The Flavin-containing Monooxygenase 2 Gene (FMO2) of Humans, but Not of Other Primates, Encodes a Truncated, Nonfunctional Protein*

Flavin-containing monooxygenases (FMOs) are NADPH-dependent flavoenzymes that catalyze the oxidation of heteroatom centers in numerous drugs and xenobiotics. FMO2, or “pulmonary” FMO, one of five forms of the enzyme identified in mammals, is expressed predominantly in lung and differs from other FMOs in that it can catalyze the N-oxidation of certain primary alkylamines. We describe here the isolation and characterization of cDNAs for human FMO2. Analysis of the sequence of the cDNAs and of a section of the corresponding gene revealed that the major FMO2 allele of humans encodes a polypeptide that, compared with the orthologous protein of other mammals, lacks 64 amino acid residues from its C terminus. Heterologous expression of the cDNA revealed that the truncated polypeptide was catalytically inactive. The nonsense mutation that gave rise to the truncated polypeptide, a C → T transition in codon 472, is not present in theFMO2 gene of closely related primates, including gorilla and chimpanzee, and must therefore have arisen in the human lineage after the divergence of the Homo and Pan clades. Possible mechanisms for the fixation of the mutation in the human population and the potential significance of the loss of functional FMO2 in humans are discussed.

the C(4a)hydroperoxide derivative of the flavin cofactor, exists stably within the protein in the absence of substrate, thereby enabling the enzyme to oxidize any soft nucleophile able to gain access to the active site. FMO was originally characterized by using a homogeneous enzyme preparation isolated from pig liver (3). Proteins with equivalent catalytic properties and substrate profiles were later isolated from the livers of several other species (4 -6). However, contemporaneous studies indicated that rabbit lung contained a form of FMO which, although clearly related to the liver enzyme, possessed distinct physicochemical properties (7) and substrate preferences (8,9). The purification of this "lung" or "pulmonary" FMO (10 -12) confirmed the existence of two distinct FMOs, distinguishable both immunochemically and by substrate preference (10 -12), and the subsequent isolation and characterization of the corresponding cDNA clones (13) demonstrated that each was the product of a separate gene.
Following the identification of "liver" and "lung" FMO as discrete enzymes, evidence accumulated indicating the presence in these tissues of additional forms of the enzyme. The existence of multiple mammalian FMOs was subsequently confirmed when new forms were identified, either by direct sequencing of purified proteins (14,15), or via the isolation and characterization of cDNA clones (16 -18). To date, five distinct forms of FMO, designated FMOs 1-5, each of which is encoded by its own gene and which exhibits approximately 50 -60% pairwise amino acid sequence similarity, have been identified in mammals (19). According to the present nomenclature (19) "liver" and "lung" FMO are designated FMO1 and FMO2, respectively, whereas the forms identified subsequently have been designated FMOs 3,4,and 5.
Although detected at other sites, such as the kidney (13,20,21), FMO2 is expressed predominantly in pulmonary tissue (10, 11, 13, 20 -23) where, in the rabbit, it has been localized to the nonciliated bronchiolar epithelial (clara) cells and endothelial type I and II cells (24). FMO2 gene expression has been demonstrated to be regulated by sex hormones in experimental animals (25,26) and putative glucocorticoid responsive elements have been identified in the 5Ј-flanking region of the rabbit FMO2 gene (27). FMO2 displays catalytic characteristics that distinguish it from other forms of FMO. For instance, although able to oxidize many typical FMO substrates, it is inactive toward certain tertiary amines, such as imipramine and chlorpromazine (9,10), that are good substrates for other forms of the enzyme. Furthermore, in contrast to other FMOs, FMO2 is capable of mediating the N-oxidation of some primary alkylamines to their oximes, via an N-hydroxylamine intermediate (28,29).
We have previously reported the isolation of cDNAs encoding human FMOs 1 (30), 4 (16) (previously designated FMO2), and 3 (31) and have determined that the corresponding genes, plus the genes encoding human FMOs 2 and 5, are all located on the long arm of chromosome 1 (16,30,32,33). In this report we describe the isolation and characterization of cDNA clones encoding human FMO2 and demonstrate that, in common with human FMOs 1 and 3 (31), expression of the corresponding gene is subject to both ontogenic and tissue-specific regulation. Furthermore, we report that, in contrast to apparently all other mammalian species, including closely related primates, in humans the major FMO2 allele encodes a truncated polypeptide which is catalytically inactive.

EXPERIMENTAL PROCEDURES
Oligonucleotide Synthesis-Oligonucleotides were synthesized on a PCR-MATE DNA synthesizer (model 391, Applied Biosystems, Warrington, UK).
