Fungal Metabolic Model for Type I 3-Methylglutaconic Aciduria*

catabolizes Leu to acetyl-CoA and acetoacetate through a pathway homologous to that used by humans. Fungal hlyA encodes a bifunctional polypeptide comprising the last two enzymes in this pathway, 3-methylglutaconyl-CoA hydratase and 3-hy-droxy-3-methylglutaryl-CoA lyase. hlyA transcription is specifically induced by Leu. A (cid:1) hlyA mutation removing the complete 3-methylglutaconyl-CoA hydratase C-ter-minal domain prevents growth on Leu but not on lactose or other amino acids and, in agreement with the predicted enzyme function, leads to Leu-dependent accumulation of 3-methylglutaconic acid in the culture supernatant. These data represent a formal demonstration in vivo of the specific involvement of 3-methylgluta-conyl-CoA hydratase in Leu catabolism. Type I 3-meth-ylglutaconic aciduria patients deficient in 3-methyl-glutaconyl-CoA hydratase show urinary excretion of 3-methylglutaconic acid and, in contrast to the other three types of methylglutaconic acidurias, 3-hydroxy-isovaleric acid argB (cid:2) prog- eny carrying the mutant hlyA allele were selected and analyzed by Southern for the presence of the (cid:1) mccB allele, using a probe specific for Neuropora crassa pyr-4 . Six of thirteen analyzed strains were shown to be double mutants. RNA Isolation and Northern Blot Analysis and Transformation— These were carried out as described (9). Mycelia for RNA isolation were pre-grown in glucose minimal medium and transferred to media with the indicated carbon sources. A cDNA library from mycelia grown under Leu catabolism-inducing conditions has been described. In Northern blots, a PCR cDNA fragment from positions 168 to 2053 of GenBank TM accession no. AY484417 was used as a hlyA- specific probe. MAD003 biA1 argB2 methG1 was used as the recipient strain in transformation experiments (20). Southern blot analysis with argB- and hlyA -specific probes was used to confirm the expected hlyA truncation/deletion event. GC / MS Analysis of Culture Supernatants— Fungal mycelia were pre-cultured as described for RNA isolation and transferred to minimal medium with one of the following carbon sources: 0.05% (w/v) lactose, 3 or 10 m M Leu. Secondary cultures were incubated for an additional 21 h. Culture supernatants were collected after removing mycelia by filtration. Organic acids were extracted, derivatized, and analyzed by GC/MS as described (9).

3-Methylglutaconyl-CoA hydratase (EC 4.2.1.18) converting 3-methylglutaconyl-CoA to 3-hydroxy-3-methylglutaryl-CoA catalyzes the fifth step in the Leu degradation pathway ( Fig.  1a; Ref. 1). In humans, deficiency of this enzyme results in type I 3-methylglutaconic aciduria (MGA) 1 (MIM 250950), a poten-tially pathogenic condition in which 3-methylglutaconic acid , accumulating as a result of the metabolic block, is excreted in urine. However, urinary 3-MG excretion is a hallmark of three additional diseases: type II MGA (Barth syndrome (MIM 302060), resulting from a deficiency of a putative phospholipid acyltransferase) (2), type III MGA or Costeff optic atrophy syndrome (MIM 258501, a deficiency of a predicted mitochondrial protein) (3), and type IV MGA or "unspecified" MGA (MIM 250951, a heterogeneous class including patients for which all of the other three syndromes have been excluded) (1). In contrast to type I MGA patients, type II, III, and IV MGA patients show abnormally elevated urinary excretion of 3-MG, but their 3-methylglutaconyl-CoA (3-MG-CoA) hydratase activity is normal. In addition to MGA patients, it has been shown that certain Smith-Lemli-Opitz patients also show abnormally increased plasma levels of 3-MG. This apparently complex scenario challenges our current understanding of the 3-MG-CoA metabolism. Remarkably, the characterization of the type II and III MGA genes gave no clues as to the biochemical mechanism underlying the abnormal accumulation of 3-MG, and it was not until the recent successful exploitation of a Saccharomyces cerevisiae model for type II MGA that direct evidence was obtained that tafazzin (the deficient protein in type II MGA) is involved in cardiolipin metabolism (4), illustrating the usefulness of microbial eukaryotic models.
