Introduction

Top Abstract Introduction Materials and Methods Results Discussion References

The microsomal NADPH-dependent flavin-containing monooxygenases (FMOs¹) E.C. (1.14.13.8) comprise a family of chemical and drug-metabolizing enzymes that, barring steric limitations, catalyze the oxygenation of a wide variety of nucleophilic heteroatom-containing compounds (Ziegler, 1993; Cashman, 1995, 2000a). Evidence for six mammalian FMO genes exist, but in humans, FMO3 appears to be the form largely responsible for FMO-mediated hepatic clearance of drugs (Lomri et al., 1992). Expression of FMO3 is species- and tissue-specific, but unlike human cytochrome P450, mammalian FMO does not appear to be inducible (Shehin-Johnson et al., 1995; Dolphin et al., 1996). The interindividual variation that could result in population-wide differences in the FMO3-dependent metabolism of drugs or xenobiotics is therefore more likely due to genetic effects and not environmental ones.

Interindividual variation of human monooxygenases, such as FMO3, may contribute to the susceptibility of humans to adverse reactions to chemicals or exaggerated clinical response. One extreme example of this is the inborn error of metabolism that causes defective N-oxygenation of the diet-derived trimethylamine (TMA) (Ayesh et al., 1993; Mitchell et al., 1997). Detoxication and deoderation of TMA to its nonvolatile and benign metabolite TMA N-oxide normally occurs with approximately 97% efficiency giving rise to urinary TMA N-oxide/TMA ratios of 97:3. To date, approximately 17 human FMO3 mutations have been reported, of which 13 mutations have been causally associated with trimethylaminuria (Hadidi et al., 1995; Cashman et al., 1997; Dolphin et al., 1997, 2000; Thithapandha, 1997; Treacy et al., 1998; Akerman et al., 1999; Zschocke et al., 1999). In addition to a severe trimethylaminuria phenotype, evidence for a milder trimethylaminuria may be related to hepatic dysfunction (Fernandez et al., 1997) or other conditions (Blumenthal et al., 1980; Zhang et al., 1996) or due to less debilitating human FMO3 polymorphisms (Cashman et al., 2000b). Regardless, it is likely that trimethylaminuria in its severe or modest forms is more prevalent in the North American population than previously estimated (Preti et al., 1995).

The possibility of other idiopathic reactions related to rare human FMO3 genes stems from the observation that individuals suffering from severe trimethylaminuria have additional symptoms, including hypertension, adverse tyramine reactions, and depression (possibly related to the symptomology) (Treacy et al., 1998). In an animal model of decreased FMO activity, the FMO-mediated tertiary amine N-oxygenation of tamoxifen and nicotine was significantly decreased (Katchamart et al., 2000). Thus, rats treated with major chemopreventative indoles present in cruciferous vegetables (i.e., indole-3-carbinol or its acid condensation products likely to be produced by the acid contents of the stomach) significantly reduced the detoxication of nicotine. This result parallels that observed for two individuals suffering from trimethylaminuria that showed impaired nicotine N-1'-oxygenation (Ayesh et al., 1988). For tamoxifen, inhibition of N-oxygenation could actually decrease the risk of humans developing toxic side effects if FMO-mediated N-oxygenation is responsible for tamoxifen-dependent covalent binding to protein (Umemoto et al., 1999). Previously, we reported that humans fed 300 g of Brussels sprouts/day led to a significant decrease in urinary TMA N-oxide formation, presumably from the inhibition of hepatic FMO3 (Cashman et al., 1999). It is possible that individuals with decreased human FMO3 activity have the potential for altering the toxicity of drugs or chemicals and inducing adverse drug reactions.

Mutations that are identified through individuals with extremely rare phenotypes may not be typical of variation in the gene in the population at large. Before determining whether variation of a gene contributes to susceptibility to disease in the population at large, the amount and organization of genetic variation in the gene must be evaluated in samples of individuals ascertained without regard to their health status. It is likely that many common genetic variants (i.e., least frequent allele greater than 1% in the general population) having smaller effects combine to determine the major fraction of variation in human disease susceptibility in the population at large (Risch and Merikangas, 1996). Previous studies examined the prevalence of two prominent human FMO3 polymorphisms in French Canadian and Australian Caucasian (Cashman et al., 2000b), European (Sachse et al., 1999), and Asian (Kang et al., 2000) populations. In the present study, we report the relative frequencies of single and multiple site allele configurations, haplotypes, and genotypes of the three major common polymorphisms of human FMO3 in samples of non-Hispanic Caucasians, non-Hispanic African Americans, Asians, and Hispanics ascertained without regard to health status from the United States. The DNA sites with the most variation in the gene were quantified using a high-throughput mutation detection method combining MassEXTEND coupled with mass measurement by matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry (Scheme 1). Our study suggests that population differences may exist in the susceptibility of humans to abnormal metabolism or adverse drug reactions for chemicals metabolized by human FMO3 due to the frequency of the variants observed.

