Structural and Biochemical Characterization of the Folyl-poly-γ-L-glutamate Hydrolyzing Activity of Human Glutamate Carboxypeptidase II

Abstract

In addition to its well-characterized role in the central nervous system, human glutamate carboxypeptidase II (GCPII, Uniprot ID Q04609) acts as a folate hydrolase in the small intestine, participating in the absorption of dietary polyglutamylated folates (FolGlu_n), which are the provitamin form of folic acid (also known as vitamin B₉). Despite the role of GCPII as a folate hydrolase, nothing has been known about the processing of polyglutamylated folates by GCPII at the structural or enzymological level. Moreover, many epidemiologic studies on the relationship of the naturally occurring His475Tyr polymorphism to folic acid status suggest that this polymorphism may be associated with several pathologies linked to impaired folate metabolism. In this study we present 1) a series X-ray structures of complexes between a catalytically inactive GCPII mutant (Glu424Ala) and a panel of naturally occurring polyglutamylated folates, 2) X-ray structure of the His475Tyr variant at 1.83 Å resolution, 3) study of the recently identified arene-binding site of GCPII through mutagenesis (Arg463Leu, Arg511Leu, and Trp541Ala), inhibitor binding and enzyme kinetics with polyglutamylated folates as substrates and 4) comparison of the thermal stabilities and folate-hydrolyzing activities of GCPII wild type and His475Tyr variants. As a result, the crystallographic data reveal considerable details about the binding mode of polyglutamylated folates to GCPII, especially the engagement of the arene binding site in recognizing the folic acid moiety. Additionally, the combined structural and kinetic data suggest that GCPII wild type and His475Tyr variant are functionally identical.

Introduction

Glutamate carboxypeptidase II (GCPII; EC 3.4.17.21) is a 750-amino-acid type II transmembrane glycoprotein [1] and a Zn²⁺-dependent metalloprotease of the M28 peptidase family (Fig. 1). This enzyme is also known as prostate-specific membrane antigen (PSMA), folate hydrolase 1 (FOLH1), folyl-poly-γ-glutamate carboxypeptidase (FGCP), and N-acetylated-α-linked acidic dipeptidase (NAALADase). These different designations reflect the various functions and tissue distribution of this protein.

Fig. 1.

Overall view of GCPII topology with the active site, “arene-binding site”, His475Tyr mutation, and natural substrates N-acetyl-L-aspartyl-L glutmate (NAAG) and folyl-poly-γ-L-glutamates (FolGlu_n) highlighted. Note that folic acid (abbreviated as FolGlu₀) is pteroic acid linked to a single glutamate residue with a peptide bond. Mono-γ-L-glutamylated folic acid is abbreviated FolGlu₁. Pteroic acid is composed of a substituted pteridine moiety linked to p-aminobenzoic acid via a nitrogen atom. The left monomer of GCPII is shown in surface representation, with saccharide moieties depicted as gray spheres, the protease domain (amino acids 57 - 116 and 352 - 590) in red, the apical domain (amino acids 117 - 351) in green, and the C-terminal domain (amino acids 591 - 750) in blue. The right monomer is shown in cartoon representation, with the arene-binding site highlighted in the upper circle and the surface amino acid His475, subject of the disputed polymorphism, in the lower circle.

The well-known and thoroughly studied enzymatic activity of GCPII is the cleavage of peptide neurotransmitter N-acetyl-L-aspartyl-L-glutamate (NAAG) into N-acetyl-L-aspartate and L-glutamate [2, 3, 4, 5, 6]. In contrast to that, nothing has been known about the GCPII's folyl-poly-γ-L-glutamate hydrolyzing activity. At the lumenal surface of the human jejunum, GCPII cleaves γ-linked L-glutamates from folyl-poly-γ-L-glutamic acids (FolGlu_n) [7], the storage form of folic acid that cannot pass through the cell membrane [8]. (Structures of the folyl-poly-γ-L-glutamic acids and their building block pteroic acid, which is composed of a substituted pteridine double-ring and p-aminobenzoyl moiety, are shown in Fig. 1.) Atlhough the structural and enzymological aspects of FolGlu_n cleavage have not been characterized at all and folate is crucial for replication of rapidly dividing cells, the physiological relevance and structural and enzymological aspects of FolGlu_n cleavage in the small intestine have not been well-characterized. Numerous reports and epidemiologic studies on the naturally occurring His475Tyr polymorphism have analyzed the potential influence of this polymorphism on physiological levels of folates and folate-related metabolites (e.g. homocysteine). Some studies involving 600 – 1900 subjects concluded that this polymorphism influences neither folate nor homocysteine levels [9, 10, 11]. In contrast, a 2700-subject study concluded that the His475Tyr polymorphism confers higher folate and lower homocysteine levels [8]. Other, much smaller studies (30 - 44 participants) have correlated the polymorphism with altered folate levels and some pathologies, such as breast cancer [12]. Importantly, all of these studies investigated only genotype-phenotype association data, which makes their results rather inconclusive. For these reasons, the folyl poly-γ-L-glutamate hydrolyzing activity of GCPII and its natural His475Tyr variant, both at structural and enzymological level, is the focus of this study.

Structurally, the extracellular part of GCPII folds into three distinct domains: the protease domain (amino acids 57 - 116 and 352 - 590), the apical domain (amino acids 117 - 351), and the C-terminal domain (also called the dimerization domain; amino acids 591 - 750) (see Fig. 1). All three GCPII domains contribute to formation of the substrate binding cavity, which consists of the S1 site, the active site, and the S1′ site.

The active site features two Zn ions coordinated by the side chains of His377, Asp387, Asp453, Glu425, and His553. In the vicinity of the S1 site, there is a highly flexible segment called the “entrance lid” (amino acids Trp541 - Gly548). In its closed conformation, the entrance lid fully shields the substrate binding cavity from the extramolecular space, as observed when small substrates such as NAAG are bound [4] (PDB ID 3BXM). When the entrance lid is in open conformation, larger inhibitors can bind [13] (PDB ID 2XEF, 2XEG, 2XEI, 2XEJ). Similar to the S1 site, the S1′ site can be shielded from the extramolecular space by a flexible amino acid segment. The so-called “glutarate-sensor” (Tyr692 - Ser704) takes on a beta-stranded conformation upon binding of the glutamic acid moiety in the S1′ position, sealing off an otherwise present small funnel (reviewed elsewhere [6, 14]).