Aspergillus nidulans is a filamentous fungus sharing with S. cerevisiae those features that make microbial eukaryotic models so valuable: a compact, haploid genome whose complete 31-Mb sequence is available, 2 relatively easy gene inactivation procedures by homologous recombination/gene replacement, meiotic recombination, and small average intron size (less than 100 bp). A. nidulans grows in a chemically defined medium containing inorganic salts and sole nitrogen and carbon sources which can be changed as appropriate (see below). In contrast to budding yeast, however, A. nidulans has a much broader metabolic versatility. For example, A. nidulans is able to use most amino acids as sole carbon sources using metabolic pathways remarkably similar to those of human hepatocytes (5), a feature that we have exploited for designing fungal models for homogentisate dioxygenase (6), fumarylacetoacetate hydrolase (7), maleylacetoacetate isomerase, (8) and 3-methylcrotonyl-CoA carboxylase (9) deficiencies. These models have been instrumental for the identification of the human genes involved in alkaptonuria (6,10,11) and methylcrotonylglycinuria (12).
The use of A. nidulans models for Leu metabolism deficiencies is particularly advantageous (9). Leu is a ketogenic amino acid in humans/mammals which is catabolized to acetyl-CoA and acetoacetate (Fig. 1a). In contrast, fungi can convert acetyl-CoA into carbohydrates through the glyoxylate bypass of the Krebs cycle (13)(14)(15). Therefore, A. nidulans grows on Leu as the sole carbon source, enabling the use of simple growth tests to determine whether mutations impair this catabolic pathway. When this work was started, the gene encoding 3-MG-CoA hydratase had not been characterized in any organism. Here we use an in silico approach to identify the A. nidulans and human genes encoding this enzyme. As two recent papers reported the identification of the human gene and the characterization of loss-of-function mutations in type I MGA patients (16,17), we focus this paper on the construction and physiological characterization of a fungal metabolic model for the disease. We provide here formal genetic evidence that 3-MG-CoA hydratase catalyzes in vivo one step in the Leu catabolic pathway shown in Fig. 1, and additionally, we demonstrate using epistasy tests that 3-MG can be synthesized without involvement of 3-methylcrotonyl-CoA carboxylase, a finding that may have implications in our understanding of the biochemical basis of those MGA types in which 3-MG-CoA hydratase is normal. Our data additionally suggest interplay between mitochondria and peroxisomes in the late steps of A. nidulans Leu catabolism.

EXPERIMENTAL PROCEDURES
Fungal Strains, Media, and Growth Conditions-A. nidulans strains carried markers in standard use (18). Complete and minimal medium were as in Ref. 19. Lactose was used at 0.05% (w/v) as the sole carbon source and amino acids at 30 mM, unless otherwise indicated. A ⌬mccB ⌬hlyA double mutant was constructed by crossing two parental strains of the complete genotype biA1 argB2 methG1 (⌬hlyA::argB ϩ ) and yA2 pabaA1 pyrG89 argB2 riboB2 (⌬mccB::pyr-4), respectively. argB ϩ progeny carrying the mutant hlyA allele were selected and analyzed by Southern for the presence of the ⌬mccB allele, using a probe specific for Neuropora crassa pyr-4. Six of thirteen analyzed strains were shown to be double mutants.
RNA Isolation and Northern Blot Analysis and Transformation-These were carried out as described (9). Mycelia for RNA isolation were pre-grown in glucose minimal medium and transferred to media with the indicated carbon sources. A cDNA library from mycelia grown under Leu catabolism-inducing conditions has been described. In Northern blots, a PCR cDNA fragment from positions 168 to 2053 of GenBank TM accession no. AY484417 was used as a hlyA-specific probe. MAD003 biA1 argB2 methG1 was used as the recipient strain in transformation experiments (20). Southern blot analysis with argB-and hlyA-specific probes was used to confirm the expected hlyA truncation/deletion event.
