The structure of human GALNS reveals the molecular basis for mucopolysaccharidosis IV A

Abstract

Lysosomal enzymes catalyze the breakdown of macromolecules in the cell. In humans, loss of activity of a lysosomal enzyme leads to an inherited metabolic defect known as a lysosomal storage disorder. The human lysosomal enzyme galactosamine-6-sulfatase (GALNS, also known as N-acetylgalactosamine-6-sulfatase and GalN6S; E.C. 3.1.6.4) is deficient in patients with the lysosomal storage disease mucopolysaccharidosis IV A (also known as MPS IV A and Morquio A). Here we report the three-dimensional structure of human GALNS, determined by x-ray crystallography at 2.2 Å resolution. The structure reveals a catalytic gem diol nucleophile derived from modification of a cysteine side chain. The active site of GALNS is a large, positively charged trench suitable for binding polyanionic substrates such as keratan sulfate and chondroitin-6-sulfate. Enzymatic assays on the insect cell-expressed human GALNS indicate activity against synthetic substrates and inhibition by both substrate and product. Mapping 120 MPS IV A missense mutations onto the structure reveals that a majority of mutations affect the hydrophobic core of the structure, indicating that most MPS IV A cases result from misfolding of GALNS. Comparison of the structure of GALNS to paralogous sulfatases shows a wide variety of active site geometries in the family, but strict conservation of the catalytic machinery. Overall, the structure and the known mutations establish the molecular basis for MPS IV A and for the larger MPS family of diseases.

Introduction

Lysosomal enzymes are responsible for the catabolism of biomolecules. Deficiencies in lysosomal enzymes result in accumulation of undegraded substrates in tissues, leading eventually to lysosomal storage diseases. The human lysosomal galactosamine-6-sulfatase (GALNS, also known as N-acetylgalactosamine-6-sulfatase and GalN6S; E.C. 3.1.6.4) removes sulfate groups from a terminal N-acetylgalactosamine-6-sulfate (or galactose-6-sulfate) in mucopolysaccharides such as keratan sulfate and chondroitin-6-sulfate (Fig. 1a and b). Defects in GALNS lead to accumulation of substrates, resulting in the development of the lysosomal storage disease mucopolysaccharidosis IV A (also known as MPS IV A and Morquio A disease).¹

Fig. 1.

Human GALNS reaction and overall structure

(a): GALNS cleaves 6-sulfate attached to Gal and GalNAc saccharides. (b): GALNS substrates include chondroitin-6-sulfate and keratan sulfate. Red arrows show the bond cleaved by GALNS. (c): The GALNS monomer structure colored from blue to red from N to C terminus and Ca²⁺ in magenta. (This coloring scheme is matched in subsequent panels.) The molecular surface in green (calculated in Povscript⁴³) is cut away to indicate the trench defining the active site. (d): An overview of the GALNS dimer structure viewed down the molecular dyad. (e): A topology diagram of the monomer shows domains boxed in grey. (f): An orthogonal view of the GALNS monomer viewed down the large β sheet in domain 1.

The autosomal recessive MPS IV A affects approximately 1 in 200,000 live births and is characterized by a wide variety of symptoms, including severe skeletal abnormalities, hearing loss, corneal clouding, heart valve disease, and other impairments.² The disease presents two phenotypes depending on its severity: a mild form, which generally allows the patient a full lifespan, and a severe form, which often results in death before the second decade of life.² However, the correlation between genotype and phenotype is not fully understood. At present, 157 different mutations have been identified in the GALNS gene in patients with MPS IV A, 120 of which are missense mutations leading to a change of a single side chain residue in the protein.³ While there is no currently approved treatment for MPS IV A, enzyme replacement therapy, where patients are injected weekly with recombinant enzyme, is in Phase III clinical trials.⁴

Despite the intense clinical interest in GALNS and the large amounts of purified protein available for decades, to date there has been no structure of the protein reported. GALNS was first purified from human placenta in 1979,⁵ and later was isolated from liver cells⁶ and fibroblasts. The recombinant enzyme used in clinical trials is purified from Chinese hamster ovary cells.⁹ Because of the importance of GALNS and its relationship to disease, those interested in the structure have resorted to homology modeling¹⁰ using the best available paralogs, the 36% and 28% identical human arylsulfatase A¹¹ (ASA) and arylsulfatase B¹² (ASB) structures.

In general, sulfatases catalyze the hydrolysis of sulfate ester bonds from a diverse range of substrates. In vitro, most sulfatases can hydrolyze synthetic substrates such as 4-methylumbelliferyl sulfate (4-MU-S), which shares only a sulfate group with native substrates. Therefore, determining native substrate specificity for sulfatases is non-trivial.

Sulfatases require a modified amino acid nucleophile for catalytic function. The polypeptide encodes a cysteine (or occasionally serine), which is then enzymatically converted to a formylglycine aldehyde by formylglycine-generating enzyme (FGE).^13–16 The ubiquitous FGE recognizes a specific sequence motif (typically CXPXR), and oxidative desulfurization of the side chain leads to an aldehyde. Hydration of the aldehyde leads to a dihydroxyalanine (DHA) nucleophile with geminal hydroxy groups.^21–23

To elucidate the molecular basis for MPS IV A, we determined the structure of human GALNS to 2.2 Å resolution. Using stably transfected insect cells, we expressed human GALNS glycoprotein and demonstrated enzyme activity comparable to endogenous enzyme. Using x-ray crystallography and mass spectrometry, we identified FGE modification of the GALNS nucleophile in our insect cell expression system. To extend our structural knowledge about GALNS to the effects seen in patients with MPS IV A, we mapped the disease-causing mutations onto the structure. The mutations fall into three categories: disruption of the active site, perturbation of overall fold, and surface exposure, suggesting potential treatments depending on a patient’s genotype. Overall, these results will improve the understanding of the molecular defects in MPS IV A and will provide insight into lysosomal storage diseases and other protein folding diseases.

