Dopa decarboxylase (DDC), or aromatic amino acid decarboxylase (AADC), is a pyridoxal 5′-phosphate enzyme responsible for the production of the neurotransmitters dopamine and serotonin. Deficit of this enzyme causes AADC deficiency, an inherited neurometabolic disorder. To date, 18 missense homozygous mutations have been identified through genetic screening in ∼80 patients. However, little is known about the mechanism(s) by which mutations cause disease. Here we investigated the impact of these pathogenic mutations and of an artificial one on the conformation and the activity of wild-type DDC by a combined approach of bioinformatic, spectroscopic and kinetic analyses. All mutations reduce the kcat value, and, except the mutation R347Q, alter the tertiary structure, as revealed by an increased hydrophobic surface and a decreased near-UV circular dichroism signal. The integrated analysis of the structural and functional consequences of each mutation strongly suggests that the reason underlying the pathogenicity of the majority of disease-causing mutations is the incorrect apo-holo conversion. In fact, the most remarkable effects are seen upon mutation of residues His70, His72, Tyr79, Phe80, Pro81, Arg462 and Arg447 mapping to or directly interacting with loop1, a structural key element involved in the apo-holo switch. Instead, different mechanisms are responsible for the pathogenicity of R347Q, a mere catalytic mutation, and of L38P and A110Q mutations causing structural-functional defects. These are due to local perturbation transmitted to the active site, as predicted by molecular dynamic analyses. Overall, the results not only give comprehensive molecular insights into AADC deficiency, but also provide an experimental framework to suggest appropriate therapeutic treatments.

INTRODUCTION

Dopa decarboxylase (DDC) (EC 4.1.1.28) is a homodimeric stereospecific α-decarboxylase that utilizes pyridoxal 5′-phosphate (PLP) as coenzyme. The enzyme is found in neural and peripheral tissues, notably liver and kidney. DDC is responsible for the conversion of l-Dopa and l-5-hydroxytryptophan (l-5HTP) to dopamine and serotonin, respectively, which are two of the major neurotransmitters of the mammalian nervous system. However, since the enzyme, at least in vitro, displays a much broader substrate specificity, it is also named aromatic amino acid decarboxylase (AADC). Naturally occurring and recombinant pig kidney and rat liver enzymes as well as human DDC in the recombinant form have been purified and characterized (13). Many structural and functional data were obtained from pig kidney, and, recently, from human DDC enzymes (4). They consist in the definition of the steady-state kinetic parameters for l-Dopa and l-5-HTP (5,6), the absorbance and dichroic features of the enzyme in the internal and external aldimine forms (5,6), and in the identification of structural elements functionally relevant for catalysis, like a loop susceptible to proteases and some residues at or near the active site (710).

Burkhard et al. (11) obtained the crystal structure of ligand-free pig kidney DDC and its complex with the anti-Parkinson drug carbidopa. Several features of DDC were evident in these structures. The overall structure of the protein is a tightly associated dimer in which each monomer consists of a large domain containing the PLP-binding site, a C-terminal small domain, and a N-terminal domain packing on the top of the large domain. In both these structures a short stretch of 11 amino acids (residues 328–339), invisible in the electron density map, represents a mobile loop important for the catalytic mechanism (7,10). The resolved structures also revealed the way in which PLP is anchored to the enzyme, which amino acid residues might be involved in the catalytic activity of the enzyme, and, importantly, how carbidopa, a substrate analog, binds DDC. The inhibitor forms a hydrazone linkage with PLP, its carboxylate moiety is nearly orthogonal to the PLP ring and its 3′ and 4′ catechol hydroxyl groups are hydrogen bonded to the hydroxyl group of the phosphate of the coenzyme and to Thr82, respectively. The resolution of the crystal structure of the apo form of human DDC is a recent achievement (12). Unexpectedly, this structure is an open conformation in which the active site becomes solvent exposed and the dimer interface is reduced to only the N-domains. The comparison of this structure with that of the holoenzyme from pig kidney allowed us to establish that a remarkable change of loop 1 (residues 66–84) transmitted to loops 2 (residues 100–110) and 3 (residues 323–357) of the adjacent subunit occurs in the transition from the apo to the holo form of DDC.

Since an inherited mutation can alter protein function in many different ways (change in the kinetic parameters and/or in the coenzyme binding, reduced protein expression level, protein misfolding or instability), it is essential to understand the molecular mechanism(s) of the effects of each disease-causing mutation. Characterization of some variants linked to AADC deficiency has begun to provide invaluable insights into their molecular defects, thus allowing to foresee the proper therapeutic treatments for patients bearing the examined mutations (3,23). However, a thorough study of the pathogenic effects of a significant number of missense pathogenic mutations is still lacking.

In the present study we have employed bioinformatic, kinetic and spectroscopic analyses to characterize all the AADC deficiency associated point mutations identified in homozygous patients. The already studied pathogenic variants (G102S, F309L and S250F) were included in this study for a more in-depth and complete molecular insights into the pathogenesis of AADC deficiency. In addition, the artificial variant F80A was also taken in consideration since Phe80, a key residue of loop 1, is involved in the switch between the apo and holo forms of the enzyme (12). All variants in the recombinant purified form were analyzed to determine their PLP binding mode and affinity, secondary and tertiary structure, surface hydrophobicity, and kinetic parameters. The results allowed us to highlight a linear correlation between the structural and functional effects of a large number of mutations, and to suggest that the affected residues are involved, even if at a different degree, into the conversion from the open apo conformation to the closed active holo form. This set of mutations does not include the catalytic mutation R347Q or the mutations L38P and A110Q, for which a catalytic effect predominant over a structural one has been observed. Bioinformatic and molecular dynamic (MD) analyses help to envisage the possible role of the residues Arg347, Leu38 and Ala110 affected by pathogenic mutations on the structure–function relationship of DDC.

RESULTS AND DISCUSSION

Location of pathogenic mutations in the crystal structure of human DDC

In the past few decades, a number of rare inborn errors due to the loss of function of enzymes requiring PLP to function were identified (24,25). The genetic defects are most frequently due to missense mutations. Among these inherited diseases, we focused our attention on homocystinuria, gyrate atrophy, primary hyperoxaluria type 1 and AADC deficiency. These are caused, respectively, by deficit of cystathionine-β-synthase (CBS), ornithine aminotransferase (OAT), alanine glyoxylate aminotransferase (AGT) and DDC. The reasons for such focusing are the following. First, a remarkable number (ranging from 8 to 150) of pathogenic mutations have been identified for these disorders. Second, the crystal structure of human CBS (26), OAT (27) and AGT (28) as well as that of pig kidney DDC (11) (∼ 90% sequence identity with human DDC) has been solved. The inspection of these structures reveals that while the pathogenic mutations of CBS, OAT, and AGT are spread out over the entire structure (Supplementary Material, Fig. S1), those of DDC are essentially grouped in a central region of each enzyme monomer. This region comprises a large portion of the large domain, two β-strands and an α-helix of the C-terminal domain as well as a limited portion of the N-terminal domain. The majority of the residues affected by mutations map to highly or moderately conserved residues among α-decarboxylases of distantly related species. This indicates that they are essential for enzyme structure and function (Fig. 1).

Figure 1.

Amino acids conservation and pathogenic mutation sites map of DDC. Ribbon representation of the dimeric structure of pig kidney holoDDC (pdb file 1js6) colored on the basis of the conservation degree as light gray, dark gray and black for low (up to 25%), medium (from 25 to 75%) and high (from 75 to 100%) conservation level, respectively, in the selected subset of DDC homologous sequence (see Material and Methods). The pathogenic mutation sites and the PLP molecules are represented as yellow and green sticks, respectively.

Figure 1.

Amino acids conservation and pathogenic mutation sites map of DDC. Ribbon representation of the dimeric structure of pig kidney holoDDC (pdb file 1js6) colored on the basis of the conservation degree as light gray, dark gray and black for low (up to 25%), medium (from 25 to 75%) and high (from 75 to 100%) conservation level, respectively, in the selected subset of DDC homologous sequence (see Material and Methods). The pathogenic mutation sites and the PLP molecules are represented as yellow and green sticks, respectively.