GC/MS Analysis of Culture Supernatants-Fungal mycelia were precultured as described for RNA isolation and transferred to minimal medium with one of the following carbon sources: 0.05% (w/v) lactose, 3 or 10 mM Leu. Secondary cultures were incubated for an additional 21 h. Culture supernatants were collected after removing mycelia by filtration. Organic acids were extracted, derivatized, and analyzed by GC/MS as described (9).

RESULTS
In Silico Identification of an A. nidulans Gene Encoding 3-Methylglutaconyl-CoA Hydratase-We predicted that 3-MG-CoA hydratase would belong to the enoyl-CoA hydratase/ isomerase enzyme family, whose members share less than 20% amino acid sequence identity. Northern blot analysis failed to identify a putative Leu-regulated, candidate gene within or in the proximity of the mccA-ivdA-mccB leucine catabolic gene cluster (9), which led us to use two complementary strategies to identify the A. nidulans 3-MG-CoA hydratase gene. We first searched the genomes of metabolically versatile soil bacteria for Leu degradation operons, using the amino acid sequences of A. nidulans MCC␣, MCC␤, and isovaleryl-CoA dehydrogenase (see Fig. 1) as in silico probes in TBLASTN searches. The genomes of Pseudomonas aeruginosa (21), Agrobacterium tumefaciens (22), Streptomyces coelicolor (23), and Mycobacterium tuberculosis (24) have, within 6.5 kb regions of their genomes, genes predictably encoding the above three proteins. We found that M. tuberculosis, P. aeruginosa, and A. tumefaciens (GenBank TM genome accession nos. NC_002755, NC_002516, and NC_003063, respectively) contain putative enoyl-CoA hydratases within these operons (see examples in Fig. 1b). Sequence alignments revealed that whereas the amino acid sequence identity between P. aeruginosa and M. tuberculosis Leu metabolism proteins ranged between 47% (in the case of isovaleryl-CoA dehydrogenase (IVD)) and 68% (in the case of MCC␤), enoyl-CoA hydratase candidates to 3-MG-CoA hydratase were only 33% identical, which illustrates the relatively low conservation of this family of proteins.
Secondly, we hypothesized that the 3-MG-CoA hydratase gene might be physically linked to the A. nidulans gene encoding 3-hydroxy-3-methylglutaryl-CoA lyase (HL), catalyzing the last step in the pathway (Fig. 1a). We used the amino acid sequence of human HL (GenBank TM accession no. P35914) to identify its A. nidulans orthologues using TBLASTN; two were Enzyme deficiencies in Leu catabolism cause the following diseases: (I) hypervalinemia/hyperleucine-isoleucinemia, (II) maple syrup urine disease, (III) isovaleric acidemia, (IV) isolated 3-methylcrotonyl-CoA carboxylase deficiency, methylcrotonylglycinuria, (V) type I methylglutaconic aciduria, (VI) 3-hydroxy-3-methylglutaric aciduria. The entry point of metabolites from the mevalonate shunt is indicated. A finding of this work is that 3-methylcrotonyl-CoA carboxylase-independent pathway(s) (indicated by a question mark and an arrow) can produce 3-MG-CoA. B, bacterial Leu degradation operons. Genes are denoted with roman numbers corresponding to the enzymes in (A), with IVa and IVb corresponding to genes encoding the ␣ and ␤ subunits of 3-methylcrotonyl-CoA carboxylase, respectively. The predicted P. aeruginosa Leu degradation operon is located within nucleotide positions 2206353-2199762 of the genome (GenBank TM accession no. NC_002516) and includes loci (from 5Ј to 3Ј) PA2015, -2014, -2013, -2012, and -2011. The predicted A. tumefaciens Leu degradation operon is located within nucleotide positions 524423-530991 of the genome (GenBank TM accession no. NC_003305) and includes loci Atu3477, -3478, -3479, -3480, and -3481.