Results

Protein expression

We purified GALNS from Trichoplusia ni (Tn5) insect cells, using both baculovirus-infected and stable-cell line approaches. The protein sequence contained the native signal sequence for secretion and a C-terminal hexahistidine tag for purification. Baculovirus-infected cells yielded about 0.2 mg purified GALNS per liter of culture while stably transfected cell lines yielded approximately 0.5 mg per liter. Monoclonal selection of the stable cell lines increased expression to 0.7 mg per liter of culture.

Overall description of the structure

The structure of the recombinant human GALNS was determined by X-ray crystallography to a resolution of 2.2 Å (Table 1). The X-ray structure reveals GALNS as a homodimeric glycoprotein with each monomer having an N-terminal domain containing the active site (domain 1), a second domain with antiparallel β-strands (domain 2), and a C-terminal meander (Fig. 1c–f). Domain 1 (residues 28-379) has an α/β topology formed from a core ten-stranded β-sheet at the center of six α-helices. Domain 2 (residues 380-481) comprises a four anti-parallel stranded β-sheet perpendicular to a long α-helix (Fig. 1e). Domain 2 is followed by a C-terminal meander (residues 482-510), which threads back into domain 1 and defines a portion of the active site. The GALNS active site contains a calcium liganded to the catalytic nucleophile and the side chains of four other residues. Each monomer contains two N-linked glycosylation sites at Asn 204 and Asn 423, the latter of which falls close to the molecular dyad, and the attached carbohydrates pack against one another. Each monomer contains three disulfide bonds (308-419, 489-518, and 501-507), and an unpaired cysteine (164). The interface between the two monomers in the dimer buries 3094 Ų of surface area, and the large interface is consistent with the dimeric GALNS used in clinical trials as enzyme replacement therapy for MPS IV A.²⁴

Table 1.

X-ray crystallographic data

PDB:	4FDI	4FDJ
Ligand Soak	-	GalNAc
Beamline	APS 24-ID-C	BNL X6A
Wavelength, Å	0.9795	1.100
Space Group	P21	P21
Cell Lengths, Å	61.4, 155.5, 62.6	60.7, 155.8, 61.8
Cell β angle, °	113.8	113.4
Resolution, Å (last shell)	50-2.2 (2.26-2.2)	50-2.8 (2.85-2.8)
Observations overall	202,615 (9,634)	71,414 (2,536)
Unique observations (last shell)	54,154 (2,676)	24,690 (1,055)
Completeness, % (last shell)	99.7 (99.3)	96.4 (83.7)
Multiplicity (last shell)	3.7 (3.6)	2.9 (2.4)
Rsym, % (last shell)	11.0 (27.7)	14.3 (27.8)
<I/σI> (last shell)	14.3 (4.2)	8.6 (2.5)
Refinement
Rwork/Rfree (%)	16.4/21.1	20.7/25.1
No. of atoms: total	8,682	8,080
Protein	7,811	7,787
Carbohydrate	56	86
Water	775	205
Other	40	2
Average B-factors, Å2	22.8	40.0
Protein	21.8	39.9
Carbohydrate	56.0	64.5
Water	29.9	33.2
Ramachandran Plot, %
Favored	96.7	95.9
Allowed	3.3	4.1
Outlier	0.0	0.0
RMS deviations
Bonds, Å	0.007	0.008
Angles, °	1.10	1.16

The formylglycine modification and enzyme mechanism

Sulfatases require maturation of a side chain residue into a catalytic nucleophile. In GALNS, the polypeptide chain encodes a cysteine at residue 79 at the start of the motif CXPXR (Fig. 2a). FGE recognizes this motif and converts Cys 79 into a formylglycine aldehyde (Fig. 2b).²⁵ Addition of Ca²⁺ in the active site (Fig. 2c) and hydration of the aldehyde by a water molecule generates the gem diol nucleophile DHA (Fig. 2d). The GALNS crystal structure contains branched electron density at residue 79, corresponding to the hydrated form of the aldehyde (Fig. 2d). In the denatured and metal-chelated GALNS protein examined by mass spectrometry, residue 79 contains an overwhelming abundance of formylglycine (MW 85) with little presence of cysteine (MW 103) and undetectable amounts of DHA (MW 102) (Figs. 2 and SF1). The equilibrium between the aldehyde and the hydrated gem diol is shifted toward the aldehyde in the denatured GALNS and toward the gem diol in the crystal structure. The insect-cell expressed GALNS demonstrates nearly 100% post-translational modification of Cys 79, as seen by both mass spectrometry and crystallography.

Fig. 2.

GALNS maturation and mechanism

The GALNS gene encodes a cysteine at position 79 (a), which formylglycine generating enzyme (FGE) enzymatically converts into a formylglycine aldehyde (b). In the crystal, the aldehyde binds the calcium cofactor (c) and hydrates into the gem diol DHA (d). Substrate binding leads to nucleophilic attack on the sulfate (e), resulting of transfer of sulfate to the enzyme (f). Finally, intramolecular hydrolysis releases the sulfate (g) and regenerates the aldehyde. The protein in the crystal appears in state (d), while denaturing the enzyme for mass spectrometry disrupts the metal binding site and shifts the equilibrium to the aldehyde seen in state (b). The electron density in (d) is a σ_A-weighted 2F_o-F_c map contoured at 1.8s of the tripeptide centered at residue 79.

Both Ca²⁺ and Mg²⁺ have been found in the active sites of sulfatases. Mg²⁺ can appear in the active site when present at high concentrations during the purification and/or crystallization process, but Ca²⁺ is the preferred ion for biochemical activity. With GALNS, crystallographic refinement of Mg²⁺ in the active site resulted in positive peaks of 4–5 σ in difference electron density maps while Ca²⁺ led to a more consistent fit to the electron density. Additionally, we checked the bond valence of the metal using the software Valence,²⁷ which showed a value of 1.92 for Ca²⁺ and a value of 0.918 for Mg²⁺, strongly suggesting that the ion in GALNS is Ca²⁺.

After the discovery of FGE and its creation of an active site aldehyde in sulfatases, it was unclear whether the sulfatase mechanism involved the sulfated substrate attacking the aldehyde on the enzyme or the DHA nucleophile on the enzyme attacking the sulfated substrate.²² Our structure lends support to the latter mechanism, that upon substrate binding, nucleophilic attack of the sulfate by DHA cleaves the substrate (Fig. 2e and f). Subsequent intramolecular hydrolysis of the sulfated enzyme intermediate uses the unique gem diol to regenerate the apoenzyme (Fig. 2g).