Isolation of RNA and Genomic DNA-Human adult and fetal tissue samples were obtained as described previously (31) except for single adult lung and kidney samples, which were obtained from the International Institute for the Advancement of Medicine (Exton, PA). Total RNA was prepared by the guanidinium thiocyanate/LiCl method (34), resuspended in diethylpyrocarbonate-treated water and stored in aliquots at Ϫ80°C. RNA concentration was determined from absorbance at 260 nm. Human genomic DNA was isolated from whole blood by the method of Lahiri and Nurnberger (35) and from solid tissue by use of a commercial DNA isolation kit (Nucleon Biosciences, Coatbridge, Scotland). Gorilla and chimpanzee genomic DNA was isolated by the method of Blin and Stafford (36) from post-mortem tissue samples.
Amplification of a Partial-length cDNA Encoding Rabbit FMO2-Reverse-transcription of rabbit (New Zealand White) lung total RNA (20 g) was primed with 200 pg of random hexamer oligonucleotides (Pharmacia Biotech, St. Albans, UK) and catalyzed by 200 units of Moloney murine leukemia virus reverse-transcriptase (Life Technologies, Paisley, Scotland) in a total volume of 20 l according to the supplier's recommendations. The reaction mix was then incubated for 5 min at 75°C and the volume increased to 100 l with water. Oligonucleotides 102 (5Ј-GACGCAGTCATGGTCTGCAGTGGC-3Ј) and 180 (5Ј-GATGTAATTGGTCTGCAGTTTCTG-3Ј), designed from the rabbit FMO2 cDNA sequence (13) and which incorporated PstI restriction endonuclease sites present within the rabbit sequence, were used to prime the amplification, by PCR, of 879 bp of coding region. The PCR was performed in a volume of 40 l containing 15 pmol of each primer, 66.7 mM Tris-HCl (pH 8.4), 1.25 mM MgCl 2 , 16.7 mM (NH 4 ) 2 SO 4 , 0.1% (v/v) Tween 20, 50 M of each dNTP, and a 5-l aliquot of the reversetranscription product. After an initial denaturation at 94°C for 1 min, 0.5 l (2.5 units) of Taq DNA polymerase (BioLine, London, UK) was added, and the reaction mix was incubated (TR1, Hybaid, Ashford, UK) for 32 cycles at 94°C for 45 s, 58°C for 45 s, and 72°C for 1 min, followed by an additional 5 min at 72°C. The PCR product was incubated with PstI, gel-purified, and ligated into PstI-digested pBS (Stratagene, Cambridge, UK) to give the plasmid pRABLUNG. The identity of the cDNA insert was confirmed by restriction mapping and partial DNA sequencing.
cDNA Library Screening-The insert from pRABLUNG was excised by incubation with PstI, gel-purified, radiolabeled by the oligonucleotide random primer method (37) to a specific radioactivity of approximately 10 9 cpm/g with [␣-32 P]dCTP (800 Ci/mmol, Amersham International, Amersham, UK), and used to screen an adult human lung cDNA library constructed in gt11 (gift of Dr. K. Reid) plated at a density of 1.5 ϫ 10 5 plaque-forming units per 20 ϫ 20-cm plate (Nalge Nunc, Hereford, UK). Duplicate filters were prehybridized, hybridized, washed, and subjected to autoradiography as described previously (30). Because of loss of the EcoRI cloning sites during library construction, insert DNAs could not be excised from positive phage clones by restriction digestion and were therefore recovered by PCR as described previously (30).
Amplification of Full-length cDNAs Encoding Human and Macaque FMO2-Reverse-transcription of human or macaque lung total RNA (10 g) and the subsequent removal of RNA with RNases A and H were performed as described previously (31), except that reverse-transcription was catalyzed by 200 units of SuperScript II reverse-transcriptase (Life Technologies) in a volume of 20 l according to the supplier's recommendations.
Oligonucleotides 392 (5Ј-CTAGAATTCTAGAATGGCAAAGAAGG-TAGCTGTGATTG-3Ј) and 394 (5Ј-TACTGGATCCTGACAAGATAATA-AAGCCCAAAG-3Ј), containing unique XbaI and BamHI restriction enzyme sites, respectively, were designed to prime the amplification of the entire protein-coding region of the human FMO2 cDNA plus a small stretch of the 3Ј-untranslated region. PCR was performed in a volume of 40 l containing 15 pmol of each primer, 10 mM KCl, 10 mM (NH 4 ) 2 SO 4 , 20 mM Tris-HCl (pH 8.8), 2 mM MgSO 4 , 0.1% (v/v) Triton X-100, bovine serum albumin (100 g/ml), 200 mM of each dNTP, and a 2-l aliquot of the reverse-transcription product. After an initial denaturation step at 94°C for 1 min, 1 l (2.5 units) of Pfu DNA polymerase (Stratagene) was added and the reaction incubated for 36 cycles at 94°C for 75 s, 52°C for 75 s, and 72°C for 4 min, and then for an additional 10 min at 72°C. The resulting PCR product was incubated with XbaI and BamHI, gel-purified, and ligated into pBS, previously digested with XbaI and BamHI, to give the plasmid pBSFMO2/2. Amplification of the entire protein-coding region of the cynomolgus macaque FMO2 cDNA, plus small stretches of both the 5Ј-and 3Ј-untranslated regions, was achieved with the forward primer 618 (5Ј-CGTTGAATTCCAAGGGAGAAAACTATTCTGTC-3Ј), designed from the sequence of the 5Ј-RACE-PCR products (data not shown), and the human FMO2 cDNA-specific reverse primer 394. PCR conditions were as described above for the amplification of the partial-length rabbit FMO2 cDNA, except that a 5-l aliquot of the reverse-transcription product was used as template and the reaction mix was incubated for 35 cycles at 94°C for 30 s, 56°C for 1 min, and 72°C for 2 min, followed by an additional 5 min at 72°C. The PCR product was purified from the reaction mix (SpinBind, FMC) and sequenced directly.
