Human α-l-iduronidase uses its own N-glycan as a substrate-binding and catalytic module

Abstract

N-glycosylation is a major posttranslational modification that endows proteins with various functions. It is established that N-glycans are essential for the correct folding and stability of some enzymes; however, the actual effects of N-glycans on their activities are poorly understood. Here, we show that human α-l-iduronidase (hIDUA), of which a dysfunction causes accumulation of dermatan/heparan sulfate leading to mucopolysaccharidosis type I, uses its own N-glycan as a substrate binding and catalytic module. Structural analysis revealed that the mannose residue of the N-glycan attached to N372 constituted a part of the substrate-binding pocket and interacted directly with a substrate. A deglycosylation study showed that enzyme activity was highly correlated with the N-glycan attached to N372. The kinetics of native and deglycosylated hIDUA suggested that the N-glycan is also involved in catalytic processes. Our study demonstrates a previously unrecognized function of N-glycans.

Asparagine-linked protein glycosylation, one of the major posttranslational modifications in eukaryotes, causes linking of an oligosaccharide chain to the Nδ atom of an asparagine in the N-glycosylation signal (Asn-Xaa-Ser/Thr, Xaa can be any amino acid except proline). N-glycosylation endows proteins with several abilities including immune recognition, ligand-receptor binding, and cell signaling, trafficking, folding, and stability (1–3). For some lysosomal enzymes such as glucocerebrosidase (4) and α-galactosidase A (5), the deglycosylation reduces the enzymes’ activities, presumably through a reduction in protein stability. However, the precise relationships between N-glycans and enzyme activities remain unknown (2, 6).

α-l-Iduronidase (IDUA; EC 3.2.1.76) is a lysosomal enzyme, and deficient activity of IDUA causes accumulation of glycosaminoglycans in lysosomes leading to mucopolysaccharidosis type I (MPS I) (7). Human IDUA is translated as 653 amino acids and N-glycosylated at six potential sites (N110, N190, N336, N372, N415, and N451), and then its N-terminal 26 residues are removed and it is processed to the mature form in lysosomes (8, 9). IDUA hydrolyses the glycosidic bond between the terminal l-iduronic acid (IdoA) and the second sugar of N-acetylgalactosamine (GalNAc)-4-sulfate/N-sulfo-D-glucosamine (GlcNS)-6-sulfate, which are the major components of dermatan/heparan sulfate (Fig. 1A). Thus, a defect of IDUA leads to excess storage of dermatan/heparan sulfate and causes a systemic disorder, MPS I, involving progressive mental retardation, gross facial features, an enlarged and deformed skull, a small stature, corneal opacities, hepatosplenomegaly, valvular heart defects, thick skin, joint contractures, and hernias (10). The prevalence of MPS I in England and Wales from 1981 to 2003 was 1.07 cases per 100,000 births (11). A recombinant human IDUA (hIDUA) expressed in Chinese hamster ovary (CHO) cells (marketed as Aldurazyme) was developed for enzyme replacement therapy (12), which is widely used for MPS I treatment.

Fig. 1.

Deglycosylation study of hIDUA. (A) Reaction scheme of dermatan sulfate hydrolysis catalyzed by hIDUA. (B) hIDUA was deglycosylated with PNGase F overnight under the indicated conditions (lanes 1–6), and then enzyme activities were measured (lanes 1–6, bottom; ND, not detected). Small amounts of the reaction mixture were mixed with ConA-Sepharose and washed three times, and then ConA-gel was separated by SDS/PAGE (lanes 7–12). The gel was stained with CBB. Nonidet P-40 prevents unfavorable aggregation, resulting in increasing activity. Human IDUA aggregates were produced due to the reaction mixture’s pH (20). (C) Deglycosylation analysis of hIDUA with Endo H. The procedure was the same as that in B other than the use of Endo H.

