Expression Analysis of Phenylketonuria Mutations

Phenylketonuria is an autosomal recessive human genetic disease caused by mutations in the phenylalanine hydroxylase (PAH) gene. In the present work we have used different expression systems to reveal folding defects of the PAH protein caused by phenylketonuria mutations L348V, S349L, and V388M. The amount of mutant proteins and/or the residual activity can be rescued by chaperonin co-overexpression in Escherichia coli or growth at low temperature in COS cells. Thermal stability profiles and degradation time courses of PAH expressed inE. coli show that the mutant proteins are less stable than the wild-type enzyme, also confirmed by pulse-chase experiments using a coupled in vitro transcription-translation system. Size exclusion chromatography shows altered oligomerization, partially corrected with chaperonins coexpression, except for the S349L mutant protein, which is recovered as inactive aggregates. PAH subunit interaction is affected in the S349L protein, as demonstrated in a mammalian two-hybrid assay. In conclusion, serine 349, located in the three-dimensional structure lining the active site and involved in the structural maintenance of the iron binding site, is essential for the structural stability and assembly and also for the catalytic properties of the PAH enzyme, whereas the L348V and V388M mutations affect the folding properties and stability of the protein. The experimental modulation of mutant residual activity provides a potential explanation for the existing inconsistencies in the genotype-phenotype correlations.

Mammalian phenylalanine hydroxylase (PAH, 1 phenylalanine 4-monooxygenase, E.C. 1.14.16.1) is a non-heme iron and tetrahydrobiopterin-dependent enzyme that catalyzes the hydroxylation of phenylalanine to tyrosine. Defects in the human phenylalanine hydroxylase gene (GenBank TM cDNA reference sequence U49897, MIM 261600) cause phenylketonuria (PKU), a recessive disorder that if not treated from birth leads to variable degrees of mental retardation. PKU is in many ways regarded as a "model genetic disease," as clinical and biochemical characteristics are well defined, an effective treatment has been successfully implemented, both the gene and the enzyme are well characterized, mutations have been identified, genotype-phenotype correlations have been established, and an animal model has been produced (1). Although PKU is a classical monogenic disorder, the associated features are complex, as pointed out by Scriver and Waters (2). From the genetic point of view, more than 380 mutations have been described associated to different populations (3). After defining the mutational spectrum of PKU in several populations, the aim of the researchers has been the study of the genotype-phenotype correlations (4 -6). These studies have addressed the assessment of the severity of the mutations by in vitro expression analysis and examination of the phenotype in homozygous or functional hemizygous patients. Up to now 57 PKU mutations have been expressed in at least one in vitro system (7), and the data obtained have allowed in most cases the prediction of the biochemical phenotype based on the genotype. Nevertheless, some discrepancies have been detected in the genotype-phenotype correlations, especially in patients bearing mutations that result in decreased immunoreactive protein and consequently decreased activity when expressed in vitro (4 -6). This is the case of many missense mutations, which are broadly referred to as causing PAH enzyme instability. There are now several reports documenting increased instability and susceptibility toward aggregation and degradation of PKU mutant proteins (8 -10), and recently, the thermodynamic stability of native wild-type PAH has been examined (11,12) analyzing the contribution of instability to PKU compared with other reasons for reduced activity. Currently, knowledge of the three-dimensional structure of PAH is also available (13)(14)(15), providing essential information to understand the effect of different mutations on the architecture of the protein. The enzyme is structured in three domains, a flexible N-terminal regulatory domain (residues 1-110), a ␣-helical rich catalytic domain (residues 111-410), and an oligomerization domain (residues 411-452), which includes a tetramerization motif at the extreme C-terminal end (residues 428 -452).
