Expression and characterization of a modified flavin-containing monooxygenase 4 from humans.

The inability to obtain flavin-containing monooxygenase 4 (FMO4) in heterologous systems has hampered efforts to characterize this isoform of the FMO gene family. Neither the human nor the rabbit ortholog of FMO4, each of which has been cloned and sequenced, has been expressed. Attempts to achieve expression of FMO4 have been made with Escherichia coli, baculovirus, yeast, and COS systems. The cDNAs encoding FMO4 have extended coding regions compared with those encoding other FMO isoforms. The derived amino acid sequences of FMO1, −2, −3, and −5 from all species examined contain about the same number of residues (531-535 residues), whereas the derived sequences of human and rabbit FMO4 contain 558 and 555 residues, respectively. We have investigated whether the elongation of the FMO4 coding region is related to the inability to achieve expression. The cDNA encoding human FMO4 has been modified by a single base change that introduces a stop codon at the consensus position. This modification allows for expression in E. coli. Lack of expression of intact FMO4 is caused by a problem that occurs following transcription, a problem that is overcome completely by relocation of the stop codon 81 bases to 5′ of its normal position. Truncated FMO4 is expressed as an active enzyme with characteristics typical of an FMO isoform. Possible functional changes resulting from altering the 3′-end of an FMO were investigated with human FMO3. Elongation of the coding region of the FMO3 cDNA to the next available stop codon (FMO3*) resulted in the expression of an enzyme with properties very similar to those of unmodified FMO3. Elongation of FMO3 lowered the level of expression in E. coli but did not eliminate it. As with FMO4, the difference in expression levels between FMO3 and elongated FMO3 (FMO3*) appears to be related to translation rather than transcription. The functional characteristics of FMO3 and FMO3* are not significantly different.

the monooxygenation of numerous nitrogen-, phosphorous-, or sulfur-containing xenobiotics, including drugs, pesticides, and industrial chemicals (1,2). FMOs convert many xenobiotics into more polar substances as a prelude to excretion but in some cases can also catalyze the formation of reactive metabolites capable of binding to cellular macromolecules. The FMOs are primarily associated with the endoplasmic reticulum of cells located in most organs of all mammalian species examined. Several endogenous compounds have been identified as substrates for FMOs (3), but the physiological role of these enzymes has not been determined. Regulation of disulfide bond formation and metabolism of dietary amines are two possibilities that have been suggested (4,5).
Although FMO1 and FMO2 purified from pig liver and rabbit lung, respectively, have been studied extensively, recombinant technology has greatly aided the characterization of FMO isoforms. Catalytically active FMO1 and FMO2 from rabbit have been expressed in several systems (20,21) and particularly high levels of expression of FMO1, -2, -3, and -5 have been noted in Escherichia coli (11,12,16,21). Our understanding of the human FMO isoforms, in particular FMO3 and FMO5, is based primarily on studies of enzymes expressed in E. coli (16,22) or insect cells (23). In contrast to these successes, attempts to express rabbit FMO4 in E. coli, yeast, or COS-1 cells (12) or human FMO4 in E. coli or insect cells (23), have proven futile. Thus, the catalytic properties and distribution of FMO4, for which no purified enzymes or antibodies are available, have not been determined, although distribution of mRNA suggests that FMO4 may play a role in the brain (24).
A comparison of the human and rabbit FMO4 transcripts with those encoding all known FMO isoforms reveals two differences that might be associated with the inability to express FMO4 in heterologous systems. First, the coding regions of the FMO4 transcripts are 60 -75 nucleotides longer than the coding regions of the other isoforms, owing to a shift in the stop codon to the 3Ј-end of the consensus position (12,15). Second, FMO4 transcripts have a premature start codon (12,15) that may be eliminated from part of the transcript population by alternative splicing in the 5Ј-flanking region (12).
The present work demonstrates that an active form of human FMO4 can be expressed in E. coli transformed with a truncated cDNA. The recombinant enzyme has been characterized and found to exhibit the functional properties associated with other members of the FMO gene family. In addition, evidence is presented in contradiction to our previous results that suggested alternative splicing in the 5Ј-region of the FMO transcript.