The active site and ligand binding

The active site of GALNS is found in a pocket at the center of the N-terminal α/β domain. The primary active site residues include Asp 39, Asp 40, Arg 83, Tyr 108, Lys 140, His 142, His 236, Asp 288, Asn 289, Lys 310, and DHA 79 (Fig. 3a). The active site contains a Ca²⁺ hexavalently coordinated by oxygens from the nucleophile at 79, Asp 39, Asp 40, Asp 288 (2 atoms), and Asn 289 (Fig. 3b). A citrate molecule from the buffer binds in the active site. The active site contains a high concentration of basic residues, leading to a considerable positive electrostatic surface potential, as expected for an enzyme that binds highly sulfated substrates (Fig. 3d). GALNS has a predicted isoelectric point of 6.3, and the electrostatic surface shows a molecule of pronounced negative potential with patches of positive potential near the active site.

Fig. 3.

GALNS active site and ligand binding

(a): Schematic of interactions in the GALNS active site. NCS-averaged interatomic distances in Å are shown in blue. (b) σ_A-weighted 2F_o-F_c map of the active site residues contoured at 1.8σ. (c): Ligand-bound structure. The electron density shows a σ_A-weighted 2F_o-F_c map contoured at 1.3σ around the GalNAc ligand. The molecular surface shows the large substrate-binding cavity with key residues indicated. (d): The GALNS electrostatic surface potential (calculated in Pymol⁴²) plotted from −57kT (red) to +57kT (blue) shows the large areas of positive charge suitable for interacting with polyanionic substrates. The active sites are indicated with arrows, and the inset shows the active site with GalNAc bound. A tetrasaccharide fragment of the keratan sulfate substrate (labeled “KS fragment”) is shown for scale.

To better understand the molecular basis for ligand binding in GALNS, we determined the crystal structure of the protein soaked with a catalytic product, the N-acetylgalactosamine (GalNAc) monosaccharide, at 2.8 Å resolution (Fig. 3c and Table 1). The sugar binds in a non-productive orientation, with the 1-OH 2.6 Å from the catalytic nucleophile and the 6-OH (which is sulfated in the substrate) pointing away from the nucleophile. GalNAc binds GALNS using a common carbohydrate-recognition motif, where the plane of the sugar makes van der Waals contacts with the aromatic side chain Tyr 108.

The GALNS active site lies at the base of a deep conical crevice surrounded by positively charged groups (Fig. 3d). Unlike lysosomal glycosidases, which have exquisite specificity for the saccharide they recognize, GALNS and other lysosomal sulfatases have substantial promiscuity in substrate recognition. For example, GALNS will efficiently hydrolyze the synthetic substrate 4-MU-S, which shares only a sulfate group with the natural substrate. This promiscuity in substrate recognition is reflected in the substrate-binding pocket, a large open depression on the surface of the enzyme, allowing a wide variety of molecules access to the catalytic site. With the natural substrate keratan sulfate, the multiple negative charges on the polysulfated substrate presumably help guide the substrate into the active site (Fig. 3d).

Enzyme kinetics and inhibition

In the lysosome, GALNS removes sulfate groups from 6-sulfated galactosides and 6-sulfated N-acetylgalactosaminides in keratan sulfate and chondroitin-6-sulfate (Fig. 1a and b). To examine the enzymatic activity of purified GALNS, we measured the kinetic parameters K_M and k_cat of the enzyme in vitro using two synthetic substrates: 4-methylumbelliferyl-β-D-galactose-6-sulfate (4-MU-Gal-6-S)³⁰ and 4-MU-S (Fig. 4). The 4-MU-Gal-6-S substrate mimics the terminal sulfated sugar found in the natural substrates, and GALNS has reported K_M values of 0.05 mM for keratan sulfate⁶ and 0.01 mM for chondroitin-6-sulfate.³¹ With the 4-MU-Gal-6-S substrate, recombinant GALNS has a k_cat of 228 ± 13 min⁻¹ and a K_M of 13 ± 2 mM (Fig. 4a), comparable to the values measured for ASA towards a synthetic substrate containing a 4-MU fluorophore.²² With the 4-MU-S substrate, GALNS has a k_cat of 4.2 ± 0.7 min⁻¹ and a K_M of 2.1 ± 0.6 mM and showed substrate inhibition (with a substrate K_i of 23 mM), suggesting a non-productive binding mode (Fig. 4c). Because the natural substrates show considerably lower K_M values than their smaller synthetic homologs,^6,31 sulfatases such as GALNS interact with more of the natural substrates beyond the terminal sulfated monosaccharide. The kinetic parameters of the insect cell-expressed human GALNS are comparable to those from GALNS purified from endogenous sources.⁶

Fig. 4.

GALNS reaction kinetics

(a) and (b): Michaelis-Menten velocity vs. substrate concentration plots of GALNS with the synthetic substrates 4-MU-Gal-6-S (a) and 4-MU-S (b). 4-MU-S showed evidence of substrate inhibition (see text). (c): GALNS cleavage of 4-MU-S is inhibited by the presence of sulfate, which shows an IC₅₀ of 0.89 mM in the assay.

To enable inhibitor testing of human GALNS, we established an assay for measuring inhibition of cleavage of the 4-MU-S substrate by GALNS. We hypothesized that sulfate (the catalytic product of the GALNS reaction) would act as a competitive inhibitor, and the inhibition assay showed an IC₅₀ of 0.89 ± 0.23 mM and a K_i of 0.57 ± 0.15 mM for sulfate inhibition of GALNS (Fig. 4c). This assay will allow us to test small molecules for inhibition and potential pharmacological chaperone activity.