found. The one showing the highest score was mapped on chromosome V. Next, we used the deduced amino acid sequence of the putative P. aeruginosa 3-MG-CoA hydratase (AAG05401) in a TBLASTN search of the A. nidulans genome, which revealed a large number of putative genes for enoyl-CoA hydratases. However, the highest score (expect 2e Ϫ16 ) was obtained for a polypeptide encoded by a region adjacent to that encoding chromosome V HL. This open reading frame was a convincing candidate to encode A. nidulans 3-MG-CoA hydratase.
Molecular Characterization of A. nidulans hlyA-Inspection of the genomic sequence in this region of chromosome V showed that the HL coding sequence was located immediately upstream of that encoding the candidate 3-MG-CoA hydratase. Notably, no stop codons separating these open reading frames were found, strongly indicating that a single gene encodes a bifunctional polypeptide carrying an N-terminal HL domain and a C-terminal 3-MG-CoA hydratase domain (Fig. 2). PCR amplification using cDNA as the template (data not shown) demonstrated the existence of transcripts encoding both domains. Sequencing of a full-length cDNA clone confirmed the existence of a single open reading frame encoding a 599-residue polypeptide that included the two domains. The length of this cDNA clone (1841 bp without the poly(A) tail) agrees with the deduced size of the transcriptional unit as determined by Northern blot hybridization (Fig. 3), which additionally gave no indication of the existence of smaller transcripts. Comparison to the genomic DNA sequence revealed that the corresponding gene, which we denoted hlyA (hydratase-lyase), is split by a short, 68-bp intron. We conclude that hlyA encodes a bifunc-  tional polypeptide catalyzing the two final steps in the Leu degradation pathway.
Although the amino acid sequence of the HL domain including residues 1 through 302 shows 48% identity to human HL (Fig. 2), the closest homologue of the predicted C-terminal 3-MG-CoA hydratase domain including residues 320 -599 was the putative P. aeruginosa 3-MG-CoA hydratase (29% amino acid sequence identity). The C-terminal HlyA 597-Ala-Lys-Val-599 tripeptide matches the peroxisome C-terminal targeting signal consensus (Prosite Databank accession no. PS00342) (25). No mitochondrial signal peptide was evident in the Nterminal region of the protein.
Northern blots (Fig. 3) demonstrated that hlyA transcription is induced by Leu and repressed by glucose. Of note, neither the branched chain amino acid Ile nor the aromatic amino acid Phe, which induce transcription of genes encoding the upstream enzymes IVD and MCC to the same extent as Leu, had any significant effect upon hlyA transcription, indicating that the transcriptional regulation of hlyA differs from that of the ivdA, mccA, and mccB genes in the chromosome III Leu catabolic cluster (9). Acetate, another gluconeogenic carbon source, had no inducing effect. The hlyA transcriptional pattern is as expected for a gene playing a specific role in Leu, but not in Ile or Phe catabolism. Although we note that deletion of mccB does not prevent induction, additional data precluded the conclusion that the inducer is a pathway metabolite located upstream of the MCC-catalyzed step (see below). The molecular basis of hlyA induction has not been investigated further.
