Introduction

Lysosomal α-galactosidase (GLA, EC3.2.1.22) catalyzes hydrolysis of terminal α-d-galactosyl residues of glycoconjugates, predominantly globotriaosylceramide (GL-3), in lysosomes. The enzyme is encoded by the GLA gene on the long arm of the X-chromosome and is synthesized on endoplasmic reticulum (ER)-bound ribosomes as a precursor form, which consists of 429 amino acid residues. Then, the enzyme is translocated into the lumen of the ER, with subsequent cleavage of the signal peptide consisting of 31 residues. Then, the enzyme is modified in the ER by the addition of N-linked oligosaccharides. The oligosaccharides are then trimmed in the ER, and the enzyme is transferred to the Golgi apparatus, where further modification of sugar chains and the addition of mannose 6-phosphate residues occur. The enzyme, having mannose 6-phosphate residues at the nonreducing ends of sugar chains, is transported to endosomes via mannose 6-phosphate receptors. Subsequently, the enzyme is transported to lysosomes, where it exerts its function as a mature form consisting of 398 residues. The native GLA from humans is thought to have a homodimeric structure.

A genetic defect causes progressive accumulation of GL-3, which results in Fabry disease (MIM 301500) (Desnick et al. 2001). This disease exhibits a wide clinical spectrum. Patients with the classic form having no GLA activity develop systemic manifestations, including pain in peripheral extremities, hypohidrosis, angiokeratoma, corneal clouding, renal failure, and cardio- and cerebrovascular disorders. On the other hand, there are also variant Fabry-disease patients with residual GLA activity and milder clinical manifestations, sometimes limited to heart disorders. So far, more than 500 genetic mutations causing Fabry disease have been reported (Desnick et al. 2001). Among them, gross alterations of the GLA gene have been identified in patients with the classic form, but missense mutations comprising the majority of mutations have been found in both classic and variant forms.

Recombinant GLAs produced in Chinese hamster ovary cells and human fibroblasts have been developed and are clinically available for enzyme replacement therapy for Fabry disease (Eng et al. 2001a, b; Schiffmann et al. 2000). Recently, another potential approach for treating Fabry disease was developed, and a clinical trial has been performed. This enzyme enhancement therapy is based on the ability of substrate analogues including galactose and 1-deoxygalactonojirimycin to improve the stability or transportation of mutant GLAs in cells, but the therapy is only efficient in a limited group of patients having specific missense mutations (Frustaci et al. 2001; Yam et al. 2006; Fan and Ishii 2007). As a high incidence of variant Fabry disease has been revealed by newborn screening (Spada et al. 2006), prediction of the clinical outcome of the disease is becoming more and more important to determine a proper schedule for treating the disease.

Previously, we built structural models of mutant GLAs resulting from 161 missense mutations by means of homology modeling with SYBYL/BIOPOLYMER (TRIPOS, St Louis, MO, USA) and examined the correlation between structural changes in GLAs and clinical and biochemical phenotypes (Matsuzawa et al. 2005).

Recently, we developed a structural analysis system for mutant proteins involving molecular modeling software, TINKER, developed by Ponder et al. (Department of Biochemistry and Molecular Biophysics, Washington University) (Ren and Ponder 2003), which is available worldwide. We applied it to investigations on lysosomal diseases including mucopolysaccharidosis type 6 (Saito et al. 2008), mucopolysaccharidosis type 1 (Sugawara et al. 2008), and Tay-Sachs disease (Ohno et al. 2008). We believe that the standardization of a structural analysis method will enable us to compare the results for different genetic disorders, which will provide us with a deeper insight into the basis of genetic disorders. Furthermore, because TINKER is free software, other researchers can easily conduct follow-up studies.

In this study, we conducted further structural investigation of Fabry disease using the same structural analysis system. We increased the number of Fabry patients for the analysis and examined structural changes in GLAs due to 212 amino acid substitutions by determining the number of atoms affected, the root-mean-square distance (RMSD), and the solvent-accessible surface area (ASA). Then, we paid attention to mutant GLAs for which substrate analogues are effective for stabilization or transportation to lysosomes and characterized their structural changes by coloring the affected atoms.

Materials and methods

Amino acid substitutions causing classic and variant Fabry disease

In this study, we analyzed 212 missense mutations (196 classic and 16 variant) responsible for Fabry disease. Amino acid substitutions, phenotypes, and references are summarized in Table 1.