Comparison to other sulfatases

We compared the structure of GALNS to the most closely related paralogs with known structures, human ASA (PDB ID: 1AUK), ASB (PDB ID: 1FSU), and arylsulfatase C (ASC; PDB ID: 1P49; Figs. 5 and SF2). Overall GALNS has with a root-mean-square (RMS) deviation of 1.31 Å with ASA (for 408 Cαs, or 83% of the total), of 1.37 Å with ASB (339 Cαs, 72%), and of 1.24 Å with ASC (381 Cαs, 70%). Domains 1 and 2 superpose better than the divergent C-terminal meander (Fig. 5c). GALNS domain 1 superposes with an RMS deviation of 1.17 Å with ASA (318 Cαs, 90%), of 1.24 Å with ASB (291 Cαs, 84%), and of 1.06 Å with ASC (312 Cαs, 75%), and domain 2 superposes with an RMS deviation of 1.17 Å with ASA (73 Cαs, 75%), of 1.53 Å with ASB (60 Cαs, 58%), and of 1.15 Å with ASC (59 Cαs, 60%). Less than 40% of the residues in the small C-terminal meander superpose across the structures. The pairwise sequence identity³² across the structures is highest in domain 1, where it approaches 40%, decreases in domain 2 to 13–25%, and drops to undetectable in the C-terminal meander (Fig. SF2).

Fig. 5.

Comparison of GALNS to other human sulfatases

(a) Surface representations of GALNS (PDB ID 4FDI), ASA (PDB ID 1AUK), ASB (PDB ID 1FSU), and ASC (PDB ID 1P49), colored as in Fig. 1 and viewed into the active sites (indicated by arrows). (b) Surfaces of GALNS (green), ASA (orange), ASB (purple), and ASC (cyan) are shown around ribbon representation of the monomers (colored as in Fig. 1). The front surfaces are cropped to show the active site in the vicinity of the metal-binding site (magenta), and insets show the surface cavities (calculated in Pymol⁴²) colored by electrostatic potentials as in Fig. 3. (c) Superposition of GALNS (green), ASA (orange), ASB (purple), and ASC (cyan) are separated into panels corresponding to domain 1, domain 2, and the C-terminal meander, with the termini of each chain indicated.

One remarkable feature of the GALNS structure is the participation of the extreme C-terminal segment of the C-terminal meander in defining the active-site pocket (Figs. 1 and 5). GALNS contains a novel conformation of its C-terminal meander, where the polypeptide reverses direction and forms a wall of the active site contributing to substrate selectivity. We compared the substrate selectivity in the sulfatase family by examining the active-site pockets of GALNS and the paralog structures. The active-site machinery in domain 1, including the requirement for a metal, the metal-binding residues, and the post-translationally modified nucleophile, is strictly conserved across the family.³³ All of the family members have positively charged active-site pockets suitable for attracting sulfated substrates, but the overall sizes, shapes, and electrostatics of the pockets vary extensively (Fig. 5a and b). For example, GALNS and ASB (which cleave sulfates from polyanionic substrates like chondroitin sulfate) contain open, flat, positively charged active sites.¹² In contrast, ASA (which cleaves sulfates off of sphingolipid sulfate esters solublized by saposin presentation molecules) contains a small active site complementary to the sulfate head group of the substrate.¹¹ ASC (a membrane bound enzyme which cleaves sulfates from 3-β-hydroxysteroid sulfates) contains an active site with both charged and hydrophobic regions.²⁶ In sulfatases, highly conserved catalytic machinery couples with diverse active site geometry, leading to cleavage of a wide variety of substrates.

MPS IV A disease mutations

To gain further insight into the defects in GALNS that lead to disease, we mapped the locations of 120 disease-causing missense mutations found in MPS IV A patients (Fig. 6 and Table 2). Surprisingly, only 6 mutations (5%) affect active site residues, while 78 (65%) map to buried residues in the core of the protein and 32 mutations (27%) that map to surface residues. Domain 1 contains a larger number of mutations that map to buried residues (68) vs. surface residues (19). In contrast, mutations that map to domain 2 are found nearly equally in buried residues (9) and surface residues (10).

Fig. 6.

GALNS mutations in MPS IV A

(a): Disease-causing mutations leading to substitutions in the protein were mapped onto the GALNS structure. In the top panel, reported mutations are colored by their disease phenotype, as severe (less than 125cm in stature, dark green), mild (over 125cm, light green), or not reported (yellow). In the bottom panel, the mutations are mapped to their location on the protein, as near the active site (black), buried in the hydrophobic core (red), or on the surface (blue). (b): GALNS residues were ranked by solvent accessibility and vertical bars indicate a substitution leading to MPS IV A disease. Most of the mutations lead to changes in buried residues, meaning MPS IV A is most often a protein-folding disease.

Table 2.