The aim of this work has been to provide more information about the effect of three point mutations (L348V, S349L, and V388M) on PAH function, structure, and subunit interaction. Mutation L348V has been reported to have 25-33% residual activity in COS cells (3) and is an example of inconsistent genotype-phenotype correlations, as it is associated with different phenotypes in functionally hemizygous patients (5). V388M is one of the most frequent mutations in Spain, and we have shown that it retains 43% activity in COS cells (16), although normal levels of mutant immunoreactive proteins were detected. This apparent catalytic effect is discrepant with the fact that V388M affects residues located outside the active site in the three-dimensional structure of the PAH enzyme (13). On the contrary, S349L is located lining the active site, but expression both in COS cells and in Escherichia coli rendered a highly unstable protein (17). To clarify and extend these results, we have used several complementary expression systems (eukaryotic and prokaryotic) and experimental conditions (co-overexpression with chaperonins, different growth temperatures), providing evidence that the mutations affect directly the folding and stability of the protein. Additionally, we observe that the S349L mutation also affects the catalytic properties of the enzyme, which is attributable to the fact that serine 349 is involved in the iron binding site. Mutant PAH subunit interaction has been examined by the two-hybrid system in mammalian cells, and altered oligomerization is documented by size exclusion chromatography of mutant PAH expressed in the E. coli system. The different experimental approaches employed allow the demonstration of a major folding defect of the mutations causing protein instability, which can be modulated experimentally, revealing a possible mechanism to account for the existing inconsistencies in the genotype-phenotype correlations.

EXPERIMENTAL PROCEDURES
Expression analysis was performed in COS cells using the pRc/CMV expression vector (Invitrogen), as described previously (16), and in E. coli using pMALc2 (Biolabs) where PAH is cloned as a fusion protein with MBP under the control of an inducible promoter (17,18). Mutations were introduced in the PAH cDNA sequence by site-directed mutagenesis using the Gene Editor kit from Promega. COS cells (4 ϫ 10 6 or 6 ϫ 10 5 ), grown at 37 or 27°C were transfected with the Lipofectin reagent (Life Technologies, Inc). The plasmid pGroESL bearing the GroES and GroEL genes and the chloramphenicol resistance marker was from DuPont. The plasmids pMALc2-PAH wild-type or pMALc2-PAH mutant were cotransformed with pGroESL into E. coli JM109, and the colonies were selected using LB plates with ampicillin (0.1 mg/ml) and chloramphenicol (0.1 mg/ml). Cells were grown at 37°C, and expression of MBP⅐PAH fusion proteins and of GroES and GroEL was induced by the addition of 1 mM isopropyl-1-thio-␤-D-galactopyranoside. At the same time, 0.2 mM ferrous ammonium sulfate was added. Bacteria were harvested 16 -21 h after induction and disrupted by sonication in Na-Hepes 20 mM, NaCl 0.2M, pH 7.0, with 1 mg/ml lysozyme and 0.2 mM Pefabloc. After centrifugation, the supernatant (crude protein extract) was used to measure PAH residual activity by monitoring conversion of 14 C-Phe to 14 C-Tyr (17). Briefly, the standard reaction mixture performed in a 50-l final volume contained 150 -300 g of total protein, catalase (10 l at a concentration 100 units/l), 14 C-Phe (0.5 Ci, Ͼ450 mCi/mmol), and 10 l of 1 mM 6-methyltetrahydropterin (synthetic cofactor, added last), in 20 mM Na-Hepes, 0.2 M NaCl, pH 7. After 1 h at 37°C, the reaction was stopped by boiling for 5 min and centrifugation at 10,000 ϫ g for 5 min. A 6-l sample of the supernatant was spotted onto a TLC plate, developed two times in chloroform:methanol:amonia (55:35:10), dried, and visualized by autoradiography.
Wild-type and mutant fusion proteins were purified in an amylose column equilibrated with 20 mM Na-Hepes, 0.2 M NaCl, pH 7.0, and eluted with buffer containing 10 mM maltose (18). PAH activity was also measured in the purified fraction, using 30 -60 g of protein.
Size exclusion chromatography of the purified fusion proteins was performed at 4°C following the conditions described (18) and using a HiLoad Superdex 200HR column (1.6 cm ϫ 60 cm) prepacked from Amersham Pharmacia Biotech. The fast protein liquid chromatography system, UV monitor, and recorder were all from Amersham Pharmacia Biotech. Assignation of the different enzyme forms to the peaks obtained in the chromatograms was done by comparison with previously published elution positions of tetramers and dimers of the fusion protein (18) and by molecular mass value estimation using calibration curves obtained by running the following standard proteins, obtained from Amersham Pharmacia Biotech: ribonuclease A (13.7 kDa), chymotrypsinogen A (25 kDa), ovalbumin (43 kDa), albumin (67 kDa), aldolase (158 kDa), catalase (232 kDa), ferritin (440 kDa), and thyroglobulin (669 kDa). Blue dextran and acetone were used to determine the void volume (V 0 ϭ 44.06 ml) and the exclusion volume (V T ϭ 114.9 ml), respectively.