Eight mutations are located at loop 1 (T69M, H70T, H72Y, Y79C, P81L), 2 (G102S, A110Q) and 3 (R347Q), which have a leading role in the conversion from the apo to the holo form of DDC. Even if not affected by a pathogenic mutation, the residue Phe80 mapping to loop 1 deserves note since, together with Tyr79, it represents a key residue in the loop 1 rearrangement (12). The other mutations that affect residues not directly involved in the apo-holo transition comprise two mutations (L38P and P47H) of residues belonging to the N-terminus, four mutations (L408I, R412W, R447H and R462P) of residues lying on the C-terminal domain and four mutations (G123R, S250F, R285W, F309L) of residues mapping to the large domain, but outside the loops (Fig. 2A).

Figure 2.

Tridimensional representation of the mutation sites. Ribbons representation of the crystal structure of pig kidney holoDDC. The large domains of both monomers are colored white except for loops 1, 2 and 3 that are colored yellow, cyan and blue, respectively. The C-terminal and the N-terminal domains are colored in red and magenta, respectively. The described color code has been used for all the panels. (A) The overall dimeric structure of holoDDC with all the mutation sites represented as white sticks. (B, C, D and E) The residues subjected to pathogenic mutations (orange stick) and their main contact residues (white sticks) belonging to loops 1, 2 and 3 (B), the C-terminal domain (C), the N-terminal domain (D), and the remaining portion of large domain (E). The PLP molecules are represented as green sticks.

Figure 2.

Tridimensional representation of the mutation sites. Ribbons representation of the crystal structure of pig kidney holoDDC. The large domains of both monomers are colored white except for loops 1, 2 and 3 that are colored yellow, cyan and blue, respectively. The C-terminal and the N-terminal domains are colored in red and magenta, respectively. The described color code has been used for all the panels. (A) The overall dimeric structure of holoDDC with all the mutation sites represented as white sticks. (B, C, D and E) The residues subjected to pathogenic mutations (orange stick) and their main contact residues (white sticks) belonging to loops 1, 2 and 3 (B), the C-terminal domain (C), the N-terminal domain (D), and the remaining portion of large domain (E). The PLP molecules are represented as green sticks.

Functional and conformational impact of amino acid replacement on 3D networks of side-chain interactions in human DDC

The position and the possible structural effects of each mutation were analyzed by the inspection of the crystal structures of pig kidney holoDDC (pdb file 1js6) and human apoDDC (pdb file 3rbf), and by in silico mutational analysis. On the basis of the position of the mutated residue, the examined mutations can be grouped in four clusters. The first cluster comprises mutations of residues belonging to loops 1, 2 or 3 of DDC (Fig. 2B). The side chain of Thr69 and His70 points far from the active site, whereas His72 side chain faces the active site and contributes to generate the hydrophobic cleft that accommodates the substrate. Mutational analysis predicted that while the T69M substitution does not produce significant steric clashes with the neighboring residues, H70T and H72Y substitution could affect both the loop 1 conformation and the active site microenvironment, and might alter the position of Trp71, a residue involved in the substrate binding (13). Tyr79 and Phe80 are located in the center of loop 1 and, as already reported, upon PLP binding both residues undergo an inversion of position with respect to the loop 1 backbone (12). In fact, in the holoDDC structure Tyr79 interacts with the Nε1 of Arg447 and the side chain of Phe103 by, respectively, a hydrogen bond and a σ-π hydrophobic interaction. The phenyl ring of Phe80 engages Arg447 and Tyr274 with a cation-π and a π–π interaction, respectively, while in the apoDDC structure the hydroxyl group of Tyr79 is hydrogen bonded to the Nδ1 of His302, and Phe80 phenyl ring interacts with the Trp71 side chain. Moreover, Tyr79 is involved in the binding of the substrate (13). On this basis, it is reasonable to say that the Y79C and F80A substitutions could cause the loss of important anchoring points that stabilize the loop 1 conformation both in the holo and in the apo form. Pro81 is located in the center of loop 1 and does not interact with any of the neighboring residues neither in the holo- nor in the apoDDC form. Its presence reduces the flexibility of the backbone, thus possibly affecting the torsion movement of loop 1 during the apo-holo transition. It is reasonable to infer that the Pro-to-Leu substitution could alter either the conformation and/or the movement of loop 1. Gly102 and Ala110 are located in proximity of the N- and C-terminal ends of loop 2, respectively. In silico analyses, together with previously reported molecular modeling studies (3), indicate that the G102S substitution could introduce a minimal steric hindrance in the cleft between loop 2 and loop 3 and indirectly alter the position of Phe103, a residue that plays a role in substrate binding (11). The A110Q substitution generates a remarkable steric hindrance between the α-helix 8 and the N-terminal α-helix 4 (55–65) that could induce a repositioning of the two helices and a local alteration of loop 2 conformation. Arg347 adopts a central position in loop 3 at the entrance of the active site cavity. It is noteworthy that Arg347 is close to the loop 328–339, a mobile segment invisible in the electron density map that is essential for the decarboxylase activity (7,10,11). Nevertheless, in silico analyses predict that R347Q substitution does not generate substantial local structural alterations.

The second cluster comprises the L408I, R412W, R462P and R447H mutations that affect residues belonging to the C-terminal domain of the protein (Fig. 2C). Arg447 and Arg462 are directly interacting with loop 1. In fact, in the holoDDC structure, the side chain of Arg447 interacts with the phenyl ring of the Phe80 through an hydrophobic σ-π linkage and, in the apoDDC structure, the Nε1 of the side chain of Arg462 is hydrogen bonded with the side chain hydroxyl group of Tyr75. Thus, it is reasonable that R447H and R462P substitutions could perturb the correct position of loop 1 and/or the conformational change of the loop during the apo-holo transition. Leu408 is located in the N-terminal end of the β-strand 9 in proximity of Arg447 and contributes in generating the hydrophobic compartment that hosts Phe80 in the holoDDC. The L408I substitution does not change the polarity of the region but could determine a moderate steric hindrance that possibly disturbs the correct position of the side chain of Phe80. Arg412 is a surface residue located far from the active site at the C-terminal end of the β-strand 9 and does not interact with loops 1, 2 and 3. The drastic substitution R412W is predicted to generate a remarkable steric hindrance between β-strand 9 and the α-helix 25, which could impair the correct folding of the overall C-terminal domain conformation.

The third cluster comprises the N-terminal domain mutations L38P and P47H (Fig. 2D). Leu38 belongs to the 3/10 helix 3 and faces the Ala110 side chain, while Pro47 is positioned in the middle of the unstructured loop 42–52. The Leu38→Pro substitution alters the backbone flexibility of the short helix 3 and could impair the hydrophobic contact with Ala110, while the Pro47→His substitution could affect the conformation of loop 42–52. Considering the distance of the mutation sites from the active site, the possible functional effects of both substitutions are expected to be only indirect and hard to predict.

The fourth cluster comprises the G123R, S250F, R285W and F309L mutations that affect residues belonging to the large domain of DDC that do not directly interact with loops 1, 2 and 3 (Fig 2E). Gly123 and Arg285 are located far from the active site on the α-helix 8 and 16, respectively. The bulky arginine side chain introduced by the G123R substitution generates a considerable hindrance between the surface helix 8 and the β-barrel structure that wraps up the active site on the re face of PLP. Moreover the R285W substitution impairs the Arg285 side chain interactions with the Gln127 and Asn289, and introduces a bulky hydrophobic side chain in a surface region. Ser250 is located on loop 243–252. As already reported (23), the S250F substitution impairs the H-bond between the Ser250 and Thr245 side chains, and could locally change the surface hydrophobicity and destabilize the conformation of loop 243–252. Considering the distance from the active site and the chemical changes introduced by the G123R, S250F and R285W substitutions, it is reasonable to suggest that these mutations probably cause structural and/or folding alterations with only indirect functional effects. Differently from the other residues belonging to the third cluster that are located far from the coenzyme, Phe309 is placed in proximity of the PLP phosphate group in loop 304–311. As previously reported (3), the reduced hindrance of Leu309 in the F309L mutant might be less efficient in delimiting the apolar cavity of the substrate binding cleft.