Effect of Morquio A Disease-Causing Mutations

Position	Mutation	Atomic acc. surf.	Effect on GALNS	Disease Phenotype	Mutation Category	Ref.
M1V	1A>G		In signal sequence	Severe	Processing	48
W10X	29G>A		In signal sequence	Severe	Processing	49
L15M	43C>A		In signal sequence	Mild	Processing	50
G23R	67G>C		In signal sequence	Mild	Processing	51
L36P	107 T>C	0.0	Buried middle of in β strand; Pro would kink strand	Mild	Buried	48
M41L	121 A>T	0.2	Little room for branched side chain	Mild	Buried	3
G42E	125 G>A	0.0	No room for side chain	Severe	Buried	50
G47R	139 G>A	0.0	No room for large side chain	Severe	Buried	51
S53F	158 C>T	3.8	Polar region near dimer interface; little room for large side chain	NR	Surface	50
D60N	178G>A	7.5	Unknown	Mild	Surface	52
R61W	181C>T	17.0	Solvent exposed hydrophobic residue	Mild	Surface	53
G66R	196G>C	0.0	No room for large side chain	Severe	Buried	50
F69V	205T>G	0.0	Loss of hydrophobic packing	Severe	Buried	48
P77R	230C>G	0.0	No room for large side chain	Severe	Buried	54
C79Y	236G>A	1.3	Loss of nucleophile	Severe	Active Site	55
S80L	239C>T	0.0	Buried in polar environment, possible disruption of formylglycine modification	Severe	Active Site	49
T88I	263C>T	0.0	Loss of buried hydrogen bonding	NR	Buried	56
R90W	268C>T	3.4	Partially buried; no room for Trp	Severe	Buried	57
R94G	280C>G	3.2	Loss of ion pair to Glu121	Mild	Buried	50
R94C	280C>T	3.2	Loss of ion pair to Glu121	Mild	Buried	3
R94L	281G>T	3.2	Loss of ion pair to Glu121	Severe	Buried	10
G96C	286G>T	0.1	Gly-specific phi/psi angles	Severe	Buried	50
G96V	287G>T	0.1	Gly-specific phi/psi angles	Severe	Buried	3
F97V	289T>G	0.0	Loss of hydrophobic packing	Mild	Buried	52
A107T	319G>A	0.0	No room for larger side chain	NR	Buried	53
Q111X	331C>T	0.7	Stop	Severe	-	49
Q111R	332A>G	0.7	Loss of buried hydrogen bonding	Mild	Buried	58
I113F	337A>T	0.0	Buried; no room for larger side chain	Severe	Buried	3
G116S	346G>A	0.0	Buried; no room for larger side chain	Severe	Buried	57
P125L	374C>T	0.0	Buried; no room for larger side chain	Severe	Buried	3
K129X	385A>T	17.6	Stop	Severe	-	51
S135R	405C>G	0.2	Buried; no room for larger side chain	Severe	Buried	59
V138A	413T>C	0.0	Loss of hydrophobic packing	Mild	Buried	50
G139S	415G>A	0.1	Gly-specific phi/psi angles	Severe	Buried	48
W141R	421T>A	0.0	Buried in hydrophobic core; near active site	Severe	Buried	60
W141C	423G>T	0.0	Buried in hydrophobic core; near active site	Severe	Buried	61
H142R	425A>G	0.7	Active site residue	Severe	Active Site	3
Q148X	442C>T	18.5	Stop	Severe	-	3
H150Y	448C>T	1.9	Buried; little room for larger side chain	Mild	Buried	52
P151S	451C>T	0.2	Buried in hydrophobic core	Severe	Buried	50
P151L	452C>T	0.2	Buried in hydrophobic core	Severe	Buried	49
G155R	463G>A	2.5	Gly-specific phi/psi angles	Severe	Surface	3
G155E	464G>A	2.5	Gly-specific phi/psi angles	Severe	Surface	50
F156S	467T>C	0.0	Buried in hydrophobic core	Mild	Buried	53
F156C	467T>G	0.0	Buried in hydrophobic core	Severe	Buried	10
W159C	477G>T	0.2	Buried in hydrophobic core	Severe	Buried	50
S162F	485C>T	0.1	Buried; no room for larger side chain	Severe	Buried	3
P163H	488C>A	0.6	Buried; little room for larger side chain	Mild	Buried	3
N164T	491A>C	0.8	Disruption of buried hydrogen bonds	NR	Buried	62
H166Q	498C>G	0.0	Disruption of buried ion pair with Asp233	Severe	Buried	53
F167V	499T>G	15.2	Near active site; mutation may affect substrate recognition or protein binding	Mild	Surface	55
G168R	502G>A	1.6	Gly-specific phi/psi angles	Severe	Buried	50
D171A	512A>C	16.7	Unknown	Mild	Surface	49
P179S	535C>T	0.0	Completely buried	Severe	Buried	56
P179H	536C>A	0.0	Buried; little room for larger side chain	Severe	Buried	60
P179L	536C>T	0.0	Buried; little room for larger side chain	Severe	Buried	57
E185G	554A>G	17.9	Unknown	Severe	Surface	50
A203V	608C>T	0.0	Buried; little room for branched side chain	NR	Buried	57
N204K	612C>G	6.0	Glycosylation site	Mild	Surface	49
Q211X	631C>T	17.8	Stop	Severe	-	49
W230G	688T>G	0.0	Loss of hydrophobic packing	Severe	Buried	60
W230X	689G>A	0.0	Stop	Severe	-	3
D233N	697G>A	0.3	Disruption of buried ion pair with His166	Severe	Buried	50
H236D	706C>G	4.7	Active site	Mild	Active Site	50
V239F	715G>T	0.0	Buried; no room for larger side chain	Severe	Buried	60
G247D	740G>A	18.0	Unknown	Severe	Surface	56
R253W	757C>T	6.6	Partially buried; little room for Trp	Mild	Surface	57
A257T	769G>A	0.0	Buried; little room for branched side chain	Severe	Buried	3
R259Q	776G>A	9.8	Unknown	Mild	Surface	49
E260D	780G>C	0.0	Loss of buried hydrogen bonding	NR	Buried	51
F284V	850T>G	0.2	Loss of hydrophobic packing	Mild	Buried	51
S287L	860C>T	0.0	Loss of buried hydrogen bonding	Severe	Buried	53
N289S	866A>G	0.0	Active site, metal binding residue	Mild	Active Site	3
G290S	868G>A	0.0	Little room for larger side chain; near active site	Severe	Buried	49
A291T	871G>A	0.4	Buried; little room for larger side chain	Severe	Buried	49
A291D	872C>A	0.4	Buried; little room for larger side chain	Mild	Buried	51
S295F	884C>T	3.5	Partly buried; little room for larger side chain	Mild	Surface	3
G301C	901G>T	3.7	Buried; little room for larger side chain	Severe	Buried	49
L307P	920T>C	0.4	Buried in hydrophobic pocket	Severe	Buried	50
G309R	925G>A	0.2	Buried; no room for larger side chain	Severe	Buried	52
K310N	930G>C	4.5	Active site	Mild	Active Site	56
T312A	934A>G	0.0	Loss of hydrophobic packing	NR	Buried	51
T312S	935C>G	0.0	Loss of hydrophobic packing	Mild	Buried	50
G316V	947G>T	0.1	Buried; no room for larger side chain	NR	Buried	60
M318R	953T>G	0.0	Buried; little room for larger side chain	Severe	Buried	51
A324E	971C>A	0.0	Buried in hydrophobic core	Mild	Buried	50
W325X	974G>A	0.0	Stop	Severe	-	63
W325C	975G>T	0.0	Buried in hydrophobic pocket	NR	Buried	52
Q338X	1012 C>T	0.2	Stop	Severe	-	3
G340D	1019 G>A	0.4	Buried in hydrophobic pocket	Severe	Buried	51
S341R	1023 C>A	0.8	Little room for larger side chain	Severe	Buried	57
M343L	1027 A>T	1.6	Buried; little room for branched side chain	Severe	Buried	64
M343R	1028 T>G	1.6	Buried; little room for larger side chain	Severe	Buried	51
D344N	1030 G>A	0.1	Loss of buried ion pair with Arg380	Severe	Buried	52
D344E	1032 C>G	0.1	Disruption of buried ion pair with Arg380	Severe	Buried	48
L345P	1034 T>C	0.7	Little room for Pro side chain within helix	Severe	Buried	48
F346L	1038 C>A	0.0	Loss of hydrophobic packing	Severe	Buried	50
A351V	1052 C>T	5.6	Little room for branched side chain	Severe	Surface	50
L352P	1055 T>C	2.0	Partly buried in hydrophobic pocket	Mild	Buried	49
P357L	1070 C>T	8.1	Unknown	Severe	Surface	3
R361G	1081 A>G	6.6	Ion pair with Asp360	Severe	Surface	48
L366F	1096 C>T	7.1	Unknown	NR	Surface	53
L369P	1106 T>C	3.6	Little room for Pro side chain within helix	Severe	Surface	3
Q374X	1120 C>T	17.0	Stop	Severe	-	63
R376Q	1127 G>A	16.6	Unknown	Severe	Surface	3
R380T	1139 G>C	1.2	Loss of buried ion pair with Asp344	Severe	Buried	56
R380S	1140 G>T	1.2	Loss of buried ion pair with Asp344	Mild	Buried	51
R386C	1156 C>T	1.1	Disruption of hydrophobic core	Severe	Buried	48
R386H	1157 G>A	1.1	Disruption of hydrophobic core	Severe	Buried	57
D388N	1162 G>A	5.1	Loss of ion pair with His103	Mild	Surface	57
M391V	1171 A>G	0.0	Disruption of hydrophobic core	Mild	Buried	65
A392V	1175 C>T	0.0	Little room for larger side chain	NR	Buried	3
L395V	1183 C>G	4.1	Unknown	Severe	Surface	57
L395P	1184 T>C	4.1	Phi-psi angles not favored for Pro	NR	Surface	48
H398D	1192 C>G	2.5	Loss of ion pair with Glu460	Mild	Surface	57
H401Y	1201 C>T	0.1	Loss of buried ion pair with Glu450	NR	Buried	53
N407H	1219 A>C	0.4	Disruption of buried hydrogen bonds	Severe	Buried	52
W409S	1226 G>C	7.5	Unknown	Mild	Surface	3
G421E	1262 G>A	2.3	Little room for larger side chain	Severe	Surface	50
Q422K	1264 C>A	4.6	Disruption of hydrogen bonds	Severe	Surface	50
Q422X	1264 C>T	4.6	Stop	Severe	-	51
E450V	1349 A>T	0.0	Loss of buried ion pair with His401	Severe	Buried	62
F452I	1354 T>A	7.7	Loss of hydrophobic packing	Severe	Surface	66
F452L	1354 T>C	7.7	Loss of hydrophobic packing	Severe	Surface	53
S470P	1408 T>C	24.3	Little room for Pro side chain in helix	NR	Surface	51
Q473X	1417 C>T	20.2	Stop	Severe	-	50
P484S	1450 C>T	4.1	Unknown	Mild	Surface	57
N487S	1460 A>G	6.9	Disruption of hydrogen bonds	Severe	Surface	50
V488M	1462 G>A	17.5	Unknown	NR	Surface	56
M494V	1480 A>G	1.8	Little room for branched side chain	NR	Buried	49

