A common SNP in ER aminopeptidase 2 induces a specificity switch that leads to altered antigen processing

Abstract

ER aminopeptidases 1 and 2 (ERAP1 and ERAP2) cooperate to trim antigenic peptide precursors for loading onto MHC class I molecules and help regulate the adaptive immune response. Common coding single nucleotide polymorphisms (SNPs) in ERAP1 and ERAP2 have been linked with predisposition to human diseases ranging from viral and bacterial infections to autoimmunity and cancer. It has been hypothesized that altered antigen processing by these enzymes is a causal link to disease etiology but the molecular mechanisms are obscure. We report here that the common ERAP2 SNP rs2549782 that codes for amino acid variation N392K leads to alterations in both the activity and the specificity of the enzyme. Specifically, the 392N allele excises hydrophobic N-terminal residues from epitope precursors up to 165-fold faster compared to the 392K allele, although both alleles are very similar in excising positively charged N-terminal amino acids. These effects are primarily due to changes in the catalytic turnover rate (k_cat) and not in the affinity for the substrate. X-ray crystallographic analysis of the ERAP2 392K allele suggests that the polymorphism interferes with the stabilization of the N-terminus of the peptide both directly and indirectly through interactions with key residues participating in catalysis. This specificity-switch allows the 392N allele of ERAP2 to supplement ERAP1 activity for the removal of hydrophobic N-terminal residues. Our results provide mechanistic insight to the association of this ERAP2 polymorphism with disease and support the idea that polymorphic variation in antigen processing enzymes constitutes a component of immune response variability in humans.

Introduction

Cytotoxic T-lymphocyte responses are driven by the recognition of antigenic peptides presented on the cell surface by specialized receptors on the Major Histocompatibility Complex (MHC). MHC class I molecules present peptides originating from the proteolytic degradation of intracellular (direct presentation) or endocytosed (cross-presentation) proteins (1, 2). Although the proteolytic cascade leading to the generation of these peptides is initiated in the cytosol by the proteasome, the last few steps of antigenic peptide customization occur in the ER or in specialized endosomal vesicles (3, 4). There, specialized aminopeptidases trim extended precursors to the mature antigenic epitopes so that the correct-length peptides can be loaded onto nascent MHC class I molecules (5). Aminopeptidase-mediated antigenic peptide customization has been shown during the last few years to be a key indispensible component to antigenic epitope generation (6).

At least two intracellular aminopeptidases, ERAP1 and ERAP2, are localized in the ER and cooperate to trim N-terminally extended peptide precursors to generate mature antigenic epitopes (7–9). A third homologous aminopeptidase, IRAP, has been recently implicated in a separate endosomal pathway of cross-presentation in specialized cells (4). Of these three, the better characterized one, ERAP1 has been shown to play the role of an antigenic peptide editor, influencing the antigenic peptide repertoire and the adaptive immune response in vivo (6, 10–12). ERAP1 achieves peptide editing by both generating many mature epitopes from elongated precursors and by destroying others by further trimming them to smaller peptides that cannot bind onto MHC class I molecules (8, 13). Less is known about the function of ERAP2, although it has been proposed that it acts as an accessory aminopeptidase, complementing ERAP1 specificity to allow the trimming of the vast number of different sequences in the antigenic peptide repertoire (9, 14, 15). ERAP1 has some key functional properties that make it suitable for antigenic peptide precursor trimming: it is specialized for relatively large peptides and can trim peptides depending on their length and internal sequence (16, 17). Furthermore, ERAP1 can be regulated by its own substrates and products, suggesting a complex and poorly understood landscape of the regulation of antigen processing (18, 19). ERAP2 and IRAP, although less studied to date, appear to share some of those properties, but also have key differences that may underlie distinct biological roles (9, 13, 15, 16).

Several large-scale population wide genetic studies have linked specific amino-acid coding single nucleotide polymorphisms (SNPs) in ERAP1 and ERAP2 with predisposition to human disease (reviewed in (20)). SNPs in ERAP1 and ERAP2 have been linked with predisposition for development of the chronic inflammatory disease Ankylosing Spondylitis, a disease with a strong hereditary autoimmune component (21–23). The original hypothesis that this link is mediated by the antigen processing function of ERAP1/ERAP2 was later supported by the demonstration of interplay between ERAP1 and ERAP2 SNPs and more importantly with the MHC class I allele HLA-B27 (24–27). Recently, ERAP1/ERAP2 haplotypes have been linked with resistance to HIV infection (28). More specifically, the ERAP2 SNP rs2549782 that codes for the variation N392K has been linked with resistance to HIV infection in homozygous individuals and with development of pre-eclampsia (28–30). ERAP1 and ERAP2 polymorphisms have been proposed to be the result of a long-standing balancing selection, indicating significant functional consequences of both polymorphic states (28, 31). Preliminary results regarding the effects of coding ERAP1 polymorphisms have pointed towards changes in both enzyme activity and specificity that may influence the antigenic peptide repertoire (19, 27, 32). Although these changes in antigenic peptide processing have been of a relatively low magnitude (generally up to 2-fold) it has been hypothesized that they can be sufficient in altering chronic immune and/or inflammatory responses (19, 27).

In this study we characterized the ERAP2 SNP rs2549782 that codes for the amino acid variation N392K. This particular polymorphism constitutes an attractive model for understanding the role of polymorphic variation in antigen processing for three reasons: i) both polymorphic states are almost equally distributed in the human population, presumably as a result of host-pathogen balancing selection (28), ii) it is the main polymorphism in ERAP2 that has been associated with disease in virtually all genetic studies and iii) its structural proximity to the enzyme’s active site makes it reasonable to expect at least some perturbation in enzymatic activity (15). We used a combination of biochemical, cell-based, structural and computational approaches to understand the effects of this polymorphic variation to ERAP2 function. We demonstrate that the polymorphic variation at position 392 affects both the activity and specificity of the enzyme up to 2 orders of magnitude, altering its antigen processing profile as well as its ability to complement ERAP1-mediated antigen processing. Structural and biochemical analysis suggests that these effects are due to changes in the catalytic site of the enzyme that affect transition-state stabilization. To our knowledge, these are the largest functional changes described for a common coding polymorphism in the antigen processing machinery. Our results provide mechanistic insight to the hypothesized causal link between this ERAP2 SNP and disease. Furthermore, our results support the hypothesis that allelic variability in antigen processing enzymes is an important component of immune system variability in natural populations.