The cleavage of PAH protein from the MBP fusion partner was performed with Xa factor (ratio protein:Xa factor 1:100) at 4°C for 3 h. The reaction was further incubated at 4 or 37°C, and aliquots were removed at different times up to 24 h. After SDS-PAGE and Coomassie Blue staining, MBP and PAH proteins were quantified by laser densitometry.
The TnT-T7 transcription-translation system from Promega was used for pulse-chase experiments. The wild-type and mutant PAH cDNAs cloned in the pRc/CMV vector were amplified using a sense primer that introduces the T7 promoter and the consensus Kozak sequences close to the ATG initiation codon (5Ј-TAATACGACTCAC-TATAGGGAGCCACCATGTCCACTGCGGTCCTGGAA-3Ј). Five microliters of the polymerase chain reaction product from the wild type and mutant cDNAs were mixed with the reticulocyte lysate and [ 35 S]methionine-cysteine (14.3 mCi/ml). After a 35-min incubation at 30°C the reaction was stopped with excess cold methionine, RNase (1 mg/ml), and DNase (1 mg/ml). The whole reaction was incubated at 37°C, and aliquots were removed at different times between 1 and 8 h. All samples were separated by denaturing polyacrylamide gel electrophoresis, and the labeled PAH protein was quantitated by laser densitometry after fluorography.
Thermal stability profiles were performed with purified fusion proteins MBP⅐PAH expressed in E. coli. Aliquots (20 l, containing 30 -60 g of purified protein) were incubated at different temperatures for 10 min and chilled on ice. PAH enzyme activity was subsequently measured as described above.
For two-hybrid analysis we have used the Mammalian Matchmaker two-hybrid assay kit (CLONTECH). The plasmid pGL-G5 (kindly provided by P. Stä heli) was used as reporter vector containing the luciferase gene under the control of the Gal4 promoter. The full-length human PAH cDNA was excised from phPAH247 (19) with SmaI and EcoRI and subcloned into pBluescript. To introduce the normal PAH as fusion protein to the GAL4 (binding domain, BD) and VP16 (activating domain, AD) proteins, the PAH cDNA was excised from pBluescript with BamHI and XbaI and ligated with the AD and BD vectors previously digested with the same enzymes. Both plasmids with the normal cDNA were sequenced using a fmol sequencing kit (Promega) to confirm the in-frame cloning. The GeneEditor in vitro site-directed mutagenesis system from Promega was used to introduce the L348V, S349L, and V388M mutations in the PAH sequence. COS cells were plated in 6-well plates at a density of 4 ϫ 10 5 cells/well. In each transfection 1 g of each plasmid (reporter vector, AD-PAH and BD-PAH vectors) was introduced using Lipofectin reagent (Life Technologies, Inc.). The cells were harvested after 72 h, and luciferase activity was measured. In the transfection experiments using pRc/CMVPAH or in the two-hybrid system, PAH proteins were detected by Western blot using PH8 anti-PAH monoclonal antibody (20).
For the homology modeling, three-dimensional models of L348V, S349L, and V388M were built on the basis of the human phenylalanine hydroxylase coordinates determined by x-ray crystallography (13,14,15) (Protein Data Bank accession codes 1PAH and 2PAH). In the respective models, the formerly introduced alanine residues at the corresponding mutation sites were exchanged for the correct residues. The models were optimized for stereochemistry and refined by energy minimization using X-PLOR (21). The energy-minimized structures were heated at 1000 K and were further refined with a slow cooling simulated annealing molecular dynamics protocol using X-PLOR.

RESULTS
As described in a preliminary report, we have overexpressed and affinity purified PAH protein as fusion protein with MBP (17). PAH with the S349L mutation produced an unstable protein undetectable in SDS-PAGE. The expression of this mutation in eukaryotic cells showed that the mutant enzyme was not detected by Western blot, and consequently no residual activity in bacteria nor in COS cells could be measured (17). In this work we have extended the expression analysis to two other mutations, L348V and V388M, which have been previously reported to retain residual activity in COS cells, 25-33% for L348V (3) and 43% for V388M (16). Based on the emerging view that many missense mutations in human disease cause defective folding resulting in protein instability (22), we have tested this hypothesis using different experimental approaches known to prevent missfolding of proteins expressed in eukaryotic and prokaryotic systems.