A Truncation/Deletion Mutation Demonstrates the Specific Involvement of hlyA in Leu Catabolism-To demonstrate the physiological involvement of hlyA in Leu catabolism, we constructed a deletion/truncation allele by transformation with a linear DNA fragment in which 823 bp corresponding to hlyA codons 83-333 (including an intron) and 1235 bp corresponding to the hlyA 3Ј-UTR were flanking a genomic DNA insert containing the A. nidulans argB ϩ gene (Fig. 4). Site-directed inte-gration of this fragment mediated by a double cross-over event results in a mutant hlyA in which the coding region is truncated after residue 333. In this mutant, the sequence encoding the C-terminal 3-MG-CoA hydratase domain (codons 334 -599 of HlyA) is replaced by the A. nidulans DNA insert containing argB ϩ (Fig. 4). An argB2 arginine-requiring strain was transformed with this fragment, and arginine-independent clones were selected and purified. About 10% of the transformants were unable to grow on Leu as the sole carbon source. Three such transformants were analyzed by Southern hybridization using argB-and hlyA-specific probes and were shown to carry the truncation/deletion mutation shown in Fig. 4, denoted ⌬hlyA. All three transformants were found to be phenotypically indistinguishable and unable to grow on Leu. Therefore, one was chosen for further characterization. Northern blot analysis of this strain confirmed the absence of full-length message and the presence of a truncated message terminating within the argB ϩ -containing DNA insert and including the region coding for the N-terminal HL domain (Fig. 3, compare lanes 3 and 7;  Fig. 4).
Although ⌬hlyA prevents growth on Leu (Fig. 5), the mutation has no effect upon the ability of the fungus to use other branched chain amino acids (Ile, Val), Pro, Thr, or lactose as the sole carbon source (data not shown). As shown below, this ⌬hlyA strain accumulates 3-MG when challenged with Leu, strongly indicating that the truncation/deletion mutation results in the specific impairment of 3-MG-CoA hydratase activity. Taken together, these data formally demonstrate the physiological, specific involvement of hlyA in Leu catabolism.
Biochemical Characterization of ⌬hlyA: a Fungal Model for 3-Methylglutaconyl-CoA Hydratase Deficiency-To establish that the ⌬hlyA mutation blocks Leu catabolism at the level of 3-MG, we used GC/MS analysis of culture supernatants (9). Mycelia of the wild-type and ⌬hlyA strains were pre-cultured on glucose minimal medium and transferred to fresh minimal medium containing either lactose (at 0.05% (w/v)) or Leu (at 3 or 10 mM) as sole carbon source. Organic acids in supernatants of these secondary cultures were extracted and analyzed by GC/MS. Fig. 6a and b show that the metabolic profiles of the wild type and ⌬hlyA strains are virtually indistinguishable upon transfer to lactose minimal medium. In contrast, the mutant specifically accumulated the E (trans) and Z (cis) isomers of 3-MG upon transfer to 3 mM leucine minimal medium (Fig. 6, c and d). 3-MG in the mutant culture supernatant derives from 3-MG-CoA, the substrate of 3-MG-CoA hydratase. The specific accumulation in the mutant culture supernatant of 3-MG was markedly more prominent upon transfer of 10 mM Leu minimal medium (Fig. 6, e and f; see E and Z), demonstrating that it is Leu-dependent. Both the wild-type and the mutant supernatants showed an accumulation of 2-hydroxyisocaproic acid resulting from Leu transamination. We have previously shown that branched chain ketoacid dehydrogenase is limiting under these growth conditions (9).
In addition to 3-MG, we noted that the mutant also accumulated 3-hydroxyisovaleric acid (3-HIVA) resulting from the re-  Fig. 2). The deletion/truncation mutation results in a mutant HlyA product truncated after Leu-333 followed by 30 additional residues (dotted line) resulting from in-frame translation of the DNA fragment containing argB until ribosomes encounter a termination codon.
FIG. 5. Leu toxicity to ⌬hlyA and ⌬mccB single and double mutants. Wild-type and mutant strains were inoculated onto minimal medium plates containing the indicated carbon sources (see "Experimental Procedures" for concentrations) and incubated at 37°C for 5 days.
versibility of the pathway to 3-methylcrotonyl-CoA and hydration (see Fig. 1; Ref. 1). This is in marked similarity to human type I MGA patients, whose urine GC/MS analysis shows, in addition to the E and Z isomers of 3-MG, a significant amount of 3-HIVA. Indeed, the presence in urine of 3-HIVA in addition to 3-MG is a diagnostic feature of type I MGA patients which helps to differentiate them from the three other MGA types (see "Discussion"). Therefore, these data additionally illustrate the remarkably similar consequences of equivalent enzyme deficiencies in humans and fungi, giving credence to our metabolic model.