Materials And Methods

Peptides

Peptides WLRRYLENGK, RLRRYLENGK, LSLYNTVATL, KSLYNTVATL, LSRHHAFSFR, RSRYWAIRTR, LSRYWAIRTR, LRSRYWAIRTR, YKRFEGLTQR, LKRVVINKDT and ASRHHAFSFR were purchased from JPT Peptide Technologies (Berlin, Germany). Peptide KSLYNTVATL was purchased from Invitrogen. All peptides were purified by RP-HPLC and were more than 95% pure.

Protein Expression And Purification

Construction of the N392K ERAP2 variant (ERAP2^K) was performed using the Quikchange II XL site-directed mutagenesis kit (Agilent technologies, Santa Clara, CA) on the pFastbac1-ERAP2^N vector (described in (14)). The primers used for the variant construction were: 5′-GGA ATG ATA TCT GGC TTA AGG AGG GTT TTG CAA AAT AC-3′ (sense) and 5′-GTA TTT TGC AAA ACC CTC CTT AAG CCA GAT ATC ATT CC-3′ (antisense). DNA sequencing was performed to validate successful mutagenesis.

Over-expression of ERAP2^N in insect cell cultures has been described before (14, 15). The pFastbac1-ERAP2^K vector was used to generate recombinant baculovirus according to the directions of the bac-to-bac baculovirus expression system (Invitrogen), as described before (17). Human recombinant ERAP2^K was produced in Hi5 cells after infection with recombinant baculovirus carrying the ERAP2 gene with a C-terminal 6-His tag. The cell supernatant was harvested 4 days post infection by centrifugation and subjected to Ni-NTA affinity chromatography purification as described before (15, 17). Briefly, after binding in the presence of buffer containing 1mM imidazole for 2 hours, protein was eluted from the column using step gradient of 3, 5, 10, 20, 40 and 150mM imidazole solutions (Supplemental Figure 1A). The fractions with highest enzymatic activity and purity were collected, dialyzed to remove the imidazole and stored in −80°C in aliquots in the presence of 10% glycerol, until needed. Each aliquot was thawed once, used in assays immediately and then discarded. Both ERAP2 variants were over-expressed and purified in parallel for purposes of comparing enzymatic activities.

Enzymatic Assays

The activity of human recombinant ERAP2 was measured by following the time-dependent increase of the fluorescence signal at 460nm of the fluorigenic substrate R-AMC (L-Arginyl-7-amido-4-methyl coumarin, Sigma). Excitation was set at 380nm. Fluorescence was followed for 5 min using a Quantamaster-4 fluorimeter (Photon Technology International, Lawrenceville, NJ) and the resulting time slope was used to calculate the rate of hydrolysis of the substrate. ERAP2 specificity for small fluorigenic substrates was analyzed using a library of fluorigenic compounds described before (14, 33). 10μM of each substrate was incubated with 8ng ERAP2 (total volume 140μl, ERAP2 concentration 0.52nM) at 25°C and fluorescence was recorded for 5min with excitation set at 380nm and emission at 460nm using a Tecan Infinite M200 microplate fluorescence reader. Inhibition data were fit to a simple binding model V=V_o*X/(K_i+X)−V_o, where V is the rate of substrate hydrolysis, X the inhibitor concentration, K_i the inhibition constant and V_o the rate of hydrolysis in the absence of inhibitor. Inhibition by the DG001 peptide was found to proceed to less than 100% and data were fit to a partial inhibition model V=V_o*X/(K_i+X)−V_o+B in which the parameter B is the maximal inhibition attained after saturation.

Synthesis, purification and characterization of pseudopeptide transition state analogs

The L-Ψ[PO(OH)CH₂]-LAFKARAF peptide (DG001) was synthesized by Fmoc-solid phase peptide synthesis using the multipin technology as previously described (34). Peptide Fmoc-AFKARAF was developed on trityl alcohol lanterns (surface: polysterene, loading 15 μmol) using the Fmoc-solid phase protocol (35). Pmc and Boc protecting groups were used for the side chain functional groups of Arg and Lys, respectively. After the deprotection of the N-terminal Ala, the phosphinic dipeptide unit was introduced by shaking the lantern-anchored peptides with a solution of 1.3 equivalent of phosphinic block BocLeu-Ψ[PO(OAd)CH₂]-LeuOH, DIC (2 equivalents) and HOBt (2 equivalents) in dichloromethane during 24 hours. The phosphinic block was synthesized based on a previously described protocol (36) (full details will be reported elsewhere). The crude peptide was obtained after treatment of lanterns with TFA/dichloromethane/H₂O/triisopropylsilane 80/15/2.5/2.5 for 2 hours and purified by reverse-phase HPLC using a Chromolith C-18 semi-prep column (Merck) and a 5–50% acetonitrile gradient in water containing 0.05% trifluoroacetic acid. Eluted peaks were collected, lyophilized and characterized by mass spectrometry.

Analysis Of Peptide Trimming Activity On Reverse-Phase HPLC

Analysis of digestion products after incubation of peptides with ERAP1 or ERAP2 was performed by RP-HPLC on a Chromolith C-18 analytical column (Merck) as previously described (17). Briefly, 20μM of each peptide were incubated at 37°C with 100 to 4,000ng (7.3 to 293.6nM) ERAP1 or ERAP2 in 50mM HEPES pH 7.0, 100mM NaCl, buffer in a total volume of 125μl for 15 min to 1 hour. After incubation, the reactions were terminated by adding 100μL 0.1% trifluoroacetic acid and stored at −20°C until analysis. Before analysis samples were centrifuged for 5 min at 10,000xg to remove precipitated protein. HPLC elution was performed using a 5 to 50% acetonitrile gradient, while following the absorbance at 220nm. In order to calculate initial reaction rates, several experiments were performed for each peptide in order to achieve conditions under which the reaction was less than 30% complete. All comparisons of the activity of ERAP1 and the two ERAP2 variants were performed in parallel. Product peaks in the HPLC chromatogram were identified by running peptide controls and by mass spectrometry as described (13).