View larger version (22K): [in this window] [in a new window]   Scheme 1.   MassARRAY MassEXTEND reaction.

	Materials and Methods

Top Abstract Introduction Materials and Methods Results Discussion References

Chemicals. All chemicals and reagents were purchased from Aldrich Chemical Co. (Milwaukee, WI) in the highest purity commercially available. Buffers and other agents were purchased from VWR Scientific, Inc. (San Diego, CA). ⁹D-Trimethylamine (99 atom D %) (⁹D-TMA) was purchased from Isotech, Inc. (Mamsburg, OH). Oasis solid-phase extraction cartridges were purchased from Waters (Milford, MA).

Genomic DNA and Study Population. Genomic DNA was obtained from blood bank donors in California and Utah. After institutional review board ethics approval, the frequencies of common human FMO3 single nucleotide polymorphisms were experimentally determined from DNA obtained from independent individuals. The ethnicity of the individuals was defined by a self-report questionnaire that indicated the race of both parents. One hundred and seventy nine non-Hispanic Caucasians (i.e., 89 males and 90 females) were from Utah (1% Amish pedigree, 2% Venezuelan pedigree, 2% French pedigree, and 95% Utah pedigree). Ninety non-Hispanic African Americans (i.e., 45 males and 45 females) were from California. The 85 Hispanic samples (i.e., 43 males and 42 females) were composed entirely of Mexican origin. The 66 Asians (i.e., 33 males and 33 females) were composed of 5% Vietnamese, 8% Asian, not further specified, 13% Korean, 31% Japanese, and 43% Chinese. The pools broadly represent the populations in the United States. Another group of 92 non-Hispanic Caucasians was examined for the presence of the rare human FMO3 P153L mutation that is the cause of trimethylaminuria. Using the MassEXTEND technology also showed that all of the samples were negative for the E305X mutation. Genotyping for FMO3 P153L and E305X served as a positive control that the samples were not from individuals with severe trimethylaminuria.

Primate DNA. Genomic DNA from baboon (i.e., Papio cynocephalus, Papio anubis, Papio ursinus, and Papio hamadryas) and genomic DNA from chimpanzee (i.e., Pan troglodytes) was generously provided by Professor David Smith (University of California, Davis, CA). cDNA from monkey (i.e., Macaca mulatta) was prepared from liver obtained from the University of Washington Primate Center (Seattle, WA) tissue-sharing program.

PCR Amplification. Oligonucleotides were synthesized at Operon, Inc. (Alameda, CA). Nonbiotin-labeled primers (-bio) were obtained unpurified, and biotin-labeled primers were obtained as high-pressure liquid chromatography-purified materials from Operon, Inc. The sequences of the PCR primers were as follows: exon 4, FMO3-329-bio: d(GCC AGT TAT GTG GCT AGC AG), FMO3-70: d(ATC GCG AAA CCA TTT GCT); exon 6, FMO3-339-bio: d(AAA GCC AGC AGG CAT ATC AC), FMO3-111: d(CAG TTC CAG AAG TGG CTC CT); exon 7, FMO3-743-bio: d(CAA AAC TGA AGG GGA CCT TG); and FMO3-341: d(CAA GCA TTC TGT GTG GCA TT). PCR was carried out with Hotstar Taq obtained from QIAGEN (Valencia, CA). The PCR cocktail was composed of 1× PCR QIAGEN Buffer [Tris-HCl, (NH₄)₂SO₄, 15 mM MgCl₂], 200 µM dNTPs, 1 U of final concentration Hotstar Taq, forward and reverse primers both at a 0.04-µM final concentration, and high-pressure liquid chromatography-purified water to a total reaction volume of 45 µl. The final addition was 5 µl of 5 ng/µl DNA template. PCR cycling was carried out in an MJ Research thermal cycler (Watertown, MA). PCR-cycling conditions were: an initial hold of 15 min at 95°C, followed by 44 cycles of 5 s at 95°C, 20 s at 56°C and 30 s at 72°C, and a final hold at 4°C. Biotinylated amplicons were transferred to a microtiter plate well containing 0.1 mg of streptavidin-coated Dyna beads (Dynal, Inc., Oslo, Norway) that usually had 15 µl of Dyna beads for 45 µl of PCR product. The bound amplicons were processed according to the specifications of the supplier in an automated instrument. After the final wash, the DNA was denatured by addition of 100 µl of 50 mM NaOH solution for 3 min, and the wells were washed three times with 200 µl of 70 mM ammonium citrate.