Discussion

Many nearly identical carbohydrate-containing substrates appear in the lysosome. For example, glucosides and galactosides differ by a single chiral center (at the 4 position of the sugar ring), α- and β- sugars differ by a single chiral configuration at the anomeric center, etc. In response, lysosomal glycosidases have evolved exquisite specificity for their substrates, as they are able to distinguish among the many similar substrates differing by a single chiral center. In contrast, lysosomal sulfatases have greater promiscuity in their substrate recognition. The differences between the lysosomal glycosidases and sulfatases are manifested in different active site geometries: exoglycosidases have small active sites and make direct interactions with every functional group on the saccharide, while sulfatases have larger and deeper active sites that are capable of binding a wide variety of substrates. The GALNS structure suggests that the specificity of lysosomal sulfatases derives from a large number of interactions with the substrate, beyond the sulfated monosaccharide proximal to the catalytic nucleophile.

The active sites of sulfatases including GALNS are often broad trenches that accommodate a variety of potential substrates (including synthetic substrates such as 4-MU-S, which share only a sulfate group with endogenous substrates). However, the promiscuity of the sulfatases is limited in vivo, since the multiple sulfatases in the lysosome are non-redundant and loss of one sulfatase cannot be compensated by the activity of a related enzyme. For example, MPS IV A symptoms develop due to the loss of one sulfatase (GALNS), but the missing activity in patients is not compensated by the presence of other fully functional sulfatases like ASA and ASB, although all three sulfatases have overlapping specificity in vitro.

MPS IV A is primarily a disease of protein folding, because the majority of disease-causing mutations result in disruption of the hydrophobic core of the protein. Domain 1 appears to be a critical scaffold for the overall structure of GALNS, as 78% of the mutations in domain 1 map to the hydrophobic core of the domain. In contrast, 53% of the mutations in domain 2 map to surface residues, which are less likely to perturb the folding of GALNS. The prevalence of disease-causing mutations mapping to surface residues in domain 2 suggests that this domain may participate in required intermolecular contacts for correct protein function. Along this line, GALNS has been reported to participate in a lysosomal complex with the β-galactosidase, sialidase-1, and protective protein/cathepsin A (PPCA) proteins.