Analysis Of Peptide Trimming Activity by mass-spectroscopy

Trimming of ASRHHAFSFR peptide by ERAP2 was followed by mass spectroscopy since the product SRHHAFSFR is poorly separated by HPLC. Reactions were set up as described above with the exception that they were stopped by adding 1% formic acid. The reactions were initially cleaned-up using pipette tips packed with reverse-phase material (C18MB, OMIX, VARIAN). The samples were dried completely in a speed-vac, dissolved in 100μl of 0.1% formic acid and 2% ACN and sonicated for 1 min. The samples were analyzed by direct infusion nano-electrospray on an LTQ Orbitrap XL mass spectrometer at a resolving power of 60,000 at m/z 400. The scan range was set to 350–450 m/z. MS/MS spectrum was acquired to confirm the sequence of the peptides. The triply charged peptides were monitored. Reaction progress was evaluated by measuring the relative peak intensity of the substrate and product peptide respectively. All reactions were evaluated at <30% completion to ensure linearity of the rate measurement.

Cell-based antigen presentation assay

Construction of plasmids, encoding for ERAP2^N and ERAP2^K variants (rs2549782, Genbank accession # NM_022350.3) was performed as previously described for ERAP1 alleles (7, 19) in pTracerCMV-RFP, where the GFP/Zeo fusion protein was replaced with red fluorescent protein ORF. Briefly, ERAP2^N ORF was subcloned into pcDNA3.1 (Invitrogen, Carlsbad, CA). pcDNA-ERAP2^K was generated using Quikchange II XL Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA), utilizing the following primers (AAT mutated into AAG codon):

• ERAP2_MUT_F 5′ GGAATGATATTTGGCTTAAGGAGGGTTTTGCAAAATAC

• ERAP2_MUT_R 5′ GTATTTTGCAAAACCCTCCTTAAGCCAAATATCATTCC

pcDNA-based ERAP2 containing plasmids were digested with SalI and XhoI and subcloned into pShuttleCMV. pShuttle-ERAP2 plasmids were digested with NotI and XbaI and subcloned into pTracerCMV-RFP. Integrity of ERAP2^N and ERAP2^K genes, cloned into pTracerCMV-RFP was confirmed by sequencing of whole length ERAP2 ORF using the following primers:

• ERAP2cl.end_SEQ-For 5′ GAACTTGGCAGCTCTCCTTCATGC

• ERAP2cl.Start_SEQ-Rev 5′ CTGACTGAAGGGTGGCATTCGTG

• ERAP2-SEQ-Rev 5′ GGTTTGGAGATAGGGCGGGATG

• ERAP2-SEQ-For 5′ CAACCCAGGCACGCATGGC

HeLa cells, stably expressing H-2K^b, HLA-B27 and Herpesvirus protein ICP47 (TAP1 blocker), HeLa-K^b-B27/47 were described previously (8). To construct plasmids expressing HLA-B27 restricted peptides (and their N-terminal precursors), nucleotide sequences corresponding to ASRHHAFSFR and SRHHAFSFR peptides were subcloned into pTracerCMV (Invitrogen, Carlsbad, CA). All peptides were designed to contain the ER-signal sequence MRYMILGLLALAAVCSA (from the Ad2E3gp19K protein) upstream of the peptide sequence (7). HeLa-K^b-B27/47 were transiently transected with HLA-B27 specific peptide and either ERAP2^N or ERAP2^K expressing plasmids for 48 hours. Surface expression of HLA-B27 in transfected cells was measured by flow cytometry, exactly as described (19). Statistical analysis was completed using one sample t-test [(mean ratio − 1)/SE)] to test whether antigen presentation in ERAP2 transfected cells deviated from mock transfections performed in parallel.

Quantitation of protein expression in HeLa cells

HeLa-K^b-B27/47 cells were transfected with pTracer-CMV-RFP-ERAP2^N (or ERAP2^K). At 24 hours post transfection, protein was harvested in lysis buffer (20mM Tris-HCl, pH 7.4, 1mM EDTA, 150mM NaCl) containing 1% Triton X-100 with protease inhibitors. Lysed samples were then centrifuged at maximum speed (16,000×g) for 15 minutes at 4°C, after which the protein concentration of the supernatant was determined using BCA method. Equivalent concentrations of protein samples (5–20 μg) were run on 8% polyacrylamide gels and transferred onto nitrocellulose membranes. Blots were then probed with the mouse-anti-hTubulin monoclonal primary antibody (Sigma-Aldrich, St. Louis, MO) at 1:30000 dilution for 20 minutes and primary polyclonal mouse-anti-hERAP2 (Abcam, Cambridge, MA) at 1:500 dilution for 3–16 hours. Following incubations, membranes were probed with fluorescent antibodies (IRDYE800 conjugated goat anti-mouse IgG (Rockland, Gilbertsville, PA) for 45 minutes at 1:3300 dilution. Blots were scanned and bands were quantified using Licor’s Odyssey scanner (models 2.1 or 3.0). For data analysis, the fluorescence of ERAP2 bands was normalized to Tubulin bands prior to quantification.