IDUA belongs to glycoside hydrolase (GH) family 39 in the CAZy database (13). To date, crystal structures of the bacterial GH39 β-xylosidase (XynB) have been reported (14, 15). In addition, a homology model of hIDUA constructed from Thermoanaerobacterium saccharolyticum XynB (PDB ID code 1PX8) has been reported (16). As the sequence homology between hIDUA and T. saccharolyticum XynB is quite low (28.4% similarity), the reliability of the model is not high. Recently, the crystal structures of apo-hIDUA, expressed in a plant seed, were solved. Nevertheless, the structure of hIDUA expressed in mammalian cells is strongly required for an insight into the basis of MPS I and the development of new therapies. In this study, we explored the functions of the N-glycans in hIDUA. We found that the deglycosylation of hIDUA with endoglycosidase H (Endo H), but not peptide-N-glycosidase F (PNGase F), reduces the enzyme’s activity. Concanavalin A (ConA) pull-down assay suggested that PNGase F–resistant N-glycans are essential for the enzyme activity of hIDUA. We also solved the crystal structures of hIDUA alone and in a complex with IdoA: they revealed that the N-glycan attached at N372 makes up one side of the substrate-binding pocket and directly interacts with the IdoA. Further, we found that the enzyme activity showed high correlation with the amount of N-glycan at N372. The kinetics of native and deglycosylated hIDUA implied that the N-glycan is also involved in the catalytic process. Our finding indicates a previously unrecognized function of N-glycans in the enzyme activity.

Results

Activity of hIDUA Is Reduced on Endo H but Not PNGase F Treatment.

To examine the influence of N-glycans on hIDUA activity, we carried out a deglycosylation study. Aldurazyme was digested with PNGase F or Endo H overnight and then subjected to the enzyme assay with 4-methylumbelliferyl α-l-iduronide (17) as the substrate. A small amount of each digested sample was subjected to ConA-Sepharose pull-down assaying to detect the residual N-glycans. The PNGase F–treated hIDUA showed no defect in enzyme activity, and some N-glycans showed PNGase F resistance (Fig. 1B, lanes 10 and 12). These glycans became sensitive to PNGase F on denaturation (Fig. 1B, lane 8). These results suggest that some N-glycans of hIDUA are so rigid or buried, and they cannot gain access to the catalytic site of PNGase F. We observed that in the presence of Nonidet P-40, the activity increased by about twofold (Fig. 1B, lanes 3 and 4). This result is presumably due to the prevention of unfavorable aggregation of hIDUA (18). On the other hand, on Endo H digestion, the hIDUA activity was reduced to 7% (Fig. 1C, lane 5). Endo H–resistant N-glycans were also observed (Fig. 1C, lanes 8 and 10); this is in agreement with the previous study showing that hIDUA expressed in CHO cells carries a complex type of N-glycans (8). These results suggest that high-mannose and/or hybrid type N-glycans have a rigid conformation and affect the activity of hIDUA.

Overall Structure of hIDUA.

To determine the structural basis of the relationships between the N-glycans and enzyme activity of hIDUA, we solved the crystal structures of apo- and IdoA-bound hIDUA at 2.3- and 2.76-Å resolution, respectively (19) (Fig. 2A; Table S1). We could build an almost full-length hIDUA (residues 27–642). There are two hIDUA subunits per asymmetric unit. Judging from the small buried surface area (548.3 Ų against the total accessible surface area of 25,127 Ų) and the results of gel filtration analysis, these two subunits are unlikely to represent a functional dimer (20). hIDUA consists of three domains: residues 42–396 form a classic (β/α)₈ triosephosphate isomerase (TIM) barrel fold, residues 27–42 and 397–545 form a β-sandwich domain with a short helix–loop–helix (482–508), and residues 546–642 form an Ig(Ig)-like domain. The latter two domains are linked through a disulfide bridge between C541 and C577. The β-sandwich and Ig-like domains are attached to the first, seventh, and eighth α-helices of the TIM barrel. A β-hairpin (β12–β13) is inserted between the eighth β-strand and the eighth α-helix of the TIM barrel, which includes N-glycosylated N372 (Fig. 2A). The topologies of the TIM barrel and β-sandwich domains of hIDUA are almost identical to that of XynB, which belongs to the same GH family 39 (12, 13). However, XynB has a shorter amino acid length than hIDUA and lacks the C-terminal Ig-like domain (Fig. S1).