Amplification of a Region of the FMO2 Gene-An equivalent 235-bp region of the human, gorilla (Gorilla gorilla), and chimpanzee (Pan troglodytes) FMO2 gene was amplified by PCR using the primers 422 (5Ј-TTCAAAGATCCTAAACTGGCTGTGAG-3Ј) and 394. Reaction mixtures containing approximately 200 ng of genomic DNA as template were incubated for 38 cycles at 94°C for 45 s, 56°C for 45 s, and 72°C for 45 s, followed by an additional 5 min at 72°C. Reaction conditions were as described above for the amplification of the partial-length rabbit FMO2 cDNA. PCR products were purified from the reaction mix (SpinBind, FMC) and sequenced directly.
DNA Sequencing-Plasmid DNA and purified PCR products were sequenced by the dideoxy chain-termination method (39) using DNA sequencing kits (Sequenase II, Amersham International; Cyclist, Stratagene) and either vector-or insert-specific primers. The insert of pBSFMO2/2 and the subcloned products of the human FMO2 3Ј-RACE-PCRs and of the human and cynomolgus macaque FMO2 5Ј-RACE-PCRs were sequenced completely on both strands ( Fig. 1), whereas the full-length cDNA encoding cynomolgus macaque FMO2 was only partially sequenced. Sequence data were analyzed with MacVector (Oxford Molecular, Oxford, UK) or Genetics Computer Group (Madison, WI) sequence analysis software.
RNase Protection Assay-The human FMO2 RNase protection plasmid, pBSFMO2/2/15, was constructed by excising a 985-bp section from within the insert of pBSFMO2/2 (by digestion with HindIII and StyI), followed by blunt-ending of the DNA with T4 DNA polymerase and self-ligation of the larger vector fragment. pBSFMO2/2/15 (10 g) was linearized by digestion with MscI, after which the reaction mix was treated with proteinase K and SDS (40), and extracted with phenolchloroform (1:1, v/v). The linearized plasmid was ethanol-precipitated and resuspended in diethylpyrocarbonate-treated water. In vitro synthesis of radiolabeled antisense RNA, probe purification, and RNase protection assays were performed as described previously (31,40,41). Comparison of the autoradiographic signal derived from the protected species with a standard curve of undigested probe permitted quantification of FMO2 mRNA in terms of molecules/g of total RNA. This was converted to molecules/cell by using the average RNA content of a mammalian cell (5 pg) (42).
Northern Blot Hybridization-RNA samples (15 g) were denatured in formaldehyde and electrophoresed through a 1% agarose gel (43). After transfer to a nylon membrane (BDH, Poole, UK) RNA was immobilized by baking the membrane at 80°C, followed by UV cross-linking (Stratalinker 1800, Stratagene). The membrane was prehybridized, hybridized, washed, and subjected to autoradiography as described above.
Construction of Recombinant Baculovirus-To construct FMO2X472 recombinant baculovirus, pBSFMO2/2 was incubated with XbaI and PstI, and the smaller of the resulting restriction fragments, comprising the 1413-bp ORF, 192 bp of associated 3Ј-untranslated region, and approximately 50 bp of pBS, was ligated to XbaI/PstI-digested pFastBac (Life Technologies) to give pFastFMO2/2/8. To construct recombinant virus containing FMO2Q472, two oligonucleotides, 576 (5Ј-CGAGCTC-GCCTTAGAGATAGGTGCG-3Ј) and 394, were used to prime the amplification of a 330-bp contiguous region of the FMO2 gene that encompassed codon 472 and the second in-frame stop signal located 192 bp downstream (see Fig. 2). PCR conditions were as described above for FMO2 gene amplification, except that genomic DNA (100 ng) isolated from an individual previously identified as being heterozygous for TAG and CAG triplets at codon 472 was used as a template. The PCR product was ligated directly into EcoRV-linearized pBS, and, after propagation in Escherichia coli, a plasmid clone containing CAG at codon 472 in the amplified region of the FMO2 gene was identified by DNA sequencing. The insert of this plasmid was excised by incubation with SacI and BamHI, gel-purified, and ligated, in a three-fragment reaction, with a SacI/XbaI restriction fragment of pBSFMO2/2 that contained the remainder of the FMO2 coding region, and XbaI/BamHI-digested pBS, to give pBSFMO2/2/16. The extended FMO2 ORF was excised from pBSFMO2/2/16, by incubation with XbaI and PstI, then gel-purified and ligated into pFastBac to give pFastFMO2/2/16.