X-ray crystallography

Crystallization of ERAP2^K was performed using conditions identical to the ones described previously for ERAP2^N (15). During the last stages of concentration, DG001 peptide was added in the protein solution at a protein:peptide ratio of 1:4, the mixture was incubated at room temperature for 4 hrs and the sample was further concentrated to 6mg/ml. Single crystals were obtained as for ERAP2^N using the Morpheus (37) protein crystallization screen (Molecular Dimensions Ltd, UK) by the sitting drop vapour diffusion technique: by mixing 200nl of ERAP2^K (6mg/ml in 150mM NaCl, 0.01% NaN₃, 25mM HEPES pH 7.0) with 100 nl of precipitant using an Oryx4 nano-drop crystallization robot and incubating against 50μl reservoir. Initial crystals appeared after 2–3 days. The crystal used for data collection was grown in a 2μl drop at 4°C, using the condition 10% PEG 8000, 20% ethylene glycol, 20mM each of glycine, DL-alanine, DL-serine, DL-lysine, Na-glutamate, 31mM MES and 69mM Imidazole, at pH 6.5. X-ray diffraction data were collected at 100°K using synchrotron radiation at the XO6DA beamline, Swiss Light Source, Paul Scherrer Institut, Villigen, Switzerland. Diffraction data up to 3.2Å resolution were processed by MOSFLM (38) and scaled with the SCALA software (39). Five percent of reflections were flagged for R_free calculations. The crystal is isomorphous with that of ERAP2^N, space group P2₁, a = 74.3 Å, b = 134.4 Å, c = 127.4 Å, and β = 90.8° (Table S1). The structure was determined by the isomorphous replacement method using the ERAP2^N coordinates (PDB ID, 3SE6). Phenix.refine (40) was used for structure refinement imposing 2-fold NCS restraints. All residues, but 16 pairs of residues, were restrained according to NCS. Alternating cycles of restrained refinement and manual fitting/building with Coot (41) resulted in an R factor and R_free of 21.36% and 26.12%, respectively. Residual electron density (more than 3σ) at the active site resembled that of ERAP2^N and was interpreted to belong to a Lysine (Lys_a) and 2-(N-morpholino) ethanesulfonic acid (MES), present at high concentrations in the crystallization mixture(15). A set of three cycles of refinement with Lys_a and MES included resulted in final R factor and R_free of 20.87 % and 26.08 %.

Besides the two crystallographically independent protein molecules, the asymmetric unit includes a total of 14 sugar residues and 106 water molecules. A lysine and a MES molecule were modeled on residual electron density close to the zinc atom of the active site of each protein molecule in the dimer. The carbohydrate moieties were found attached to the following Asn N-glycosylation sites: A85, A103, A219, A405, A650, B85, B219 and B431. All occupancies of protein and carbohydrate atoms as well as water molecules were set to one, except for the disordered amino acid side-chains and the two MES molecules. Two amino acid side-chains in one of the molecules (A) were disordered (Arg366 and Lys839) and refined with two alternative conformations. There was no assignable density before residue 54 of chain A and residue 55 of chain B. Also missing were residues 127–129, 503–527 and 570–580 of chain A and 126–132, 503–531 and 571–582 of chain B. Figures were made using Pymol(42).

Atomic coordinates and structure factors have been deposited in the Protein Data Bank as entry 4E36.

Computational Methods

Continuum electrostatic calculations were performed using the finite difference Poisson–Boltzmann method, implemented in Delphi v5 (43, 44). The individual residue contributions to the electrostatic interaction energies were calculated as described previously (45, 46). Briefly, hydrogen atoms and missing side-chains of the two X-ray structures (ERAP2^K and ERAP2^N) were built and optimized with AMBER v10 (47). Energy minimizations were performed in implicit solvent with the Generalized Born GBn method (48) by applying harmonic constraints of 100 Kcal mol⁻¹ Å⁻² to heavy atoms so that their crystallographic positions were marginally perturbed. Atomic radii and charges were taken from the param22 parameter set of the CHARMM program (49), except for Zn that was assigned a charge of +2.0 and a radius of 1.4Å. The interior of the enzyme and water were modeled with dielectric constants of 4 and 80, respectively, and boundary conditions were approximated by the sum of Debye–Hückel potential of all the charges. The ionic strength of the solution was set to 0.15M, the solvent probe radius to 1.4Å and the ion exclusion radius to 2.0Å. Calculations were performed using the linearized PB equation until the maximum change of the grid potential converged to within 10⁻⁴k_BT, where k_B is the Boltzmann constant and T is the absolute temperature. Each calculation was performed 10 times with the protein translated systematically on the grid, and 10 separate times by varying the percentage fill of a lattice with resolution of 3.0 grids per Å. The reported values are the average of 20 runs with the standard deviation being less than 1%, representing the uncertainty due to the granularity of the grid.

Results

ERAP2^K is less active in hydrolyzing small fluorigenic substrates

Small fluorigenic substrates such as arginyl-7-amido-4-methyl-coumarin (R-AMC) have been used previously to characterize the activity and specificity of M1 aminopeptidases (14). To compare the enzymatic activity of the two ERAP2 variants we measured the rate of hydrolysis of R-AMC by ERAP2^N and ERAP2^K. The two variants had significantly different activity, with R-AMC being hydrolyzed at 78±1 mmol product/mol enzyme sec⁻¹ by ERAP2^N and 4.8±3 mmol/mol sec⁻¹ by ERAP2^K. Such a difference in activity is at least 10-fold higher compared to previously characterized ERAP1 polymorphisms (19, 32). To investigate whether this change in catalytic activity was due to changes in substrate recognition or catalytic turnover rate we performed a standard Michaelis-Menten analysis (Figure 1). According to this analysis the polymorphic state at position 392 of ERAP2 affected substrate affinity by less than 2-fold (ERAP2^N: K_M=30±9μM; ERAP2^K: K_M=17±5μM) but affected catalytic turnover rate to a large degree, lowering the k_cat constant by about 18-fold (ERAP2^N: k_cat=104±9×10⁻³ sec⁻¹; ERAP2^K: k_cat=5.8±0.4×10⁻³ sec⁻¹). This large change in catalytic efficiency brought about by a common polymorphism is, to our knowledge, the largest described for this class of enzymes to date.

Figure 1.

Panel A Michaelis-Menten analysis of R-AMC hydrolysis by ERAP2^N (filled circles) and ERAP2^K (open circles) variants. Panel B: Double-reciprocal plot of the data.

To investigate whether the change in catalytic turnover was substrate specific, we used a previously characterized collection of fluorigenic substrates bearing all natural amino acids as their scissile N-termini (14, 33). Consistent with previous observations on the specificity of ERAP2, both polymorphic states presented a highly similar pattern of N-terminus selectivity with only positively charged side chains processed efficiently (Supplemental Figure 1B). This finding suggests that the N392K polymorphism affects the catalytic efficiency but not the specificity of ERAP2, at least towards these small fluorigenic substrates.