MassEXTEND Primer Reactions. To each well containing immobilized single-stranded templates, the composition of the MassEXTEND cocktail contained: 1× reaction buffer from Amersham Pharmacia Biotech (Arlington Heights, IL) (26 mM Tris-HCl, pH 9.5, 6.5 mM MgCl₂), the appropriate dNTPs in a final concentration of 50 µM from Invitrogen (Carlsbad, CA), and the appropriate ddNTPs each in a final concentration of 50 µM from Boehringer Mannheim (Indianapolis, IN) and 2.5 U of Thermosequenase (32 U/µl; Amersham Pharmacia Biotech). For the template-specific MassEXTEND reactions, 10 to 20 pmol of the following primers were used: FMO3 P153L: d(CGG ACA TCA TGT GTA TC); FMO3 E158K: d(CCC AAC CTA CCA AAA); FMO3 V257 M: d(ATCTCTGACTGGTTGTAC); FMO3 E308G: d(AAC GTG AAG GAA TTC ACA G). The reactions were carried out on an MJ Research thermal cycler using the following cycling conditions: 80°C for 30 s, 40°C for 15 s, 72°C for 30 s, and hold for 25°C. The paramagnetic beads were then washed with 0.7 M ammonium citrate followed by 0.05 M ammonium citrate. Sequencing products were denatured from the template by heating the beads at 80°C in 2 µl of 50 mM NaOH for 2 min, and the supernatant was used in the MALDI-TOF mass spectrometry (MS).

MALDI-TOF MS. Matrix preloaded Spectro chips from Sequenom, Inc. (San Diego, CA) were spotted with 5 to 12 nl of sample with a piezoelectric nanoplotter. MALDI-TOF mass spectrometry was carried out on a Sequenom instrument. Data acquisition and mass spectrometry and genotyping were done concomitantly in an automated mode with SpectroTyper software from Sequenom, Inc. (San Diego, CA). The sequence data were assembled using external calibration, and accurate mass values were determined using internal calibration from two of the preidentified sequence peaks.

LCMS. The profile of TMA and TMA N-oxide was examined by LCMS analysis in selected human serum samples. Onto a freshly prepared Oasis C₁₈ Cartridge was placed 200 µl of serum or an internal standard in water. The sample was washed with 700 µl of water followed by 700 µl of methanol/water (5:95, v/v). The analyte was eluted with 700 µl of acetonitrile/water (80:20, v/v). The eluant was separated and analyzed for the presence of TMA and TMA N-oxide with a Hewlett Packard 1100 LC/MSD system (Wilmington, DE). The system was fitted with a C₁₈ reverse-phase analytical Altima column (4.6 × 250 mm) from Alltech (Deerfield, IL). The mobile phase consisted of H₂O/MeOH/HCOOH (95:5:0.1), and the flow rate was 1.0 ml/min. Detection was with atmospheric pressure electrospray ionization. The mass spectrometer was set in the scan mode, positive ionization, with a m/z range of 50 to 78. ⁹D-Trimethylamine was used as an internal standard.