In order to bring the translational initiation codon of each of the inserted cDNAs closer to the position of that of the original polyhedrin gene, approximately 50 bp were excised from the multiple cloning sites of both pFastFMO2/2/8 and pFastFMO2/2/16 by digestion with EcoRI and XbaI. After religation of the blunt-ended vector portions to give pFastFMO2X472 and pFastFMO2Q472, the respective recombinant baculoviruses, AcFMO2X472 and AcFMO2Q472, were generated by transfection of Spodoptera frugiperda (Sf) 9 cells with the corresponding recombinant bacmid DNA, obtained via site-specific transposition using the Bac-to-Bac system (Life Technologies) according to the supplier's recommendations.
Baculovirus-mediated Expression of FMO2X472 and FMO2Q472-Sf9 and Tricoplusia ni insect cells were grown and passaged in shaker cultures using Sf-900II (Life Technologies) and Excell 405 (JRH Biosciences-Europe, Marlow, UK) serum-free media, respectively, containing amphotericin B (5 g/ml), penicillin G (100 units/ml), and streptomycin sulfate (50 g/ml). For amplification of virus, Sf9 cells were infected at a multiplicity of infecton of Յ0.1. For expression, 100 ml of T. ni cells, grown to a density of 1 ϫ 10 6 cells/ml in a 500-ml Erlenmeyer flask, were infected with virus at a multiplicity of infecton of 7 and incubated by shaking at 130 rpm in an orbital shaker (New Brunswick Scientific (UK) Ltd., Hatfield, Herts) at 28°C for 60 h. Cells were pelleted, resuspended in 30 ml 0.154 M KCl, 50 mM Tris-HCl (pH 7.4), 0.2 mM phenylmethylsulfonyl fluoride, and subjected to two 30-s bursts of sonication (Dynatech, model 150) on ice. The cell lysate was centrifuged at 10,000 ϫ g for 15 min at 4°C. The microsomal fraction was obtained by centrifugation of the resulting supernatant at 100,000 ϫ g for 1 h at 4°C. Microsomal pellets were resuspended in 5 ml of 0.154 M KCl, 10 mM HEPES (pH 7.5), 1 mM EDTA, 20% (v/v) glycerol, and stored in aliquots (150 l) at Ϫ80°C until use. Protein concentration was determined by the BCA method (Pierce & Warriner (UK) Ltd., Chester, UK) using bovine serum albumin as standard.
Western Blotting-Western blotting was performed as described previously (44). Blots were incubated with goat anti-(rabbit FMO2) serum (1 in 3000 dilution; gift of R. Philpot) and a rabbit anti-(goat immunoglobulin G)-alkaline phosphatase conjugate (Sigma Chemical Company (UK), Poole, UK), and antigen visualized (Color Development Kit, Bio-Rad Laboratories Ltd., Hemel Hempstead, UK). The concentration of each expressed FMO2 was determined by scanning densitometry (Bio-Rad, model GS-670) using a standard curve of authentic rabbit FMO2 (gift of R. Philpot).
FMO Assays-FMO activity was determined using the substrate methimazole according to the method of Dixit and Roche (45). The assay was performed in a 1-ml cuvette, held at 37°C, containing 0.1 M Tris-HCl (pH 8.4), 1 mM EDTA, 0.02 mM dithiothreitol, 0.06 mM 5,5-dithiobis-(2-nitrobenzoate), 0.1 mM NADPH, 2 mM methimazole, and microsomal membranes. When used, MgCl 2 was added to both sample and reference cuvettes. The reaction was started by addition of methimazole (in 10 l of buffer) to the sample cuvette and 10 l of buffer to the reference cuvette. Reaction buffer was aerated for 30 min before use, and reaction mixtures were allowed to equilibrate for 2 min after addition of substrate. The progress of the reaction was monitored by measuring the decrease in the concentration of nitro-5-thiobenzoate, as revealed by decreased absorbance at 412 nm.