Trimming of antigenic peptide precursors

Since the trimming of small fluorigenic substrates does not adequately describe the in vivo function of the enzyme, we tested the trimming of antigenic peptide precursors by the two ERAP2 alleles. ERAP2^N was able to efficiently trim the leucine residue from the precursor peptide with the sequence LSRHHAFSFR and generate the mature epitope SRHHAFSFR (from the aggrecan protein, a critical component for cartilage structure and function of the joints, normally presented by the HLA-B2705 MHC class I allele) (Figure 2A). This result was rather surprising given the postulated role of ERAP2 in removing positively charged amino acids(9, 14) and may be due to other factors including the recognition of side-chains distal from the N-terminus allowing the processing of peptides with suboptimal N-termini, a property established for ERAP1 (19) (also see below). In contrast to ERAP2^N, ERAP2^K was much less efficient in processing this antigenic peptide precursor and generated the mature epitope with a 37-fold slower rate (Figure 2A,B). This change in trimming rate was not due to substrate recognition since the K_M for this peptide was similar for both alleles (33±9μM vs 19±5μM) as calculated by competition titrations (Figure 2C). We conclude that the difference in peptide processing kinetics is due to a change in the catalytic turnover rate of the enzyme for this particular substrate (Figure 2D).

Figure 2.

Panel A, RP-HPLC chromatograms of products of the enzymatic digestion of peptide LSRHHAFSFR (substrate) by the two ERAP2 variants. The produced epitope SRHHAFSFR (product) is indicated. Two characteristic reaction conditions are shown, using 200ng (14.7nM, molar ratio peptide to enzyme=1361) of enzyme for 15min or 1hr and 2μg (147nM, molar ratio peptide to enzyme=136) enzyme for 1hr. Panel B, epitope generation rate for each ERAP2 variant. Panel C, the rate of hydrolysis by each ERAP2 variant of the fluorigenic substrate R-AMC was plotted versus the concentration of competing substrate LSRHHAFSFR and data were fit to a simple competition model to allow the calculation for the affinity for the competing substrate (57). Panel D, turnover number (k_cat) and specificity constant (k_cat/K_M) for the production of the SRHHAFSFR epitope by the two ERAP2 variants.

ERAP2 has been described previously as an accessory aminopeptidase that “assists” ERAP1 in trimming positively charged residues from antigenic peptide precursors, residues for which ERAP1 has reduced activity (9, 14). To test whether ERAP2 allelic variation affects the trimming of a peptide precursor with a positively charged N-terminus, we measured the kinetics of trimming of a peptide with the sequence KSLYNTVATL, a precursor to the epitope SLYNTVATL (derived from the HIV-1 p17 Gag protein and presented by the HLA-A0201 MHC class I allele). In sharp contrast to the SRHHAFSFR epitope, SLYNTVATL was produced equally effectively by both ERAP2 alleles (Figure 3). This surprising result suggested that the change in catalytic turnover rate between ERAP2^N and ERAP2^K may be substrate-dependent for larger, more physiologically relevant, peptidic substrates.

Figure 3.

Panel A, RP-HPLC chromatograms of production of the SLYNTVATL epitope from a single residue extended precursor carrying a positively charged amino acid (KSLYNTVATL). Panel B, trimming rate of peptide KSLYNTVATL by the two ERAP2 variants. Panel C, R-AMC competition titrations for the KSLYNTVATL peptide as in Figure 2 reveal similar affinity for both ERAP2 variants.

To test the generality of changes in substrate specificity between the two ERAP2 alleles we measured the trimming rates for 8 additional antigenic peptide precursors in vitro. These precursors were designed based on the protein sequence upstream of antigenic peptides that are known to be generated by the action of intracellular aminopeptidases. The N-terminal amino acid was in every case either a positively charged residue or a hydrophobic residue. The trimming rates for all precursors tested are compared in Figure 4. In most cases there was a significant difference between the trimming rates for the two alleles; these differences were much more potent for precursors with a hydrophobic N-terminal amino acid. For example, the precursors RSRYWAIRTR and KSLYNTVATL were processed with almost identical rates by both ERAP2 alleles. In contrast, the precursor WLRRYLENGK was processed with a rate difference of 165-fold (330±34 pmol nmol⁻¹ sec⁻¹ for ERAP2^N versus 2.0±0.04 pmol nmol⁻¹ sec⁻¹ for ERAP2^K). We concluded from this limited screen that the polymorphic variation in position 392 of ERAP2 can affect peptide specificity, primarily for peptides that carry a hydrophobic amino acid as their N-terminus. This is in sharp contrast to what we discovered for small non-peptidic substrates (Figure 1 and Figure S1) in which case we could only demonstrate a difference of catalytic turnover between the two alleles but not in specificity. This apparent paradox may signify fundamental differences between recognition of small versus larger natural substrates. Such a distinction has recently been described for ERAP1 and has been hypothesized to constitute the main mechanism for efficient large peptide trimming necessary for its biological function (18).

Figure 4.

Trimming rates of model epitope precursors by ERAP2^N (black bars) and ERAP2^K (white bars). 20μM of each indicated epitope precursor were incubated with an appropriate amount of ERAP2 to produce a limited amount of mature epitope as described in the methods section. Error bars correspond to standard deviation calculated from 3 separate experiments.

In an effort to isolate the importance of the N-terminal residue of the peptide-substrate in the trimming rate by the two ERAP2 alleles we compared the trimming of identical peptide backbones that only differ in the N-terminal residue. The ratio of the trimming rate of the peptide carrying an N-terminal extension of a positively charged amino acid versus the trimming rate of the same peptide carrying a hydrophobic N-terminal extension, defines a specificity ratio (SR) that indicates the preference of ERAP2 for positive versus hydrophobic N-termini. The calculated SR value for three antigenic epitopes is shown in Figure 5A. In all cases ERAP2^K showed a clear preference for positively charged residues at the N-terminus of the peptide (SR>1, maximal SR=92), although the magnitude of this value depended on the remaining peptidic sequence, indicating that the whole peptide sequence is important for substrate recognition (as has been demonstrated for ERAP1, see reference (17)). Additionally, in all cases, the SR value for ERAP2^N was significantly lower indicating that the strong preference for positively charged N-termini is not shared by this allele that can efficiently process both hydrophobic and positively charged N-termini. These differences in specificity are highly significant and can lead to qualitative differences in trimming between the two alleles. As shown in Figure 5B, trimming of the precursor RSRYWAIRTR is essentially identical for both alleles, whereas trimming of the precursor LSRYWAIRTR (same sequence but for the N-terminus) is drastically different, with ERAP2^K essentially failing to produce detectable amounts of mature epitope under the specific experimental conditions. Overall, these results suggest that the polymorphism at position 392 of ERAP2 can induce changes in specificity when trimming natural peptidic substrates.