Expression and purification of the variants

All mutations under study were inserted in the DDC cDNA by site-directed mutagenesis, and the variants were expressed in Escherichia coli and purified. The purified variants were homogeneous as indicated by a single band on SDS–PAGE electrophoresis with a mobility identical to that of the wild-type. Yields of the mutant enzymes after standard purification were quite variable, ranging from 5 to 90% with respect to that of wild-type. In comparison with wild-type DDC, the total expression level of each variant in E. coli was (i) equal, within experimental error, for the L38P, T69M, R347Q and L408I variants, (ii) ≥75% for the H70T, H72Y, Y79C, F80A, G102S, A110Q, G123R and F309L variants (iii) between 50 and 75% for the P47H, P81L, S250F, R285W, R412W, R447H and R462P variants. Moreover, the soluble fraction of the lysate contained ∼50% of the total protein for all the enzymatic species except the P47H, R285W, R412W and R447H variants, for which this percentage dropped to ∼20%. The effects of each amino acid substitution were analyzed by means of bioinformatic, spectroscopic and kinetic analyses of the variants.

PLP binding of variants: mode and affinity

Titration analysis of the apomutants with PLP fitted to the appropriate equation yielded the equilibrium dissociation constant for PLP (KD(PLP)) values reported in the first column of Table 1. All variants examined, except P47H, T69M, G102S, G123R, R347Q, R412W, L408I and S250F, showed, even if to a different extent, an increased value of the KD(PLP), ranging from 2- to at least 43-fold that of wild-type DDC (Table 1). In most cases, these data are in line with the in silico analyses. For the variants L38P and A110Q we could not explain the effect of the mutation on the coenzyme binding affinity by in silico mutagenesis, but we referred to MD analyses (see below). Human DDC displays absorbance maxima at 420 and 335 nm (ratio A335nm/A420nm ≈ 2) associated with positive dichroic bands centered at the same wavelengths, and emits fluorescence at 384 and 504 nm upon excitation at 335 nm and at 504 nm when excited at 420 nm. On the basis of these spectral features and of their dependence on pH, the absorbance bands at 420 and 335 nm have been attributed to the ketoenamine and the enolimine tautomers of the internal aldimine (5). The absorbance and dichroic properties of all variants examined were generally similar to those of the wild-type. However, with respect to the wild-type, the following differences were observed: (i) the absorbance maxima at 420 nm is 1–14 nm blue shifted for all variants but F309L and R347Q (Supplementary Material, Fig. S2), (ii) the ratio between the absorbance and dichroic bands of the enolimine and the corresponding ones of the ketoenamine is lower for all variants but R347Q and (iii) the optical activity (mdeg/absorbance unit) measured at the maximum absorbance of the ketoenamine changes for all variants, with the exception of T69M, G102S, G123R, S250F, R347Q and L408I. Altogether, these data indicate that the chiral environment of the PLP-binding site of the majority of the examined variants is altered, at varying degree.

Table 1.

Equilibrium dissociation constants for PLP (KD(PLP)) and steady-state kinetic parameters for l-dopa of wild-type DDC and variants

Enzyme Location of mutated residue KD(PLP) (nmkcat (s−1Km (mmkcat/Km (mm s−1
Wild type  43 ± 12 7.6 ± 0.1 0.11 ± 0.01 70.6 ± 8.4
T69M LOOP 1 100 ± 13 3.5 ± 0.2 0.46 ± 0.07 7.6 ± 1.2
H70T 510 ± 90 0.34 ± 0.02 0.69 ± 0.09 0.49 ± 0.07
H72Y 2145 ± 409 0.20 ± 0.01 1.38 ± 0.23 0.14 ± 0.02
Y79C 487 ± 110 0.32 ± 0.01 3.98 ± 0.37 0.080 ± 0.007
F80A 1520 ± 280 0.24 ± 0.01 1.03 ± 0.06 0.23 ± 0.04
P81L 390 ± 63 0.50 ± 0.01 0.19 ± 0.02 2.6 ± 0.5
G102Sa LOOP 2 61.7 ± 9.2 1.2 ± 0.1 1.2 ± 0.1 1.0 ± 0.1
A110Q 829 ± 60 0.015 ± 0.01 3.6 ± 0.4 0.004 ± 0.002
R347Q LOOP 3 54 ± 10 0.087 ± 0.005 0.49 ± 0.08 0.16 ± 0.06
L38P N-Terminal domain 460 ± 58 n.d. n.d. n.d.
P47H 100 ± 24 1.70 ± 0.06 0.60 ± 0.05 2.8 ± 0.2
L408I C-Terminal domain 105 ± 33 0.78 ± 0.06 1.78 ± 0.26 0.44 ± 0.07
R412W 84 ± 23 1.45 ± 0.05 0.27 ± 0.03 5.37 ± 0.61
R447H 1010 ± 160 0.38 ± 0.02 0.79 ± 0.13 0.41 ± 0.07
R462P 544 ± 50 0.40 ± 0.01 0.25 ± 0.03 1.60 ± 0.19
G123R Large domain (outside loops) 101 ± 13 3.30 ± 0.14 0.74 ± 0.07 4.46 ± 0.45
S250Fb 62 ± 24 2.1 ± 0.1 0.22 ± 0.04 9.3 ± 1.7
R285W 335 ± 39 2.08 ± 0.08 0.16 ± 0.03 13.0 ± 2.5
F309La 225 ± 26 0.47 ± 0.03 4.8 ± 0.6 0.097 ± 0.013
Enzyme Location of mutated residue KD(PLP) (nmkcat (s−1Km (mmkcat/Km (mm s−1
Wild type  43 ± 12 7.6 ± 0.1 0.11 ± 0.01 70.6 ± 8.4
T69M LOOP 1 100 ± 13 3.5 ± 0.2 0.46 ± 0.07 7.6 ± 1.2
H70T 510 ± 90 0.34 ± 0.02 0.69 ± 0.09 0.49 ± 0.07
H72Y 2145 ± 409 0.20 ± 0.01 1.38 ± 0.23 0.14 ± 0.02
Y79C 487 ± 110 0.32 ± 0.01 3.98 ± 0.37 0.080 ± 0.007
F80A 1520 ± 280 0.24 ± 0.01 1.03 ± 0.06 0.23 ± 0.04
P81L 390 ± 63 0.50 ± 0.01 0.19 ± 0.02 2.6 ± 0.5
G102Sa LOOP 2 61.7 ± 9.2 1.2 ± 0.1 1.2 ± 0.1 1.0 ± 0.1
A110Q 829 ± 60 0.015 ± 0.01 3.6 ± 0.4 0.004 ± 0.002
R347Q LOOP 3 54 ± 10 0.087 ± 0.005 0.49 ± 0.08 0.16 ± 0.06
L38P N-Terminal domain 460 ± 58 n.d. n.d. n.d.
P47H 100 ± 24 1.70 ± 0.06 0.60 ± 0.05 2.8 ± 0.2
L408I C-Terminal domain 105 ± 33 0.78 ± 0.06 1.78 ± 0.26 0.44 ± 0.07
R412W 84 ± 23 1.45 ± 0.05 0.27 ± 0.03 5.37 ± 0.61
R447H 1010 ± 160 0.38 ± 0.02 0.79 ± 0.13 0.41 ± 0.07
R462P 544 ± 50 0.40 ± 0.01 0.25 ± 0.03 1.60 ± 0.19
G123R Large domain (outside loops) 101 ± 13 3.30 ± 0.14 0.74 ± 0.07 4.46 ± 0.45
S250Fb 62 ± 24 2.1 ± 0.1 0.22 ± 0.04 9.3 ± 1.7
R285W 335 ± 39 2.08 ± 0.08 0.16 ± 0.03 13.0 ± 2.5
F309La 225 ± 26 0.47 ± 0.03 4.8 ± 0.6 0.097 ± 0.013

n.d., not determined.

aFrom reference (3).

bFrom reference (23).