Recently, Zhang et al. [13] identified and structurally characterized an exosite of GCPII that binds aromatic moieties. This exosite, termed the “arene-binding site” (ABS), is formed by the first amino acid of the entrance lid, Trp541, together with Arg511 and Arg463. Adding a dinitrophenyl moiety with a length-optimized linker to a GCPII inhibitor significantly enhances the inhibitor's affinity toward GCPII via the avidity effect, i.e., by allowing it to bind to GCPII in a bidentate mode.

In this paper we set out to analyze i) how GCPII recognizes and cleaves polyglutamylated folates, ii) the role of the “arene-binding site” in the recognition and binding of folyl-poly-γ-L-glutamic acids and iii) if the His475Tyr polymorphism alters the folyl-poly-γ-L-glutamate hydrolyzing activity. To this end, we solved a series of X-ray structures of complexes between the inactive GCPII mutant, Glu424Ala, and several naturally occurring polyglutamylated folates (FolGlu_1-6). We also analyzed the 3D structure of the GCPII His475Tyr variant and compared its activity to the wild type using a novel assay, and performed structure-activity studies of the ABS.

Results

X-ray structures of recombinant human GCPII-Glu424Ala-FolGlu_1/2/3/4/5/6 complexes

To shed light on the binding mode of individual poly-glutamylated folates in the substrate-binding pocket of GCPII, we determined X-ray structures of seven complexes of the hydrolytically inactive recombinant human GCPII (further abbreviated as rhGCPII) Glu424Ala mutant [4] and FolGlu_1/2/3/4/5/6 substrates. The FolGlu_1-3 structures were refined at resolutions between 1.65 and 2.00 Å, with good crystallographic parameters (see Table 1). Their overall folds are virtually identical. The only major differences were observed in the conformation of residues Arg463, Arg511, and Trp541 (forming the arene-binding site) implicated in substrate binding. For FolGlu_1-2, the complete substrate molecules were defined in the electron density. For FolGlu₃, only the C-terminal di-γ-L-glutamyl-glutamate and pteridine moieties were defined in the electron density map (Fig. 2a). In the structures obtained uding FolGlu_0/4/5/6 crystallization conditions, only di-γ-L-glutamyl-glutamate moieties could be modeled, and the remaining parts of these substrates were not defined in the electron density (data not shown, see also Discussion).

Table 1. Data collection and refinement statistics for the x-ray models of rhGCPII-Glu424Ala-FolGlu_1-3 and rhGCPII-His475Tyr-Glu complexes, PDB ID 4MCP, 4MCQ, 4MCR, 4MCS, respectively.

The number of ions includes two zinc, one calcium and one chlorine atom. Ramanchandran plot outliers are: Gly335, Val382, and Ser547 in FolGlu₁ structure, Gly335, and Val382 in FolGlu₂ structure, Val382 in FolGlu₃ structure, Gly335, and Val382 in His475Tyr structure. All outliers were firmly defined by the electron density maps.

	rhGCPII Glu424Ala - FolGlu1	rhGCPII Glu424Ala - FolGlu2	rhGCPII Glu424Ala - FolGlu3	rhGCPII His475Tyr - Glu
Data collection statistics
Wavelength [Å]	1.0000	0.91840	1.0000	0.91840
Temperature [K]	100	100	100	100
Space group	I222	I222	I222	I222
Unit cell parameters a, b, c [Å]	101.5, 130.2, 158.8	101.6, 130.1, 159.9	101.5, 129.8, 158.8	101.5, 130.4, 159.0
Resolution limits	40.0–1.65(1.71–1.65)	50.0–2.00(2.07–2.00)	40.0–1.65(1.71–1.65)	30.0–1.83(1.90–1.83)
Number of unique reflections	119,318(7644)	72,031(6798)	114,830(6220)	86,816(8128)
Redundancy	6.9(3.8)	5.0(4.4)	7.0(4.1)	4.0(2.9)
Completeness	94.5(61.1)	99.3(94.8)	91.1(49.9)	93.3(88.3)
I/σ(I)	22.6(2.8)	30.1(5.0)	24.2(3.1)	16.5(2.1)
Rmerge	0.077(0.320)	0.051(0.226)	0.076(0.299)	0.067(0.429)
Refinement statistics
Resolution limits [Å]	30.0–1.65(1.69–1.65)	22.96–2.00(2.05–2.00)	29.5–1.65(1.69–1.65)	28.5-1.83(1.88-1.83)
Total number of reflections	117,931(5137)	69,042(4495)	113,362(4236)	84,871(5872)
Number of reflections in the working set	116,731(5078)	67,854(4431)	112,222(4185)	83,135(5750)
Number of reflections in the test set	1200(59)	1188(64)	1140(51)	1736(122)
R/Rfree	0.150/0.164(0.220/0.265)	0.137/0.172(0.146/0.206)	0.133/0.167(0.221/0.260)	0.159/0.191(0.251/0.284)
Total number of non-H atoms	6405	6267	6365	6384
Number of ligand atoms	41	50	59	19
Number of ions	4	4	4	4
Number of water molecules	668	521	610	584
Average B-factor [Å2]
Protein atoms	28.8	27.1	26.1	30.7
Water molecules	36.0	31.3	32.1	39.6
Substrate	38.1	40.6	67.0	31.1
Wilson B factor	29.6	27.6	26.1	31.5
r.m.s.d.
Bond lengths [Å]	0.015	0.021	0.011	0.019
Bond Angles [°]	1.49	1.66	1.38	1.61
Planarity [Å]	0.008	0.008	0.007	0.009
Chiral centers [Å3]	0.110	0.118	0.092	0.120
Ramanchandran plot [%]
Favored	97.1	97.2	97.5	97.4
Poor rotamers	1.36	1.02	1.19	1.36
Outliers	0.44	0.29	0.15	0.29
Gaps in the structure	44–55, 654–655	44–55, 654–655	44–55, 654–655	44–54, 541–543, 654–655

Values in parentheses correspond to the highest-resolution shell.

