Characterization of a Fungal Maleylacetoacetate Isomerase Gene and Identification of Its Human Homologue*

We have previously used Aspergillus nidulans as a fungal model for human phenylalanine catabolism. This model was crucial for our characterization of the human gene involved in alcaptonuria. We use here an identical approach to characterize at the cDNA level the human gene for maleylacetoacetate isomerase (MAAI, EC 5.2.1.2), the only as yet unidentified structural gene of the phenylalanine catabolic pathway. We report here the first characterization of a gene encoding a MAAI enzyme from any organism, the A. nidulans maiA gene.maiA disruption prevents growth on phenylalanine (Phe) and phenylacetate and results in the absence of MAAI activity in vitro and Phe toxicity. The MaiA protein shows strong amino acid sequence identity to glutathione S-transferases and has MAAI activity when expressed in Escherichia coli. maiA is clustered with fahA and hmgA, the genes encoding the two other enzymes of the common part of the Phe/phenylacetate pathways. Based on the high amino acid sequence conservation existing between other homologous A. nidulans and human enzymes of this pathway, we used the MaiA sequence in data base searches to identify human expressed sequence tags encoding its putative homologues. Four such cDNAs were sequenced and shown to be encoded by the same gene. They encode a protein with 45% sequence identity to MaiA, which showed MAAI activity when expressed in E. coli. Human MAAI deficiency would presumably cause tyrosinemia that would be characterized by the absence of succinylacetone, the diagnostic compound resulting from fumarylacetoacetate hydrolase deficiency in humans and fungi. Culture supernatants of an A. nidulansstrain disrupted for maiA are succinylacetone-negative but specifically contain cis and/or trans isomers of 2,4-dioxohept-2-enoic acid. We suggest that this compound(s) might be diagnostic for human MAAI deficiency.

We have previously used Aspergillus nidulans as a fungal model for human phenylalanine catabolism. This model was crucial for our characterization of the human gene involved in alcaptonuria. We use here an identical approach to characterize at the cDNA level the human gene for maleylacetoacetate isomerase (MAAI, EC 5.2.1.2), the only as yet unidentified structural gene of the phenylalanine catabolic pathway.
We report here the first characterization of a gene encoding a MAAI enzyme from any organism, the A. nidulans maiA gene. maiA disruption prevents growth on phenylalanine (Phe) and phenylacetate and results in the absence of MAAI activity in vitro and Phe toxicity. The MaiA protein shows strong amino acid sequence identity to glutathione S-transferases and has MAAI activity when expressed in Escherichia coli. maiA is clustered with fahA and hmgA, the genes encoding the two other enzymes of the common part of the Phe/phenylacetate pathways.
Based on the high amino acid sequence conservation existing between other homologous A. nidulans and human enzymes of this pathway, we used the MaiA sequence in data base searches to identify human expressed sequence tags encoding its putative homologues. Four such cDNAs were sequenced and shown to be encoded by the same gene. They encode a protein with 45% sequence identity to MaiA, which showed MAAI activity when expressed in E. coli.
Human MAAI deficiency would presumably cause tyrosinemia that would be characterized by the absence of succinylacetone, the diagnostic compound resulting from fumarylacetoacetate hydrolase deficiency in humans and fungi. Culture supernatants of an A. nidulans strain disrupted for maiA are succinylacetone-negative but specifically contain cis and/or trans isomers of 2,4dioxohept-2-enoic acid. We suggest that this compound(s) might be diagnostic for human MAAI deficiency.
The catabolism of phenylalanine and tyrosine in humans is both of intrinsic and clinical interest. The enzymatic steps of this pathway were definitively established in the '50s by the work of Knox and colleagues (see Fig. 1A; Ref. 1). However, two of its structural genes remained uncharacterized. We recently used a novel approach based on the development of a fungal model to characterize one of them (2)(3)(4). Here we report our successful application of this approach to the characterization of the other and address by reverse genetics the consequences of the corresponding enzyme deficiency in our model organism.
An enzyme deficiency in any of the steps of this pathway causes in humans a known metabolic disease. For example, a deficiency in phenylalanine hydroxylase causes phenylketonuria (reviewed by Scriver et al. (5)). Enzyme deficiencies in four other steps (those labeled as II, III, and VI in Fig. 1A) cause different hypertyrosinemias (reviewed by Mitchell et al. (6)), and absence of homogentisate dioxygenase (IV, see Fig. 1A) causes alkaptonuria (4,7). Although the historical interest in the later is notable as it enabled Archibald Garrod to coin the term of "inborn error of metabolism" (8,9), the gene had not been characterized until recently (3,4,10). Crucial for the isolation and characterization of this gene was our establishment of a fungal model for human phenylalanine catabolism based on the filamentous ascomycete Aspergillus nidulans (2). We cloned its homogentisate dioxygenase gene (the first gene encoding this enzyme identified for any organism) and used its derived amino acid sequence as a probe to identify in similarity searches of the human expressed sequence, tag data base (EST) 1 cDNAs encoding its human homologue (3).