Structural features of the variants

We first investigated whether the examined mutations affect the secondary structure composition of DDC. We measured the circular dichroism (CD) signal of each variant in the far-UV region and compared it to that of the wild-type. The overall shape of the CD spectrum of the variants in the region 190–260 nm was essentially identical to that of the wild-type indicating that the mutations do not affect the secondary structure of the protein. To examine whether the mutations previously investigated and under study here affect the tertiary structure of holoDDC, the near-UV CD spectra and the 1-anilino naphthalene sulfonic acid (ANS) emission spectra were acquired for all variants in the holo form and compared with the corresponding ones of the wild-type. As previously reported (3,12), the holo DDC displays a positive dichroic band in the 280 nm region, and binding of ANS to the holoenzyme results in an ∼4-fold increase in its fluorescence emission intensity accompanied by a 27-nm blue shift consistent with ANS binding to hydrophobic sites of the enzyme. In fact, it is well known that binding of ANS to the exposed hydrophobic clusters of a protein results in both an increased intensity of the emission of the dye and a blue shift in its maximum emission fluorescence. For all variants but one (R347Q) we observed a reduction, at varying degree, of the magnitude of the CD signal in comparison with that of the wild-type (Supplementary Material, Fig. S3). Moreover, upon addition of ANS to each of the holo variants we saw in comparison with the wild-type (i) no significant changes in the ANS emission spectrum of R347Q and (ii) an enhanced emission fluorescence intensity, even if at a different extent (from 7- to 35-fold) and a blue shift (in the range 14–34 nm) for all other variants (Fig. 3). When we plotted the magnitude of the near-UV CD signal versus the ANS emission intensity, we noticed for all the variants, except L38P and P47H (both bearing mutations of residues located on the N-terminal domain), a linear relation between the variables whose values span over a wide range with R347Q having equivalent values to those of wild-type (Fig. 4A). From this plot it can be derived that (i) the mutation of Arg347 does not significantly affect the tertiary structure of holoDDC and (ii) the H70T, H72Y, Y79C, F80A, P81L, A110Q, R447H and R462P mutations cause a structural alteration more pronounced than that due to the T69M, G102S, G123R, F309L, S250F, R285W, L408I and R412W mutations.

Figure 3.

ANS emission fluorescence spectra of wild-type DDC and variants in the holo form. Fluorescence emission spectra of 1 μm wild-type and the indicated variants in the presence of 20 μm PLP incubated with 1 μm ANS at 25°C for 1 h and registered upon excitation at 365 nm. All measurements were made in 100 mm potassium phosphate buffer, pH 7.4.

Figure 3.

ANS emission fluorescence spectra of wild-type DDC and variants in the holo form. Fluorescence emission spectra of 1 μm wild-type and the indicated variants in the presence of 20 μm PLP incubated with 1 μm ANS at 25°C for 1 h and registered upon excitation at 365 nm. All measurements were made in 100 mm potassium phosphate buffer, pH 7.4.

Figure 4.

Correlation between the ANS emission fluorescence intensity and the magnitude of the near-UV of wild-type and 19 variants. The ANS emission fluorescence intensity of the holoforms is plotted (A) versus the magnitude of their near-UV CD signal or (B) versus the difference between the magnitude of the near-UV CD signal of the holoenzyme and the corresponding apoenzyme. The diagonal lines correspond to a linear fit, with R2 values of 0.86 and 0.76, for (A) and (B), respectively.

Figure 4.

Correlation between the ANS emission fluorescence intensity and the magnitude of the near-UV of wild-type and 19 variants. The ANS emission fluorescence intensity of the holoforms is plotted (A) versus the magnitude of their near-UV CD signal or (B) versus the difference between the magnitude of the near-UV CD signal of the holoenzyme and the corresponding apoenzyme. The diagonal lines correspond to a linear fit, with R2 values of 0.86 and 0.76, for (A) and (B), respectively.

In order to verify if the alteration of the structural integrity is only limited to the holoforms, the near-UV CD spectra of the variants in the apo form were acquired. The data obtained indicate that (i) the variants T69M, H70T, H72Y, P81L and R447H display a positive dichroic signal whose magnitude is remarkably reduced (≤50%) in comparison with that of the wild-type, (ii) the variants L38P and R412W show a slightly negative band and (iii) the remaining variants do not exhibit significant changes in their dichroic spectra, or, as is the case for R347Q, any change at all (Supplementary Material, Fig. S3). The finding that the positive dichroic band of the apo wild-type is ∼2.8-fold lower than that of the holo is consistent with the conformational change accompanying the transition from the apo-open form to the holo-closed one of DDC (12). Thus, the extent of the difference between the dichroic bands of the holo and the apo forms could represent an indication of the impact of a mutation on the apo-holo conversion. When we plotted the difference of the near-UV CD signal between the holo and the apo form against the ANS emission intensity of each variant (Fig. 4B), a linear relation similar to that of Figure 4A could be observed. Thus, it seems reasonable to suggest that in the apo-holo transition (i) the mutations H70T, H72Y, Y79C, F80A, P81L, A110Q, R447H and R462P have more marked impact than the mutations T69M, G102S, G123R, F309L, S250F, R285W, L408I and R412W, (ii) the R347Q mutation does not have any effect, and the mutations L38P and P47H appear not to be involved. This could imply that 16 residues affected by mutations are, yet at varying degree, structural elements involved in the achievement of the correct conformation of holoDDC.

Catalytic features of the variants

In order to understand if all mutations taken in consideration in this study affect the decarboxylase activity, the steady-state kinetic parameters of the variants were measured (Table 1). All variants exhibited a decrease in the kcat/Km value spanning from ∼6-fold to ∼17 000-fold. Reduced kcat values were found for variants T69M (46%), G123R (43%), S250F and R285W (27%), P47H (22%), and G102S (16%). The remaining variants showed a more remarkable reduction in kcat value (0.2–10%). For most variants, the Km value was slightly increased (from 1.5- to 16-fold). Only Y79C, A110Q and F309L show a remarkably decreased l-Dopa affinity with an ∼35-fold increase in Km in comparison with that of the wild-type. The in silico analysis predicted a decrease of the substrate binding affinity for the substitution of Tyr79 and Phe309 by Cys and Leu, respectively, but did not explain the high Km value of A110Q. The kinetic parameters of the L38P variant for l-Dopa were difficult to obtain because of the very slow reaction rate. In fact, when 10 μm variant was allowed to react with 2 mml-Dopa in the presence of 50 μm PLP, we found that dopamine was linearly produced with time, reaching after 40 min a value of 13 nmol, which corresponds to an initial rate of 2.1 × 10–3 s−1. Assuming that this value is close to the kcat value of decarboxylation, the rate of this reaction in L38P is ∼3800-fold lower than the corresponding reaction in wild-type. Taken together, these results indicate that none of the examined mutations involves residues essential for the catalytic activity of DDC. However, they open the question if the loss of decarboxylase activity is related or not to the incorrect apo-holo transition. In attempting to clarify this point, we plotted the ratio kcat(wild-type)/kcat(variant) value as a function of the ratio ANSemission fluorescence intensity (variant)/ANSemission fluorescence intensity (wild-type) (or the difference of the near-UV CD dichroic signal between the holo and apo form) of the enzymatic species, which fit the linear correlation shown in Figure 4A and B. With the exception of A110Q and R347H, a linear relation statistically significant could be observed for 15 variants (Fig. 5). This implies that for these variants the following relationship holds: high extent of conformational change → increase of the exposure of hydrophobic clusters → decrease of decarboxylase activity. Thus, on the basis of all our data, the mutations can be arranged in two categories with a different degree of structural and functional effects within each category. One group comprises the holo forms of the variants H70T, H72Y, Y79C, F80A, P81L concerning residues mapping to loop 1, and those of R462P and R447H relative to residues mapping to the C-terminal domain. These variants showed a near-UV CD signal similar or close to that of the corresponding apo form, the highest increase in the additional hydrophobic regions, a relevant change in the PLP binding mode and affinity, and a remarkably decreased catalytic activity (from 15- to 38-fold) in comparison with the wild-type. This is in line with the in silico prediction showing that except T69M, all mutations of residues belonging to loop 1 or residues (Arg462 and Arg447) directly interacting with loop 1 could have a remarkable impact on the architecture of the active site. The other group includes the holo forms of T69M, G102S, G123R, S250F, R285W, F309L, L408I and R412W which are characterized by a near-UV-CD signal higher than that of the corresponding apo forms, a mild increase in the accessible hydrophobic surface areas, a modest change in the PLP binding mode and affinity, and a slightly decreased catalytic activity (from 2- to 9-fold) with respect to the wild-type. Consistent with these data, mutations of residues Thr69 and Gly102 located on the N-terminal end of loops 1 and 2, respectively, are predicted to induce a mild local perturbation at the active site microenvironment. Moreover, mutations of residues Gly123, Ser250, Arg285 and Phe309 located on the large domain as well as mutations of residues Leu408 and Arg412 located on the C-terminal domain are predicted not to directly affect both the active site and loops residues.