Fig. 2.

Binding modes of FolGlu_1/2/3 and ARM-P4 (PDB ID 4MCP, 4MCQ, 4MCR, 2XEG, respectively) to rhGCPII Glu424Ala. (a) FolGlu_1-3 substrates and the ARM-P4 inhibitor are shown in stick representation, while the selected part of GCPII is shown in surface representation. Residues forming the arene-binding site (ABS) are shown in stick representation and are colored red (Arg463), blue (Arg511) and yellow (Trp541). Notice the similar positioning of the distal pteroatee (FolGlu_2/3) and dinitrophenyl (ARM-P4) functionalities in the ABS. In contrast, the glutamylation status of FolGlu₁ is too low to allow the distal pteroate moiety to fully engage the ABS. FolGlu_1/2/3 substrates and arene-binding site residues are shown with their corresponding 2F_o–F_c electron density maps contoured at 1σ. Note the weak electron density maps of the Trp541 residue, reflecting its inherent flexibility (and allowing to model only one conformation for FolGlu₁ and two conformations for FolGlu_2-3). Similarly, when FolGlu₃ substrate is bound, weak electron density map is observed for the p-aminobenzoic acid and the most distal glutamic acid. (b) Superposition of FolGlu_1/2/3 and ARM-P4. Substrates/inhibitor are shown in stick representation with carbon atoms colored magenta (FolGlu₁), cyan (FolGlu₂), green (FolGlu₃), and pink (ARM-P4). The active-site zinc atoms are shown as purple spheres. For clarity, amino acids of the S1′ site, S1 site and arene-binding site are shown only by their names. Notice the structural overlap among the ultimate and penultimate glutamates and the variability of the distal substrate/inhibitor parts. (c) Superposition of FolGlu₁ (magenta) and FolGlu₂ (cyan) structures. Substrates are in stick representation; selected GCPII residues are shown as lines. Notice the flexibility of the arene-binding site residues (Arg463, Arg511, and Trp541), which is crucial for engagement of the distal pteroate moiety of the substrates by GCPII, while the S1 site residues (Arg 534, Arg536, Tyr700) take up the same conformation in both structures. The active-site zinc atoms are shown as purple spheres; selected hydrogen bonds are shown as broken lines. For clarity, the double conformation of the Arg536 in the FolGlu₁ structure is omitted.

As expected, the ultimate (C-terminal) glutamate, the scissile peptide bond, and the penultimate (S1-bound) glutamate overlap in all seven structures (data not shown). The C-terminal glutamate is placed in the S1′ pocket in a manner identical to that described in previously reported structures [13, 15], as evidenced e.g. by the superposition of FolGlu_1-3 structures with ARM-P4 structure (Fig. 2b; for structure of ARM-P4, see Fig. 5a). The carbonyl oxygen of the scissile peptide bond is polarized by the Zn1 ion (2.5 Å, not visible in Fig. 2b), and the polarization/positioning is further assisted by its interactions with the Tyr552 hydroxyl group (2.7 Å) and His553 Nε2 (3.1 Å). The peptide amide group donates a hydrogen bond to the main-chain carbonyl oxygen of Gly518 (3.0 Å). The α-carboxylate of the penultimate glutamate directly engages positively charged Arg534 (2.6 Å) and Arg536 (3.1 Å). The above-mentioned interactions are structurally conserved for both folate-based and NAAG substrates [4].

Fig. 5.

Inhibition studies with inhibitors ARM-P2, ARM-P4, ARM-P8 and MeO-P4. (a) Structures of the inhibitors ARM-P2, ARM-P4, ARM-P8 and MeO-P4. Inhibitor structures are adapted with permission from [13]. Copyright (2010) American Chemical Society. (b) Comparison of K_i values determined for inhibitors ARM-P2, ARM-P4, ARM-P8 and MeO-P4 towards GCPII wild type and Trp541Ala mutant with FolGlu₁ as a substrate. Trp541 is an important element in interacting with arene-binding-site-targeted aromatic moieties. For ARM-P2, ARM-P8 and MeO-P4 the FolGlu₁ substrate concentration was 80 nM, while for ARM-P4, it was 160 nM. (c) K_i values for ARM-P4 toward GCPII wild type and Arg463Leu, Arg511Leu and Trp541Ala mutants. Error bars indicate standard deviation derived from a sigmoidal fit of a single inhibition curve.

More structural diversity was observed for the distal (pteroate) parts of individual substrates. In the case of FolGlu₁ (see Fig. 2a), the “single glutamate linker” is too short for the benzoate/pteroate moieties to reach fully into the ABS. Instead, the p-amino benzoyl moiety of FolGlu₁ is engaged in a staggered π-π stacking interaction with the S1′ & S1 side chain of Tyr700 (3.7 Å) (Fig. 2c), while the pteridine ring system interacts in a π-cation fashion with the side chain of Arg463 (Fig. 2a). Another consequence of the substrate being too short is that the electron desity map allows the Trp541 to be modeled in only one conformation (while for FolGlu_2-3 complexes, two conformations are possible). The Trp541 residue is in a conformation of π-π stacking with the pteroate moiety. For FolGlu₂ (Fig. 2a), which features a one-Glu-residue-longer linker, π-cation and T-shaped stacking interactions were observed between the p-aminobenzoyl group of the substrate and side chains Arg463 (3.3 Å) and Trp541 (3.2 Å), respectively (Fig. 2c). The pteridine ring is wedged between the side chains of Arg511 and Trp541 with its plane parallel to the indole of Trp541 (3.3 Å) and the guanidinium group of Arg511 (3.3 Å). We observed two possible conformations for Trp541, as well as higher B-factors for its side chain. A simple overlay of FolGlu₁ and FolGlu₂ structures can be found in Fig. 2c, with amino acids of the ABS (Arg463, Arg511, Trp541) and the S1 site (Arg 534, Arg536, Tyr700) shown as lines. This overlay illustrates, that the S1 site residues (Arg 534, Arg536 and Tyr700) are essentially in the same conformation (except for a double conformation of Arg536 in the FolGlu₁ structure), while the ABS residues (Arg463, Arg511, Trp541) adopt different conformations according to how long is the γ-glutamyl chain of the folic acid. For FolGlu₃ (Fig. 2a), the binding mode and positioning of the pteroate moiety are almost identical to those observed for FolGlu_2, with Trp541 present in two conformations. However, the 2F_o–F_c electron density for the distal parts of FolGlu₃ was much weaker, and the B-factors of the final model were much higher compared to those calculated for the structure of the FolGlu₂ complex. As already mentioned above, FolGlu₃ substrate has only the C-terminal di-γ-L-glutamyl-glutamate and pteridine moieties defined in the electron density map, so the p-aminobenzoic acid moiety and the most distal glutamic acid residue are modeled rather only sterically. These factors point toward the conformational flexibility of FolGlu₃ and the limited contribution of its distal part to GCPII binding. In general, the side chain of Trp541 was defined by weak 2F_o–F_c electron density in our structures (see Fig. 2a), and the variety of conformations observed indicates inherent flexibility of this residue. In summary, the structures of FolGlu_1-6 complexes reported here clearly suggest that His475 is a surface residue too distant from the binding cavity and cannot play any role in GCPII's interaction with folyl-poly-γ-L-glutamic acids.