Table 1 Fabry mutations, structural changes in α-galactosidase, and phenotypes
Full size table

Development of a structural analysis system for mutant proteins

We developed a structural analysis system for mutant proteins to examine their structural changes responsible for genetic diseases from various viewpoints. This system comprises six stages: (1) modeling mutant proteins, (2) determining the number of atoms affected by amino acid substitutions, (3) determining the RMSD values of all atoms in the mutant proteins, (4) determining ASA values of amino acid residues in the mutant proteins, (5) statistical analysis, and (6) coloring the atoms affected in the mutant proteins based on the differences between wild-type and mutant ones. Then, we applied the system to elucidation of the basis of Fabry disease.

Structural modeling of mutant GLAs responsible for Fabry disease and determination of the number of atoms affected by amino acid substitutions

Structural modeling of mutant GLAs was performed using molecular modeling software TINKER (Kundrot et al. 1991; Dudek and Ponder 1995; Kong and Ponder 1997; Pappu et al. 1998; Ren and Ponder 2003). The crystal structure of human GLA (Garman and Garboczi 2004) (PDB: 1R46) was used as a template, and energy minimization was performed. The root-mean-square gradient value was set at 0.05 kcal/mol Å. Each mutant model was then superimposed on the wild-type GLA structure based on Cα atoms by the least-square-mean fitting method (Kabsch 1976, 1978; Sakuraba et al. 2000, 2004). In this study, we defined that the structure was affected by an amino acid substitution when the position of an atom in a mutant differed from that in the wild type by more than the cutoff distance (0.15 Å) based on total RMSD, as described previously (Matsuzawa et al. 2005). Then, we determined the numbers of atoms affected in GLA main chain and side chain.

Determination of RMSD values of all atoms in mutant GLAs

RMSD values of all atoms in mutant GLAs were determined according to Weiner’s method (Weiner et al. 1984) to predict the degrees of GLA structural changes, and the average RMSD values for the classic and variant Fabry groups were determined and compared with each other, as described previously (Sugawara et al. 2008).

Determination of ASA values of amino acid residues in mutant GLAs

To predict the position of a substituted amino acid residue in the GLA molecule, the ASA value of each residue in the wild-type GLA was calculated using ACCESS (McDonald and Thornton 1994). The average ASA values of the residues for which a substitution had been found in the classic and variant Fabry groups were determined and compared with each other, as described previously (Saito et al. 2008; Sugawara et al. 2008).

Statistical analysis

Statistical analysis to determine differences in the numbers of atoms affected, RMSD values, and ASA values between classic and variant Fabry groups was performed using the F test and then Welch’s t test, it being taken that there was a significant difference if P < 0.05.

Coloring the atoms affected in mutant GLAs for which substrate analogues are effective

To determine the influence of amino acid substitutions geographically and semiquantitatively, coloring the affected atoms in the three-dimensional structure of GLA based on the distances between the wild-type and mutant ones was performed, followed by determination of the numbers of affected atoms, RMSD values, and ASA values. We analyzed amino acid substitutions including E59K, E66Q, M72V, I91T, A97V, R112H, F113L, A156V, L166V, N215S, G260A, Q279E, M296I, M296V, R301Q, R356W, and G373S, for which substrate analogues are effective for stabilization or transportation of mutant enzymes to lysosomes (Okumiya et al. 1995b; Yam et al. 2006; Ishii et al. 2007).

Results

Localization of amino acid substitutions responsible for Fabry disease

According to the crystallographic structure of human GLA (Garman and Garboczi 2004), the enzyme unit comprises two domains: an N-terminal (β/α)8-barrel domain and a C-terminal antiparallel β-sheet domain. The active-site pocket is localized in the C-terminal of the β-sheet of the N-terminal domain. We determined the locations of residues of which amino acid substitutions have been identified in Fabry disease patients in the homodimeric enzyme structure (Fig. 1a). Then, we localized the residues of which amino acid substitutions are responsible for classic (Fig. 1b, c) and variant (Fig. 1d, e) Fabry groups in the GLA subunit to compare them with each other. In the classic Fabry group, amino acid substitutions were distributed all over the enzyme protein molecule, including the active-site pocket. On the other hand, in the variant Fabry group, they were located far from the active-site pocket, and most of them were localized on the molecular surface of the protein.

Fig. 1
figure 1

Localization of amino acid substitutions responsible for Fabry disease in the α-galactosidase (GLA) structure. Secondary structures in GLA are shown as a tube drawing. Locations of amino acid substitutions identified in the classic and variant Fabry groups are shown in yellow and green, respectively. N The N-terminal (β/α)8-barrel domain. C The C-terminal antiparallel β-sheet domain. An arrow indicates the active-site pocket. a Homodimeric GLA structure, classic and variant Fabry groups. b Front view of the GLA subunit, classic Fabry group. c Side view of the GLA subunit, classic Fabry group. d Front view of the GLA subunit, variant Fabry group. e Side view of the GLA subunit, variant Fabry group

Full size image

Numbers of atoms affected by amino acid substitutions responsible for Fabry disease

We built structural models of the mutant GLAs and calculated the number of atoms affected by the amino acid substitution for each mutant model (Table 1), the results being summarized in Fig. 2.