Genomic DNA Sequencing. Individual exons from chimpanzee and baboon were amplified from genomic DNA using primers (below) and Hotstart mix. The PCR conditions were as follows: 94°C 5-min hot start, 35 cycles of 94°C for 30 s, 58°C for 30 s, 72°C for 2 min, and finally 72°C, with a 5-min extension. The PCR fragments were fractionated, gel purified, and sequenced. The exons and product sizes (nucleotide) were as follows (primers): Exon 2, product size: 498 (tcaaactcctgggctcaagt and tttccaacctgctcttgaca); Exon 3, product size: 581 (cagattcaacccaccattga and ttcttcagcattatgacaagagc); Exon 4 and 5, product size: 651 (atctgccaaaaccatttgct and acgagagtcacccgagtacc); Exon 6, product size: 408,(agccaaagcaatggtcaaaa and aaagccagcaggcatatcac); Exon 7, product size: 429(tccaataattgtctctgttttcca and ttcatcttcgcaatccatga); Exon 8, product size: 473 (tgaatttggtgtctgtctgaaaa and aggcatggagggctaaaact); Exon 9, product size: 475 (gcgagccattttctctgttc and cccctgtctgggtattgtca). Sequence was analyzed with Sequencher Software (Gene Codes Corporation, Ann Arbor, MI) by procedures that could resolve heterozygotes under reliable quality-control conditions.

Monkey FMO3 DNA Sequencing and Comparison with Human. The cDNA sequence of the FMO3 gene from monkey (M. Mulatta) was obtained through reverse transcription-PCR from RNA prepared from monkey liver. PCR was done with the same primer pairs as that for the human FMO3 gene (i.e., 5' FMO3 GATCGGATCCATGGGGAAGAAAGTGGCC and 3' FMO3 GATCAAGCTTAGGTCAACACAAGGAAAAC). The PCR fragment was cloned into pMal-2c (New England Biosciences, Beverly, MA). Sequencing of individual clones was done with fluorescent dye termination cycle sequencing. The sequence was deposited into GenBank. Of the 1599 nucleotides (532 amino acids) scored in human FMO3, the monkey homolog was aligned and scored for amino acid changes at 31 of these sites. The number of synonymous and nonsynonymous substitutions was compared between the human and the monkey. K_s and K_a were calculated and compared using options available in the frequency estimates of nucleotide substitution (FENS) program as outlined previously (Zhang et al., 1999).

Data Analysis and Statistics. The statistics D and D' were used to compare the summary statistics. Under neutrality, the two estimates should be equal and D = 0. Pair-wise linkage disequilibrium was measured with the linkage disequilibrium parameter D, calculated as D = P_ij  p_ip_j, where P_ij is the frequency of the most common gametic type for a pair of sites and p_i and p_j are the frequencies of the nucleotides in the haplotype (Hartl and Clark, 1997).

	Results

Top Abstract Introduction Materials and Methods Results Discussion References

The MassEXTEND primer experiments to quantify potential sequence variations in exons 4, 6, and 7 were developed for the human FMO3 gene. Scheme 1 shows a schematic representation of the procedure. For example, in exon 4 the antisense strand of the PCR product served as a template, and the nucleotides for the primer extension are shown in bold letters. The specific termination reactions (i.e., ddA; Table 1) were used to study known mutations at codons 158, 257, and 308 of human FMO3.

Figure 1 shows mass spectra of MassEXTEND reaction products for DNA from an individual that was homozygous-normal at each of FMO3 codons 158, 257, and 308. In the ddA reaction, the 4459-Da FMO3 158-primer was converted to a single product consistent with the addition of A, generating a signal at 4759 Da. A lower molecular weight side product apparently arose from the MassEXTEND reaction but this did not interfere with the quantitation. The data of Fig. 1 were consistent with a homozygous common human FMO3 gene because a heterozygote would have given two products of 4756 and 5085.4 Da. Simultaneous detection of the codon 153 and 158 gene mutations by MassEXTEND showed that the 153 mutation was not observed. The simultaneous detection of the codon 305 truncation and 308 gene mutation by MassEXTEND also showed the 305X truncation was not observed. It has previously been observed that the human FMO3 P153L mutation is rare and is linked to the rare inborn error of metabolism called trimethylaminuria (Dolphin et al., 1997; Treacy et al., 1998).

View larger version (18K): [in this window] [in a new window]   Fig. 1.   MALDI-TOF MS of the human FMO3 gene with a common sequence at codon 158.