Structural characterization of recombinant human GCPII His475Tyr

To elucidate the putative effects of the His475Tyr substitution on the 3D structure of GCPII and on the enzyme's substrate binding and processing, we co-crystallized the rhGCPII His475Tyr variant with the natural substrate NAAG. Similar to our previous observations [3] for the wild type enzyme, glutamate (the product of the enzymatic reaction) was observed in the S1′ site of the final model, which was refined at 1.83 Å resolution, with crystallographic R-factors equal to 0.159 (R_free = 0.191; for complete data collection and refinement statistics, see Table 1).

The His475Tyr/glutamate structure (PDB ID 4MCS) was analyzed and compared to the corresponding complex of glutamate with the wild type enzyme [3] (PDB ID 2C6G). The overall fold is nearly identical for both the His475Tyr and wild type proteins, as illustrated by a root-mean-square deviation of 0.19 Å (for the 661 equivalent Cα pairs). Additionally, no differences were observed in the arrangement of the internal substrate-binding cavity and residues in the active site or in the positioning and binding mode of glutamate. Most importantly, both proteins share a near-identical arrangement of amino acids in the vicinity of residue 475, and the imidazole ring of wild type His475 and the benzene ring of Tyr475 spatially overlap (Fig. 3).

Fig. 3.

Superposition of wild type GCPII (PDB ID 2C6G) and His475Tyr variant (PDB ID 4MCS). (a) Detailed view of amino acids in the vicinity (5.0 Å) of the polymorphism site. Atoms are colored according to element, with carbons colored green and yellow for the wild type and His475 variants, respectively. His and Tyr 475 are depicted in ball-and-stick representation; surrounding residues (within 5Å) are shown as lines. The main chain (cartoon in gray) is from the His475Tyr structure. (b) and (c) Detailed pictures of Tyr475 (B) and His475 (C) residues showing the F_o–F_c electron density contoured at 3.0σ.

Site-directed mutagenesis study of the arene-binding site (ABS)

The structural data presented here indicate that at least some FolGlu_n substrates interact with GCPII in a bidentate mode: the C-terminal and penultimate glutamates engage residues of the S1′ and S1 sites, respectively, while the pteridine double-ring and p-aminobenzoyl moiety interact with the residues of the recently identified ABS [13]. To investigate the role of the ABS in folate binding and hydrolysis, a series of GCPII variants with mutations in the ABS (Arg463Leu, Arg511Leu, and Trp541Ala) was prepared by site-directed mutagenesis, expressed in Drosophila S2 cells, and purified to homogeneity. The activities of these enzymes were characterized in terms of their hydrolysis of NAAG and FolGlu₄ (substrates with structurally no and a poorly defined interaction with the ABS, respectively) and FolGlu_1-3 (substrates with structurally well-defined interactions of the pteridine double-ring with the ABS). We reasoned that if interactions between FolGlu_1-3 and the ABS influence the kinetic parameters of substrate hydrolysis, a decrease in substrate affinity (i.e., an increase in K_M value) should be observed for the GCPII ABS mutants. While no major differences between the three mutants and the wild type enzyme were observed for the kinetic parameters of FolGlu₁ hydrolysis, there was a slight decrease in K_M values for both NAAG and FolGlu_2-4 (Fig. 4). However, as the binding of NAAG does not engage residues of the ABS [13], the observed changes cannot be linked to the interaction of the pteroate with the ABS.

Fig. 4.

Comparison of the kinetic parameters of FolGlu_n hydrolysis by GCPII wild type and ABS mutant variants. This comparison shows differences between wild type and the mutant variants Arg463Leu, Arg511Leu and Trp541Ala in K_M (a) and k_cat (b). The reaction buffer was 25 mM Tris pH 7.5. Error bars indicate standard deviation derived from a simple hyperbolic fit of a single saturation curve. 1 indicates extrapolation of K_M (K_M value lower than the lowest substrate concentration).

To analyze the relevance of the ABS for interactions with small molecule ligands, we compared the inhibition profiles of four GCPII-specific inhibitors [13] (Fig. 5) towards the Trp541Ala mutant and the wild type enzyme. Three of these inhibitors (ARM-P2, ARM-P4, and ARM-P8) have been reported to engage the ABS. On the other hand, the MeO-P4 compound, which lacks the distal dinitrophenyl (DNP) moiety, is incapable of any interactions with ABS residues [13]. The inhibition data are summarized in Fig. 5. As expected, we observed no difference between the mutant and wild type proteins in the K_i values for MeO-P4. In contrast, two inhibitors featuring the distal DNP moiety linked with an optimal linker (and thus capable of ABS interactions; ARM-P4, ARM-P8) show several-fold higher affinity for the wild type protein in comparison to Trp541Ala (Fig. 5b), thus confirming the relevance of the ABS (namely Trp541) for the binding of aromatic moieties. Moreover, for ARM-P4, we observed an approximately one-order-of-magnitude difference in K_i values for Trp541Ala and Arg463Leu in comparison with wild type (see Fig. 5c; for structure of GCPII in complex with ARM-P4, see Fig. 2a).