Type 1 hereditary tyrosinemia (HT1, hepatorenal tyrosinemia, McKusick 276700) is the most severe disease in human Phe catabolism, affecting liver, kidney, and peripheral nerves. HT1 patients surviving infancy develop chronic liver disease with a high incidence of hepatocellular carcinoma (6). HT1 results from fumarylacetoacetate hydrolase (FAAH) deficiency (11). It is generally accepted that fumarylacetoacetate and its spontaneous reaction product, succinylacetone (the diagnostic compound of the disease), are toxic due to their considerable reactivity with key cellular molecules (6,11), and fumarylacetoacetate has been shown to be mutagenic in Chinese hamster cells (12). In agreement with this, growth of an A. nidulans strain disrupted for the FAAH-encoding gene is prevented by phenylalanine even in the presence of an alternative carbon source (2). succinylacetone is accumulated in culture supernatants of this strain, illustrating the equivalent consequences of a FAAH deficiency in humans and A. nidulans (2).
The clinical consequences of a MAAI (MAAI, EC 5.2.1.2; step V in Fig. 1A) deficiency in humans are largely unknown. It is predicted that this deficiency should also lead to HT1, as maleylacetoacetate has similar reactivity to fumarylacetoacetate (for example, see Ref. 13). By contrast, it is thought that it * The work was supported by the Spanish Comisión Interministerial de Ciencia y Tecnología (BIO94/0932). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  should not result in the presence of succinylacetone in plasma and urine, as the latter compound is likely to be formed from succinylacetoacetate resulting from in vivo reduction of maleyl and fumarylacetoacetate (6,11). Succinylacetoacetate is efficiently degraded by FAAH (1), and its hydrolysis would prevent succinylacetoacetone formation. Only one such succinylacetone-negative patient showing type 1 tyrosinemia with nondetectable levels of MAAI but normal levels of FAAH in liver has been described (14).
Mammalian MAAI has been little studied since its original characterization (1,15,16), possibly due, among other possible reasons, to the instability of the substrate (6). The gene encoding MAAI has not been cloned from any organism, and it is therefore the only structural gene of the Phe/Tyr degradation pathway that remains uncharacterized, precluding the analysis of the molecular basis of succinylacetone-negative type I tyrosinemia. Here we successfully use our fungal model to identify cDNAs encoding human MAAI. The liver enzyme requires glutathione (1,15,16) as does the equivalent bacterial enzyme that has been purified to homogeneity (17). Our characterization of fungal and human MAAI cDNAs revealed strong amino acid sequence identity of their derived protein sequences to glutathione S-transferases, in agreement with the proposed mechanism of the isomerization (18). We also extend the work of Edwards and Knox (16) and demonstrate MAAI activity by an in vitro complementation assay using extracts from a recombinant fungal strain deficient for MAAI. Notably, we detected no succinylacetone in culture supernatants of this strain.

EXPERIMENTAL PROCEDURES
Fungal Strains, Media, and Growth Conditions-A. nidulans strains carried markers in standard use (19). A biA1 strain was used as a source of cDNA and wild type protein extracts. The biA1, methG1, ⌬fahA strain has been described (2). A biA1, methG1 strain was used as the wild type in growth tests. Standard media for A. nidulans (20) were used for strain maintenance, growth tests, and transformation. Culture conditions inducing high levels of expression of the fahA/maiA/hmgA genes, which were routinely used to grow mycelia for protein extraction, have been described (21).
Identification of the A. nidulans maiA Gene-Genomic maiA sequences were identified by Southern analysis of DNA from EMBL4 clones carrying the fahA and hmgA genes, using a subtracted cDNA probe representing genes induced by phenylacetate ("plus" probe) and a cDNA probe from glucose-grown mycelia ("minus" probe). A 2.5-kb EcoRI fragment contiguous to the fahA transcription unit (see Fig. 1B) showed strong differential hybridization to the plus probe. In addition, cDNA library screening using the F2 fragment (see Fig. 1A) resulted in the isolation, in addition to fahA cDNA clones, of a second class of cDNA clones representing maiA transcripts. The cDNA library enriched in transcripts induced by phenylacetate and the subtracted plus cDNA probe have been described (2,3). The A. nidulans genomic library was a standard EMBL4 library constructed from the wild type strain. RNA isolation and Northern analysis followed (22). Equal loading of the different samples was confirmed using an actin probe.