MATERIALS AND METHODS
Cloning of Human FMO3 and FMO4 -A human cDNA library constructed from hepatic mRNA isolated from the liver of an adult male has been described previously (16). The titer of the library was 1 ϫ 10 6 plaques/l. The library was screened with a 620-base pair 5Ј-fragment (EcoRI) of the cDNA for rabbit FMO3 (12) random labeled (25) with [␣-32 P]dCTP (Boehringer Mannheim). Plaque lifts were prehybridized and hybridized in 6 ϫ SSC, 100 g/ml salmon sperm, 4 ϫ Denhardt's solution, 0.5% SDS, and 50% formamide. Subsequent to overnight hybridization at 37°C, the lifts were washed twice under conditions of low stringency (1 ϫ SSC and 0.1% SDS) for 30 min at 37°C. Fifteen positive plaques were isolated to purity and characterized by restriction mapping based on published sequences (14 -16) and end sequencing with T3 and T7 primers. Seven clones containing complete coding regions were identified, four for FMO3, one for FMO4, and two for FMO5.
Sequencing of Human FMO3 and FMO4 -Oligonucleotide primers were used to obtain sequences from both strands of three full-length FMO3 clones and one FMO4 clone (26). Primers (19 for FMO3 and 10 for FMO4) were based on published sequences (14,15). Primers for these experiments and others described below were synthesized in our laboratory (381A PCR MATE; Applied Biosystems, Foster City, CA). Sequence data were analyzed and aligned with the software package from Genetics Computer Group, Inc. (27) and by comparison with published data (14,15).
Preparation of E. coli Expression Vector (pJL2)-FMO cDNA Constructs-A 1715-base fragment of the cDNA encoding FMO3 (coding region plus 115 bases of the 3Ј-flank and a single base (C) 5Ј of the start codon) was excised from pBluescript (Stratagene, La Jolla, CA) with NcoI and ScaI and inserted into pJL-2 (FMO3-pJL) restricted with NcoI and EcoRV. The vector, pJL-2, is a derivative of pKK223-2 (Pharmacia Biotech Inc.) in which the origin of replication is changed, and a translation enhancer sequence is inserted between the ribosome binding site and the start codon (28). The coding regions of elongated FMO3 (FMO3*), normal FMO4, and truncated FMO4 (FMO4*) were isolated by PCR (GeneAmp kit and thermocycler; Perkin-Elmer) and inserted into the pJL-2 vector for expression in E. coli. The ends of the cDNA fragments were modified by the PCR procedure to contain 5Ј-NcoI (FMO4 and FMO4*) or XbaI (FMO3*) and 3Ј-HindIII restriction sites. For amplification of FMO4, the sense primer was 5Ј-GCGGCCATGGC-CAAGAAAGTTGCAGTG-3Ј and the antisense primer was 5Ј-GCG-CAAGCTTTCATCCTCGCCAAAGAC-3Ј. FMO4* was generated with the same sense primer and an antisense primer (5Ј-GCGCAAGCTT-TCAACAGATAAGTAGAA-3Ј) that altered codon 531 from AAA (lysine) to TGA (stop). The FMO4 and FMO4* amplified fragments were digested with NcoI and HindIII and inserted into expression vector pJL2 (FMO4-pJL and FMO4*-pJL) restricted with the same enzymes. These constructs encoded proteins of 558 (FMO4-pJL) and 530 (FMO4*-pJL) amino acids.
The FMO3* fragment was obtained with a 5Ј-GCGCTCTAGAAT-GGGGAAGAAAGTGGCC-3Ј sense primer and a 72-base antisense primer corresponding to the sequence extending from the normal stop codon 72 bases toward the 3Ј-end, the position of the next naturally occurring stop codon. This antisense primer was also used to alter the normal stop codon from TAA (stop) to TAT (tyrosine). The FMO3* amplified fragment was digested with XbaI and HindIII and ligated into pJL-2 (FMO3*-pJL) digested with the same enzymes.