Fig. 2.

Crystal structure of IdoA-bound hIDUA. (A) Domain organization of hIDUA. N-glycosylation sites observed in the crystal structure are indicated as yellow hexagons. N336, another N-glycosylation site, but with no glycans, is also indicated. The disulfide bond between C541 and C577 is indicated as S-S below the bar. The overall structure of hIDUA complexed with IdoA (side and end views) is shown. Human IDUA (subunit A) is shown as a cartoon representation with the TIM barrel domain (pink), β-sandwich domain (green), and Ig-like domain (blue). N-glycans (yellow), IdoA (blue), and phosphate are shown as stick models. The positions of N-glycosylation sites are indicated. (B) A stereo image of the omit map of the high mannose type N-glycan attached to N372(subunit A) in an apo-state crystal. The map is contoured at 2.5σ. N372 and catalytic glutamates (E182 and E299) are also shown. (C) Schematic drawing of the N-glycan structure at N372. The cleavage sites for Endo H and PNGase F are also indicated. Glycan linkage patterns are denoted as follows: β4, β1–4; α2, α1–2; α3, α1–3; α6, α1–6.

We observed at least five N-glycans in the electron density map other than that at N336 (Table 1). The loop region including N336 exhibited slightly poor electron density, and we could not place any sugars there. Although only one or two sugars were detected in most of the N-glycans, we observed long oligosaccharide chains at N190 (subunit B) and N372 (subunits A and B). N190 has complex type oligosaccharides, as previously predicted (8). The GlcNAc₃Man₂Gal₁Fuc₁ structure (Fig. S2) was visible at N190 of subunit B. This oligosaccharide interacts with the 590–592 loop region of symmetry-related subunit B.

Table 1.

N-glycan structures observed in the IDUA crystal structures

Subunit	Position	apo-IDUA	holo-IDUA
Subunit A	N110	GlcNAc	GlcNAc
	N190	Not observed	Not observed
	N336	Not observed	Not observed
	N372	(GlcNAc)2(Man)8	(GlcNAc)2(Man)8
	N415	(GlcNAc)2	(GlcNAc)2
	N451	Not observed	Not observed
Subunit B	N110	GlcNAc	GlcNAc
	N190	(GlcNAc)2(Fuc)(Man)2(GlcNAc)(Gal)	(GlcNAc)(Fuc)
	N336	Not observed	Not observed
	N372	(GlcNAc)2(Man)7*	(GlcNAc)2(Man)7*
	N415	(GlcNAc)2	(GlcNAc)2
	N451	GlcNAc	Not observed

^*Man9 was not observed.

The most remarkable feature is the N-glycan attached to N372, which is of the high-mannose type (8), and we clearly observed a GlcNAc₂Man₈ oligosaccharide chain in subunit A (Fig. 2B). The N-linked oligosaccharide chain is tightly bound to the surface of the TIM barrel. The tip of mannose residue (Man7) reaches the active site, and constitutes a part of the substrate-binding pocket (Fig. 2B). The N-glycan at N372 interacts with a protein through many polar- and water-mediated contacts (Fig. S3) including the side chains of H58, W306, S307, and Q370 and the backbone carbonyls of P54, L56, and H356. In addition, a hydrophobic interaction between Y355 and Man3 was observed. The structural characteristics of N-glycan at N372 were very similar in both subunits. A similar N-glycan interaction with an enzyme has been reported for Trichoderma reesei β-galactosidase. In the structure of the latter, the tip of the N-glycan attached at N930 comes near the active site; however, it seems too far for any direct interactions with a substrate (21).