Disruption of the maiA Gene-Transformation followed Tilburn et al. (23). For disruption of maiA we used a 4.2-kb linear DNA fragment in which the sequence between maiA codons 140 -226 had been replaced by a 3.2-kb XbaI fragment carrying argB ϩ . A genomic fragment carrying maiA sequences from an XbaI site at position Ϫ133 (relative to the initiation codon) to an XhoI site at position ϩ655 (relative to the stop codon) was subcloned in pBS-SK ϩ (Stratagene). Substitution of an internal 0.26-kb SalI-EcoRI fragment by the above 3.2-kb A. nidulans genomic fragment (whose XbaI ends had been previously converted to EcoRI and XhoI) removed maiA sequences between codons 140 and 226 to yield pBS-⌬MAI. The transforming fragment was isolated from this plasmid after digestion with XbaI and XhoI.
Enzyme Assays-Maleylacetoacetate, which is not commercially available, was synthesized enzymatically from homogentisate (1, 15) using homogentisate dioxygenase from A. nidulans extracts or from Escherichia coli cells overexpressing the human enzyme. The procedure used to obtain mycelial protein extracts from ⌬fahA, ⌬maiA, and wild type A. nidulans strains and the conditions for the homogentisate dioxygenase reaction have been described (21). For in vitro complementation assays, the initial homogentisate concentration was 100 -125 M. Maleylacetoacetate formation was monitored spectrophotometrically at 330 nm. When the reaction reached a plateau (with usually more than 80% of the substrate converted to maleylacetoacetate), 150 M reduced glutathione was added to allow the MAAI-dependent isomerization of maleylacetoacetate to fumarylacetoacetate (1,16), which is then a substrate for FAAH. Complementation of ⌬maiA extracts was used to detect MAAI activity in crude lysates of E. coli cells overexpressing fungal or human MAAI, as described in the corresponding figure legends.
Absorption Spectra of Intermediates-Maleylacetoacetate was synthesized in separate reactions using 250 -500 M substrate and an excess of recombinant human HGO. When the reactions reached a plateau, they were stopped and deproteinized after the addition of 0.1 vol of 10% metaphosphoric acid, incubation for 10 min in ice, and centrifugation at 13,000 ϫ g for 10 min. The supernatants were neutralized with KOH to pH 7.5-8 and immediately used in in vitro reactions in the presence of wild type or mutant A. nidulans protein extracts. These reaction mixtures were essentially as for homogentisate dioxygenase but contained GSH as a cofactor for MAAI. Reactions were carried out for 15 min at room temperature, deproteinized with metaphosphoric acid, and neutralized with KOH as above. The absorption spectra of these samples were determined and compared with those of duplicate reaction mixtures for which the final neutralization step was omitted.
Overexpression of Proteins in E. coli-High levels of protein expression were achieved using the pD1 vector (a gift of E. Espeso). This is a modified pET19b (Novagen) derivative that was engineered to introduce a single BamHI site allowing in-frame fusion of the desired coding region to an N-terminal His tag. Details of this vector will be described elsewhere. Proteins overexpressed in this system carry the sequence MGHHHHHHHHHHSSGHIDDDDKHMGS at their amino termini. The MaiA coding region was amplified using the following pair of primers (underlined sequences add or modify restriction sites): 5Ј-CGG-GATCCCCCGCACCGGTCAAGATCTC-3Ј (upper) and 5Ј-CGGAAT-TCAACACCTAAATTCCGTTGGTG-3Ј (lower). The fusion protein contains the complete MaiA sequence with four further extra residues (PAPL) between the above N-terminal tag and the MaiA initiation methionine. The corresponding recombinant plasmid was denoted pD1::MaiA. The human MAAI coding region was amplified using EST 265310 (5Ј) as template and primers 5Ј-CAGGGATCCAAGCCCATC-CTCTATTCC-3Ј (upper) and 5Ј-CAGGAATTCGGAGCTAGGCCCTC-3Ј (lower). The recombinant gene fused the above N-terminal tag to residues 5-216 of the protein. The corresponding plasmid was denoted pD1::HSMAAI. A pD1::HSHGO plasmid (a gift from M. C. Estébanez), driving high level expression of human HGO, will be described elsewhere.
Recombinant plasmids were selected in E. coli DH1, purified, and transformed into E. coli BLB21(DE3)pLysS. Primary transformants were selected on LB plates containing ampicillin (100 g/ml) and chloramphenicol (35 g/ml) and directly used to inoculate LB liquid cultures that were grown at 37°C until A 600 reached 0.8 -0.9 units. Expression of T7 RNA polymerase was induced after the addition of 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside and further incubation for 2.5 h. 0.5 ml samples were taken before and after induction, and bacteria were collected by centrifugation and resuspended in SDS-polyacrylamide gel electrophoresis loading buffer. Samples were boiled for 3 min before loading appropriate aliquots onto a 12% SDS-polyacrylamide gel alongside Bio-Rad wide-range protein markers. Proteins were detected by Coomassie staining.