Human GALNS tolerates a variety of active site ligands. In addition to the cellular substrates keratan sulfate and chondroitin-6-sulfate, the active site can cleave synthetic substrates like 4-MU-Gal-6-S and 4-MU-S and can bind inhibitors including GalNAc and sulfate. Multiple sugar-binding possibilities occur within the GALNS active site, as revealed by catalytically productive binding of 4-MU-S, by non-productive binding of 4-MU-S (i.e. substrate inhibition), and by non-productive GalNAc binding (in the cocrystal structure). Other sulfatases including ASB show similar non-productive binding of monosaccharides.¹² In GALNS, the key active-site interaction appears to be sulfate recognition, because sulfate acts as an effective inhibitor and 4-MU-S (lacking a sugar) is a competent substrate. Critical ionic interactions between GALNS and a sulfated ligand explain previous observations of GALNS inhibition by sulfate, phosphate, and chloride,³¹ and citrate binding in the unliganded crystal structure.

Pharmacological chaperone therapy, or the use of a small molecule to stabilize a lysosomal enzyme as it transits from the ER to the lysosome, has been proposed and tested for many lysosomal storage diseases. Pharmacological chaperones have been developed for lysosomal glycosidases, but not yet for lysosomal sulfatases. The highly specific substrate recognition performed by lysosomal glycosidases can be exploited to make tight-binding pharmacological chaperones from substrate and product analogs. However, the larger and less specific substrate-binding pockets found in lysosomal sulfatases make pharmacological chaperone design more complex. Additionally, the preference of sulfatases for anionic substrates could present pharmacokinetic problems. The structure of GALNS reported here reveals a large and convoluted substrate-binding pocket well suited to the rational design of ligands that might serve as pharmacological chaperones or mechanism-based inhibitors.

In enzyme replacement therapy for MPS IV A, recombinant enzyme is injected into patients to replace the missing GALNS enzymatic activity. In an effort to improve delivery of recombinant GALNS to the bones of MPS IV A patients, one proposed variant of GALNS contains six additional glutamate residues at the N terminus.³⁶ The GALNS structure indicates that modification of the N terminus is likely to be well tolerated by the protein, but additions to the C terminus could perturb the substrate-binding properties of the variant.

Purification of GALNS from endogenous sources led to proteins that were processed into smaller fragments, thought to be a requirement for full enzymatic activity. The structure lends no support to the processing requirement for activity, because the insect-cell expressed GALNS is unprocessed and shows comparable enzymatic activity to GALNS purified from endogenous sources. It appears that endogenous GALNS may be occasionally nicked by a cellular protease, presumably at a site or sites between the 308-419 disulfide, because the fragments remain disulfide-linked.

In summary, the crystal structure reported here provides for the first time a molecular basis for accurately predicting the effect of a MPS IV A mutation on the GALNS protein. Additionally, the high-resolution structure reveals a large, convoluted active site highly suited to rational drug-design experiments. The human GALNS structure furthers our understanding of the molecular defects that underlie the family of inherited metabolic defects known as lysosomal storage disorders.

Materials and Methods

Molecular biology

Human GALNS cDNA (NCBI Sequence ID: NM_000512.4) was purchased from Open Biosystems. The open reading frame, including the 26-residue native signal sequence (for secretion into the media) and a C-terminal hexahistidine affinity tag, was amplified by PCR using Hot Start Phusion polymerase (NEB). For transfection into insect cells, the PCR product was gel purified, incubated with Taq polymerase to add 3′ A-overhangs, and cloned into the pIB/V5-His-TOPO TA vector per the manufacturer’s protocol (Invitrogen). Top 10 E. coli were transformed with the GALNS plasmid DNA; single clones were isolated, analyzed by restriction enzyme digestion, and sequenced.

Insect cell protein expression and purification

GALNS plasmid DNA was used to transfect Trichoplusia ni (High Five) insect cells. 72 h post transfection, 100 mg/mL blasticidin was added for 2 weeks to select for GALNS integration. Subsequently, cells at 50–60% confluence were diluted to 1 cell/well into 96-well plates containing 10 μg/mL blasticidin and then grown to confluence. Supernatants were tested by immunoblotting to assess GALNS expression, and the best-expressing monoclonal cell was amplified for large-scale suspension culture. Typically 5–10 L cultures were grown in multiple 3 L baffled Fernbach flasks (Corning). Cultures were inoculated at 5×10⁵ cells/mL in fresh serum-free media and grown for 3 days until viability dropped below 95%. Cells were removed by centrifugation, and the supernatant clarified by centrifuging for 45 min at 4700g prior to concentration and exchange into Buffer A (50 mM sodium phosphate, pH 7.0, and 250 mM sodium chloride) by tangential flow filtration.

GALNS was purified by nickel affinity chromatography and cation exchange chromatography. Clarified supernatant was loaded into a 5 mL Ni-NTA column and eluted with a gradient of Buffer B (50 mM sodium phosphate, pH 7.0, 250 mM sodium chloride, 400 mM imidazole) over 70 column volumes on an FPLC at 4°C. Fractions with GALNS were pooled and exchanged into sodium acetate, pH 5.1 prior to loading onto a 5 mL SOURCE-S cation-exchange column (GE Healthcare) in Buffer C (20 mM sodium acetate, pH 5.1). Proteins were eluted with a linear gradient of Buffer D (20 mM sodium acetate, pH 5.1, and 1 M sodium chloride) over 100 column volumes at 4°C. Fractions containing GALNS were pooled, exchanged into 20 mM Bis–Tris, pH 6.0, and concentrated to 10 mg/ml for crystallography.