Enzymatic characterization of recombinant human GCPII wild type and His475Tyr

A single report describes an approximately 50% decrease in the folate-hydrolyzing activity of the His475Tyr variant compared to the wild type enzyme [16]. To provide a detailed enzymatic characterization and side-by-side comparison of the wild type and His475Tyr variants, we developed a novel UPLC-based assay and measured the kinetic parameters of hydrolysis of a panel of natural GCPII substrates by purified recombinant proteins (Fig. 6). Overall, K_M and k_cat values for hydrolysis of the substrates by wild type and His475Tyr GCPII were very similar. However, we observed a nearly one-order-of-magnitude difference in affinity (difference in K_M values) for both enzymes for FolGlu_n substrates with more than one C-terminal glutamates (compare FolGlu₁ and FolGlu_2-6 substrates, Fig. 6a). This clearly suggests additional interaction(s) between the FolGlu_2-6 substrates and the protein or, alternatively, lesser steric hindrance. Somewhat surprisingly, we found some difference between the wild type and His475Tyr variants when NAAG was used as a substrate (approximately the same K_M, but different k_cat, see Fig. 6b); however, this difference was not statistically significant.

Fig. 6.

Comparison of the kinetic parameters of rhGCPII wild type and rhGCPII His475Tyr for hydrolysis of a panel of natural substrates (NAAG and FolGlu_1-6). This comparison shows differences between wild type and the His475Tyr variant in K_M (a) and k_cat (b). Error bars indicate standard deviation derived from a simple hyperbolic fit of a single saturation curve. 1 indicate extrapolation (K_M value lower than the lowest substrate concentration).

Thermal stability of recombinant human GCPII wild type and His475Tyr

Although the wild type and His475Tyr variants are virtually indistinguishable at the structural level, we investigated the possibility that the His to Tyr substitution influences the thermal stability of the His475Tyr variant. To this end, we used a Thermofluor assay to define the temperature midpoint for the unfolding transition, T_m. The T_m value determined for the wild type protein (67.2 ± 0.118 °C) is comparable to that determined for the His475Tyr variant (69.2 ± 0 °C), arguing against a biologically significant (de)stabilization effect of the mutation. The kinetic parameters obtained for folate-based substrates, together with the data obtained by 3D structure analysis and thermal stability experiments, suggest that the His475Tyr variant is equivalent to the wild type enzyme at both the structural and enzymological levels.

Discussion

The aims of this study were to investigate the molecular recognition of polyglutamylated folates by GCPII and clarify whether the reported His475Tyr polymorphism might alter the binding and/or turnover of polyglutamylated folates, potentially leading to altered levels of folates and related metabolites in humans. Furthermore, we set out to analyze the role of the “arene-binding site”, a recently identified putative exosite in the GCPII structure, in the molecular recognition of polyglutamylated folates and, more generally, in the binding of ligands containing an aromatic moiety capable of interacting with this exosite.

Although extensive epidemiologic studies on the effect of the His475Tyr polymorphism have been conducted, the possible role of this mutation in the folate uptake and metabolism of metabolically related molecules remains unclear. Folates are essential for the C1 metabolism of cells, i.e., for the biosynthesis of nucleobases, and they are important for the replication of rapidly dividing cancer cells. Folate insufficiency might result in altered levels of folate-related metabolites and/or in altered susceptibility to certain types of cancer. However, it is difficult to speculate how a single amino acid mutation in a region distant from the active site (see Fig. 1; His475 is located 27 Å from the active site Zn1) could influence the catalytic activity of the enzyme. Nevertheless, there have been reports of long-range rearrangements of an enzyme's substrate binding cleft caused by very distant mutations, for example in HIV-1 protease (reviewed elsewhere [17]). Thus, the idea that the His475Tyr mutation could influence GCPII activity was worth exploring. However, we found that the enzymatic properties of recombinant human GCPII His475Tyr are very similar to those of the wild type enzyme (almost identical K_M values and turnover numbers toward all tested folyl-poly-γ-L-glutamate substrates). Furthermore, 3D structure analysis of the protein by X-ray crystallography did not reveal any significant structural changes caused by the mutation. The very subtle differences that we observed cannot account for the reported phenotype of the polymorphism [8, 9, 10, 11, 12]. It is relevant to mention that all recombinant proteins discussed in this manuscript were prepared in Drosophila S2 cells and comprise the full ectodomain of human GCPII in which the enzymatic activity resides. We have previously shown that the enzymatic activity of this recombinant protein is indistinguishable to that of native, full-length GCPII [18].

All seven structures solved as part of the present work provide structural explanation as to why the His475Tyr mutant behaves virtually identically to the wild type enzyme. Our structural analysis showed that the bound FolGlu_1-6 substrates do not interact with His475. Additionally, the structure of GCPII His475Tyr exhibits no significant re-arrangement of the substrate binding pocket. Although the enzymologic comparison of rhGCPII wild type and His475Tyr found no significant differences in cleavage of FolGlu_n, thermal stability, or structure, the possibility that the His475Tyr mutation causes altered folic acid levels and folate-related metabolites in humans remains open. The His475Tyr mutation may influence folate levels by other mechanisms, e.g., by changing the trafficking of GCPII, by changing its expression level or by changing the enzyme's interaction with potential signaling partners. These potential mechanisms of action have yet to be investigated.

There was one peculiarity about the X-ray structure of the complex obtained from the crystallization condition with the inactive GCPII mutant (Glu424Ala) and FolGlu₀. As already mentioned, this structure contains a di-γ-L-glutamyl-glutamic moiety, even though no compound containing such a moiety has been added to the crystallization drop. A possible explanation is that some higher polyglutamylated folates (FolGlu_>4) coming from the cultivation medium may have remained bound throughout the purification process (only a one-step affinity purification), not allowing the FolGlu₀ to bind. In line with such a hypothesis, folic acid did not inhibit the cleavage of FolGlu₃ (data not shown), suggesting that the folate (the product) has lower affinity to the enzyme than polyglutamylated folate (the substrate).