Fig. 2
figure 2

Numbers of atoms in the a main chain and the b side chain of the α-galactosidase (GLA) protein affected by amino acid substitutions. Classic classic Fabry group. Variant variant Fabry group. Mutations located in the active-site pocket are colored red, others being colored black. Boxes indicate mean ± standard errors of mean

Full size image

The classic Fabry group showed a wide distribution. Averages for the affected atoms in the main chain and side chain were 108 and 130, respectively. In particular, regarding the former, 108 of the 196 classic cases (56%) had 50 atoms or more affected. There were 36 amino acid substitutions located in the active-site pocket, all of which cause the classic Fabry phenotype. These cases are colored red for their differentiation from other cases, which are colored black.

In contrast, the number of affected atoms in the variant Fabry group was low, and the distribution was narrower. These cases are colored black. Averages of the affected atoms in the main chain and the side chain were 18 and 21, respectively. In particular, regarding the main-chain atoms, 14 of the 16 variant Fabry cases (88%) had 49 atoms or less affected.

The F test showed that the distribution exhibited unequal variance (P < 0.05) between the classic and variant Fabry groups, and thus, Welch’s t test was performed. Results revealed that there were significant differences in the numbers of affected atoms in both the main chain and side chain between the two groups (P < 0.05).

RMSD values for amino acid substitutions responsible for Fabry disease

The RMSD values for the classic and variant Fabry groups were determined. Results are shown in Fig. 3. The average RMSD values in the classic and variant Fabry groups were 0.089 and 0.029 Å, respectively. In 115 of the 196 classic Fabry cases (59%), the RMSD value was ≥0.05 Å. On the other hand, it was <0.05 Å in 14 of the 16 variant Fabry cases (88%). Results of the F test followed by Welch’s t test showed that there was a significant difference in the RMSD values between the two groups.

Fig. 3
figure 3

Root-mean-square distance (RMSD) values for classic and variant Fabry mutations (Å). Boxes indicate mean ± standard errors of mean

Full size image

ASA values of amino acid substitutions responsible for Fabry disease

To determine and compare locations of amino acid residues in the GLA molecule associated with the classic (126 residues) and variant (15 residues) Fabry cases, the ASA values of the residues in the wild-type GLA structure were calculated, the results being shown in Fig. 4 (the result for each residue is presented in “Supplementary data No. l”). In the classic Fabry group, the average ASA value for the 126 residues analyzed was 13.3 Å2, 93 of them (74%) being <20 Å2. In the variant Fabry group, the average ASA value for the 15 residues analyzed was 25.1 Å2, eight of them being ≥20 Å2 (53%).

Fig. 4
figure 4

Solvent-accessible surface area (ASA) values of amino acid residues associated with classic and variant Fabry disease (Å2). Boxes indicate mean ± standard errors of mean

Full size image

The F test followed by Welch’s t test revealed that P was 0.09. Results suggest that the residues associated with classic Fabry mutations tend to be less solvent-accessible than those associated with variant ones, although this could not be confirmed statistically.

Coloring the affected atoms due to amino acid substitutions for which substrate analogues are effective

Regarding amino acid substitutions responsible for Fabry disease, we paid attention to 17 mutations for which substrate analogues improved the stability or transportation of mutant GLAs in cells, i.e., E59K (classic), E66Q (variant), M72V (variant), I91T (variant), A97V (classic), R112H (variant), F113L (variant), A156V (classic), L166V (classic), N215S (variant), G260A (classic), Q279E (variant), M296I (variant), M296V (variant), R301Q (variant), R356W (classic), and G373S (classic). Coloring the affected atoms in the three-dimensional GLA structure was performed for these mutations, the results being shown in Fig. 5.

Fig. 5
figure 5

Coloring the atoms in the three-dimensional structure affected by amino acid substitutions for which substrate analogues are effective. The degrees and distributions for E59K, E66Q, M72V, I91T, A97V, R112H, F113L, A156V, L166V, N215S, G260A, Q279E, M296I, M296V, R301Q, R356W, and G373S, for which substrate analogues are effective, are shown. Each atom is colored according to the distance between the atom in the mutant and the corresponding atom in the wild-type structure. The colors of the atoms show the distances as follows: blue <0.15 Å, 0.15 Å ≤ cyan < 0.30 Å, 0.30 Å ≤ green < 0.45 Å, 0.45 Å ≤ yellow < 0.60 Å, 0.60 Å ≤ orange < 0.75 Å, and red ≥ 0.75 Å. Arrows indicate the active-site pocket