Expression of FMO3, FMO3*, FMO4, and FMO4* in E. coli-E. coli were transformed with recombinant FMO3-pJL, FMO3*-pJL, FMO4-pJL, or FMO4*-pJL or nonrecombinant pJL-2 vector and grown at 37°C in LB medium plus ampicillin (50 g/ml) to an absorbance of 0.4 -0.5 at 600 nm. Isopropyl ␤-D-thio-galactopyranoside (IPTG) was then added to a final concentration of 1 mM, and the cells were grown overnight at 30°C. Cells were harvested by centrifugation at 2,000 ϫ g for 5 min (all centrifugation steps were done at 4°C unless specified otherwise) and resuspended in 10 ml of ice-cold lysis buffer (100 mM KCl, 50 mM KP i , pH7.4, and 1 mM EDTA) containing lysozyme (1 mg/ml). After incubation for 30 min on ice with occasional gentle inversions, the cells were harvested (2,000 ϫ g for 5 min), resuspended in 2 ml of lysis buffer, and sonicated (Branson 185 Sonifier Cell Disrupter with microtip) five times (30 s at full power) with 30 s cooling between each treatment). Cell debris was removed (2,000 ϫ g for 12 min), and membrane fragments were recovered from the supernatant fractions (100,000 ϫ g for 30 min). The membrane fractions were suspended in lysis buffer (3 ml) and repelleted (100,000 ϫ g for 30 min). The final membrane fraction was suspended in 50 mM KP i , pH 7.4, containing 20% glycerol and 1 mM EDTA. Samples were stored at Ϫ70°C.
Isolation of RNA from E. coli-E. coli transformed with recombinant FMO3, FMO3*, FMO4, or FMO4* or nonrecombinant pJL2 were incubated overnight at 30°C with IPTG (1 mM) and ampicillin (50 g/ml) in LB medium. Total RNA was isolated with PUREscript RNA isolation kits (Gentra Systems, Inc., Minneapolis, MN). The purity of the isolated RNA was assessed by determination of the A 260 -280 ratio and by visualization with ethidium bromide following electrophoresis on agarose.
Analysis of E. coli RNA-Samples of RNA were subjected to electrophoresis on 1% agarose gels containing methyl mercury (29) and transferred to nylon membranes (Nytran; Schleicher & Schuell). The membranes were washed with 5 ϫ SSC for 5 min at room temperature and treated with UV irradiation (Stratalinker 2290; Stratagene) three times at 120,000 mJ/cm 2 (30 s each time with 30-s intervals). After 2 h of prehybridization at 42°C, hybridization was continued overnight at 42°C. Then, blots were washed twice with 0.1 ϫ SSC and 0.1% SDS at room temperature for 30 min, followed by overnight exposure at Ϫ70°C for autoradiography.
Analytical Methods-Membrane proteins were electrophoresed on polyacrylamide gels in the presence of SDS (30) and analyzed by staining with Coomassie Blue or by immunoblotting following transfer to nitrocellulose (31). Immunoblots were developed by the method of Towbin et al. (31) as modified by Domin et al. (32) and were stained with antibodies (goat IgG) to rabbit FMO3 purified from E. coli by methods described previously (16). These antibodies, prior to back-absorption with E. coli particulate fractions containing expressed FMOs, crossreact with FMOs 1, 2, and 5.
Activity of the expressed enzymes was determined spectrophotometrically (Aminco DW2A UV-visible spectrophotometer) in 0.1 M Tricine, pH 8.4, and 1 mM EDTA with methimazole as the substrate (33). Flavinadeninedinucleotide contents were determined in washed particulate fractions by the method of Faeder and Siegel (34), and protein concentrations were determined by the method of Lowry et al. (35). Calculations of active enzyme concentrations were done on the basis of a 1:1 molar ratio for protein and flavin.