Inspection of the kinetic properties of the wild type and His475Tyr variants revealed an interesting observation. Folate substrates harboring more than two C-terminal glutamates bound to the enzyme more effectively (by almost one order of magnitude), as reflected by differences in their K_M values (compare FolGlu₁ and FolGlu_2-6, Fig. 6a). This difference in K_M is in good agreement with previously published observation that di-γ-L-glutamate moiety inhibited the NAAG-hydrolysing activity of GCPII twenty times less effectively than tri-γ-L-glutamate moiety, while tri, tetra- and penta-γ-L-glutamate moieties inhibited comparably [19].

To further analyze the contribution of the ABS to substrate recognition, we performed site-directed mutagenesis of the key residues of the exosite (Arg463, Arg511, and Trp541). Kinetic analysis of the recombinant mutant proteins did not show a major influence of the ABS residues on polyglutamyl-folate binding and turnover, only a modest, approximately three-fold increase in NAAG binding affinity. As NAAG cannot engage the ABS, the direct effects of these mutations on substrate recognition can be excluded and more intricate/subtle (unidentified) contributions (e.g. the flexibility of the entrance lid) may play role.

The binding of selected inhibitors targeting the arene-binding site, on the other hand, has been previously shown to be influenced by their interaction with the residues forming the ABS both structurally and using inhibition studies [13]. Our study tested those inhibitors with rhGCPII wild type and Trp541Ala variants [13] (structures of the inhibitors are shown in Fig. 5a). Inhibitors that had been previously reported to interact with the ABS showed a significant loss of binding to the GCPII Trp541Ala variant, in which the ABS is presumably disrupted (see Fig. 5b), confirming the importance of this structural feature for inhibitor binding. It is somewhat surprising that on one hand, the crystallographic and site-directed mutagenesis data of the ABS provide a good mechanistic explanation for changes in inhibitor affinity, while on the other hand there is a disconnection between the structural data (ABS engages the pteroate) and kinetic studies (mutations in the ABS mostly do not change kinetic parameters of substrate hydrolysis). We can speculate that these differences can be attributed to differences in the kinetics of substrate vs. inhibitor interactions with GCPII. The crystallographic data represent the most preferred conformation of a given GCPII complex achieved in a time-scale of several days (crystal growth). In addition, when a typical inhibition constant is determined, the inhibitor is usually preincubated with the enzyme for several minutes (5 - 13 in this case), giving the inhibitor enough time to adopt the most energetically favorable conformation. On the other hand, k_cat values greater than 1 s^-1 indicate that a given substrate is bound and hydrolyzed in less than a second. Although the high-affinity C-terminal part of the substrate is docked into (and released from) the active site of GCPII within this time scale, the flexible distal part might not be able to engage the ABS and thus has limited (or no) contribution to the overall affinity. This conjecture is supported by binding kinetics observed for GCPII interactions with a fluorescently labeled GCPII inhibitor/probe [20]. We have observed that the BODIPY fluorophore engages the ABS (unpublished work, C. Bařinka) and that this engagement leads to quenching of fluorescence intensity. The fluorescence quenching is not instantaneous but rather gradual, taking tens of seconds. This points toward a bidentate mode of substrate/inhibitor binding, in which the C-terminal (docking) part of the inhibitor binds “immediately” and the distal part somewhat more slowly. Another possibility could be that this exosite is inherently a lower-affinity site to allow to bind all forms of folic acid moiety (folyl/5-methyl-dihydrofolyl/5-methyl-tetrahydrofolyl). The natural diet contains predominantly 5-methyl-tetrahydrofolyl-poly-γ-L-glutamates [21], which are less stable than their fully oxidized folyl-forms and may easily be oxidized to 5-methyl-dihydrofolyl-poly-γ-L-glutamates or may even degrade to the folyl-forms. The same may apply to the folate receptor α isoform (FRα), which has been recently co-crystallized with the folic acid, but not with 5-methyl-tetrahydrofolic acid [22] (PDB ID 4LRH), although this isoform of folate receptor is expressed in kidney in renal tubular cells to resorb 5-methyl-tetrahydrofolic acid, a major form of folic acid found in blood.

Another indication that the arene-binding site is a rather low-affinity binding site provides a recently published x-ray structure of GCPII in complex with a non-hydrolyzable methotrexate analogue of the FolGlu₁ substrate (PDB ID 3BI1) [23]. In this structure, only the γ-D-glutamyl-L-glutamate moiety with spatial arrangement analogous to what is seen in FolGlu_1-3 structures, was defined by the electron density map, while the pteridine ring was disordered and no stacking with Trp541 could be observed [23].

Measurements of kinetic parameters also revealed that the mechanism, by which polyglutamylated folates are cleaved, is probably distributive, i.e. the γ-linked L-glutamates from folyl-poly-γ-L-glutamic acids are cleaved sequentially from C-terminus. For higher FolGlu_n turnovers (10 -20 %), we have always observed a small amount (≈1 - 2 %) of the FolGlu_n-2 product (data not shown; the contribution of the FolGlu_n-2 product to the kinetic data could thus be neclected).

Data presented in this study might be a starting point for the design of a homolog-specific ligand or inhibitor of GCPII and GCPIII. Although these two homologous proteins share 81% similarity at the amino acid sequence level [24] and their 3D structures are very similar [25], some reports suggest that GCPIII might play a distinct enzymatic role in specific tissues, such as the testes [26]. As shown by Zhang et al. [13], the choice of a proper arene-binding-site-targeting moiety can improve inhibitor potency by several orders of magnitude. Since GCPIII does not seem to have a corresponding ABS in its structure, an appropriate inhibitor targeting this exosite in GCPII might represent a useful tool in distinguishing both enzymes at the protein level.

Materials and Methods

Preparation of recombinant human GCPII

Recombinant human GCPII (rhGCPII) wild type and mutants were heterologously overexpressed in Drosophila Schneider 2 (S2) cells [27] using the pMT/Bip/AviTEV/rhGCPII vector, as recently described [28]. In this expression system, the soluble extracellular part of GCPII N-terminally fused with a so-called AviTEV tag is secreted into the medium. The AviTEV tag comprises an in vivo biotinylated Bir ligase recognition sequence (AviTag™, Avidity, Aurora, CO, USA) separated from the target protein with a tobacco etch virus (TEV) recognition sequence. Proteins were purified using the Streptavidin Mutein Matrix™ (Roche Diagnostics, Basel, Switzerland) [28].