Prokaryotic Expression Studies-In prokaryotes, PAH was expressed as a fusion protein with MBP. Expression of the mutant proteins compared with the wild type resulted in a variable but lower yield of fusion protein. When we performed the co-overexpression of the plasmid pGroESL, there was a considerable increase in the amount of fusion protein, both for wild type and mutant PAH, although the effect was more pronounced for the mutant proteins, especially for PAH harboring S349L, which as described before, was undetectable without chaperonin coexpression (Fig. 1). Thus, high levels of GroES and GroEL have a clear stabilizing effect on the mutant proteins, revealing a primary defect in folding and/or oligomer assembly. The increase in mutant protein correlated with an increase in residual catalytic activity with chaperonin coexpression, except for the S349L mutation, for which no enzyme activity is rescued (Table I).
The oligomeric state of the fusion proteins was analyzed by size exclusion chromatography. It has previously been described that the oligomeric structure of PAH is similar as fusion protein with MBP and as isolated enzyme (18,23), and our results also show that the wild-type fusion protein is resolved into three main components, a fraction of high molecular mass aggregates, eluting at the column void volume, a major fraction corresponding to tetramers, and a minor component of dimers. With chaperonin coexpression, there is no substantial change in the oligomeric profile of the wild-type protein (Fig. 2). Regarding the mutant proteins, L348V and V388M fusion proteins show a much lower proportion of tetramers and increased amounts of aggregated forms. When V388M is coexpressed with GroES and GroEL, a major peak corresponding to the tetrameric form is observed with a concomitant decrease in aggregated forms. For L348V with chaperonins the proportion of tetramers also increases, although there is still a considerable amount of aggregates (Fig. 2). In contrast, the S349L protein rescued by chaperonin coexpression is exclusively recovered as aggregates.
In addition to impaired folding, the mutations could also be affecting the stability of the assembled enzyme. Therefore, we analyzed the thermal inactivation profiles for the L348V and V388M fusion proteins purified in the pMalc2 expression system. The curves of the mutant enzymes are clearly shifted to lower temperature, demonstrating a reduced stability (Fig. 3). The half-denaturation temperatures were 59°C for the wildtype enzyme, 50°C for L348V, and 51°C for the V388M mutant protein. Similar denaturation profiles were observed if the enzymes were expressed with or without chaperonins.
Another approach used to analyze the effect of the PAH mutations on the stability of the protein was performed after digestion with factor Xa of the purified normal and mutant fusion proteins. The cleaved proteins were subsequently incubated at 4 or 37°C up to 24 h, and the Coomassie Blue-stained bands were quantified by laser densitometry. After cleavage, MBP and wild-type PAH are essentially stable up to 24 h, the ratio PAH/MBP is close to 1 up to 24 h. In contrast, immediately after cleavage, the amount of detectable mutant PAH forms, expressed as PAH/MBP ratio, is reduced to 50% (for V388M), 40% (L348V), or 20% (in the case of S349L). This remaining mutant protein is stable up to 24 h. Similar data are obtained if the fusion proteins are coexpressed with or without chaperonins (data not shown).
Eukaryotic Expression Studies-To test the relevance of these results obtained in the E. coli expression system, the mutations were also expressed in COS cells at different temperatures, 27 and 37°C. At low temperature, S349L mutant protein could be detected by Western blot analysis, reaching near normal levels, although no activity was rescued. Both L348V and V388M showed an increase in residual activity at 27°C, from 38 to 77% and from 43 to 78%, respectively. Western blot analysis revealed similar levels of immunoreactive protein for the wild-type and mutant proteins expressed at both temperatures (Table I).