The Human FMO2 Gene Contains a Premature In-frame
Termination Codon-A screen of 6 ϫ 10 5 plaques of a human lung cDNA library with the insert of pRABLUNG, which comprised an 879-bp fragment of rabbit FMO2 cDNA, yielded four positive clones, L1, L2, L3, and L8, containing inserts of 0.6, 2.0, 1.6, and 0.9 kb, respectively. Sequence analysis confirmed that the inserts of all four clones contained sequences that encoded the human ortholog of FMO2. However, two of the clones, L2 and L3, also contained sequences that were not related to FMO2. A search of the GenBank TM data base revealed that the latter sequences encode pulmonary surfactantassociated protein and mitochondrial 12 S ribosomal RNA, respectively. In each case, the FMO2 cDNA sequence was found to be joined to the unrelated sequence by an EcoRI restriction site, indicating that both clones were chimeric artifacts formed during library construction.
A consensus sequence for the human FMO2 cDNA derived from these four clones contained a short stretch of 5Ј-untranslated region and almost 1100 bp of protein-coding region (Fig.  1). The latter region was considerably shorter than the corresponding regions of FMO2 cDNAs of rabbit and guinea pig and did not contain an in-frame stop codon. To obtain sequence encoding the carboxyl-terminal portion of human FMO2 we carried out 3Ј-RACE-PCR. Two independent 3Ј-RACE-PCR products (Fig. 1) were found to have identical sequences, which contained the remainder of the protein-coding region and 45 bp of an apparent 3Ј-untranslated region. To confirm and extend the sequence of the 5Ј-untranslated region we performed 5Ј-RACE-PCR. The products of two independent 5Ј-RACE-PCRs (Fig. 1) were each found to contain 60 bp of 5Ј-untranslated region, the sequences of which were identical with each other and with the overlapping regions of the shorter stretches of 5Јuntranslated region contained within the clones L3 and L8. A cDNA, FMO2/2 (Fig. 1), containing the entire protein-coding region of FMO2, plus a short stretch of 3Ј-untranslated region, was obtained by reverse-transcription PCR.
The sequence of human FMO2 cDNA (Fig. 2) was assembled from the partial sequences of the phage clones L1, L2, L3, and L8, the complete sequences of the 5Ј-and 3Ј-RACE-PCR products and the complete sequence of the full-length cDNA obtained by reverse-transcription PCR. The sequence contains 60 bp of 5Ј-untranslated region, followed by an ORF of 1413 bp, a TAG translational termination codon and 237 bp of an apparent 3Ј-untranslated region that lacks a consensus polyadenylation signal. The high degree of sequence identity with FMO2 cDNAs of rabbit (87%) (13) and guinea pig (86%) (46) confirmed the identity of the human clone. However, in contrast to FMO2 proteins of rabbit (13), guinea pig (46) and rhesus macaque (23), each of which contains 535 amino acid residues, the ORF of the human cDNA that we have isolated encodes a polypeptide of only 471 residues, with a calculated molecular mass of 53,639 daltons. Comparison of the human FMO2 cDNA sequence with the corresponding sequences of rabbit, guinea pig, and macaque (data not shown) revealed that the TAG codon terminating the human ORF aligns with a CAG codon, encoding Gln-472, in the nucleotide sequences of FMO2s of the other three species. The human cDNA contains another in-frame TAG triplet, 192 bp downstream of that at codon 472. This second in-frame stop codon aligns exactly with the termination codons of FMO2s of rabbit, guinea pig, and rhesus macaque. As the upstream TAG triplet was present in both of the 3Ј-RACE-PCR products and in the full-length FMO2 cDNA obtained by reverse-transcription PCR it is unlikely to be the result of a cloning artifact.
To confirm the presence of the apparent premature stop codon in the human FMO2 gene, a region of the gene encompassing both stop codons and part of the 3Ј-untranslated region was amplified from each of 12 unrelated individuals and the resulting PCR products directly sequenced. All 12 were found to be homozygous for a TAG triplet at codon 472 ( Fig. 3A and  data not shown). Therefore, in humans, the FMO2 gene con-tains a premature stop codon and thus encodes a polypeptide that is 64 amino acid residues shorter than FMO2s of other mammalian species such as rabbit, guinea pig, and macaque monkey. Southern blot hybridization of human genomic DNA revealed that the full-length cDNA hybridized to fragments of 7.0, 6.4, 4.7, 3.7, and 3.1 kb in EcoRI-digested DNA (Fig. 4, lane  1), and to fragments of 9.2, 7.2, and 6.9 kb in HindIII-digested DNA (Fig. 4, lane 2). This gives an estimated minimum size for the human FMO2 gene of approximately 23 to 25 kb, similar to that of the human FMO3 gene (47) and suggests that humans contain only one FMO2 gene.