For preparation of protein extracts, bacteria from a 50-ml culture were collected by centrifugation, washed in 100 mM potassium phosphate buffer, pH 7.0, resuspended in 4 ml of the same buffer, and lysed by sonication. Lysates were clarified by centrifugation at 10,000 rpm and 4°C for 20 min in an SS34 rotor. Protein concentrations were estimated by the Bradford assay (24).
GC-MS Analysis of Culture Filtrates-Fungal mycelia pregrown on 0.6% glucose (w/v) as the sole carbon source were transferred to appropriately supplemented minimal medium with 20 mM phenylacetate as the sole carbon source (see Ref. 21) and incubated for 20 h at 37°C. Culture filtrates were ether-extracted and derivatized with bis(trimethylsilyl)trifluoroacetamide as described (2). TMS derivatives were analyzed by GC-MS in a fused silica capillary column SBP-1 (30 m ϫ 0.25 mm; 0.2-mm film thickness) with a temperature program from 80 to 280°C (4°C/min), and a Q-MASS (Perkin-Elmer) mass detector. Identification of peaks was carried out by comparison of sample spectra with reference spectra from the NIST/EPA/NIH mass spectral data base.
DNA Sequencing-Genes and cDNAs were sequenced using a Dye Terminator Cycle sequencing kit (Perkin-Elmer) and Taq FS DNA polymerase with universal and custom primers. Sequencing reactions were resolved on an ABI Prism 377 automatic sequencer and analyzed with the ABI analysis software (Version 3.1). Genomic and cDNA versions of maiA and human EST cDNA clones 265310 (5Ј), 290219 (5Ј), 683733 (5Ј), and 52677(5Ј) encoding human MAAI were completely sequenced in both strands. cDNA clones of the IMAGE consortium (25) were purchased from Genome Systems Inc., (St. Louis, MO). 2

A Cluster of Three Genes Encoding Enzymes of the Common
Part of the A. nidulans PhAc/Phe Catabolic Pathway-The ascomycete fungus A. nidulans can use either Phe or PhAc as sole carbon source (Fig. 1A). Both compounds were catabolized to homogentisate, which was then converted to fumarate and acetoacetate through the action of three enzyme activities, HGO, MAAI, and FAAH. We have previously reported that the A. nidulans fahA and hmgA genes encoding FAAH and HGO, respectively, are closely linked and divergently transcribed from a 414-base pair intergenic region. fahA and hmgA gene transcription is strongly inducible by PhAc (or its structural relatives) or Phe and partially repressible by glucose. Neither of these genes is expressed on glucose as the sole carbon source. No gene encoding MAAI has yet been characterized from any organism. Southern blot hybridization of EMBL4 phage clones carrying the fahA and hmgA genes with a substracted cDNA probe representing transcripts induced by PhAc revealed the presence of a third linked gene strongly hybridizing to this probe. As fahA and hmgA, this third gene, designated maiA, was not expressed on glucose. Clustering of genes encoding activities of the same catabolic pathway is not unusual in A. nidulans. Genomic and cDNA nucleotide sequencing of the region encoding this new transcript confirmed the presence of a third gene 3Ј from fahA, transcribed in a tail-to-tail orientation (Fig. 1B). The transcribed region contains an intron-less ORF encoding a putative 230-residue polypeptide (Fig. 2) whose stop codon is 486 base pairs downstream from that of fahA. The nucleotide sequence of this three-gene cluster has been submitted to the DDBJ/EMBL/GenBank TM data bases under accession number AJ001836.
A. nidulans maiA Encodes Maleylacetoacetate Isomerase-Data base searches revealed that the predicted product of maiA shows strong amino acid sequence identity to glutathione Stransferases. For example, FASTA searches of the Swiss-Prot data base revealed that among the 30 protein sequences showing the highest alignment scores, 27 were glutathione S-transferases (data not shown). Similar results were obtained after searching the conceptual translation of EMBL ϩ GenBank TM nucleotide sequence data bases with TBLASTN. The highest FASTA score corresponded to glutathione S-transferase 1 from Diantus caryophyllus, which showed 33.3% identity to MaiA in a 228-amino acid overlap including the complete sequence of both proteins (data not shown). The deduced molecular mass for MaiA (25,129 Da) is similar to the 25-kDa size of glutathione S-transferases.