Figure 5.

Correlation between the decrease of the kcat and the increase of the ANS emission fluorescence intensity of 17 variants with respect to wild-type. The decrease in the kcat value with respect to wild-type was plotted versus the increase in the ANS emission fluorescence intensity. The diagonal line corresponds to a linear fit with a R2 value of 0.72.

Figure 5.

Correlation between the decrease of the kcat and the increase of the ANS emission fluorescence intensity of 17 variants with respect to wild-type. The decrease in the kcat value with respect to wild-type was plotted versus the increase in the ANS emission fluorescence intensity. The diagonal line corresponds to a linear fit with a R2 value of 0.72.

Besides the variants of these two groups, two variants (A110Q and R347Q) have different characteristics. In fact, the R347Q variant mapping to loop 3 exhibited a severely reduced catalytic activity despite its conformation being comparable to that of wild-type, while the A110Q variant mapping to loop 2 disclosed signs of loss of catalytic activity more pronounced than signs of impairment of structural integrity of its holo form. While the catalytic effect of the R347Q mutation should be related to the proximity of Arg347 to the flexible catalytic loop 328–339, that of A110Q is hard to explain. A more deep kinetic characterization should be required to identify the step(s) along the catalytic pathway of decarboxylation altered in each of these variants. Finally, it was surprising to find a different impact of the mutations L38P and P47H of residues located on the N-terminal domain and far from the active site. Indeed, both the variants showed similar extents of alteration of their tertiary structure (Fig. 4A). However, while the P47H variant displayed a modestly reduced catalytic activity, the L38P variant unexpectedly exhibited an extremely low decarboxylase activity such that the kinetic parameters values could not be determined.

MD

To gain insights into the functional effect of L38P and A110Q pathogenic mutations, MD simulations were run on DDC for wild-type, L38P and A110Q mutants. A structural analysis of the wild-type protein reveals that Leu38 and Ala110, located on two opposite alpha helices that run parallel to each other, form a hydrophobic interaction (Fig. 2D).

Three MD simulations of 50 ns were performed, totalizing 150 ns of total simulation time. We have first calculated the root mean square deviation (RMSD) on the backbone atoms of the three simulations in order to assess the stability of the three systems. Indeed, it can be appreciated (Supplementary Material, Fig. S4) that all systems became equilibrated after the first 10 ns of MD simulations. The simulated time span was long enough to relax and equilibrate the molecules, as indicated by the leveling of the RMSD from the initial conformations to values of ∼0.3 nm (Supplementary Material, Fig. S4). Moreover, since no significant differences could be observed between the mutants and the wild-type forms, we can assume that the mutations do not destabilize the folding of the enzyme in an extensive manner, in agreement with results obtained with purified proteins (see above).

We then calculated the root mean square fluctuations (RMSF) on the C-alpha atoms of the three systems in order to characterize the flexibility of the three simulated systems and to assess putative differences at the backbone level due to the mutations. Indeed, RMSF represents the standard deviation of atomic positions in the trajectory after fitting to a reference frame. Important differences on the backbone flexibility only localized in the neighborhoods of the mutated residues could be appreciated (Supplementary Material, Fig. S5). We thus concentrated in analyzing differences in the active site. With this aim, we have analyzed the ψ/Φ angle variation of residues involved in PLP-carbidopa binding during the MD simulation. In particular, we considered residues located on the three loops near the active sites, i.e. loops 1, 2 and 3. While in loops 1 and 3 there were not clear differences in the distribution of the angle values between the three simulations (data not shown), a significant difference could be appreciated for residues belonging to loop 2. Indeed, it can be clearly seen that Ile101 (Fig. 6A) and Phe103 (Fig. 6B) explore different regions of the Ramachandran plot upon mutations compared with the wild-type. In this regard, it can be observed that (i) the L38P mutant showed the most relevant variation with respect to the wild-type, and (ii) the A110Q mutant showed disperse values that localize in the middle between the values obtained for the wild-type and for the L38P mutant. It is interesting to note that Ile101 and Phe103 were already identified as involved in substrate binding (11,13).

Figure 6.

Distribution of the ψ/Φ dihedral angles values along the molecular dynamic trajectories. Dihedrals angle distributions for residues Ile101 (A) and Phe103 (B) in the wild-type form (black squares), the A110Q mutant (green squares) and the L38P mutant (red squares). The corresponding 3D graphic representations are squared in black for the wild-type, in green for the A110Q mutant and in red for the L38P mutant.

Figure 6.

Distribution of the ψ/Φ dihedral angles values along the molecular dynamic trajectories. Dihedrals angle distributions for residues Ile101 (A) and Phe103 (B) in the wild-type form (black squares), the A110Q mutant (green squares) and the L38P mutant (red squares). The corresponding 3D graphic representations are squared in black for the wild-type, in green for the A110Q mutant and in red for the L38P mutant.

In summary, the MD simulations revealed that, although the L38P and the A110Q mutations do not change the conformation of DDC structure in a macroscopic way, they create local perturbations at the backbone level extending, through loop 2, towards the active site. Moreover, differences in the dynamic properties of the A110Q mutant compared with the wild-type were less evident than those of the L38P mutant. Although these predictions are of course not an experimental evidence of the structural effects caused by these mutations, they are consistent with the finding that L38P mutation has an impact on the catalysis higher than that of A110Q.

Catalytic intermediates of variants

In order to ascribe the catalytic consequence of the amino acid substitutions under study to a particular reaction step, the effects of each mutation on the formation of the external aldimine were investigated by means of CD analyses. The CD features of the external aldimine of DDC wild-type with l-Dopa cannot be monitored because the kcat value is 7.6 s−1. This implies that, at 6 μm enzyme concentration, the DDC wild-type-l-Dopa complex remains close to saturation for only ∼45 s, a time too short to register a CD spectrum. Since the kcat value of the wild-type for l-5HTP is 1.0 ± 0.1 s−1 (Km = 0.05 ± 0.01 mm), we used this substrate to perform these measurements. As reported previously for wild-type human DDC (3), the binding of l-5HTP to the enzyme leads to the inversion of the 420 nm CD signal, i.e. the disappearance of the original positive CD band and its replacement by a negative CD band shifted to 440 nm and the increase of the positive 335 nm dichroic band. In contrast, when l-5HTP was added to each of the variants examined, the CD spectrum immediately registered was characterized by positive dichroic bands at 420 and 335 nm, the latter with a magnitude unaltered with respect to that of the corresponding holoenzyme. The only exceptions were relative to the R285W and R412W variants, for which the addition of l-5HTP caused a slightly negative dichroic band at ∼420 nm and did not change the 335-nm positive dichroic band (Fig. 7). Taken together, these data indicate that the binding of l-5HTP to each of the variants causes changes in the orientation of the coenzyme, with respect to the neighboring residues, different from that of the wild-type. It is of interest to note that this occurs in the case of mutated residues at the active site or interacting with the active site as well as in the case of residues located away from the active site. This suggests that the effect of each of these mutations is possibly communicated throughout the protein by disruption of functionally crucial network of amino acid interactions.