Analysis of Human FMO2 mRNA-The sequence flanking the proposed translational initiation codon of human FMO2 mRNA, GAGCUGAUGG (initiation codon underlined) (Fig. 2), conforms poorly to the consensus sequence, GCC(AG)C-CAUGG, for efficient initiation of translation of vertebrate mRNAs (48). Of particular note is the lack of a purine at the Ϫ3 position (the A of the AUG codon is designated ϩ1), as pyrimidines at this position are rarely associated with functional initiation codons (48,49). However, the sequence contains a G at the ϩ4 position, which has been shown experimentally (50) to compensate for the otherwise unfavorable presence of a pyrimidine at Ϫ3. The remaining 5Ј-untranslated sequence of the mRNA contained no additional AUG triplets which might serve as alternative translational initiation codons (Fig. 2). The region flanking the initiation codon of the FMO2 mRNA of cynomolgus macaque, determined from the sequence of a 5Ј-RACE-PCR product (data not shown), also has a poor match to the consensus Kozak sequence, as it differs from the human sequence at only the Ϫ4 position, at which an A replaces a G. In contrast, the Kozak regions of mRNAs encoding FMO2s of rabbit, GAGACGAUGG, and guinea pig, GGGGCAAUGG, conform better to the consensus and contain a purine at the important Ϫ3 position. Northern blot hybridization of total RNA isolated from the lungs of three human individuals revealed the presence of a single species of FMO2 mRNA of 5.7 kb (Fig. 5, lanes 1-3), which is similar in size to the mRNAs encoding FMO2 of macaque (5.0 kb) (Fig. 5, lane 4) (23), rabbit (a single major transcript of 4.8 kb and minor transcripts of 6.0, 2.6, and 2.4 kb) (13), guinea pig (6.0 kb) (46), and pig (6.0 kb) (46). The human FMO2 mRNA thus contains a 3Ј-untranslated region of approximately 4 kb. The much shorter length of 3Ј-untranslated sequence in the 3Ј-RACE-PCR products that were obtained ( Figs. 1 and 2), together with the absence of a consensus polyadenylation signal, suggests that the reverse-transcription step of the 3Ј-RACE procedure was primed from an internal adenosine-rich sequence rather than from the poly(A) tail of the mRNA. In support of this, both rabbit and guinea pig FMO2 cDNAs contain adenosine-rich sequences at the corresponding positions within their respective 3Ј-untranslated regions (13,46).
The Human FMO2 Gene Is Subject to Developmental and Tissue-specific Regulation-The tissue distribution and developmental expression of the mRNA encoding human FMO2 was investigated by quantitative RNase protection. A representative autoradiogram is shown in Fig. 6A and demonstrates that a discrete protected RNA, of the expected size, was obtained from the FMO2 antisense RNA probe. FMO2 mRNA was expressed relatively abundantly (13-36 molecules/cell) in adult lung samples from four different individuals (Fig. 6B). It was also expressed, but in lower abundance (3 molecules/cell), in one of two adult kidney samples analyzed (Fig. 6A and data not  shown), but was not detected in the other kidney sample or in either of two adult liver samples, even after prolonged exposure. In fetal tissues, FMO2 mRNA was not detected in samples of liver, kidney or brain (data not shown) but was present in moderate abundance (5-12 molecules/cell) in each of four lung samples (Fig. 6B).
The Nonsense Mutation in the Human FMO2 Gene Occurred after the Pan and Homo Lines Diverged-The FMO2 gene of rhesus monkey encodes a polypeptide of 535 amino acid residues (23), the same length as the orthologous protein of rabbit (13) and guinea pig (46). Analysis of the sequence of cDNA for FMO2 of the cynomolgus monkey, a close relative of the rhesus monkey, revealed that a stop codon was not present at position 472 (data not shown). The mutation that gave rise to the premature stop codon present in the human FMO2 gene must therefore have occurred within the Hominoidea, sometime after this primate superfamily, which includes apes and humans, diverged from the Cercopithecoidea superfamily, which includes the Old World monkeys. To determine at what stage during the evolution of the Hominoidea the mutation occurred, a 235-bp region of the FMO2 gene encompassing codon 472 was amplified by PCR from genomic DNA of gorilla and chimpanzee. Within this region the sequence of the FMO2 gene of these species is 98% identical to that of the corresponding region of the FMO2 gene of humans (data not shown). Both of these non-human hominoids were homozygous for a CAG triplet, encoding Gln, at codon 472 of FMO2 (Fig. 3B and data not shown) and contained a translational stop codon at position 536 (data not shown), corresponding to that present in FMO2 of rabbit, guinea pig, and rhesus and cynomolgus monkeys. The mutation that gave rise to the truncated FMO2 of present day humans must therefore have occurred in the human lineage sometime after the divergence, some 4 -5 million years ago, of humans from their closest relative, the chimpanzee. Subsequent analysis, with a PCR-restriction enzyme assay, of individuals of various racial and ethnic backgrounds, including 27 European Caucasians, 18 Orientals (10 Japanese, 8 Chinese), 41 of African descent (20 Africans, 16 African-Americans, 5 UK Afro-Caribbeans), 6 New Guinea Aboriginals, 2 Indians and 2 Maoris, revealed that the allele encoding the truncated FMO2, FMO2X472, occurs at a frequency of essentially 100% in all groups investigated, with the exception of populations of African descent, in which an allele containing a CAG triplet at codon 472 (Q472) is present at a frequency of approximately 4%. 