RESULTS
Sequences of FMO4 and FMO3-The sequence we obtained for human FMO4 differs with that published by Dolphin et al. (15) at only two positions: C for T in the second base of codon 323 (changes valine to alanine) and C for T in the second base of codon 502 (changes valine to alanine). These differences could easily be related to several factors, including allelic variation, polymerase infidelity, and sequencing errors. In contrast, our sequence for human FMO3 differs markedly from the sequence of human FMO3 obtained from cDNA (14) 2 and a PCR product used for expression by Lomri et al. (36). Alignment of these sequences shows 14 base substitutions and 17 positions where alignment requires introduction of a gap into one of the sequences (Table I). These changes alter the derived amino acid sequence at 20 positions and shorten the protein from 533 to 532 residues. Comparison of our sequence with a second sequence of human FMO3 found in GenBank (sequence Z47552, submitted by C. T. Dolphin, T. E. Cullingford, E. A. Shephard, R. L. Smith, and I. R. Phillips) shows none of these differences, although four other base substitutions are noted: T for C in the third base of codon 43 (retains phenylalanine), G for A in the first base of codon 158 (changes glutamine to lysine), T for C in the third base of codon 239 (retains phenylalanine), and A for G in the third base of codon 486 (changes valine to isoleucine). The alignment of our sequence with that of rabbit FMO3 (12) is also free of gaps (Table I).
Expression of FMO4 and FMO4*-Fractions from E. coli transformed with FMO4, FMO4*, or pJL-2 vector alone (pJL) were examined for evidence of FMO expression. Samples of protein electrophoresed on polyacrylamide gels were stained with Coomassie Blue. The pattern and intensity of the protein bands in the relevant region (50 -60 kDa) obtained with the pJL or FMO4 samples were virtually identical (Fig. 1A, lanes 1  and 4). In contrast, the sample from the FMO4* preparation contained a band of protein corresponding to ϳ60 kDa (Fig. 1A, lane 5) that was not evident in the samples of pJL or FMO4 (Fig. 1A, lanes 1 and 4). Antibodies to FMO3 that cross-react with FMO1, -2, and -5 were used for immunoblot analysis of the FMO4 and FMO4* samples. These antibodies detected clearly the protein of ϳ60-kDa mobility observed with the FMO4* sample (Fig. 1B, lane 5). In contrast, no immunoreactive protein was evident with the pJL or FMO4 samples (Fig. 1B, lanes  1 and 4).
Expression of FMO3 and FMO3*-Bands of protein not present in the pJL sample were detected by staining the FMO3 and FMO3* samples with Coomassie Blue (Fig. 1A, lanes 2 and 3). However, expression of the ϳ57-kDa protein in the FMO3 (lane 2) sample was clearly greater than expression of the ϳ59-kDa protein in the FMO3* sample (lane 3). The intensities of staining obtained with antibodies to FMO3 on immunoblots pointed to a similar difference between the amounts of FMO3 and FMO3* expressed (Fig. 1B, lanes 2 and 3).
Metabolism of Methimazole by FMO4 and FMO4*-The 100,000 ϫ g particulate fractions were used to examine the activities of FMO4 and FMO4* with methimazole as the substrate (Table II). With FMO4 or pJL preparations, no activity was detected. The activity of recombinant FMO4* preparations at a methimazole concentration of 1 mM was 1.15 Ϯ 0.08 nmol of product min Ϫ1 mg of protein Ϫ1 (n ϭ 4), and recombinant FMO4* had a specific activity of 2.3 nmol of product min Ϫ1 nmol of FMO4* Ϫ1 . Metabolism of methimazole catalyzed by FMO4* conformed to Michaelis-Menten kinetics. The apparent K m for the reaction was 3.3 mM and the V max was about 4 nmol of product min Ϫ1 nmol of FMO4* Ϫ1 . The specific activity and V max calculations were based on a flavinadeninedinucleotide content of 746 pmol/mg protein in the FMO4* preparation compared with 241 pmol/mg in the pJL preparation.