Crystallography

We crystallized fully glycosylated GALNS by vapor diffusion against 10–15% PEG 6000, 0.1M citric acid, pH 5.0. Despite extensive optimization efforts, all crystals grew as stacked plates, which necessitated mechanical separation. Harvested crystals were transferred to cryoprotectant buffer (20% PEG 6000, 0.1M citric acid, pH 5.0, and 25% glycerol) and cooled in liquid nitrogen. GalNAc ligand data were collected from crystals soaked in cryoprotectant supplemented with 50 mM GalNAc. Diffraction data were collected at beamline 24-ID-C at the Advanced Photon Source and at beamline X6A at Brookhaven National Lab. X-ray data were integrated and scaled using the HKL2000 software,³⁸ and high mosaicity, poor spot shape, and anisotropy in the diffraction led to poor merging statistics. Several resolution limits were tested, and the reported limits were based upon the quality of the electron density maps. Molecular replacement in the CCP4 program AMoRe³⁹ using protein atoms from the 34% identical human arylsulfatase A (PDB ID: 1N2K) led to an initial R_factor of 51.0% with the next best solution at 53.3%. The structure was completed from repeated cycles of model building in Coot⁴¹ and refinement in REFMAC5.³⁹ Alternate interpretations of DHA (as multiple cysteine conformations) and Ca²⁺ (as other ions) led to poorer fits to the electron density. Molecular figures were made in Pymol⁴² and Povscript.⁴³ Sequence alignments were calculated using ClustalW⁴⁴ and Blastp,⁴⁵ and Ramachandran statistics using Rampage.⁴⁶ Structural superpositions were calculated in LSQMAN³² using a distance cutoff of 3.5 Å for Cα positions.

Mass spectrometry

Liquid chromatography/mass spectrometry (LC/MS) was used to confirm the post-translational modification of GALNS residue 79. GALNS was denatured by addition of guanidinium HCl (GuHCl) and ethylenediaminetetraacetic acid (EDTA) to a final concentration of 50 μM GALNS, 6 M GuHCl, and 1 mM EDTA. The protein was then reduced in 20 mM dithiothreitol for 30 minutes and cysteines carboxymethylated in 20 mM iodoacetic acid for 30 minutes in the dark. The protein was exchanged into 100 mM Tris HCl, pH 8.0 on a NAP-5 desalting column prior to trypsin incubation for five hours with immobilized trypsin (Promega). After trypsin removal, GALNS peptides were injected into a NanoLC system (Dionex) for separation on a C18 reverse phase column and analysis by a QStar XL hybrid quadrupole time-of-flight mass spectrometer (ABI/Sciex). In parallel with LC/MS, the more abundant peptide ions were isolated and subjected to tandem mass spectrometry (MS/MS) for fragmentation by collision-activated dissociation (CAD). This method allowed peptide identification and measurement of the mass of the amino acid at position 79.

Kinetic assays

Kinetic assays measured the hydrolysis of the fluorescent reporter 4-MU from two sulfated substrates: 4-MU-S (Sigma) and 4-MU-Gal-6-S (Glycosynth and Toronto Research Chemicals). For the 4-MU-S assays, 4-MU release was measured directly by fluorescence excitation at 362 nm and emission at 448 nm. The starting GALNS stock (calculated using an extinction coefficient of 1.6 mg/ml protein/OD₂₈₀) contained 2.4 μM GALNS in 10 mM Bis-Tris, 10 mM sodium chloride, pH 6.0. The enzyme stock was diluted 8-fold into 100 mM sodium acetate buffer pH 4.5 at eight 4-MU-S substrate concentrations from 0.25 to 15 mM. Aliquots were taken every 5 minutes for 30 minutes and diluted into 250 mM glycine, pH 10.2 prior to reading 4-MU fluorescence in a microplate reader.

For the 4-MU-Gal-6-S reactions, cleavage of substrate was monitored with a coupled reaction where GALNS first removed the sulfate and then β-galactosidase hydrolyzed the glycosidic linkage, releasing the fluorescent product 4-MU. GALNS was diluted to 0.3 μM and incubated in 100 mM sodium acetate, pH 4.5 at eight substrate concentrations from 0.50 to 50 mM. Aliquots were diluted 11-fold into 0.1 mg/ml E. coli β-galactosidase (Sigma) in 100 mM sodium potassium phosphate, pH 6.2 for 30 minutes. Aliquots were then diluted 30-fold into 250 mM glycine, pH 10.2, prior to reading in a microplate reader. Control experiments lacking either of the two enzymes showed no fluorescence above background. Fluorescence of free 4-MU at 448 nm was converted to 4-MU concentration using a standard curve. Error bars represent the standard deviations of triplicate measurements. Velocities were calculated from linear fits of [4-MU] vs. time plots, and Michaelis-Menten hyperbolae were fit in KaleidaGraph to obtain K_M and k_cat. The 4-MU-S data showed substrate inhibition, so K_M, K_i, and k_cat fit better to a modified Michaelis-Menten function⁴⁷:

$${\text{V}}_0=\frac{k_{\text{cat}}{[\text{E}]}_{\text{T}}[\text{S}]}{K_{\text{M}}+[\text{S}]\left(1+\frac{[\text{S}]}{K_{\text{i}}}\right)},$$

where V₀ is the initial velocity, [E]_T and [S] are concentrations of enzyme and substrate, respectively, and K_i is the equilibrium constant for non-productive binding of substrate.

The inhibition assay used sub-inhibitory concentrations of 4-MU-S substrate. 0.3 μM GALNS, 1.2 mM 4-MU-S, and 100 mM sodium acetate, pH 4.5 were combined with 0 to 10 mM ammonium sulfate. Samples were taken every 5 minutes for 25 minutes and diluted 30-fold into 250 mM glycine, pH 10.2 prior to fluorimetry. The IC₅₀ was fit in KaleidaGraph from the sigmoidal dose-response curve according to

$$\text{normalized}\text{activity}=\frac{\text{maximal}\text{activity}}{1+{\left(\frac{[\text{I}]}{{\text{IC}}_{50}}\right)}^{\text{b}}},$$

where b is the Hill slope. The IC₅₀ was converted into the K_i for a competitive inhibitor⁴⁷ according to

$${\text{IC}}_{50}=K_{\text{i}}\left(1+\frac{[\text{S}]}{K_{\text{M}}}\right).$$

Mutation analysis

Accessible surfaces areas were calculated in the CCP4 program AREAIMOL,³⁹ and the average accessible surface area per side chain atom was calculated in Excel (Microsoft). Non-glycine side chains were considered buried if the average accessible surface area per side chain atom was less than 2 Ų.

Highlights.

• We have determined the crystal structure of human GALNS

• The structure reveals the molecular basis of MPS IV A (Morquio A) disease

• The structure shows post-translational modification of an active site cysteine

• The structure reveals that MPS IV A (Morquio A) is a protein-folding disease

• Human GALNS is a promising target for rational drug design