Expression by in Vitro Transcription-Translation-To confirm a folding defect causing protein instability, we studied the effect of the three mutations on the stability of the PAH protein combining pulse-chase methods with protein expression in a For each MBP⅐PAH fusion protein, the following amount was loaded on the column: wild-type, 380 g; wild-type (ϩ GroESL), 400 g; V388M, 54 g; V388M (ϩ GroESL), 170 g; L348V, 13 g; L348V(ϩ GroESL), 120 g. For ease of comparison the sensitivity of the detector in the chromatograms of V388M and L348V without chaperonins was 5-fold greater than for the rest of the chromatograms. cell-free expression system. In vitro transcription and translation of the mutant cDNAs produced a labeled protein in a similar amount to the wild type. The labeled normal and mutant proteins were incubated at 37°C over a 7-h period. Densitometric quantification showed that the mutant proteins are degraded slightly more rapidly than the wild-type enzyme, the effect is more pronounced for the S349L protein (Fig. 4).
Two-hybrid Assays-The effect of the point mutations on the protein-protein interactions was analyzed using the two-hybrid assay in mammalian cells. PAH protein was expressed fused to the GAL4 DNA binding domain (BD-PAH) and the VP16 activation domain (AD-PAH). Co-transfection of the wild-type proteins GAL4-PAH and VP16-PAH leads to the expression of cotransfected luciferase marker. No substantial change in luciferase expression was observed for the L348V and V388M proteins, indicating that these mutations do not affect the interaction between PAH subunits. In contrast, the luciferase activity obtained after cotransfection with S349L fusion proteins was reduced to 10% (Table I). Western blot studies of extracts from cotransfected COS cells were performed, and mutant proteins expressed as fusion protein of BD and AD were detected at similar levels to wild-type, demonstrating that the decrease in luciferase activity when the S349L mutation is present is not because of reduced levels of fusion proteins.
Three-dimensional Structural Analysis and Molecular Modeling-The structural implications of the L348V, S349L, and V388M mutations were based on the crystal structure of the dimeric (residues 117-424) and tetrameric (residues 118 -452) forms of PAH (PDB accession codes 1PAH and 2PAH). Serine 349 is located lining the active site, forming a hydrogen bond with histidine 285, which is one of the six coordinating ligands of the iron atom (Fig. 5A). Leucine 348 is situated on the inside of the protein, in close proximity to the iron, at a distance of 10 Å. Valine 388 is located in a hydrophobic ␤-sheet, away from the active site, surrounded by leucines 333 and 365. DISCUSSION Currently, most of the PKU missense mutations that have been characterized are believed to destabilize the protein structure. More than 50 mutations have been expressed at least in one in vitro system, and most of them have reduced levels or no immunoreactive protein, correlating with reduced or absent residual activity and not affecting mRNA levels (7). The degradation of PAH proteins harboring missense mutations could be promoted by a defect in folding or in oligomer assembly, as has been suggested in recent studies documenting altered oligomerization and/or increased aggregation of mutant PAH proteins (8,10,24). In this work, we report the effect of the L348V, S349L, and V388M mutations on the stability of the PAH protein, using different experimental approaches to reveal possible folding defects.
Our previous work demonstrated that the S349L mutation expressed in E. coli as a fusion protein with MBP produces an unstable protein not detectable in SDS-PAGE. This is in accordance with the situation in COS cells where S349L is also unstable and was not detected by Western blot (17). However,

TABLE I Expression analysis of PKU mutations in different systems
Data obtained from the expression analysis of wild-type and mutant forms of PAH in E. coli with and without chaperonins, in COS cells at different temperatures and in a two-hybrid system. Oligomeric state and proportion (percentage of total) of fusion proteins in E. coli were estimated by size exclusion chromatography (see Fig. 2 and "Experimental Procedures"). In the two-hybrid assay, the relative strength of interaction between PAH subunits was measured as luciferase activity, 100% interaction is the luciferase activity when wild-type PAH cDNA is present on both activating and binding domain; in the rest of the assays, the mutations are present on the activation domain. The data represent the mean of three independent experiments.  we show that the mutant protein can be rescued by high chaperonin levels in E. coli or low temperature in COS cells, which are known to alleviate folding defects, as has been described in other diseases involving mitochondrial enzymes such as medium chain acyl-CoA dehydrogenase (25), short chain acyl-CoA dehydrogenase (26), or E1 decarboxylase of the branched-chain ␣-keto acid dehydrogenase complex (27). Nevertheless, the catalytic activity of the mutant protein was not rescued in either system, as could be predicted from the result obtained after size exclusion chromatography of the fusion protein expressed in E. coli, which consists exclusively of inactive aggregates. This suggests that apart from a structural alteration there is also a catalytic effect. In the three aromatic amino acid hydroxylases, PAH, tyrosine hydroxylase, and tryptophan hydroxylase, the catalytic iron is coordinated by three highly conserved amino acids (28). In PAH it is coordinated by two histidines (residues 285 and 290) and one glutamic acid (residue 330), and serine 349 forms a hydrogen bond with histidine 285 participating directly in the structural maintenance of the iron binding site (13). The maintenance of the iron binding site is crucial for the structural integrity of the monomer allowing correct folding/ oligomerization of the protein (29). In this way, the replacement of this important residue would affect the iron binding site (Fig. 5B) and subsequently the correct assembly of the PAH dimers and therefore the hydroxylation capacity of the enzyme.