Metabolism of Methimazole by FMO3 and FMO3*-The 100,000 ϫ g particulate fractions were used to examine the  The sequence of human FMO3 (HuSeq-1) from the present study is aligned with the sequence for human FMO3 (HuSeq-2) published by Lomri et al. (14) and the sequence for rabbit FMO3 (RbSeq) reported by Burnett et al. (12). For HuSeq-2 and RbSeq, only bases that differ from those in the sequence of HuSeq-1 are shown. Base substitutions between the human sequences are pointed out with the number symbol, and insertions and deletions, which are designated by dashes, are pointed out with asterisks. Bases are numbered starting with the first base of the start codon as 1. Base 1 above is base 74 of the cDNA for HuSeq-1, base 137 of the cDNA for HuSeq-2, and base 42 of the cDNA for RbSeq. Differences between HuSeq-1 and HuSeq-2 are also seen at base 330 of the coding region (A/C; retention of asparagine) and base 893 of the coding region (T/A; substitution of serine for threonine). metabolism of methimazole catalyzed by FMO3 and FMO3* ( Table II) Characterization of FMO4*, FMO3, and FMO3*-The effects of a number of factors on the metabolism of methimazole catalyzed by FMO4*, FMO3, and FMO3* were compared. The optimum pH for the reaction catalyzed by FMO4* was found to be near 10.2. The responses of FMO3 and FMO3* to pH were nearly identical, with optimum pH near 9.5 (Fig. 2). The activities of all three enzymes were found to be moderately temperature sensitive (more stable than has been noted for FMO1 but less stable than FMO2). FMO3 and FMO3* responded to heat in a similar fashion and reached 50% loss of activity with an exposure to 45°C of between 2 and 3 min (Fig. 3). FMO4* was somewhat less labile and reached 50% inhibition after ϳ4 min of heat exposure (Fig. 3).
The effects of n-octylamine or sodium cholate on the activities of FMO3 and FMO3* were very similar; FMO3 and FMO3* were inhibited 39 and 31%, respectively, by n-octylamine and 60 and 55%, respectively, by sodium cholate (Fig. 4). However, FMO3* was inhibited somewhat less than FMO3 by magnesium chloride (33 versus 18%) and was not activated to the same extent by imipramine (4 versus 16%). The effects of all four agents on the activity of FMO4* differed from those observed with either FMO3 or FMO3*: less inhibition (25%) by n-octylamine, greater inhibition by magnesium chloride (54%), nearly complete inhibition by sodium cholate (91%), and inhibition (14%) rather than activation by imipramine (Fig. 4).
Transcription of FMO3 and FMO4 in E. coli-Lack of expression of FMO4 in E. coli and reduced expression of FMO3 following elongation to form FMO3* were examined to determine whether the fault could be localized to problems associated with transcription or translation. RNA was isolated from E. coli transformed with FMO3, FMO3*, FMO4, FMO4*, or pJL vector alone and examined with 32 P-labeled cDNA probes for FMO3 and FMO4 (Figs. 5 and 6). Patterns and amounts of RNA detected were very similar when results obtained for FMO3 and FMO3* or FMO4 and FMO4* were compared. No evidence of abnormal or absent transcript was seen in the case of FMO4, and no decrease in transcript amount was apparent with FMO3*.