Very recently, two crystal structures of hIDUA were released in the Protein Data Bank with the space groups of R3 (PDB ID code 4JXO) and P2₁ (PDB ID code 4JXP). The structures also contain a high-mannose type of N-glycan chain at N372; however, they have six sugars at most and lack Man7, possibly owing to the plant seed expression system used (22).

N-Glycan Is Involved in the Substrate Interaction.

In the IdoA-bound hIDUA structure, we clearly observed the electron density of the IdoA molecule at the center of the TIM barrel (Figs. 2A and 3A). There is little structural difference between the apo and IdoA-bound forms (rmsd = 0.253 Å, over 580 Cα atoms); only the side chain of D187 is flipped toward the active site, which forms hydrogen bonds between the Oδ2 atom and the backbone oxygen and Nδ atom of N181 (Fig. S4A). This change presumably yields a tight hydrogen network between the O2 atom of IdoA and Nδ of N181 (Fig. 3B).

Fig. 3.

Substrate binding site of hIDUA for IdoA. (A) Molecular surface representation around the substrate-binding pocket of hIDUA (subunit B). Mannose and basic residues are colored yellow and blue, respectively. The IdoA molecule and phosphate are drawn as stick models. An omit map of the IdoA (contour at 3σ) is also shown. (B) Details of the interactions between the IdoA and hIDUA (subunit B) are shown as a stereo image. The hydrogen bonds between IdoA and other related residues are indicated by black and cyan dashed lines, respectively.

There are 19 polar contacts between IdoA and the protein and 2 more contacts between IdoA and Man7 of the N-glycan (Fig. 3B; Table S2). The two oxygen atoms of the carboxyl group of IdoA interact with the backbone amide between G305 and W306 and the side chains of K264 and R363. The O4 atom of IdoA interacts with D349 and R363. The O3 atom also interacts with H91, D349, and Man7. In addition, the O2 atom interacts with H91, N181, and E299, a catalytic glutamate. The O1 atom interacts with Man7 and E182, another catalytic glutamate. The holo-hIDUA structure also provides us with information about the substrate specificity of the enzyme. For example, β-d-glucuronic acid, an epimer of IdoA, may have less affinity than IdoA for the binding site, as the hydrogen bonding between O5 and K264 Nζ found in IdoA (3.18 Å) will be lost in β-d-glucuronic acid (3.94 Å; Fig. S4B).

The O1 atom of IdoA faces the open side of the binding pocket, suggesting that downstream of the dermatan/heparan sulfate chain stretches out of this side (Fig. 3A; Fig. S5). We observed a phosphate ion between the side chains of H185 and H226. A putative hIDUA and dermatan sulfate complex model suggests that the sulfate moiety of IdoA-2-sulfate at the +2 position overlaps the phosphate (Fig. S5). Thus, H185 and H226 presumably interact with the sulfate moiety of the substrate sugar chain.

Amount of Glycan at N372 Correlates with the Enzyme Activity.

Because our structural study indicated that the N-glycan at N372 interacts with an IdoA molecule, we focused on N372. Aldurazyme was incubated with Endo H for varying times at 37 °C and subjected to enzyme assay. We reduced the amount of Endo H to see the effect of partial deglycosylation. The activity of hIDUA gradually decreased in an incubation time-dependent manner, the activity being almost completely lost after 48 h (Fig. 4C). The k_cat values of the native and deglycosylated (+Endo H, 48 h) hIDUA were 210 ± 3 and 9.50 ± 0.04 s⁻¹, and the K_m values were 290 ± 10 and 180 ± 0.4 µM, respectively (Fig. 4D). To determine whether the activity reduction was due to unfolding of the protein, we measured circular dichroism spectra of the native and deglycosylated hIDUA. The spectra were almost identical, suggesting that no aggregation or denaturation had occurred on Endo H treatment (Fig. S6).