Northern analysis showed that, as determined for fahA and hmgA, transcription of maiA was induced by either Phe or PhAc and was absent on glucose or gluconeogenic carbon sources (Fig. 3), strongly suggesting that maiA was a gene for Phe/PhAc catabolism and that its clustering with fahA and hmgA reflected its involvement in the same catabolic pathway. The only as yet unidentified gene encoding an enzyme essential for both Phe and PhAc catabolism is that encoding MAAI. The likely mechanism of this enzyme involves transfer of enzymebound GSH to C2 of maleylacetoacetate (26). Therefore, all the above data indicated that maiA might encode A. nidulans MAAI.
To confirm this, we replaced by transformation the wild type maiA gene by a mutant version in which the sequence encoding MaiA residues 140 -226 had been substituted by a genomic DNA fragment containing the argB ϩ gene (Fig. 4A). Transformants were selected in an argB2 background for arginine- independent growth and purified by repeated streaking of conidiospores on minimal medium lacking arginine. Two independent transformants showing the expected maiA replacement were selected after Southern analysis. Both showed an identical phenotype, being unable to grow on either phenylacetate or phenylalanine as the sole carbon source. This confirmed that maiA is a gene of the common part of the Phe/PhAc pathways. MAAI assays were carried out with protein extracts from mycelia of the disrupted strains grown on glucose and transferred to PhAc, which showed them to be deficient for MAAI activity (Fig. 5). By contrast, these extracts showed normal levels of either FAAH or HGO (data not shown). Maleylacetoacetate can be synthesized in vitro by the homogentisate dioxygenase activity present in mycelial extracts from the wild type strain or from either a mutant ⌬fahA strain (lacking FAAH (2)) or a mutant ⌬maiA strain (presumably lacking MAAI) and detected by its absorption at 330 nm (1,15,16). In the absence of GSH, an obligate cofactor of MAAI, maleylacetoacetate is not isomerized to fumarylacetoacetate, thereby providing an enzymatic method to obtain the isomerase substrate (15,16). On addition of GSH, wild type extracts catalyze the isomerization of maleylacetoacetate to fumarylacetoacetate and the conversion of the latter to fumarate and acetoacetate. As neither of these two latter compounds shows the characteristic absorption of diketoacids in the near ultraviolet region, this coupled enzyme reaction can be monitored by the decrease of A 330 (see Fig. 5  lates with the ⌬fahA extract) have similar ⑀ 330 at neutral pH (15). By contrast, a mixture of the two mutant extracts should provide both the isomerase and the hydrolase, and their coupled action should result in conversion of maleylacetoacetate to fumarate and acetoacetate. Fig. 5 shows that these predictions were fulfilled by mutant extracts. We concluded that the ⌬fahA extract provided an activity required for catabolism of maleylacetoacetate that was missing in the ⌬maiA extract. Sequential addition of extracts instead of their simultaneous mixing resulted in substantially identical results (data not shown). Taken in the context of the enzymes known to be required for homogentisate degradation (Fig. 1A), these data further indicated that MaiA is a fungal MAAI.
In contrast to fumarylacetoacetate, which shows a similar ⑀ 330 at both pH 1 and pH 7.5, maleylacetoacetate shows a 10-fold lower ⑀ 330 at pH 1 than at pH 7.5 (15). Therefore, a fall in A 330 upon acidification of the homogentisate dioxygenase reaction product is diagnostic of maleylacetoacetate (Fig. 6, A  and B; it should be noted that the absorption maximum for the above compounds is at approximately 320 nm (see also Ref. 15). This was used to confirm its accumulation with the ⌬maiA but not with the ⌬fahA extract. Maleylacetoacetate was prepared using recombinant human HGO expressed in E. coli as described under "Experimental Procedures" and incubated in the presence of the wild type or mutant extracts and GSH. Incubation with the wild type extract resulted in the disappearance of compounds absorbing in the near ultraviolet region at either acidic or neutral pH (Fig. 6, D and H). By contrast, the characteristic absorption of maleylacetoacetate at neutral but not at acidic pH was still detectable upon incubation with the ⌬maiA extract (Fig. 6, B and F). Incubation of the substrate with the ⌬fahA extract (which contains the isomerase but not the hydrolase) produced a compound (fumarylacetoacetate) showing strong absorbance in the above region both at acidic and neutral pH (Fig. 6, C and G). Expression of A. nidulans MaiA in E. coli-To definitively establish that MaiA is A. nidulans MAAI, we expressed the polypeptide as a fusion protein (see "Experimental Procedures") in E. coli under the control of a T7 RNA polymerasedependent promoter. Promoter induction resulted in the synthesis of a markedly abundant 32-kDa protein that was absent from extracts of induced cells carrying the expression vector with no insert. This protein, whose electrophoretic mobility was roughly consistent with the M r predicted for the MaiA fusion protein, remained soluble upon clarification of the cell lysate (data not shown). In vitro complementation of A. nidulans ⌬maiA and ⌬fahA protein extracts described above suggested the use of ⌬maiA extracts in a coupled enzyme assay for MAAI. ⌬maiA extracts are unable to transform homogentisate to fumarate and acetoacetate unless supplemented with MAAI enzyme. Fig. 7 shows that an extract from E. coli cells overexpressing MaiA complemented this MAAI deficiency in vitro. The coupled action of bacterially expressed MAAI and endogenous A. nidulans FAAH resulted in the efficient degradation of maleylacetoacetate. This coupled degradation obligately required GSH and did not occur with an extract of E. coli cells carrying the expression vector with no insert (Fig. 7).