Identification of the Human Gene Homologue: Mutation Screen-The Human Genome databases were screened for candidate homologues using the HlyA amino acid sequence in TBLASTN searches. Relatively low sequence conservation among enoyl-CoA hydratases precluded the unambiguous identification of candidate genes but allowed our selection of two strong candidates, the 5-exon FLJ20909 gene and the 10-exon AUH gene. AUH encodes an RNA-binding protein with an enoyl-CoA hydratase domain. We screened these candidate genes in two type I MGA-affected brothers and detected a C589T (R197stop) truncating mutation in AUH. During our screen, Ijlst et al. (17) reported that AUH was the human 3-MG-CoA hydratase gene and the presence of the R197stop mutation in the same probands that we genotyped. This and a second report (16) (also including these siblings in the analysis) describe a total of five causative mutations in this gene, unambiguously demonstrating that type I MGA results from loss-offunction mutations in AUH. genotyped by three different groups; see above). The disorder presents as a neurometabolic disease with marked clinical heterogeneity. Thus, whereas some patients suffer from severe psychomotor impairment, in others, clinical manifestations are limited to delayed speech development. This suggests that some affected individuals having the biochemical phenotype might not come to clinical attention because of their (nearly) asymptomatic clinical phenotype and that the metabolites accumulating as a result of the type I MGA metabolic block are less toxic than in other metabolic disorders. We used our fungal model to test the possibility that Leu metabolites accumulating as a result of the enzyme deficiency are toxic. Fig. 5 shows that whereas growth of the wild-type and ⌬hlyA strains is virtually indistinguishable when lactose was the sole carbon source, growth of the mutant but not of the wild type is markedly inhibited on lactose plus Leu, indicating the formation of toxic metabolites from Leu as a result of the metabolic block. Of note, Fig. 5 shows that the inhibition caused by Leu metabolites in the strain deficient in 3-MG-CoA hydratase (Fig. 5, lactose plus Leu) is markedly less pronounced than that resulting from MCC deficiency (9), strongly indicating that the metabolites accumulating in the former are less toxic to the cells.

Metabolites Accumulating in a Strain Deficient for 3-MG-CoA Hydratase Are Toxic, Although Less so than in a Strain
Unexpected Additivity of a Null mcc Mutation with ⌬hlyA-MCC is a heterodimeric enzyme composed of ␣ and ␤ subunits. ⌬mccB is a null allele of the A. nidulans gene encoding MCC␤, phenotypically indistinguishable from a double mutant carrying null alleles in both genes encoding MCC subunits (9). A double ⌬hlyA ⌬mccB strain was constructed by meiotic recombination (see "Experimental Procedures"). Because the MCCcatalyzed step immediately precedes the 3-MG-CoA hydratase step in the Leu degradation pathway and the ⌬mccB allele is a null allele, the finding that Leu inhibition was markedly more pronounced in the double mutant than in either of the singlemutant parents (Fig. 5, lactose plus Leu) was unexpected. This phenotypic additivity of the ⌬hlyA and ⌬mccB mutations un-veiled by the Leu toxicity test strongly indicated the possibility that an MCC-independent reaction(s) leading to 3-MG-CoA was operating in A. nidulans.
We addressed this possibility by GC/MS analysis of supernatants of single-and double-mutant mycelia transferred to 3 mM Leu minimal medium (Fig. 7). In addition to 2-hydroxyisocaproic acid (see above), the single ⌬mccB mutant markedly accumulated 3-HIVA (Fig. 7c), the diagnostic compound of MCC deficiency (9), in the culture supernatant. 3-MG isomers were undetectable in this supernatant. In contrast, as described above, the ⌬hlyA strain accumulated large amounts of 3-MG isomers and lower levels of 3-HIVA (Fig. 7b). Like the single-mutant ⌬mccB strain, the double-mutant strain accumulated 3-HIVA (Fig. 7d). However, in contrast to ⌬mccB, the double mutant also accumulated 3-MG isomers to a significant extent (Fig. 7d), demonstrating that the null mccB mutation is not completely epistatic to the ⌬hlyA mutation and strongly supporting the above proposal that Leu metabolites can be converted to 3-MG in an MCC-independent manner. The relatively low levels of 3-MG isomers as compared with 3-HIVA determined by GC/MS in the double-mutant supernatant would suggest that this alternative reaction(s) is of lesser relative importance than the MCC-catalyzed step of the Leu catabolic pathway.