2 The Truncated Protein Encoded by the Human Gene Is Catalytically Inactive-FMO2X472, the truncated form encoded by the major FMO2 allele of humans, and FMO2Q472, a "fulllength" form containing 535 amino acid residues (see Fig. 2), were produced via heterologous expression of the corresponding cDNAs in insect cells via the baculovirus expression system. Western blotting with antibody against rabbit FMO2 detected proteins of 53 and 57 kDa, respectively, in microsomal membranes isolated from cells infected with virus containing cDNA encoding FMO2X472 (AcFMO2X472) (Fig. 7A, lane 5), or FMO2Q472 (AcFMO2Q472) (Fig. 7A, lane 6). Authentic rabbit FMO2 (Fig. 7A, lane 7) had an estimated molecular mass of 56 kDa. The concentrations of heterologously expressed FMO2X472 and FMO2Q472 were 14 and 19 pmol/mg microsomal membrane protein, respectively. Immunoreactive protein was undetectable in microsomal membranes isolated from noninfected insect cells (Fig. 7A, lane 1) or from cells infected with wild-type virus, AcNPV (Fig. 7A, lane 2), or in the cytosolic fractions of cells infected with AcFMO2X472 (Fig. 7A, lane 3) or AcFMO2Q472 (Fig. 7A, lane 4).
The function of FMO2X472 and FMO2Q472 was investigated by determining their ability to catalyze the S-oxidation of methimazole, an excellent substrate for FMO2s of other mammalian species (Fig. 7B). Microsomal membranes isolated from cells infected with AcFMO2Q472 catalyzed the S-oxidation of methimazole with essentially maximal specific activity of 0.50 nmol of methimazole oxide formed/min/mg microsomal protein ([S] ϭ 2 mM; K m ϭ 411 M) (51) (k cat app Ϸ 27 min Ϫ1 ). In the presence of 100 mM Mg 2ϩ , a known effector of FMO2 (7), this was increased approximately 6-fold to 2.9 nmol/min/mg microsomal protein (k cat app Ϸ 155 min Ϫ1 ). In contrast, microsomal membranes isolated from noninfected insect cells, or from cells infected with either AcNPV or AcFMO2X472, failed to catalyze methimazole oxidation. Thus, although the full-length human FMO2 that we have produced can catalyze the S-oxidation of methimazole and respond to Mg 2ϩ in similar ways to FMO2s of other mammals, the truncated form of the protein encoded by the major FMO2 allele of humans appears to be catalytically inactive. DISCUSSION Our results demonstrate that the major FMO2 allele of humans encodes a polypeptide that, in comparison with the orthologous protein of rabbit (13), guinea pig (46), and rhesus macaque (23), lacks 64 residues from its carboxyl terminus. This is due to the presence in the human gene of an in-frame TAG translational termination triplet at codon 472, 64 codons upstream of a second in-frame termination signal the position of which corresponds exactly with that present in FMO2 of rabbit, guinea pig, and macaque. As the latter species all contain a CAG triplet, encoding glutamine, at codon 472, it appears that the premature stop codon present in the human FMO2 gene arose as the result of a C to T transition at the first position of codon 472.
The presence of nonsense mutations within prokaryotic and eukaryotic genes is frequently associated with decreased abundance of the corresponding mRNA, due to an increase in the turnover rate of the mutant transcript (52). This process, termed nonsense-mediated mRNA decay (53), is dependent upon the relative position of the mutation within the proteincoding region; mutations located within the region encoding the amino-terminal two-thirds of a polypeptide accelerate degradation, whereas those within regions encoding sequences closer to the carboxyl terminus may have little or no effect upon mRNA stability (54). Consequently, as the nonsense mutation present in the FMO2 mRNA is located in the 3Ј-most one-third of the protein-coding region, it would be expected to exert little effect on the stability of the mRNA. In support of this, we found no evidence of degraded FMO2 mRNA in any of the human lung samples analyzed by Northern blot hybridization, and, furthermore, RNase protection analyses demonstrated that FMO2 mRNA was moderately abundant in all samples of human lung examined. Thus, the primary transcript of the human FMO2 gene is apparently correctly processed in pulmonary tissue to produce a stable and abundant mRNA that should be available for potential translation. The conservation throughout mammalian evolution of a long 3Ј-untranslated region in FMO2 mRNA suggests that it may be important for some aspect of the metabolism or function of the mRNAs, such as stability.