Figure 5.

Panel A, specificity index (defined as the trimming rate for a peptide with a positively charged N-terminus divided by the trimming rate for the same peptide carrying a hydrophobic N-terminus) for three different antigenic epitope templates using the two ERAP2 variants. Y-axis is broken to enhance visibility of large numerical differences. Panel B, RP-HPLC chromatograms of trimming of the precursor RSRYWAIRTR by ERAP2^N and ERAP2^K. Notice the similar amount of epitope generation. Panel C, RP-HPLC chromatograms of trimming of the precursor LSRYWAIRTR by ERAP2^N and ERAP2^K. Notice the absence of a visible epitope peak produced by ERAP2^K.

ERAP2^N can supplement ERAP1 mediated trimming of hydrophobic N-termini

ERAP2 has been proposed to act as an accessory aminopeptidase to ERAP1, assisting it in trimming peptide sequences for which ERAP1 has sub-optimal affinity such as peptide carrying positive charged N-terminal amino acids (9). Our finding that ERAP2^N can excise hydrophobic N-termini much more efficiently than ERAP2^K prompted us to investigate how well ERAP2 activity compares to ERAP1. We measured the rate of generation of the SRHHAFSFR epitope from a precursor carrying an N-terminal leucine residue, by ERAP1, ERAP2^N and ERAP2^K (Figure 6A). Since ERAP1 has been reported to also be able to destroy antigenic epitopes, we measured the rate of destruction of the same epitope by quantitating smaller peptides accumulated after treating it with each of the three enzymes (characteristic chromatograms of epitope generation and destruction are shown in Supplemental Figure 2). All three enzymes, generated the epitope faster than they destroyed it, suggesting that they have the inherent ability to accumulate it (13). This effect was more pronounced for ERAP1 and ERAP2^N, since both enzymes generated the epitope at least 100-fold faster than they destroyed it. Of the three enzymes, ERAP1 was by far the most efficient in generating the mature epitope, although ERAP2^N was also competitive. Comparing the relative rate of ERAP1 to ERAP2^N mediated epitope generation revealed that ERAP2^N can account for about 16% of the ERAP1 activity for this peptide whereas the ERAP2^K activity is negligible compared to that of ERAP1 (Figure 6B). Taken together, these findings suggest that ERAP2^N but not ERAP2^K can supplement ERAP1 in trimming hydrophobic N-termini.

Figure 6.

Panel A, , Epitope generation versus epitope destruction: Trimming rates of epitope precursor LSRHHAFSFR (gray bars) and mature epitope SRHHAFSFR (white bars) by ERAP1, ERAP2^N and ERAP2^K. Panel B, ERAP2^N can trim the precursor LSRHHAFSFR with a rate corresponding to ~16% of that of ERAP1, whereas ERAP2^K is much less efficient. Bars represent the ratio of ERAP2/ERAP1 trimming rates for each ERAP2 variant. Panel C, HeLa cells, stably expressing HLA-B27, as well as TAP1 ER-transporter blocker (ICP47), were transfected with an ERAP2 variant and an ER-targeted miniprotein that after signal sequence cleavage gives rise to an HLAB27–specific peptide precursor with the sequence ASRHHAFSFR. HLAB27 cell-surface translocation was followed by flow cytometry using an MHC specific antibody (MIF=Mean Intensity of Fluorescence). Panel D, in vitro trimming rate of the ASRHHAFSFR by ERAP2 alleles.

The substrate specificity of ERAP1 has been shown to correlate well with antigen processing and presentation in cultured cells, suggesting that in vitro specificity differences are relevant to in vivo antigen processing (19, 50). To extend these findings to ERAP2 and its allelic content, we utilized a cell-based antigen presentation assay that has been previously established for ERAP1 (8). We modified this assay to report the effects of ERAP2 allele transfection by co-transfecting with a plasmid coding for the ERAP2 enzyme carrying the desired polymorphism and a mini gene encoding a miniprotein that after signal sequence cleavage gives rise to the antigenic peptide precursor ASRHHAFSFR (technical limitations of this assay necessitate the existence of an N-terminal alanine residue) (Supplemental Figure 3). Removal of the N-terminal alanine residue from this peptide generates the mature epitope that can bind onto the nascent MHC class I allele HLA-B27 constitutively produced by this cell line. We measured the changes in epitope presentation upon transfection of ERAP2^N or ERAP2^K compared to control transfected cells that perform processing only by endogenous ERAP1/2. ERAP2^N was able to enhance SRHHAFSFR epitope generation to a greater degree than ERAP2^K (Figure 6C). In vitro digestion of the same peptide by ERAP2 showed that ERAP2^N generated the mature epitope about 2-fold faster than ERAP2^K, consistent with the enhanced epitope presentation seen in the cell-based assay (Figure 6D). This finding suggests that ERAP2 allelic variation can affect antigen presentation in cells. Interestingly, the relative effect of ERAP2^N is of a magnitude similar to the effects of ERAP1 allelic variation described before using a similar assay, suggesting that ERAP2 allelic variation can be of equal importance regardless of the accessory role of this enzyme (19).