To exclude the possibility that the ⌬hlyA mutation resulted in activation of an otherwise cryptic 3-methylcrotonyl-CoA carboxylating enzyme, we measured MCC activity in wild-type and mutant protein extracts. MCC assays (Table I) negated this possibility. Although ⌬hlyA extracts showed a slight increase in MCC activity as compared with the wild type, this increase was mccB-dependent (Table I). However, very low levels of MCC activity were detectable in protein extracts from the single ⌬mccB and double ⌬mccB ⌬hlyA mutants (Table I;  inefficiently. We cannot exclude the possibility that the enzyme(s) involved in this reaction might account for the partial bypass of the MCC step that we observe in vivo with the ⌬mccB ⌬hlyA double mutant. However, the fact that levels of this MccB-independent activity are almost negligible as compared with those attributable to 3-methylcrotonyl-CoA carboxylase strongly argues against a significant physiological role for this alternative enzyme(s) in 3-MG-CoA biosynthesis.

DISCUSSION
In contrast to other genetically amenable eukaryotic microbial models such as S. cerevisiae, A. nidulans is able to use most amino acids as the sole carbon source (5), using pathways remarkably similar to those in human cells (6 -10, 12). A. nidulans is particularly useful for studying Leu (or other ketogenic amino acid) catabolism because acetyl-CoA resulting from its oxidation can be converted by the fungus (but not by human cells) into carbohydrates using the glyoxylate bypass of the Krebs cycle. Therefore, the fungus can grow on Leu as the sole carbon source. We have previously exploited this feature for the physiological characterization of the MCC-catalyzed step in the Leu catabolic pathway (9,12).
We report here the physiological characterization of A. nidulans 3-MG-CoA hydratase, the enzyme deficient in patients suffering from type I MGA. We used the amino acid sequence of the fungal enzyme to identify human homologues in sequence similarity searches. In two affected siblings, we detected a truncating mutation in a gene denoted AUH, encoding an enoyl-CoA hydratase. While this work was in progress, Ijlst et al. (17) and Nga Ly et al. (16) reported that ours and additional loss-of-function mutations in AUH cause type I MGA, formally demonstrating that AUH encodes 3-MG-CoA hydratase.
Evidence that this enzyme catalyzes a step of the Leu degradation pathway was based on the finding that excretion of 3-MG (and additional diagnostic metabolites, see below) by type I MGA patients increases 2-to 3-fold with a high protein diet, fasting, or a Leu challenge (26). We report here formal demonstration of the involvement of this enzyme in Leu catabolism: a mutant deleted for the complete 3-MG-CoA hydratase coding region is unable to grow on Leu (although its growth on Ile, Val, Phe, Pro, Thr, or carbohydrates is unaffected), and it shows dose-dependent 3-MG accumulation in the culture supernatant after a Leu challenge. This conclusion is relevant in view of the fact that 3-MG in the three other syndromes would appear not to proceed from Leu catabolism (Ref. 1, and see below). Our conclusion is further buttressed by the presence of hlyA homologues within bacterial Leu degradation operons (Fig. 1b).