Site-directed mutagenesis and cloning

The Glu424Ala and His475Tyr mutations were introduced by cutting the coding sequence out of the corresponding pMTNAEXST vector [4] using the restriction enzymes BglII/XhoI (New England Biolabs, Ipswitch, MA, USA) and XhoI/XcmI (New England Biolabs), respectively, and re-ligating into the pMT/Bip/AviTEV/rhGCPII vector using T4 DNA ligase (New England Biolabs). The mutations Arg463Leu, Arg511Leu, and Trp541Ala were introduced by mutating the pMT/Bip/AviTEV/rhGCPII vector using the QuikChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA, USA). Nucleotide sequences (5′ to 3′) of the primers were as follows: GAAGGAAACTACACTCTGTTAGTTGATTGTACACCGCTGATG (forward primer, Arg463Leu), CATCAGCGGTGTACAATCAACTAACAGAGTGTAGTTTCCTTC (reverse primer, Arg463Leu), GTTCAGTGGCATGCCCCTGATAAGCAAATTG (forward primer, Arg511Leu), CAATTTGCTTATCAGGGGCATGCCACTGAAC (reverse primer, Arg511Leu), GTATACTAAAAATGCCGAAACAAACAAATTCAG (forward primer, Trp541Ala), and CTGAATTTGTTTGTTTCGGCATTTTTAGTATAC (reverse primer, Trp541Ala). Before synthesis, all primer sequences were checked using OligoCalc [29]. Sequences of the resulting plasmids were verified by dideoxynucleotide termination sequencing. Yields of pure recombinant proteins were 0.63 mg/100 ml media for the wild type enzyme [28], 83 μg/100 ml for His475Tyr, 9.5 μg/100 ml for Arg463Leu, 6.3 μg/100 ml for Arg511Leu, and 0.10 mg/100 ml for Trp541Ala.

Crystallization and data collection

Crystals were grown by the hanging drop vaporization technique at 291 K as reported previously [30], with minor modifications. The crystallization drops consisted of 1 μL rhGCPII Glu424Ala mixed with an equal volume of the reservoir solution [33% (v/v) pentaerythritol propoxylate PO/OH 5/4 (Hampton Research, Aliso Viejo, CA, USA), 0.5% (w/v) PEG 3350 (Sigma Aldrich, St. Louis, MO, USA), 0.10 M Tris-HCl (Promega, Madison, WI, USA), pH 8.0] pre-mixed with 10 mM folyl-(0/1/2/3/4/5/6)-γ-L-glutamic acid (Schircks Laboratories, Jona, Switzerland) in a 9:1 (v/v) ratio. The His475Tyr variant was co-crystallized with NAAG at a final concentration of 2 mM in the same setup. Crystals usually appeared within one day and grew to a final size of approximately 0.40 × 0.40 × 0.2 mm during subsequent few weeks. Diffraction data were collected at 100 K using synchrotron radiation at the SER-CAT ID/22-BM beamlines (Argonne, IL, USA; 1.00 Å; MARMOSAIC 225; complexes with FolGlu_1/3/4) and at the MX 14.2 beamline (BESSYII, Helmholtz-Zentrum, Berlin, Germany; 0.9184 Å; MX-225; the His475Tyr variant and complexes with FolGlu_2/5/6). Complete datasets were collected from single crystals, and data were processed using either the HKL2000 software package [31] or XDSAPP [32]. The complex with FolGlu₂ was tested in-house at 120 K on the MAR345 detector (MAR Research, Hamburg, Germany) using Cu Kα wavelength (1.5418 Å).

Structure determination and refinement

Structures of GCPII Glu424Ala complexes were determined by molecular replacement method using the coordinates of the GCPII-Glu424Ala-NAAG complex (PDB ID 3BXM) as a starting model [4]. Refinement calculations and manual rebuilding were performed with the programs Refmac 5.5 [33] and Coot 6.1 [34], respectively. Approximately 1.0 - 1.7 % of the data (corresponding to 1140 - 1736 reflections) were excluded from calculations and used instead to calculate R_free value (R_free sets were independent). Models for individual substrates together with associated restraints were prepared using the PRODRG server [35] and were placed into the positive F_o–F_c electron density maps, located in the substrate binding cavity of rhGCPII. An isotropic refinement protocol was used throughout all stages of the refinement. The stereochemical quality of the final models was evaluated using MolProbity [36]. Data collection and structural refinement statistics are shown in Table 1.

GCPII activity assay

Reactions with the folyl-(1/2/3/4/5/6)-γ-L-glutamic acid substrates (Schircks Laboratories) and NAAG (Sigma Aldrich) were performed in 25 mM Tris-HCl, pH 7.5, in a total volume of 250 μl. Concentrations of substrate stock solutions were determined by amino acid analysis. All components were pipetted at 0 °C, and reactions were started by adding enzyme and mixing. The reactions were placed in a 37 °C thermo-block (Bioer MB-102, China, or Eppendorf Thermomixer Comfort, Hamburg, Germany) for 20 min. These reaction conditions were designed to yield 10 - 20 % conversion of substrate. Reactions were stopped by adding 3.5 μl stopping solution [72 μM 2-PMPA, 7.2 mM 2-mercaptoethanol (Sigma Aldrich), 25 mM Tris, 0.48 M phosphoric acid (Penta, Praha, Czech Republic)]. Typically, nine data points were plotted for each substrate (with each point representing the average of duplicate experiments), and saturation curves were fitted using GraFit [37].