Full size image

Coloring the affected atoms clearly allowed visualization of structural changes. Determining the numbers of atoms affected and RMSD and ASA values confirmed the results. In most mutant GLAs, the predicted structural changes were small (numbers of affected atoms in both the main chain and side chain <50, and RMSD < 0.05: E66Q, M72V, I91T, A97V, F113L, L166V, L215S, G260A, Q279E, M296I, M296V, and G373S), or localized on the molecular surface, although the structural changes were not small (ASA ≥ 20: E59K, R112H, and R301Q), regardless of phenotype. There were only two exceptions, A156V and R356W, and their clinical phenotype was classic. None of the amino acid substitutions for which substrate analogues are effective caused any structural changes in the active site.

Discussion

Considering the results of newborn screening, the incidence of Fabry disease is unexpectedly high (1 in 3,000–4,000 male newborns), especially the variant form (Spada et al. 2006). It is very important to examine the structural changes in the enzyme protein responsible for the different phenotypes to elucidate the basis of Fabry disease and to predict disease outcome. Garman and Garboczi calculated the side-chain-accessible surface area and revealed that the residues involved in Fabry mutations tend to be less solvent-accessible than the typical residues and that most of them lead to disruption of the hydrophobic core of the protein (Garman and Garboczi 2004; Garman 2007). However, there is little structural information on defective GLA proteins, although a large number of gene mutations responsible for Fabry disease have been reported so far.

In this study, we constructed structural models of Fabry mutant GLAs and examined their structural changes from various aspects by determining the number of atoms affected, as well as RMSD and ASA values, using our structural analysis system. The results revealed that structural changes in the classic Fabry group are generally large and tend to be in the core region of the protein. About 85% (116/196) of the amino acid substitutions leading to classic Fabry disease satisfied one of the conditions given below; number of affected atoms in the main chain ≥50, number in the side chain ≥50, RMSD ≥ 0.05 Å, or ASA < 20 Å2. They seriously affected protein folding or intracellular transport, leading to a deficiency of enzyme activity. All amino acid substitutions causing structural changes of the active-site pocket resulted in the classic phenotype. In such cases, structural changes would seriously affect expression of GLA activity, even if the number of influenced atoms is relatively small.

On the other hand, the predicted structural changes in GLA are generally small or localized on the surface of the molecule far away from the active site in the variant Fabry group. In such cases, a small amount of enzyme having GLA activity would be protected from the ER’s quality control system and transported to lysosomes, resulting in residual enzyme activity.

The number of affected atoms calculated using TINKER differs from that calculated using SYBYL/BIOPOLYMER. This is not surprising, because the minimized structure depends on the minimization algorithm, force field, and computational implementation. As shown in supplementary data No. 2 and No. 3, the number of affected atoms calculated using TINKER was generally larger than that using SYBYL/BIOPOLYMER. The mutant model obtained with TINKER was well optimized compared with that with SYBYL/BIOPOLYMER, which indicates that the mutant model constructed in this study is improved compared with the previous one (Matsuzawa et al. 2005). However, the supplementary data suggest that the number of affected atoms calculated using SYBYL/BIOPOLYMER was correlated with that using TINKER. Therefore, the discussion in the previous study is thought to remain correct.

Furthermore, we focused on the structural changes due to amino acid substitutions for which substrate analogues are effective and examined them by determining the affected atoms, RMSD values, and ASA values, followed by coloring the affected atoms. Results revealed that they cause small structural changes in GLA or are localized on the molecular surface, except for a couple of exceptions. None of them affected the active site. These results suggest that binding of a substrate analogue to a mutant enzyme protein in which a small structural change has occurred on the surface of the molecule reduces its folding defect and increases its stability in cells. Previously, we expressed mutant GLAs including M72V, L156V, L166V, Q279E, and R301Q in COS-1 cells and Sf9 cells and examined their biochemical characteristics (Ishii et al. 1993; Okumiya et al. 1995a, 1998; Kase et al. 2000). The expressed products had GLA activity, but they were unstable and easily lost their activity in vitro, suggesting that their structural changes are located far from the active site and that the degree of the changes is not so large. The results are well correlated with those of the structural analysis performed this time.

In conclusion, we investigated the structural changes in GLA responsible for Fabry disease. Results showed a correlation between the three-dimensional structural changes and clinical phenotypes, and they also revealed the characteristics of the structural changes in mutant enzyme proteins for which substrate analogues are effective. Structural investigation is useful for elucidation of the basis of Fabry disease, and it will increase our ability to determine a proper therapeutic schedule for this disease.