The amino acid sequence of A. nidulans 3-MG-CoA hydratase and that of HL, the enzyme converting 3-hydroxy-3methylglutaryl-CoA to acetyl-CoA and acetoacetate, are C-and N-terminal domains, respectively, of a bifunctional polypeptide encoded by the hlyA gene. Transcription of hlyA is specifically induced by Leu, which agrees with the physiological role or both enzyme activities in Leu catabolism. Leu degradation occurs in mitochondria (1). The predicted hlyA bifunctional polypeptide would appear not to contain a potential mitochondrial targeting sequence but has a C-terminal peroxisomal targeting sequence. This result was unexpected. Although human HL is located in both mitochondria and peroxisomes through N-terminal mitochondrial leader and C-terminal peroxisomal targeting peptides, respectively (27), the AUH gene product (i.e. human 3-MG-CoA hydratase) has been reported to be mitochondrial (28); human MCC is mitochondrial (12), and A. nidulans MccA and MccB have predicted mitochondrial leader peptide sequences in their N termini. The biochemical phenotype of ⌬hlyA strongly suggests that at least a proportion of HlyA acts in mitochondria (see below). Mitochondrial protein import is post-translational, and certain mitochondrial proteins such as cytochrome c contain internal mitochondrial targeting sequences. Perhaps HlyA is one such protein.
A distinctive hallmark of type I MGA is that urinary excretion of 3-MG is accompanied by a marked increase in levels of 3-HIVA which is not found in any of the other three MGA syndromes (1). We have previously emphasized the remarkable similarities in the metabolic consequences of equivalent amino acid degradation blocks between A. nidulans and humans (6 -9). Our GC/MS analyses of culture supernatants demonstrates that a ⌬hlyA strain also accumulates 3-HIVA, in addition to 3-methylglutaconic acid. This demonstrates that, from fungi to humans, these two metabolites are diagnostic markers of 3-MG-CoA hydratase deficiency, and that our metabolic model faithfully resembles the human biochemical phenotype.
We used our fungal type I MGA model to demonstrate that metabolites accumulating as a result of the enzyme deficiency are toxic (Fig. 5). However, this toxicity seems to be less pronounced than that resulting from MCC deficiency. We note that in humans, but not in our relatively isogenic fungal strains cultured under standardized conditions, other genetic or environmental factors almost certainly contribute to the clinical phenotype (for discussion, see Ref. 29). Such relatively low toxicity might be a possible explanation of the mild clinical phenotype found in certain type I patients (1). We speculate that the apparent low frequency of the disease reflects that individuals deficient for 3-MG-CoA hydratase do not always come to clinical attention.
An important finding of this paper which challenges our current understanding of Leu catabolism is that ⌬mccB and ⌬hlyA mutations affecting two sequential steps in the catabolic pathway showed phenotypic additivity in our Leu toxicity test. Unexpectedly, GC/MS analysis unambiguously demonstrated that the double mutant accumulates 3-MG (resulting from 3-MG-CoA hydratase deficiency) in addition to 3-HIVA (resulting from MCC deficiency).
3-MG accumulation in type II and III MGA patients is modest (1, 2, 30). However, high 3-MG excretion has been reported for certain type IV patients (1). Although a deficient mitochondrial function underlies the biochemical phenotype of type II (4) and possibly of type III (3) patients, the primary enzymatic or molecular defect in type IV patients remains enigmatic. It has been suggested (30) that in type IV MGA patients increased levels of sterol/isoprenoid intermediates may be the source, through the mevalonate shunt proposed by Edmond and Popjá ck (31), of the abnormally elevated 3-MG levels. However, the "classical" mevalonate/isoprenoid shunt involves MCC (9) and, therefore, this cannot be the source of 3-MG found in our double mutant. It has also been suggested that abnormally elevated 3-MG plasma levels found in certain patients with Smith-Lemli-Opitz syndrome (a deficiency of cholesterol biosynthesis) result from increased flux through this shunt. However, this explanation does not explain why this increased flux would result in increased 3-MG levels without a detectable increase in 3-HIVA (which would reflect an overload of MCC). Our findings formally demonstrate that, at least in our metabolic model, 3-MG-CoA can be generated through an MCC-independent reaction(s), which might be responsible for increased 3-MG levels characteristic of type IV MGA patients.