The human FMO3 E158K genotype distribution and allele frequency of the major common polymorphisms of human FMO3 were determined in DNA samples obtained from a population of male and female non-Hispanic Caucasians, non-Hispanic African Americans, Hispanics, and Asians sampled from a United States population. The common polymorphisms at alleles 158, 257, and 308 were examined by the MassEXTEND method described above. Table 2 lists the analysis of relative allele frequencies for sites 158, 257, and 308 of the human FMO3. For the human FMO3 158 and 257 alleles, Caucasians, African Americans, and Hispanics had similar allelic frequencies, but Asians were significantly different. For the FMO3 308 allele, Caucasians, Asians, and Hispanics possessed similar allelic frequencies, but African Americans were significantly different (Table 2). For male and female Caucasians, African Americans, and Hispanics, the 158KK genotype frequencies were 0.17, 0.14, and 0.13, respectively, whereas Asians had a considerably lower genotype frequency; only 0.02 were homozygous for FMO3 158KK (Table 3). The human FMO3 257 M allele frequency was 0.01 and further underscored the fact that the 257 MM allele frequency was rare. The 257 M allele frequency for Caucasians, African Americans, and Hispanics was 0.07, 0.07, and 0.08, respectively, but Asians had a much greater relative allele frequency (i.e., 0.20) (Table 2). The human FMO3 V257 MM genotype was only observed in Caucasians, whereas for African Americans, Hispanics, and Asians, the 257 MM genotype was not observed (Table 3).

The final common human FMO3 polymorphism examined was at allele 308. The human FMO3 E308G allelic frequency for Caucasians, Asians, and Hispanics was 0.16, 0.14, and 0.12, respectively (Table 2). In African Americans for the population examined, the 308G allele frequency was 0.04 and showed that the African American 308G allelic frequency was low. Table 3 showed that the 308GG allele frequency was 0.05, 0.02, and 0.02, respectively, for Caucasians, Asians, and Hispanics. For the African American samples examined, there was no human FMO3 308GG genotype observed.

We examined the prevalence of the eight haplotypes for the three human FMO3 polymorphisms described above. The results were listed in Table 4. Two of the eight possible haplotypes (i.e., E-M-G, and K-M-E) were not observed in the populations examined. The K-M-G haplotype was only observed in Hispanics. Of the populations examined, a fourth haplotype (i.e., E-V-G) was quite rare, appearing only in African Americans and Hispanics with relative haplotypes of 0.0056 and 0.0059, respectively. Compared with the wild type haplotype (E-V-E), some of the haplotypes were quite prevalent (Table 4). Thus, the relative frequencies of the E-M-E haplotype was 0.0726, 0.0667, 0.1969, and 0.07059 in Caucasians, African Americans, Asians, and Hispanics, respectively. In contrast to the E-M-E relative haplotype frequency that was large in Asians, the relative frequency of the K-V-E haplotype was quite low in Asians (i.e., 0.00757) but quite prevalent in the other populations examined (i.e., 0.229, 0.3722, and 0.2353 in Caucasians, African Americans, and Hispanics, respectively). The relative frequency of the K-V-G haplotype also varied considerably as a function of ethnic group. Thus, the prevalence of K-V-G was 0.1648, 0.0333, 0.14393, and 0.1117 for Caucasians, African Americans, Asians, and Hispanics, respectively.

Given the haplotype frequency differences, linkage disequilibrium was also investigated. D was calculated for all pairs of sites in all four ethnic groups examined (Table 5). For each pair of sites, we calculated the statistic D, defined as P_ij  p_ip_j, where P_ij are the frequencies of the haplotype with nucleotides i and j at the two sites, and p_i and p_j are the frequencies of the nucleotides i and j, respectively. D values are dependent on allele frequencies. We also calculated the statistic D'. D' is standardized and is independent of allele frequencies, and the value is from 0 to 1.0. A test of overall linkage disequilibrium, based on the number of observed and expected signs of D, showed a significant excess of cases of disequilibria in which rare alleles were associated.

The comparison of monkey FMO3 gene sequences to human was an important control because we wanted to compare human allelic variation with nucleotide variation in monkey. The homologous genomic exon region in monkey (M. mullatta) was compared with human and showed 31 sequence differences between the human consensus (GenBank accession NM006894) and monkey. The sequence differences included 54 nucleotide substitutions. The monkey FMO3 gene shared 94.2% sequence identity with that of the human (Figs. 2 and 3). The associated estimate of net sequence divergence between human and monkey that took into account human polymorphic variation was not statistically significant. Although the large number of nonsynonymous substitutions (i.e., 31) with a K_a value of 0.02498 ± 0.00492 compared with the number of synonymous substitutions (K_s value of 0.0668 ± 0.01486) appeared to be suggestive, there was no evidence for positive selection on these sequences. Because K_s was a good deal greater than K_a, it is possible that a modest purifying selection was at work.