Alternative 5Ј-Splicing of FMO4 -Previous results from our laboratory indicated that FMO4 undergoes alternative splicing in the 5Ј-region of the transcript (12). We have repeated the PCR analysis of cDNA synthesized from rabbit liver mRNA with the same primers altered to introduce restriction sites for subcloning into pBluescript and subsequent sequencing. The primers correspond to the extreme 5Ј-end and bases 273-291 of the FMO4 cDNA. Results of PCR amplification again produced two bands, the expected one of ϳ300 base pairs and a predicted one of ϳ230 base pairs (12). The larger band was identified by sequence as corresponding exactly to the first 291 bases of the FMO4 cDNA. In contrast, the sequence of the smaller band was found to be unrelated to FMO4. A second antisense primer (bases 292-309), used with the same sense primer, also formed a band of ϳ300 base pairs but did not produce a smaller band. These findings indicate that FMO4 does not undergo alternative splicing at the 5Ј-end and that our previous conclusion was based on an artifactual result (12). DISCUSSION An inability to achieve heterologous expression of FMO4 has been experienced with both the human and rabbit orthologs of this enzyme (12,24). This has frustrated efforts to characterize  the FMO4 isoform and to obtain purified protein for production of antibodies. These results set FMO4 apart from other FMO isoforms, all of which have been expressed at relatively high levels, particularly in E. coli. FMO4 is also distinguished from the other members of this gene family by the location of its stop codon (12,15), which in the case of the human enzyme, is some 81 bases to the 3Ј-end of the consensus position (15). We have now shown that these two differences, lack of expression in E. coli and an extended 3Ј-coding region, appear to be related. Human FMO4 cDNA was altered by site-directed mutagenesis so that a stop codon was placed in the consensus position, 81 bases to the 5Ј-end of the one that occurs naturally. This modified cDNA was used for the expression of a truncated FMO4 in E. coli. About 500 pmol of FMO4*/mg of 100,000 ϫ g protein were consistently obtained. This compares favorably with the levels of expression we have obtained for other FMO isoforms (11,12,16,21) and with that for human FMO3 reported here (ϳ650 pmol/mg protein). Thus, the problems encountered with the expression of FMO4 are completely overcome by removal of 81 nucleotides from the 3Ј-end of the FMO4 coding region.
Attempts to define the reasons for the lack of expression of FMO4 are now underway. The present findings indicate quite clearly that the problem is not associated with transcription and is likely a function of translation. Data based on RNA modeling suggest that a high potential for hybridization involving the 3Ј-end of the FMO4 coding region could theoretically interfere with translation. 3 Whether this impediment to translation is encountered in vivo and is overcome by some regulatory process remains to be determined.
Results with FMO3 were not so profound. Elongation of FMO3 to the next available stop codon (72 bases to the 3Ј-end) did decrease expression by 3-4-fold, but FMO3* was still produced in amounts that were at least 250 times the level of immunochemical detection. Differences of this magnitude are often seen when different vectors (pJL-2 versus pKK) or E. coli cell types (XL-1 versus JM-109) are used (11). Also, the expression of FMO3* reported here is actually about five times greater than that of FMO3 reported by Lomri et al. (37), who used a pTrc99A vector and E. coli strain NM522. Examination of transcript levels, however, does indicate that decreased expression of FMO3 following elongation is associated with translation and is not due to transcriptional differences related to vector and cDNA or construct and cell strain compatibility. FMO4* expressed in E. coli is detectable as an active enzyme as well as a protein. Characterization of recombinant FMO4* shows a number of general properties exhibited by other FMO isoforms, with the exception of FMO2: sensitivity to heat, MgCl 2 , sodium cholate, and n-octylamine. The enzyme catalyzes the oxidation of methimazole, but with very little enthusiasm (K m(app) , 3.3 mM). It is possible that unmodified FMO4 would be significantly more active than FMO4*, but our experience with COOH-terminal modifications of other FMO isoforms argues against this. As seen in the present work, elongation of FMO3 has little if any effect on activity. Results with FMO2 also indicate that the COOH-terminal region is not important to the catalytic activity of FMO isoforms (21).
Although methimazole is among the best substrates for every known FMO isoform, it is metabolized efficiently only by FMO1 (K m(app) , 3 M) and FMO3 (K m(app) , 30 M). The marginal activities of FMO4* (K m(app) , 3.3 mM) and FMO5 (K m(app) , Ͼ5 mM) or, for that matter FMO2 (K m(app) , 300 M) brings into question their classification as "drug-metabolizing enzymes." On the other hand, the widespread distribution and high degree of structural conservation exhibited by these enzymes does suggest that they are of some functional importance. Further studies on FMO4, particularly with respect to its expression in the brain (24), will be greatly aided by the development of monospecific antibodies made possible by the successful expression of FMO4* in E. coli.