochem J common genetic polymorphisms that alter activity. Mol Pharmacol 1961; 80: 324–332. 1998; 53: 708–717. 44 Cleland WW. Computer programmes for processing enzyme kinetic 29 Tai H-L, Fessing MY, Bonten EJ, Yanishevsky Y, d’Azzo A, Krynetski EY data. Nature Lond 1963; 198: 463–465.

The Pharmacogenomics Journal