View larger version (69K): [in this window] [in a new window]   Fig. 2.   FMO3 nucleotide sequence alignment for monkey (i.e., M. mulatta, mFMO3) and human (i.e., consensus sequence, hFMO3). Differences between monkey and human are indicated with an asterisk.

View larger version (60K): [in this window] [in a new window]   Fig. 3.   FMO3 amino acid sequence alignment for monkey (i.e., M. mulatta, mFMO3) and human (i.e., consensus sequence, hFMO3). Differences between monkey and human are indicated with an asterisk.

Comparison of the chimpanzee FMO3 gene to the human FMO3 gene showed that at least 12 single nucleotide variations and one large deletion were observed in the immediate intron regions. Five silent mutations were present in the exon regions (N285, L288, E424, I486, and I525), and only one amino acid change was identified (i.e., H116R, CAT to CGT). However, for the five individual baboon DNA samples that we examined using the same conditions for the chimp, we could only amplify exon 8. In baboon FMO3 exon 8, one amino acid variant (K418R, AAA to AGA) was identified, and 20 single nucleotide variations were observed in the intron adjacent to exon 8. We conclude that the intronic regions of baboon FMO3 differ substantially from that of human and chimpanzee.

	Discussion

Top Abstract Introduction Materials and Methods Results Discussion References

Variation in genes that encode enzymes responsible for drug or chemical metabolism generally result from single-nucleotide substitutions or small deletions/insertions that may lead to alteration in function of the enzyme (Motulsky, 1991; Kalow and Grant, 1995). Most genetic variation can be accounted for by a few prevalent alleles (i.e., alleles with a frequency of greater than 1% in the population), although the frequency of these alleles can vary wildly across populations. Interindividual variation of enzymatic metabolic activity can result in significant differences in the biotransformation of drugs or xenobiotics. For example, genetic polymorphism of CYP2D6-mediated debrisoquine 4-hydroxylation varies considerably across different ethnic groups. For example, "poor metabolizers" make up approximately 5 to 10% of the Caucasian population but only about 0.1% of the Asian population (Wolf et al., 1992). For individuals with a particular CYP2D6 variant, the polymorphism may cause an exaggerated clinical response and side effects of debrisoquine administration (Tucker et al., 1977). Another example stems from the CYP2C19-mediated 4'-hydroxylation of (S)-mephenytoin. In Caucasians, the prevalence of the CYP2C19 poor-metabolizer phenotype is approximately 0.2% but is significantly larger in Asians (i.e., 15-20%) (Goldstein and de Morais, 1994).

Human FMO3 is recognized as the prominent form of FMO present in the adult liver that contributes to the oxygenation of nucleophilic heteroatom-containing chemicals, drugs, and endogenous materials. Human FMO3 has a broad substrate specificity, and recent studies have shown that, barring steric limitations, human FMO3 is capable of oxygenating numerous N- and S-containing xenobiotics (Cashman, 2000a). In addition, human FMO3 has been shown to N-oxygenate several endogenous and dietary amines of significant physiological importance including biogenic amines (Lin and Cashman, 1997a,b). In humans, polymorphisms of human FMO3 were recognized and characterized after observations about abnormal metabolism of TMA (Dolphin et al., 1997; Treacy et al., 1998). In normal individuals, TMA is very efficiently N-oxygenated to the polar and nonodorous metabolite TMA N-oxide to the extent of at least 97%. Individuals with trimethylaminuria have a diminished ability to N-oxygenate TMA and excrete a large amount of odorous TMA. Individuals with a severe phenotype possess defective human FMO3 and are afflicted with trimethylaminuria or "fish-like odor syndrome" that can also lead to psychosocial consequences. Trimethylaminuria has been documented in the British (Dolphin et al., 1997), Australian (Treacy et al., 1998), and North American (Akerman et al., 1999) populations. Although rare, some mutations segregate with the most severe trimethylaminuria phenotype, and these include the P153L mutation and the E305X truncation mutation (Cashman et al., 1997; Dolphin et al., 1997; Treacy et al., 1998).