Transition-state analogs exert distinct inhibitory effects on the two ERAP2 alleles

Since the difference in catalytic efficiency between ERAP2^N and ERAP2^K was due to differences in the catalytic turnover of the enzyme (k_cat) suggesting changes in the recognition of the transition state, we hypothesized that such differences may also apply in the recognition of transition state analogs. Phosphinic pseudopeptide transition-state analogs have been described before as potent aminopeptidase inhibitors (51). We tested the ability of a phosphinic pseudopeptide of the sequence L-Ψ{PO₂CH₂}-LAFKARAF (DG001 peptide) carrying a hydrophobic side chain at its N-terminus to inhibit each ERAP2 allele (D. Georgiadis et al., a more thorough characterization of this class of inhibitors will be described elsewhere). This compound carries a phosphinic pseudo-peptide bond between the first two amino acids, resembling the tetrahedral intermediate of the cleavage pathway for the first peptide bond (51, 52). The sequence of this peptide was designed based on known specificity determinants for ERAP1 and ERAP2 (14, 17). The enzymatic hydrolysis of R-AMC by each ERAP2 allele was measured as a function of inhibitor concentration in the solution (Figure 7A). The DG001 peptide was found to inhibit ERAP2 with good affinity but displayed partial inhibition (Figure 7A). Strikingly, the DG001 peptide had a much higher affinity for ERAP2^N (K_i=54±8nM) than for ERAP2^K (K_i=524±107nM) (Figure 7A). This 10-fold difference in inhibitor affinity is consistent with the lower activity of ERAP2^K for substrates with a hydrophobic N-terminus and further supports the hypothesis that the difference between the two alleles lies in the recognition of the transition state. In contrast to these results, inhibition of the two ERAP2 alleles by the substrate analog Amastatin was nearly identical (Figure 7B). This distinction between a transition-state analog and a substrate analog further supports the idea that the differences between the two ERAP2 alleles lie in the mechanism of transition state stabilization and not substrate affinity. Overall, the differential recognition of a transition state analog pseudopeptide by the two ERAP2 alleles indicates differences in the catalytic mechanism and suggests that ERAP2 allelic variation should be taken into account for any pharmacological approaches that involve modulation of ERAP2 trimming activity.

Figure 7.

Inhibition of ERAP2^N and ERAP2^K by a transition-state analog (DG001, panel A) and a substrate analog (Amastatin, panel B). The rate of hydrolysis of the fluorigenic substrate R-AMC was followed in the presence of increasing concentrations of inhibitor for each enzyme. Experimental data were fit to a simple binding model accounting for full or partial inhibition (see materials and methods). The calculated constants of inhibition (K_i) are shown.

Atomic basis for different trimming between the two ERAP2 alleles

In an effort to understand the atomic basis of the functional differences between the two alleles we solved the crystal structure of the ERAP2^K allele at 3.2Å and compared it with the recently determined crystal structure of ERAP2^N (Figure 8 and Supplemental Table 1) (15). Asn392 in ERAP2^N is highly conserved amongst homologous aminopeptidases (Figure 8A). Although the DG001 pseudopeptide was present in the crystallization mixture, the residual electron density was not significantly different than that of ERAP2^N and as a result the residual electron density was interpreted to belong to a Lysine present at high concentrations in the crystallization mixture. The two structures were found to be overall identical with only minor structural deviations (RMSD=0.251Å for homologous residues). Superimposition of key catalytic or specificity residues in ERAP2^K and ERAP2^N is shown in Figure 8. Lys392 in ERAP2^K assumes a distinct conformation compared to the Asn residue in ERAP2^N and approaches the N-terminus of the bound ligand (Lys_a) making interactions with the Zn-coordinating Glu393 residue and residues Glu337 and Glu200, both key residues that stabilize the N-terminus of the substrate (distances 2.6, 2.9 and 2.6Å, respectively, Figure 8B). The NZ-atom of Lys392 in ERAP2^K is located in the position of a water molecule in ERAP2^N and is very close (3.7 Å) to the N-terminus of the bound ligand (Lys_a). Computational analysis of the electrostatic free energy of interaction of the Lys_a with ERAP2 suggests that the ERAP2^K allele introduces a highly unfavorable electrostatic interaction between Lys392 and the N-terminus of the bound Lys_a (Figure 8C). This unfavorable interaction may interfere with transition-state stabilization resulting in reduced catalytic efficiency for ERAP2^K.

Figure 8.

Crystal structure of ERAP2^K allele suggests the structural basis for changes in activity. Panel A, sequence alignment of ERAP2 and homologous aminopeptidases showing conservation of the polymorphic residue 392 (in bold); the adjacent aminopeptidase motif (HELAH) is also indicated. Panel B, key catalytic residues in the ERAP2^K structure are shown in stick representation and superimposed with equivalent residues in ERAP2^N (3SE6). Note the relative positioning of Lys392 with respect to the catalytic Glu residues as well as the N-terminus of the bound ligand (Lys_a). Electron density 2|F_o|−|F_c| at 2σ is indicated around Lys392. Panel C, pairwise electrostatic interaction energies (ΔE_inter in Kcal/mol) calculated between the polymorphic residue 392 and the Lys_a ligand. The reported values are the average of the 20 runs with a standard deviation of less than 1%, which represents the uncertainty due to the granularity of the grid. Panel D, superimposition of ERAP2 residues that cap the S1 specificity pocket of the enzyme. |F_o|−|F_c| electron density (at 2.5σ) calculated in the absence of the ligand is shown around Lys_a. 2|F_o|−|F_c| electron density at 2σ is indicated around Glu177.

Structural differences between ERAP2^N and ERAP2^K were also evident in the region capping the S1 pocket (Figure 8D): (a) the NZ-atom of Lys_a forms a strong salt bridge with the carboxylic group of Asp198 (2.8Å) and (b) Glu177 assumes a different conformation, away from Lys_a, interacting strongly via H-bonding (2.6Å) with the carboxylic group of Asp888 of domain IV; a water molecule is found close to the previous location of the side chain of Glu177, H-bonding both to NZ of Lys_a and Asp198. These observations suggest subtle changes in the stabilization/recognition properties of the S1 pocket that may underlie changes in specificity. Furthermore, the observed interaction between Glu177 and Asp888 observed only in the case of ERAP2^K, suggests differences in the interactions between domain II and IV. Reorientation of the relative positions of domains II and IV have been previously reported to be crucial for the catalytic mechanism of the highly homologous ERAP1(18).