For inhibition and kinetic studies with FolGlu_1-6 as a substrate, reactions were performed in 25 mM bistrispropane-HCl (Sigma Aldrich), pH 7.5. Reactions were carried out either in a total volume of 215 μl in a 96 well plate immersed in a water bath (Grant Instruments, Shepreth, United Kingdom) or in a total volume of 250 μl in 1.5 ml test tubes in a thermo-block (Bioer MB-102). All components except the substrate were pipetted at 0 °C. After a 5 - 13 min preincubation at 37.0 °C (5 min preincubation in the case of the 1.5 ml test tubes), substrate was added. The reaction proceeded for 20 min and was terminated by the addition of 10.0 μl stopping solution [0.44 M phosphoric acid and 21.3 μM 2-PMPA (96 well plate) or 3.48 μl of 1.5 M phosphoric acid and 71.3 μM 2-PMPA (1.5 ml test tubes)]. The identity of the inhibitors was verified by Q-TOF Micro (Waters, Milford, MA, USA) and LTQ Orbitrap XL (Thermo Fisher Scientific, Waltham, MA, USA). Each inhibition curve was fitted using GraFit [37] into at least 12 data points typically acquired in duplicate.

Novel UPLC method for assaying glutamate carboxypeptidase activity using folyl-(1/2/3/4/5/6)-γ-L-glutamic acids as substrates

FolGlu_n reaction mixtures were analyzed on a C18 Acquity UPLC HSS T3 2.1 × 100 mm column with 1.8 μm particles (Waters), guarded by a 0.22 μm pre-filter (Waters) and VanGuard pre-column (Waters), coupled to an Agilent 1200 Series UPLC instrument (Agilent Technologies, Santa Clara, CA, USA). Mobile phase A was 25 mM sodium phosphate buffer, pH 6.0 (Penta), supplemented with 0.02 % (w/v) sodium azide (Sigma Aldrich); mobile phase B was acetonitrile (VWR International, Radnor, PA, USA). Elution of individual FolGlu_1/2/3/4/5/6 molecules and their cleavage products was performed isocratically at 2.0 % / 1.5 % / 1.1 % / 0.4 % / 0.2 % / 0.0 % acetonitrile, respectively. The column temperature was set to 50.0 °C. The HPLC runs consisted of 1.8 min isocratic flow at 0.0 - 2.0 % B, 0.1 min transition to 10.0 % B, 1.1 min at 10.0 % B, 0.1 min transition back to 0.0 - 2.0 % B, and 7.4 min re-equilibration. Analytes were detected at 281 nm and 354 nm. Inhibition reactions with FolGlu₁ were analyzed in the same manner except that the percentage of acetonitrile was 2.1 % instead of 2.0 %. The substrate turnover was quantified as the ratio of the substrate and product peak areas. The sum of the areas of the substrate and the product served as an internal standard. The limit of quantification was at least 10 nM (when calculated as baseline height multiplied by 10).

Novel UPLC method for assaying glutamate carboxypeptidase activity using NAAG as a substrate

The NAAG reaction mixtures were lyophilized (for at least 6 hours at 20 μbar using Christ Beta 2-8 LD plus, Osterode am Harz, Germany, or Labconco instruments cat. no. 7753511, Kansas City, MO, USA) and re-dissolved in 25.0 μl MilliQ water (Merck Millipore, Billerica, MA, USA). Then, surpassing the need for a radioactivelly labeled substrate traditionally used in assessing the NAAG hydrolysing activity of GCPII [2], a modified fully automated ortho-phthalaldehyde-derivatization [38] step was performed: 11.0 μl of each re-dissolved reaction was manually transferred into a 96 well plate, and the plate was inserted into an autosampler set to 4 °C. A 99.0 μl aliquot of the derivatization solution [40.6 mM 2-mercaptopropionic acid (Sigma Aldrich), 33.0 mM ortho-phthalaldehyde (Sigma Aldrich) in 200 mM sodium borate (Pharmacia, Uppsala, Sweden), pH 10.0] was added by the autosampler. After mixing, the sample was injected onto a C18 Acquity UPLC HSS T3 2.1 × 100 mm column with 1.8 μm particles (Waters), guarded by a 0.22 μm pre-filter (Waters) and VanGuard precolumn (Waters), coupled to an Agilent 1200 Series UPLC instrument (Agilent Technologies). Ortho-phthalaldehyde-Glu derivative was eluted isocratically with 96.0 % mobile phase A [25 mM sodium phosphate buffer, pH 6.0 (Penta), supplemented with 0.02 % (w/v) sodium azide (Sigma Aldrich)] and 4.0 % mobile phase B (acetonitrile). The column temperature was 70.0 °C. The method consisted of 2.7 min isocratic flow at 4.0 % B, 0.1 min transition to 80.0 % B, 4.7 min at 80.0 % B, 0.1 min transition back to 4.0 % B, and 8.4 min re-equilibration. Glutamate eluted after approximately 1.4 min. Analytes were detected by fluorescence at 230/450 nm. The substrate turnover was quantified as the ratio of the peak areas of the glutamic acid (product) and the the total cleavage reaction. The limit of quantification was estimated as x_B + 10σ_B (average of the blank measurement plus 10 times the standard deviation of the blank) and was usually around 20 - 30 nM.

Thermofluor assay

To determine the temperature midpoint for the protein unfolding transition, T_m, a thermal shift assay using SYPRO® Orange (Invitrogen, Carlsbad, CA, USA) dye [39] was performed on a Roche LightCycler® 480 II instrument (Roche Diagnostics, Penzberg, Germany) in Roche LightCycler® 480 Multiwell Plates 96. Each well contained 3.9 μg protein and 2.5 μl 40 × SYPRO® Orange in a final volume of 50 μl 25 mM Tris, pH 7.5. During pipetting, all components were kept on ice.

Abbreviations

GCPII

human glutamate carboxypeptidase II

PSMA

prostate-specific membrane antigen

FOLH1

folate hydrolase 1

FGCP

folyl-poly-γ-glutamate carboxypeptidase

NAALADase

N-acetylated-α-linked acidic dipeptidase

NAAG

N-acetyl-L-aspartyl-L-glutamate

FolGlu_n

folyl-n-γ-L-glutamic acid

ABS

arene-binding site

rhGCPII

recombinant human glutamate carboxypeptidase II

DNP

dinitrophenyl

2-PMPA

2-(Phosphonomethyl)-pentanedioic acid

ARM-P2/4/8

urea based inhibitor containing the arene-binding site targeted dinitrophenyl moiety linked via 2/4/8 ethyleneglycol units (see Fig. 5)

MeO-P4

urea based inhibitor with 4 ethyleneglycol units containing no arene-binding site targeted moiety (see Fig. 5)