Figure 7.

CD spectra of wild-type DDC and variants in the presence of l-5HTP. CD spectra of wild-type DDC and the indicated variants at a concentration of 9 μm in 100 mm potassium phosphate buffer, pH 7.4, in the presence of saturating concentration of PLP, registered immediately after the addition of l-5HTP to a final concentration of 5 mm.

Figure 7.

CD spectra of wild-type DDC and variants in the presence of l-5HTP. CD spectra of wild-type DDC and the indicated variants at a concentration of 9 μm in 100 mm potassium phosphate buffer, pH 7.4, in the presence of saturating concentration of PLP, registered immediately after the addition of l-5HTP to a final concentration of 5 mm.

Therapeutic implications

The current clinical management of AADC deficiency is aimed at potentiating monoaminergic transmission, and usually includes the administration of: (i) l-Dopa and pyridoxine to increase the DDC activity; (ii) MAO inhibitors to slow down the dopamine degradation; and (iii) dopamine agonists to mimic the dopamine effect (29,30). Because of the scarcity of information about the molecular defects associated to each pathogenic mutation causing the AADC deficiency, the aforementioned drugs are almost always administered in combination, and in the absence of specific treatment guidelines. The presented results correlate each patient genotype with a specific enzymatic phenotype and allow us to group the DDC mutations on the basis of the possible responsiveness to the available pharmacological therapies. All mutants display at least one of the following functional defects: (i) high KD(PLP), (ii) high Km and (iii) low kcat. In particular, among the variants with a modest reduction of kcat (≤10-fold) the P47H, T69M, G102S, G123R, S250F, L408I and R412W mutations only cause a decrease of the substrate binding affinity, while the F309L and R285W mutations induce an increase in both the KD(PLP) and the Km values. The patients carrying the former group of mutations could be responsive to therapy with l-Dopa, whereas the patients harboring the second group of mutations could be responsive to the combined administration of pyridoxine and l-Dopa. Anyway, MAO inhibitors could be a useful treatment for patients bearing each of these eight mutations. It could be also envisaged, on the basis of our data on the expression level in E. coli, that pyridoxine administration to patients with R285W and R412W mutations could result in a chaperoning effect, as already demonstrated for S250F (23).

The mutations L38P, P47H, H70T, H72Y, Y79C, P81L, A110Q, R347Q, R447H and R462P cause a consistent reduction of the kcat value (from 15 to ∼3600-fold) with respect to the wild-type DDC. Therefore, although some of them have a remarkable impact on the KD(PLP) and/or the Km values, the administration of l-Dopa and pyridoxine to patients carrying these mutations may be useless. On the other hand, the administration of dopamine agonists and/or MAO inhibitors could be the most reasonable therapy, except for the patients carrying the mutations L38P and A110Q that cause an extreme reduction of the DDC catalytic activity (500- and 3600-fold). For these patients the only suitable therapeutic management appears to be the administration of dopamine agonists since MAO inhibitors may be ineffective. Unfortunately, considering the limited number of patients with AADC deficiency and the low frequency of each pathogenic mutation, the available data concerning clinical symptoms, disease course and family history are, at present, not sufficient to propose a correlation between the mutational impact on DDC functionality and the clinical severity of the disease, i.e. an enzymatic-clinical phenotype relationship.

CONCLUSIONS

To explore the molecular basis of the pathogenesis of AADC deficiency we studied the structural and functional effects caused by all the mutations linked to AADC deficiency applying an integrated strategy which uses bioinformatic, spectroscopic and kinetic analyses. To our knowledge, these measurements constitute the first comprehensive functional view of AADC deficiency mutants. The first point deriving from our study is that all variants either concerning residues located at/near the active site or far from the active site are characterized by a reduced catalytic activity, even if at a different degree, and a different microenvironment of the external aldimines with respect to the wild-type. Moreover, for all but one (R347Q) we observed a perturbation of the topography of the active site. The second finding is that the majority of the variants show, to varying extents, an alteration of the tertiary structure of the holoDDC linearly related to an enhanced exposure of hydrophobic sites, to an impairment of the proper apo-holo transition, known to be a conversion from an open to a closed active conformation (12), and to a decrease of the decarboxylase activity. This structure–function relationship allows us to identify the residues which play the most relevant role in the apo-holo conversion, i.e. His70, His72, Tyr79, Phe80 and Pro81 mapping to loop 1 as well as Arg462 and Arg447 directly interacting with loop 1. The data on F80A, an artificial variant, substantiate the essential role of loop1 in this process. Following this view, it can be suggested that residues Pro47, Thr69, Gly102, Gly123, Ser250, Arg285, Phe309, Leu408 and Arg412 are involved, even if to a less extent, in the achievement of the catalytic active holoform of DDC. Thus, the hindrance of a proper apo-holo conversion might be sufficient to induce pathogenicity. A different mechanism must be envisaged to explain the DDC malfunction for mutations that do not fit in this set of mutations. R347Q is a catalytic mutation since, albeit no conformational alteration was observed for this variant, it is characterized by a severe loss of decarboxylase activity. On the other hand, replacement of Ala by Gln at position 110 or of Leu by Pro at position 38 appears to have an impact on decarboxylase activity more pronounced than that on the structural integrity of the enzyme. On the basis of the comparative MD study of the putative structure of wild-type, A110Q and L38P, it can be speculated that the decrease in catalytic activity could be due to the rearrangements occurring around the mutated residue transmitted to the active site, in particular to Phe103 and Ile101 mapping to loop 2.

Finally, the finding that, among all the mutations, several of them affect the expression level and solubility of DDC in the E. coli expression system suggests folding defects of the variants harboring these mutations. However, the elucidation of possible folding defects (aggregation and/or cellular degradation) requires, as already done for S250F (23), more detailed and laborious expression studies using mammalian cellular systems, which more closely reflect the true situation.

In conclusion, our results for the first time indicate that a large percentage of the variants associated with AADC deficiency arising from point mutations shares, irrespective of their location, conformational changes that, at various degrees, do not allow a proper apo-holo transition and a full catalytic activity. This suggests that mutations occurring in distinct regions of a molecule can cause similar conformational changes with similar biological consequences. These events could not be predicted simply based on gene sequences. On the other hand, L38P, A110Q and R347Q result to be mainly or only catalytic mutations, for which additional kinetic features must be identified and understood to explain the kinetic reason(s) underlying their pathogenicity. Overall, these data allow us to acquire a complete knowledge of the enzymatic phenotype associated with each pathogenic mutation and to group patients with AADC deficiency into different therapeutic categories. This general picture of structural and/or functional effects of each mutation linked to AADC deficiency mutations will be a help to guide therapeutic decision and prevent inappropriate use of treatment regiments, thus improving the clinical outcome of this disease.

MATERIALS AND METHODS

Materials

PLP, l-Dopa, l-5HTP, 2,4,6-trinitrobenzene-1-sulfonic acid, isopropyl-β-d-thiogalactopyranoside, protease inhibitor cocktail were purchased from Sigma. ANS was purchased from Molecular Probes. All other chemicals were of the highest purity available.

Site-directed mutagenesis

The pDDChis plasmid, expressing the wild-type human DDC equipped with a C-terminal hexahistidine tag, and the expression vectors of the G102S, F309L and S250F DDC variants have been previously obtained (3,23). The expression vectors of the 16 DDC variants listed in Supplementary Material, Table S1 have been constructed using the Quick-Change II site-directed mutagenesis kit (Agilent Technologies) using the pDDChis plasmid as template and the oligonucleotides reported in Supplementary Material, Table S1. Successful mutagenesis was verified by DNA sequencing of the entire open reading frame.