Fig. 4.

Characterization of N-glycosylation at N372. (A) LC-MS profile of the native 369–383 peptide obtained on trypsin digestion of hIDUA. The spectra of the trivalent-ionized glycosylated (Upper) and deglycosylated (Lower) peptides are shown. The peak corresponding to m/z = 674.7 is magnified 100 times for clarity. (B) LC-MS profile of the Endo H–treated 369–383 peptide obtained on trypsin digestion of hIDUA, with the same representation as in A. The 1,050–1,300 range of the horizontal axis is magnified 1,000 times for clarity. (C) Ratio of the residual N-glycans during Endo H treatment determined by MS. The data were calculated using the peak area of peptide-GlcNAc₂Man_7–9 or peptide-GlcNAc of triply and quadruply charged ions. The enzyme activities of Endo H–treated (+Endo H) and nontreated (−Endo H) hIDUA are also plotted. (D) Michaelis-Menten plots for Endo H–treated (+Endo H, 48 h) and nontreated (−Endo H) hIDUA. The data are means of three repeated experiments ± SD. (Inset) Kinetic parameters of Endo H–treated and nontreated hIDUA.

To clarify whether the Endo H treatment indeed caused deglycosylation at N372, we digested the Endo H–treated hIDUA with trypsin, followed by analysis by LC-MS. As a result, we clearly detected triply and quadruply charged ions corresponding to the N372-containing fragment (residues 364–383) with GlcNAc and GlcNAc₂Man_7–9 (Fig. 4 A and B; Table S3). The ratio of the fragment (364–383) with GlcNAc₂Man_7–9 to the total fragment (364–383) was highly correlated with the enzyme activity (correlation: 0.989; Fig. 4C).

Furthermore, we examined the amounts of the other N-glycans attached to N336, N415, and N451, which could be cleaved by Endo H (8). The hIDUA treated with Endo H was digested with chymotrypsin and the peptide GlcNAcs, the Endo H products, were monitored (Fig. S7; Table S4). The amounts of the peptide GlcNAc at N336, N372, N415, and N451 increased in a time-dependent manner. However, the degrees of the inverse correlations between the enzyme activities and the amount of N372-GlcNAc were higher (correlation: −0.915) than those for the other N-glycosylation sites (N336, −0.616; N415, −0.615; and N451, −0.788) (Fig. S7). These results suggest that the N-glycan at N372, which forms a part of the substrate binding site, is involved in the enzymatic activity.

Discussion

Our structural and deglycosylation studies indicated that the N-glycan at N372 is essential for hIDUA activity. The deglycosylation study showed the different effects of Endo H and PNGase F (Fig. 1 B and C). The presence of PNGase F–resistant glycans was also observed in the previous study using CHO cell–expressed hIUDA (8). Furthermore, a report has described that treatment of Aldurazyme with PNGase F decreased enzyme activity by 50%, but significant activity still remained (23). The crystal structure revealed that the tightly bound oligosaccharide chain was linked to N372, which can be explained by the PNGase F resistance of the N-glycan. The k_cat and K_m values of the WT and deglycosylated hIDUA suggested that deglycosylation affects enzyme catalysis and substrate binding. As seen in the classical GH-A clan, the O1 atom of IdoA is the target for hydrolysis, and the carboxyl group of E182 would act as a proton donor to O1 (24, 25). We could observe the interaction between the O1 atom of IdoA and the O3 atom of Man7 in subunit B (Fig. 3B; Table S2). Thus, Man7 may somehow influence protonation of the O1 atom of IdoA.