In addition to mutations that cause severe trimethylaminuria, we previously reported evidence for allelic variation within the human FMO3 that influence drug responsiveness (Cashman et al., 2000b). However, in the present study, no TMA could be detected in the plasma by highly sensitive LCMS analysis, and this prevented a correlation of phenotype to genotype because urine samples were unavailable. The lack of TMA present in the plasma underscores the apparent efficiency of urinary excretion of TMA, and the extremely high reuptake of TMA in the kidney.

The first major conclusion from the data presented here is that human FMO3 shows a frequency of single nucleotide variation that is somewhat higher than the average reported values previously reported (Cashman et al., 2000) for Canadian and Australian populations. For populations with a significant number of poor metabolizers, it is possible that individuals may be more susceptible to adverse drug reactions or exaggerated clinical response for drugs that are metabolized by FMO3. Altered substrate activities have been observed for human FMO3 and may be responsible for mild trimethylaminuria (Zschocke et al., 1999). For example, the N-oxidative detoxication of amphetamine and methamphetamine by human FMO3 may be under pharmacogenetic control (Cashman et al., 1999). In a previously studied cohort of Australian trimethylaminuria patients, several patients manifested hypertension and adverse reactions from tyramine, other amines, and sulfur-containing medications (Treacy et al., 1998).

A second major conclusion from the data in this report is that the frequency of single nucleotide variants differ widely across ethnic groups. The human FMO3 haplotypes described here for Utah Caucasians have similarity to that of a German population (Sachse et al., 1999) and a Canadian population (Cashman et al., 2000b) previously reported. In addition, the data reported here are in agreement with that of a study of a Korean population that used ranitidine as a test substrate to correlate FMO3 genotype and phenotype (Kang et al., 2000). The observation of differences of human FMO3 haplotypes between ethnic groups is intriguing because it may suggest that FMO3-dependent metabolism may vary considerably across ethnic groups. Larger and more diverse populations must be examined before a definitive statement along these lines can be made. Ultimately, however, if the metabolism and detoxication of a drug or chemical is dependent on human FMO3, an individual with impaired FMO3 may be at risk for exaggerated clinical response, and this could lead to adverse reactions.

Knowledge of FMO3 polymorphisms could lead to the development of high-precision DNA diagnostics for pharmacogenetics purposes. For large numbers of samples, the MALDI-TOF technology provides a high-throughput and analytically robust method for assessing polymorphism. We developed MassEXTEND assays and optimized them for the three common polymorphic variants of human FMO3, and the products were analyzed by MALDI-TOF mass spectrometry. The MassEXTEND assay clearly differentiates known codon variations of different human FMO3 genotypes because the mass values are specific. In some cases, multiplexing was possible, and a duplex reaction was established for the simultaneous analysis of variation in codons 153 and 158. High-quality data presented here can provide information about possible forces that acted to shape the genetic variation observed.

Evolutionary comparisons with sequences from a near species relative, the monkey, can illuminate features about human genetic variability. The monkey sequences indicated which nucleotides that segregate in humans might be ancestral. Such sites may be conserved by chance or because the variation was not deleterious enough to be eliminated from the population via selection. Although it is tempting to speculate that selection has maintained certain genetic variation in some populations, our statistical analysis does not suggest any evidence for positive selection of these genes. It is possible, of course, that demographic forces have played a role in creating haplotype differences rather than selection. It is interesting to note that, compared with human FMO3, the monkey FMO3 variation was spread somewhat uniformly over the exons with the exception of the amino terminus and some of the consensus sequences of flavin adenine dinucleotide and NADPH-binding domains at positions 9 to 14 and 190 to 196, respectively. In addition, the region between amino acids 240 and 290 was also devoid of variation with the exception of the common polymorphism at position 257. It is possible that the monkey was heterozygotic at position 257. This is relatively common in human populations and probably without significantly deleterious metabolic consequences. Of note is that the monkey possesses the FATGY sequence at position 327 that is present in all mammalian and plant (Zhao et al., 2001) FMOs observed to date.

In summary, the use of the high-throughput MassEXTEND methodology to rapidly genotype populations of healthy individuals from Californians and Utah has provided considerable insight into the relative allelic frequencies of human FMO3 haplotypes in individuals ascertained without regard to their health status. Human FMO3 may play a role in the ethnic group-dependent metabolism of drugs and chemicals, and a complete description of the metabolic distribution of such a chemical should take this into account.