In addition, we synthesized maleylacetoacetate using recombinant human homogentisate dioxygenase. It showed the diagnostic high absorbance in the near ultraviolet region under neutral but not under acidic pH conditions (see above). Incubation of this compound in the presence of the E. coli extract overexpressing MaiA and GSH resulted in its conversion to fumarylacetoacetate, as shown by the characteristic high absorbance of the latter in this region both under acidic and neutral pH conditions. This conversion did not take place when extracts overexpressing the protein were substituted by control extracts (data not shown). As a whole, these results provide formal evidence that A. nidulans maiA encodes MAAI.
Analysis of A. nidulans Culture Supernatants for Diagnostic Compounds of MAAI Deficiency-An A. nidulans strain disrupted for the gene encoding FAAH accumulates succinylacetone, the diagnostic compound for HT1 (caused by FAAH deficiency) in the urine of human patients. Growth of a ⌬fahA strain on lactose (which is unaffected by the mutation) is strongly inhibited by Phe ((2); see Fig. 4B). Phe toxicity in this mutant background is due to fumarylacetoacetate and/or its spontaneous reaction product, succinylacetone, which accumulate(s) as a result of the enzyme deficiency. Notably, growth of a strain disrupted for maiA was also inhibited by Phe (Fig. 4B), although clearly to a lesser extent than that of the ⌬fahA strain, strongly suggesting that different catabolites with distinct toxicities accumulate in each disrupted strain (but see "Discussion").
No specific diagnostic method is available for a deficiency in human MAAI. We therefore investigated by GC-MS the compounds accumulated in a culture supernatant of our ⌬maiA strain transferred to phenylacetate and detected four peaks almost certainly corresponding to cis and/or trans isomers of 4,6-dioxohept-2-enoic acid (Fig. 8), as shown by comparison of their mass spectra with those for the above compound(s) deposited in the data bases. These four peaks presumably represent different TMS derivatives of the above compound(s) (see legend to Fig. 8). Neither of the two possible cis-trans isomers of 4,6-dioxohept-2-enoic acid, maleylacetone and fumarylacetone, had been detected in culture supernatants of either the wild type or a ⌬fahA strain (2), strongly suggesting that its (their) presence specifically results from the ⌬maiA mutation. Notably, no peaks corresponding to succinylacetone were detected.
Identification of Human EST Clones Encoding Homologues of A. nidulans MaiA-We next used the fungal MaiA amino acid sequence to screen the EST data bases for human and murine ESTs encoding putative MaiA homologues, in analogy with the protocol already established for the AKU gene (3,4). BLAST searches identified a number of these human ESTs. The 10 highest scores were obtained with the derived amino acid sequences of the following human ESTs (with the source of RNA for each EST cDNA clone in parentheses): 683733 (germinal B cells), 290219 (multiple sclerosis), 290775 and 265310 (melanocytes), 52677 (infant brain), 156401 (breast), 309975 (senescent fibroblasts), 240726 and 246479 (fetal liver/spleen), and 66e04 (skeletal muscle). These (partial) amino acid sequences showed more than 40% identity to that of A. nidulans MaiA, strongly suggesting that they represent its human homologue(s). The fact that only minor differences in the sequence (presumably resulting from automated sequencing errors) were found between these derived human proteins strongly suggests that all these cDNAs correspond to a single gene. Tissue-specific expression of such a gene does not appear to be as restricted as that of HGO (3,4), and only two of the 10 cDNA clones represented liver transcripts (see above). In addition to the above human protein sequences, data base searches detected MaiA homology to derived protein products of mouse and Arabidopsis thaliana EST clones (not shown).