Discussion

Two main ERAP2 SNPs have been associated with disease and proposed to be the result of balancing selection possibly through host-pathogen interactions (28, 31). SNP rs2248374 is a “loss of function” variant, since it leads to RNA instability and loss of ERAP2 expression(31). This variant leads to a clear functional defect, namely reduced surface MHC class I expression in B cells, but in contrast to ERAP1 SNPs is not associated with the inflammatory disease Ankylosing spondylitis (53). We show here, that the second ERAP2 SNP, rs2549782, which codes for a single amino acid switch near the enzyme’s active site, is not a “loss of function” variant but rather a “change in function” variant that leads to substrate-specific changes in enzymatic activity, essentially allowing the enzyme to alter its specificity profile for peptidic substrates. This unique SNP-related functional change along with the association of this SNP with several human diseases, most notably with resistance to HIV infection, supports the hypothesis that it has arisen by host-pathogen balancing selection and is a component of the immune system variability in natural populations.

The functional consequences of the ERAP2 allelic variation described here come in sharp contrast to the effects described for coding SNPs in ERAP1, which have been of much lower magnitude (19, 27, 32). It may, at first view, be difficult to understand how such large differences in activity between ERAP2 alleles are so common in the population (the two ERAP2 alleles are almost equally represented in the human population (28)). It is possible that the proposed “accessory” nature of ERAP2-mediated antigen processing can account for this apparent discrepancy. If indeed ERAP1-dependent antigen processing is the dominant activity in the cells, large changes in ERAP2 activity may be easier to tolerate without severely compromising the function of adaptive immunity. Interestingly, the magnitude of antigen presentation changes we report in the cell-based antigen presentation assay is similar between ERAP1 SNPs (as reported in (19)) and between the two ERAP2 variants analyzed here. The accessory role of ERAP2 may also present the immune system with an opportunity for evolving new specificities that can modify the antigenic peptide repertoire without severely damaging antigen processing. However, since disease-related immune responses can be dominated by a very small number of antigenic epitopes, it is possible that even an accessory aminopeptidase activity, such as ERAP2, can be dominant in some cases. Unfortunately, the lack of ERAP2 in rodents (although ERAP2 has been hypothesized to have been present in a primate-rodent common ancestor) has limited in vivo experimentation efforts regarding the role of ERAP2 in adaptive immunity in humans.

Modifications of enzyme activity or specificity by mutation can be a useful evolutionary tool for altering metabolic or regulatory pathways and cancer cells have been shown to utilize this strategy to gain growth advantages (54). In a notable example of such a change in function, histone methyltransferase EZH2 mutants increase histone methylation in human lymphomas (55). Similar change in function mutations in components of the adaptive immune response may be the result of positive selection of individuals that present enhanced responses to specific pathogens during human development. HIV resistance of individuals with a 392Lys/Lys ERAP2 phenotype could represent an example of such positive selection. In this context, the continuously evolving human adaptive immunity can enhance its polymorphic responses to extend away from the well-established variability in MHC binding specificity for antigenic peptides into cellular pathways that are responsible for the generation of those peptides.

Although biochemical and structural analysis suggest a specific mechanism of reduced activity of the ERAP2^K allele though reduced stabilization of the N-terminus of the peptide, it is less clear how these effects lead to altered specificity. One possibility is that the recognition of substrates by the S1 specificity pocket may be different between the two alleles. ERAP2 recognizes in its S1 pocket positively charged side chains (specifically Lys and Arg) by forming strong electrostatic interactions with residue Asp198 (15, 56). Comparison of the local environment of the S1 specificity pocket of the two ERAP2 variants shows changes in atomic interactions that could affect the energetics of substrate catalysis: (a) The NZ-atom of the bound Lys_a in ERAP2^K is coming to a closer proximity to Asp198 compared to ERAP2^N by 0.7Å, (b) Glu177 assumes a different conformation, away from Lys_a, interacting strongly via H-bonding (2.5 Å) with the carboxylic group of Asp888 leading to a novel interaction with domain IV and (c) a water molecule is found, close to the previous location of the side chain of Glu177, to interact with Asp198 and Lys_a via H-bonds (2.8 and 3.0 Å, respectively). Although the importance of Glu177-Asp888 interaction is difficult to evaluate in terms of substrate recognition or catalytic efficiency of the enzyme, it should be noted that structural rearrangements between domains II and IV have been demonstrated to be important for catalysis in the homologous ERAP1 (18). An alternative or complementary explanation may lie in the sub-optimal recognition of hydrophobic side-chains by S1 pocket or ERAP2 that leads to unfavorable interactions(14). It is possible that these unfavorable interactions can lead to altered transition-state stabilization in the case of ERAP2^K (in which the stabilization of the N-terminus is already reduced), making the enzyme more sensitive to optimal S1 pocket occupation and as a result more specific. Interestingly, ERAP2^K is not the only M1 aminopeptidase with a lysine at position 392; the homologous enzyme aminopeptidase N also has a lysine residue at that location, suggesting that this may be a more general strategy for controlling the stringency of S1 specificity in this family of aminopeptidases (51). The observation, however, that the changes in specificity only appear for the larger, physiologically relevant, peptides and not for small fluorigenic substrates suggests differences in the mechanism of recognition between the two types of substrates. A similar phenomenon has also been described for the highly homologous ERAP1 (17) (18).

In summary, we demonstrate that a common coding polymorphism in the antigen processing aminopeptidase ERAP2 significantly alters its trimming function and provide atomic-level insights on the mechanism behind this effect. The nature and magnitude of the observed functional changes in combination with the wide genetic distribution of this variation suggest that this polymorphic variation constitutes an integral part of the variability of antigen processing in individuals, a variability that can complement the well-established variability in antigen binding and presentation by MHC class I alleles. We furthermore propose that ERAP2^N and ERAP2^K are treated as distinct aminopeptidase activities in antigen processing when evaluating epitope generation in humans. The importance of polymorphic variation in gene products that regulate the antigen processing and presentation pathway is only now emerging as a key component of the natural variability of the adaptive immune response. A systematic analysis of the functional consequences of these naturally occurring polymorphisms in key components of this pathway can be a valuable tool in both establishing diagnostic tools for disease predisposition and for developing individualized immunotherapies.