Enzyme purification and assay

Purification of his-tagged wild-type DDC and of the variants was performed following the procedure previously described (3). Protein concentration was determined using the εM = 142 000 m−1 cm−1 at 280 nm. The PLP content of the variants was determined by releasing the coenzyme in 0.1 m NaOH and by using εM = 6600 m−1 cm−1 at 388 nm. The decarboxylase activity toward l-Dopa of wild-type DDC and the variants in the purified form was measured by the spectrophotometric assay described by Sherald et al. (31), and modified by Charteris and John (32). Measurements were performed in the presence of 50 μm PLP in 100 mm potassium phosphate buffer, pH 7.4. Data of enzymatic activity as a function of substrate concentration were fitted to the Michaelis–Menten equation.

Apoenzyme preparation and determination of KD(PLP)

For the apoenzyme preparation, each variant was incubated with 10 mm hydroxylamine in 0.5 m potassium phosphate buffer, pH 6.9 at 25°C for 3 h and the mixture was loaded on a desalting 26/10 column (GE Healthcare) pre-equilibrated with the same buffer without hydroxylamine (33).

The KD(PLP) value from the variants was determined by measuring the quenching of the intrinsic fluorescence of the apoenzyme (0.15 µm) in the presence of PLP at a concentration range of 0.01–10 µm in 100 mm potassium phosphate buffer, pH 7.4 and by fitting the data to the following equation:

$Y=YMAX[E]t+[PLP]t+KD(PLP)−([E]t+[PLP]t+KD(PLP))2−4[E]t[PLP]t2[E]t$

where [E]t and [PLP]t represent the total concentrations of the enzyme and PLP, respectively, Y refers to the intrinsic quenching changes at a PLP concentration, [PLP], and Ymax refers to the aforementioned changes when all enzyme molecules are complexed with coenzyme.

Molecular modeling and conservation analysis

Starting from the available structure of pig kidney holoDDC (pdb file 1JS6) (11) and human apoDDC (pdb file 3RBF) (12), the inspection of the 3D structures and the prediction of the possible steric clashes, loosing of interactions and local structural rearrangements caused by the analyzed mutations have been performed using the Molecular Operating Environment (MOE) software by Chemical Computing Group (34). The evolutionary conservation score of each amino acid of DDC has been calculated using the ConSurf server (35,36). The closed homologous sequences of DDC (150) have been collected from UNIREF90 using PSI-BLAST algorithm and the multiple sequence alignment was constructed using T-COFFEE.

MD simulations

The MD simulations were performed using the Gromacs program (37), and the structure of DDC enzyme carrying carbidopa inhibitor [pdb file 1JS3 (11)]. Since in this homodimeric structure the crystallographic coordinates between residues 328–339 are missing, a loop modeling script of the Modeller 9.8 software (38) was constructed. The mutations in the DDC enzyme were prepared using VMD software (39). The carbidopa ligand was parameterized with Prodrg software (40). Proteins were simulated in explicit aqueous solution inserted into a tetrahedral box of water molecules, ensuring that the solvent shell would extend for at least 1.0 nm around them. Gromos96 43a1 fields (41), in combination with the explicit simple point charge (spc/e) force field, were used for the simulation. Long-range electrostatic interactions were treated with the particle mesh Ewald (PME) method using a grid with a spacing of 0.12 nm (42). The time step was set to 2 fs. The cut-off radius for the Lennard–Jones interactions, as well as for the real part of the PME calculations, was set to 1 nm. The LINCS algorithm was used to constrain all bond lengths involving hydrogen atoms, and the time step used was 2 fs. The three systems were energy-minimized imposing harmonic position restraints of 1000 kJ mol_1 nm_2 on solute atoms, allowing the equilibration of the solvent without distorting the solution structure. After an energy minimization of the solvent and the solute without harmonic restraints, the systems were equilibrated and temperature was gradually increased from 0 to 300 K in 2 ns of simulation. The system was finally simulated for 50 ns for the analysis shown in this report. The RMSD/RMSF analysis and the Ramachandran plot analysis were performed through Gromacs tools. The 3D Ramachandran plots were obtained using VMD analysis tools.

Spectral measurements

All the spectral measurements were carried out using 100 mm potassium phosphate, pH 7.4, at 25°C. Absorption spectra were recorded with a Jasco V-550 spectrophotometer. CD measurements were made with a Jasco J-710 spectropolarimeter at a protein concentration of 6 μm. CD spectra were recorded at a scan speed of 50 nm/min with a band width of 2 nm and averaged automatically except where indicated. The ANS fluorescence measurements were taken with a FP750 Jasco spectrofluorimeter using 5 nm excitation and emission bandwidths at 1 μm protein and 15 μm ANS concentration.

Statistical analysis

Data analysis was performed by linear and non-linear regression curve fitting using Origin® 7.03 (Origin Lab) and the errors indicated result from fitting to the appropriate equation.

SUPPLEMENTARY MATERIAL

Supplementary Material is available at HMG online.

FUNDING

This work was supported by grants from M.I.U.R and the Consorzio Interuniversitario per le Biotecnologie CIB (IT).

ACKNOWLEDGEMENT

We thank the AADC research Trust (UK) for its interest in our research.

Conflicts of Interest statement. None declared.

REFERENCES

1
Moore
P.S.
Dominici
P.
Borri Voltattorni
C.
Cloning and expression of pig kidney dopa decarboxylase: comparison of the naturally occurring and recombinant enzymes
Biochem. J.

1996
315
Pt 1
249
256
2
Dominici
P.
Tancini
B.
Barra
D.
Voltattorni
C.B.
Purification and characterization of rat-liver 3,4-dihydroxyphenylalanine decarboxylase
Eur. J. Biochem.

1987
169
209
213
3
Montioli
R.
Cellini
B.
Borri Voltattorni
C.
Molecular insights into the pathogenicity of variants associated with the aromatic amino acid decarboxylase deficiency
J. Inherit. Metab. Dis.

2011
34
1213
1224
4
Bertoldi
M.
Mammalian dopa decarboxylase: structure, catalytic activity and inhibition
Arch. Biochem. Biophys.

2014
546
1
7
5
Montioli
R.
Cellini
B.
Dindo
M.
Oppici
E.
Voltattorni
C.B.
Interaction of human Dopa decarboxylase with L-Dopa: spectroscopic and kinetic studies as a function of pH
Biomed. Res. Int.

2013
2013
161456
6
Voltattorni
C.B.
Minelli
A.
Dominici
P.
Interaction of aromatic amino acids in D and L forms with 3,4-dihydroxyphenylalanine decarboxylase from pig kidney
Biochemistry

1983
22
2249
2254
7
Bertoldi
M.
Frigeri
P.
Paci
M.
Voltattorni
C.B.
Reaction specificity of native and nicked 3,4-dihydroxyphenylalanine decarboxylase
J. Biol. Chem.

1999
274
5514
5521
8
Bertoldi
M.
Voltattorni
C.B.
Multiple roles of the active site lysine of Dopa decarboxylase
Arch. Biochem. Biophys.

2009
488
130
139
9
Dominici
P.
Moore
P.S.
Castellani
S.
Bertoldi
M.
Voltattorni
C.B.
Mutation of cysteine 111 in Dopa decarboxylase leads to active site perturbation
Protein Sci.

1997
6
2007
2015
10
Bertoldi
M.
Gonsalvi
M.
Contestabile
R.
Voltattorni
C.B.
Mutation of tyrosine 332 to phenylalanine converts dopa decarboxylase into a decarboxylation-dependent oxidative deaminase
J. Biol. Chem.

2002
277
36357
36362
11
Burkhard
P.
Dominici
P.
Borri-Voltattorni
C.
Jansonius
J.N.
Malashkevich
V.N.
Structural insight into Parkinson's disease treatment from drug-inhibited DOPA decarboxylase
Nat. Struct. Biol.