Molecular Characterization of Human cDNAs Encoding Maleylacetoacetate Isomerase-We fully sequenced four of the above ESTs. (Fig. 9A). Nucleotide sequencing showed that they were indeed encoded by the same gene, despite the fact that they had been isolated from different tissues. EST 265310 (5Ј) (melanocytes) is the longest of these cDNAs. It is 1155 base pairs long (excluding the poly(A) tail) and contains a 216-codon ORF whose derived protein product (M r 24,083) shows 45% identity in amino acid sequence to A. nidulans MAAI (Fig. 9B). This represents nearly definitive evidence that this cDNA encodes a human MAAI (but see below). The complete nucleotide sequence of this cDNA has been submitted to DDBJ/EMBL/ GenBank TM data bases (accession number AJ001838) The 3Ј-UTR of this transcript was remarkably long (400 nucleotides, i.e. 30% of the transcript size). ESTs 290219 (5Ј) (multiple sclerosis lesions) and 683733 (5Ј) (germinal B-cells) represented cDNAs incomplete at their 5Ј ends, starting at codons 18 and 49, respectively, of the human MAAI ORF. The precise site of polyadenylation and the sequence of the 400-nucleotide 3Ј-UTR of EST 290219 were identical to those of the longest cDNA. Polyadenylation of the 683733 cDNA occurred two nucleotides upstream of the above site, but no other nucleotide sequence difference was observed either in the 3Ј-UTR or in the coding region. Finally, we detected no differences between the nucleotide sequences of EST 52677 (brain) and EST 265310 up to position 1045, where the former is prematurely polyadenylated as compared with the latter. This strongly suggests that the 5Ј ends of these two transcripts represent a transcription start site for the human MAAI gene.
To confirm that the ORF of these transcripts encodes a human MAAI, the protein product encoded by the EST 265310 (5Ј) ORF was overexpressed in E. coli as above. High levels of a protein with the expected mobility for this polypeptide product were detected by SDS-polyacrylamide gel electrophoresis when recombinant E. coli cells in which its expression was driven by a T7-polymerase promoter (see "Experimental Procedures") were grown under inducing conditions (data not shown) and were absent from these cells grown under noninducing conditions. Extracts from E. coli cells overexpressing this protein showed strong maleylacetoacetate isomerase activity, as shown by its ability to complement an A. nidulans extract deficient for the enzyme (Fig. 10). This strong MAAI activity was dependent on the presence of GSH (not shown). By contrast, control cells showed no MAAI activity. These results establish that the product of the EST 265310 ORF (and, by extension, of the above four cDNAs) is a human maleylacetoacetate isomerase. DISCUSSION We describe here the molecular characterization of maiA, a gene encoding an enzyme of the common part of the A. nidulans Ether-extractable compounds were analyzed as TMS derivatives by GC-MS. Identified peaks were assigned to a compound by computer comparison of their mass spectra with reference spectra of the NIST/EPA/NIH mass spectra data base. A selected portion of the chromatogram including three peaks (denoted with roman number I) that were identified as di-TMS derivatives of 4,6-dioxohept-2enoic acid is shown at the bottom. Peaks labeled with ? represent unidentified compounds. Peak II, 2-hydroxyphenylacetate; III, ethylvanilline (internal standard, ϳ0.2 g); IV, 3-hydroxyphenylacetate; V, possibly octadecenol. The mass spectra of the major peak of 4,6-dioxohept-2-enoic acid (di-TMS) (indicated with a filled arrowhead in the chromatogram) is shown. A fourth peak (not shown in this portion of the chromatogram) was identified as the tri-TMS derivative of 4,6-dioxohept-2enoic acid.
FIG. 9. Characterization of human MAAI cDNAs and derived amino acid sequence of human MAAI. A, four EST cDNAs encoding putative MaiA homologues were fully sequenced. The corresponding four mRNAs appear to be transcribed from the same gene (see text for details). The limits of these four cDNAs with respect to the human MAAI coding region are shown. B, amino acid sequence of human MAAI (HS) and its comparison to its fungal homologue (AN). The amino acid sequence of the 216-residue-derived protein encoded by the EST 265310 (5Ј) ORF (and the three other EST ORFs above) is shown compared with that of A. nidulans MaiA (this work). Vertical bars indicate identity. The alignment was carried out using BESTFIT (GCG package). phenylalanine/phenylacetate pathways and provide definitive biochemical and genetic evidence that this gene encodes a maleylacetoacetate isomerase. This paper represents the first characterization of a gene encoding MAAI from any organism. Compelling evidence for the above conclusion can be summarized as follows: (i) maiA is clustered with fahA and hmgA, the two other structural genes of this common part of the pathways; (ii) transcription of the gene is induced by phenylalanine or phenylacetate as sole carbon sources; (iii) the deduced amino acid sequence of its encoded protein shows identity to glutathione S-transferases, as expected for a MAAI enzyme; (iv) disruption of the gene prevents growth on either Phe or phenylacetate; (v) protein extracts from this disrupted strain convert homogentisate to maleylacetoacetate but cannot catabolize this compound further; (vi) mixing ⌬maiA and ⌬fahA extracts results in reciprocal complementation of the corresponding enzyme deficiencies required for maleylacetoacetate catabolism to fumarate and acetoacetate; (vi) expression of maiA in E. coli results in bacterial protein extracts showing MAAI activity. Two technical developments were crucial to obtaining some of the above evidence. First, we used either recombinant HGO enzyme (4) or fungal extracts showing high HGO activity (21) to efficiently synthesize maleylacetoacetate. Second, we used a complementation assay for MAAI based on a protein extract from our A. nidulans strain deleted for maiA. This extract converted homogentisate to fumarate and acetoacetate only when supplied with GSH and a source of MAAI.