The results of RNase protection assays demonstrate that in humans expression of the FMO2 gene in both the adult and fetus is essentially restricted to the lungs, with the gene being inactive in the liver. The complete absence of hepatic expression of FMO2 has been demonstrated for other species (13,21,23) and confirms that FMO2 is essentially a pulmonary-specific FMO. The higher concentration of the FMO2 mRNA present in adult, compared with fetal, human lung indicates that the gene is also regulated developmentally. Thus, in humans, FMO2, in common with FMO1 and FMO3 (31), is subject to both ontogenic and tissue-specific regulation. The similarities in the size and pattern of expression of FMO2 mRNAs of humans and other species of mammals indicate that the human FMO2 gene has suffered no mutations that affect either the expression of the gene or the processing or stability of the corresponding mRNA.
Heterologous expression studies showed that the loss of 64 amino acid residues from the carboxyl terminus of human FMO2 had no effect on targeting of the protein to the membranes of the endoplasmic reticulum, but abolished its catalytic activity. This is in accord with site-directed mutagenesis studies of rabbit FMO2 (55,56), which showed that, although the removal of up to 26 amino acid residues from the carboxyl terminus of the enzyme had no effect on catalytic activity, a deletion of 40 or more residues inactivates the enzyme. It is possible that the truncated human FMO2 has some other, unidentified, catalytic activity. However, even if this were so, Williams et al. (57) found that FMO2, although abundant in the lungs of rhesus macaque, was undetectable, by Western blotting, in all but one of 29 human lung samples. The results of our Northern blot hybridization and RNase protection experiments demonstrate that FMO2 mRNA is moderately abundant in human lung. Consequently, the absence of FMO2 protein in this tissue cannot be due to deficient gene transcription or RNA processing, or to instability of the mRNA. The relatively poor conformity of the Kozak region of the human FMO2 mRNA to the consensus sequence may compromise the translational efficiency of the mRNA. However, this seems unlikely as the FMO2 of rhesus macaque, which is abundant in the lungs (23,57), is encoded by a mRNA that contains a Kozak region that, apart from a single base difference at a noncritical position, is identical to that of the human mRNA. A more likely explanation for the lack of pulmonary expression of FMO2 in man is that, owing to the lack of 64 residues from its carboxyl terminus, the truncated polypeptide is unable to fold correctly and is thus detected by cellular surveillance systems, such as the ubiquitin pathway (58), and rapidly degraded.
The absence of FMO2 in human lung is in marked contrast to the situation in all other species of mammals investigated, in which it represents the major, if not only, form of FMO in adult lung (10, 11, 20 -23). In humans, the only member of the FMO family present in adult lung is FMO5 (31,59). However, as FMO5 has a very restricted substrate range that is quite distinct from that of any other FMO (18,60), it would be unable to substitute for FMO2. The lung plays a significant role in the metabolism of certain foreign compounds (61,62). Although the pharmacological or toxicological significance of the absence of FMO2 in human lung remains to be established, it is clear that caution should be exercised when extrapolating pulmonary drug metabolism data from experimental animals to man if an FMO-mediated metabolic pathway is suspected.
Recent reports (63,64) have indicated the presence in human brain of an FMO that is immunoreactive with, and catalytically inhibited by, antibodies raised against rabbit FMO2. However, as we have demonstrated here, almost all individuals have two FMO2X472 alleles and would thus express a truncated form of the enzyme, which would be nonfunctional and, quite likely, rapidly degraded. The FMO detected by these workers is therefore unlikely to be FMO2. This is supported by the observation that it is able to effectively catalyze the N-oxidation of the non-FMO2 substrate imipramine (63,64). The precise identity of this FMO is thus unclear.
Our results indicate that the FMO2X472 allele, which encodes a truncated, nonfunctional protein, arose as the result of a point mutation that occurred in the human lineage sometime after the divergence of the Homo and Pan clades took place some 4 -5 million years ago and has subsequently spread to attain a frequency of close to 100% in the present day human population. FMO2 thus represents a very unusual case of a gene that has become non-functional in humans but not in other primates. The average time (in generations) for fixation of a neutral mutation is approximately 4N, where N is the effective population size (65). For humans, N has been estimated to have been approximately 10,000 for most of the last 1 million years (66). Assuming a generation time of about 15 years, the average time required for fixation of a neutral mutation in humans is thus 4 ϫ 10,000 ϫ 15 ϭ 6 ϫ 10 5 years. The FMO2X472 allele could therefore have become virtually fixed in the human population merely due to the effects of random genetic drift on a neutral mutation, i.e. one that conferred no, or very little, selective advantage or disadvantage. An alternative, more intriguing, possibility is that the nonsense mutation may have conferred an evolutionary advantage, resulting in a rapid spread of the mutant allele due to a directional selective sweep.