2001
8
963
967
12
Giardina
G.
Montioli
R.
Gianni
S.
Cellini
B.
Paiardini
A.
Voltattorni
C.B.
Cutruzzola
F.
Open conformation of human DOPA decarboxylase reveals the mechanism of PLP addition to Group II decarboxylases

2011
108
20514
20519
13
Daidone
F.
Montioli
R.
Paiardini
A.
Cellini
B.
Macchiarulo
A.
Giardina
G.
Bossa
F.
Borri Voltattorni
C.
Identification by virtual screening and in vitro testing of human DOPA decarboxylase inhibitors
PLoS One

2012
7
e31610
14
Ren
J.
Zhang
Y.
Jin
H.
Yu
J.
Zhou
Y.
Wu
F.
Zhang
W.
Novel inhibitors of human DOPA decarboxylase extracted from euonymus glabra Roxb
ACS Chem. Biol.

2014
9
897
903
15
Brun
L.
Ngu
L.H.
Keng
W.T.
Ch'ng
G.S.
Choy
Y.S.
Hwu
W.L.
Lee
W.T.
Willemsen
M.A.
Verbeek
M.M.
Wassenberg
T.
et al.
Clinical and biochemical features of aromatic L-amino acid decarboxylase deficiency
Neurology

2010
75
64
71
16
Tay
S.K.
Poh
K.S.
Hyland
K.
Pang
Y.W.
Ong
H.T.
Low
P.S.
Goh
D.L.
Unusually mild phenotype of AADC deficiency in 2 siblings
Mol. Genet. Metab.

2007
91
374
378
17
Hyland
K.
Clayton
P.T.
Aromatic amino acid decarboxylase deficiency in twins
J. Inherit. Metab. Dis.

1990
13
301
304
18
Maller
A.
Hyland
K.
Milstien
S.
Biaggioni
I.
Butler
I.J.
Aromatic L-amino acid decarboxylase deficiency: clinical features, diagnosis, and treatment of a second family
J. Child Neurol.

1997
12
349
354
19
Swoboda
K.J.
Hyland
K.
Goldstein
D.S.
Kuban
K.C.
Arnold
L.A.
Holmes
C.S.
Levy
H.L.
Clinical and therapeutic observations in aromatic L-amino acid decarboxylase deficiency
Neurology

1999
53
1205
1211
20
Lee
N.C.
Chien
Y.H.
Hu
M.H.
Liu
W.S.
Chen
P.W.
Wang
W.H.
Tzen
K.Y.
Byrne
B.J.
Hwu
W.L.
Treatment of congenital neurotransmitter deficiencies by intracerebral ventricular injection of an adeno-associated virus serotype 9 vector
Hum. Gene Ther.

2014
25
189
198
21
Chtarto
A.
Bockstael
O.
Tshibangu
T.
Dewitte
O.
Levivier
M.
Tenenbaum
L.
A next step in adeno-associated virus-mediated gene therapy for neurological diseases: regulation and targeting
Br. J. Clin. Pharmacol.

2013
76
217
232
22
Hwu
W.L.
Muramatsu
S.
Tseng
S.H.
Tzen
K.Y.
Lee
N.C.
Chien
Y.H.
Snyder
R.O.
Byrne
B.J.
Tai
C.H.
Wu
R.M.
Gene therapy for aromatic L-amino acid decarboxylase deficiency
Sci. Transl. Med.

2012
4
134ra161
23
Montioli
R.
Oppici
E.
Cellini
B.
A.
Dindo
M.
Voltattorni
C.B.
S250F variant associated with aromatic amino acid decarboxylase deficiency: molecular defects and intracellular rescue by pyridoxine
Hum. Mol. Genet.

2013
22
1615
1624
24
Cellini
B.
Montioli
R.
Oppici
E.
Astegno
A.
Voltattorni
C.B.
The chaperone role of the pyridoxal 5′-phosphate and its implications for rare diseases involving B6-dependent enzymes
Clin. Biochem.

2013
47
158
165
25
Clayton
P.T.
B6-responsive disorders: a model of vitamin dependency
J. Inherit. Metab. Dis.

2006
29
317
326
26
Meier
M.
Janosik
M.
Kery
V.
Kraus
J.P.
Burkhard
P.
Structure of human cystathionine beta-synthase: a unique pyridoxal 5′-phosphate-dependent heme protein
EMBO J.

2001
20
3910
3916
27
Shen
B.W.
Hennig
M.
Hohenester
E.
Jansonius
J.N.
Schirmer
T.
Crystal structure of human recombinant ornithine aminotransferase
J. Mol. Biol.

1998
277
81
102
28
Zhang
X.
Roe
S.M.
Hou
Y.
Bartlam
M.
Rao
Z.
Pearl
L.H.
Danpure
C.J.
Crystal structure of alanine:glyoxylate aminotransferase and the relationship between genotype and enzymatic phenotype in primary hyperoxaluria type 1
J. Mol. Biol.

2003
331
643
652
29
Manegold
C.
Hoffmann
G.F.
Degen
I.
Ikonomidou
H.
Knust
A.
Laass
M.W.
Pritsch
M.
Wilichowski
E.
Horster
F.
Aromatic L-amino acid decarboxylase deficiency: clinical features, drug therapy and follow-up
J. Inherit. Metab. Dis.

2009
32
371
380
30
Pons
R.
Ford
B.
Chiriboga
C.A.
Clayton
P.T.
Hinton
V.
Hyland
K.
Sharma
R.
De Vivo
D.C.
Aromatic L-amino acid decarboxylase deficiency: clinical features, treatment, and prognosis
Neurology

2004
62
1058
1065
31
Sherald
A.F.
Sparrow
J.C.
Wright
T.R.
A spectrophotometric assay for Drosophila dopa decarboxylase
Anal. Biochem.

1973
56
300
305
32
Minelli
A.
Charteris
A.T.
Voltattorni
C.B.
John
R.A.
Reactions of DOPA (3,4-dihydroxyphenylalanine) decarboxylase with DOPA
Biochem. J.

1979
183
361
368
33
Voltattorni
C.B.
Minelli
A.
Vecchini
P.
Fiori
A.
Turano
C.
Purification and characterization of 3,4-dihydroxyphenylalanine decarboxyase from pig kidney
Eur. J. Biochem.

1979
93
181
188
34
Molecular Operating Environment (MOE), Chemical Computing Group Inc.
2013
1010 Sherbooke St. West, Suite #910, Montreal, QC, Canada, H3A 2R7
35
Ashkenazy
H.
Erez
E.
Martz
E.
Pupko
T.
Ben-Tal
N.
ConSurf 2010: calculating evolutionary conservation in sequence and structure of proteins and nucleic acids
Nucleic Acids Res.

2010
38
W529
W533
36
Celniker
G.
Consurf: using evolutionary data to rise testable hypotheses about protein function
Isr. J. Chem.

2013
53
199
206
37
Van Der Spoel
D.
Lindahl
E.
Hess
B.
Groenhof
G.
Mark
A.E.
Berendsen
H.J.
GROMACS: fast, flexible, and free
J. Comput. Chem.

2005
26
1701
1718
38
Eswar
N.
Webb
B.
Marti-Renom
M.A.
M.S.
Eramian
D.
Shen
M.Y.
Pieper
U.
Sali
A.
Comparative protein structure modeling using Modeller
Curr. Protoc. Bioinformatics

2006
Chapter 5, Unit 5 6
39
Humphrey
W.
Dalke
A.
Schulten
K.
VMD: visual molecular dynamics
J. Mol. Graph.

1996
14
33–38, 27–38
40
Schuttelkopf
A.W.
van Aalten
D.M.
PRODRG: a tool for high-throughput crystallography of protein-ligand complexes
Acta Crystallogr. D Biol. Crystallogr.

2004
60
1355
1363
41
Van Gunsteren
W.F.B.
Biomolecular Simulation: the GROMOS96 Manual and User Guide

1996
Zurich/Groningen
42
Ulrich
E.
A smooth particle mesh Ewald method
J. Chem. Phys.

1995
103
8577
8593

Author notes

R.M. and C.B.V. are both senior authors of the paper.