We next used the MaiA-derived sequence to identify human, mouse, and plant ESTs encoding proteins showing high amino acid sequence identity to A. nidulans MAAI. Four such human cDNAs were fully sequenced and shown to encode a protein with 45% identity to MaiA. Although they were isolated from different tissues, these four cDNAs (and the other ESTs detected in our searches) almost certainly represent transcripts of the same gene. The protein encoded by this transcript(s) has MAAI activity when expressed in E. coli. Our electronic screening of the human EST data base would suggest that expression of this gene would be more ubiquitous than that of HGO, whose expression is largely restricted to liver, kidney, colon, small intestine, and prostate. This apparently less restricted pattern of expression might be related to its ability to use other compounds, in addition to maleylacetoacetate, as substrates (26), which might suggest a detoxification function (6). Our characterization of human MAAI cDNAs represents the identification of the only as yet undescribed gene of the human Phe catabolic pathway. These results further confirm the validity of our fungal metabolic model and open the possibility of analyzing at the molecular level the predicted disease (a possible variant of HT1) resulting from MAAI deficiency. The incidence of this inborn error of metabolism is presently unknown, perhaps due to the absence of clear biochemical and/or molecular diagnostic criteria. A single patient with a putative MAAI deficiency has been reported in an abstract (14). Notably, this patient suffered from severe hepatorenal and brain damage. Our results with the fungal model show that the metabolite(s) accumulated as a result of a MAAI deficiency is indeed toxic for Aspergillus, but their toxicity is detectably lower than the toxicity of those accumulated as a result of a FAAH deficiency. Jorquera and Tanguay (12) have reported that, in contrast to fumarylacetoacetate, maleylacetoacetate was not mutagenic in Chinese hamster cells. We have not yet addressed if either of the above deficiencies is mutagenic in Aspergillus.
We detected no succinylacetone in culture supernatants of the A. nidulans ⌬maiA strain. This would be expected from the presumed origin of succinylacetone from decarboxylation of succinylacetoacetate, as normal levels of FAAH in this strain would degrade the latter (1). Therefore, this absence of succinylacetone might, at least in part, account for the lower Phe toxicity found in an A. nidulans strain deficient for MAAI as compared with a strain deficient for FAAH. Analysis of culture filtrates of the A. nidulans ⌬maiA strain specifically detected the presence of 4,6-dioxohept-2-enoic acid. This chemical structure would be consistent with maleylacetone (cis isomer) and/or fumarylacetone (trans isomer). These isomers cannot be reliably distinguished by the methodology used here. Taking into account the remarkable similarities in the consequences of equivalent metabolic blocks in human and fungal Phe catabolism, we suggest cis and/or trans isomers of 4,6-dioxohept-2enoic acid as possible diagnostic compound(s) for MAAI deficiency in humans. FIG. 10. The product of the human EST 265310 (5) ORF has MAAI activity. Detection of MAAI activity in cleared lysates of E. coli cells that overexpressed the protein encoded by this ORF under the control of a T7 polymerase-dependent promoter (see "Experimental Procedures"). The assay for MAAI, based on complementation of an A. nidulans ⌬maiA extract, was as in Fig. 7. The time points at which reduced glutathione (150 M) and protein extracts (ϳ50 g) were added is indicated. Filled circles, a reaction to which a bacterial extract overexpressing the product of the EST 265310 ORF was added. The levels of expression achieved for this protein (not shown) were similar to those for A. nidulans MaiA. Triangles, an extract from induced E. coli cells carrying the expression vector with no insert. Open circles, a second, negative control reaction in which an extract from a recombinant bacterial clone that showed no overexpression of the desired EST 265310 protein (as determined by SDS-polyacrylamide gel electrophoresis) was added. O.D. 330 , A 330 .