The results of a molecular phylogenetic analysis of IDUA orthologs suggest the importance of the N-glycan at N372. Multiple sequence alignment showed that in all of the IDUAs, i.e., the Ciona to human ones, the positions corresponding to N372 and T374 are conserved (Fig. 5; Fig. S8), whereas five other N-glycosylation sites are not (Fig. S8). Additionally, identical or similar residues are clustered at the reaction center and along the interface of the N-glycan at N372 (Fig. S9A). These findings strongly suggest not only the conservation of this asparagine residue but also N-glycosylation throughout multicellular animals, and thus, the N-glycosylation at N372 should play a significant role in the function of IDUA.

Fig. 5.

The loop inserted between β8 and α8 (348–384) is highly conserved. (A) The loop inserted between β8 and α8 (348–384) is colored magenta. N372 is drawn as a stick model. (B) Alignment of sequences corresponding to the loop inserted between β11 and α10 (348–384) of human IDUA. Sequences were aligned using Clustal X ver.2 (34), and colored with ESPript (35). The names of the species and protein IDs are given in the legend to Fig. S8. N372 and T374 are indicated by pink arrowheads.

We could observe a phosphate ion around H185, H226, and R230 (Fig. 3A; Fig. S5). These residues are conserved among vertebrates (Fig. S8). The surface electrostatics showed that there is another positively charged patch, comprising H226 and R263, just outside the exit to the substrate binding cleft (Fig. S9B). These positively charged residues are also highly conserved among vertebrates (Fig. S8). Such positively charged patches are likely to contribute to binding of the sulfated glycosaminoglycan of dermatan/heparan sulfate. These results suggest that the vertebrate IDUAs have adapted to dermatan sulfate, the main component of skin.

More than 55 disease-associated missense mutations in the human α-L-iduronidase (IDUA) gene have been identified (Human Gene Mutation Database, www.hgmd.cf.ac.uk/). Among them, we paid attention to the W306 to Leu (W306L) gene (26) as a possible mutation that affects the N-glycan at N372. Structural analysis revealed that W306 interacts with the N-glycan at N372, and we examined whether this mutation would lead to a dysfunction of enzyme activity by molecular modeling and energy minimization. The rmsd between the WT and W306L is 0.023 Å, and the locations of S307 and F352 in the IDUA molecule were predicted to move slightly with the amino acid substitution (Fig. 6). This finding suggests that the amino acid substitution does not affect the catalytic center or, if at all, just a little. The W306L mutation may cause MPS I through the effect on the conformation of the N-glycan at N372.

Fig. 6.

Molecular modeling of the Trp306 to Leu mutant. (A) Polar contact between W306 and Man7. (B) Superposition of energy-minimized WT (cyan) and W306L (green). The side chains of W306 (Leu306), S307, and F352 are indicated as a stick model.

Most crystallographers remove the glycans when they try to crystallize glycoproteins to improve their homogeneity. However, the functionally important glycans tend to bind to a protein tightly and are resistant to the processing in the endoplasmic reticulum and Golgi apparatus (21, 27–29). Our study suggests that it is better to retain the high-mannose type glycans to not overlook the essential functions.

In conclusion, we first determined the structure of hIDUA, and then structural and biochemical studies revealed that the N-glycan at N372 was used as a substrate-binding module. Furthermore, the results of a kinetic study suggested that the N-glycan is directly involved in enzyme catalysis. These findings will be useful not only for elucidation of the molecular basis of MPS I but also for the development of new drugs for this disease.

Materials and Methods

The methods are described in full in SI Materials and Methods.

Samples and Chemicals.

The recombinant hIDUA expressed in CHO cells (Aldurazyme) was purchased from Genzyme Japan. IdoA was purchased from Carbosynth (United Kingdom).

Crystallization, Structure Determination, and Model Refinement.

We crystallized and solved the structure of hIDUA by the single isomorphous replacement with anomalous scattering (SIRAS) method as described previously (19). We automatically built an initial model using RESOLVE (30) and subsequently fixed it by hand with COOT (31), and then refined the apo-hIDUA structure with CNS (32) and REFMAC5 (33).