The results obtained with L348V and V388M show that they are structural mutations affecting folding and assembly to active tetramers. When expressed in E. coli as fusion proteins, the amount of protein and the residual activity increases with GroES and GroEL coexpression. GroES and GroEL are the prokaryotic homologues of eukaryotic Hsp60/Hsp10, which harbor an ATPase domain and are thought to assist polypeptide folding by partially unfolding nonfunctional conformations that thus escape degradation reinitiating the folding process (30). Increasing the pool of chaperonins will therefore increase the fraction of protein that acquires a functional conformation. In addition to impaired folding, the thermal inactivation experiments reveal that the half-denaturation temperatures of the mutant proteins are clearly shifted to lower temperatures, showing a reduced stability of the assembled enzyme and irrespective of whether the protein has been expressed with or without chaperonins. The half-denaturation temperature obtained for the wild-type fusion protein (59°C) correlates with the results obtained by IR spectroscopy for the isolated PAH enzyme (11).
Size exclusion chromatography demonstrate an oligomerization defect as most of the mutant protein is present as aggregated forms, and the proportion of tetramers, which are the most active form (specific activity Ϸ2, 5-fold compared with dimers) (23), are clearly diminished compared with the wildtype protein. In the case of V388M, coexpression of chaperonins clearly shifts the oligomeric profile to a predominant fraction of tetramers. For L348V, the effect of chaperonins appears to be mainly an increase in the amount of protein. In COS cells we observe near normal amounts of mutant immunoreactive protein at 37°C associated with a reduced activity (38% for L348V and 43% for V388M). This could be explained by the fact that the capacity of the folding system in COS cells appears to be not so rate-limiting as in E. coli, which is more sensitive to reveal folding mutations, as has been argued in medium chain acyl-CoA dehydrogenase deficiency (31). In COS cells, mutant PAH proteins are apparently trapped in a relatively stable conformation, which could represent a folding intermediate or, in any case, not a correctly assembled fully active tetrameric form. Lowering the cultivation temperature has a clear positive effect in the yield of mutant protein acquiring a functional conformation, as reflected by the increase in residual activity to near normal levels. The three-dimensional structure analysis shows no evident consequences for mutations L348V and V388M, as no drastic change is predicted. However, all the results obtained in both the prokaryotic and the eukaryotic expression systems are clear and point to a genuine effect on folding for the L348V and V388M mutations.
We have also used the two-hybrid assay in mammalian cells as a powerful approach for the analysis of the effect of PKU mutations on assembly/oligomerization of the PAH dimers. In this experiment the reporter gene expression implies the association between two different PAH subunits. The dramatic reduction in luciferase activity when the COS cells were transfected with the mutant S349L protein indicates the effect of this mutation on the oligomerization of the enzyme. On the contrary, L348V and V388M show no effect on the proteinprotein interactions in the two-hybrid assay, consistent with the presence of some dimers and tetramers in the prokaryotic expression system. The results correlate with the observations in the eukaryotic system where there is some residual activity present.
In conclusion the present study demonstrates that the L348V, S349L, and V388M mutations affect the folding properties of the PAH protein, although S349L also shows a catalytic effect, because this residue is directly involved in the iron binding site. In addition, we demonstrate that L348V and V388M also affect the stability of the enzyme once assembled. Most significantly, we show that experimentally the amount of protein can be modulated depending on chaperonin and temperature conditions, providing clues to explain the genotypephenotype inconsistencies described for some mutations. These results obtained with a cytoplasmic protein extend similar in vitro studies of mitochondrial enzymes (25)(26)(27), revealing a common mechanism of many genetic disorders.