Modulation of natural HLA-B*27:05 ligandome by ankylosing spondylitis-associated endoplasmic reticulum aminopeptidase 2 (ERAP2)

The human leucocyte antigen (HLA)-B*27:05 allele and the endoplasmic reticulum-resident aminopeptidases are strongly associated with ankylosing spondylitis (AS), a chronic inflammatory spondyloarthropathy. This study examined the effect of endoplasmic reticulum aminopeptidase 2 (ERAP2) in the generation of the natural HLA-B*27:05 ligandome in live cells. Complexes of HLA-B*27:05-bound peptide pools were isolated from human ERAP2-edited cell clones and the peptides were identified using high throughput mass spectrometry analyses. The relative abundance of thousand ligands was established by quantitative tandem mass spectrometry and bioinformatics analysis. The residue frequencies at different peptide position, identified in presence or absence of ERAP2, determined structural features of ligands and their interactions with specific pockets of antigen binding site of HLA-B*27:05 molecule. Sequence alignment of ligands identified with species of bacteria associated with HLA-B*27-dependent reactive arthritis was performed. In the absence of ERAP2, peptides with N-terminal basic residues, and minority canonical P2 residues are enriched in the natural ligandome. Further, alterations of residue frequencies and hydrophobicity profile at P3, P7, and P  positions were detected. In addition, several ERAP2-dependent cellular peptides were highly similar to protein sequences of arthritogenic bacteria, including one human HLA-B*27:05 ligand fully conserved in a protein from Campylobacter jejuni. These findings highlight the pathogenic role of this aminopeptidase in the triggering of AS autoimmune disease. the HLA-B*27:05 ligands were assessed by multiple t tests. P values < 0.05 were considered to be statistically significant.


INTRODUCTION
Self and pathogenic proteins are proteolytically degraded by proteasomes and other peptidases in the cytosol to generate an extremely diverse pool of peptides, both in sequence and in length. The transporter associated with antigen-processing (TAP) molecules specifically translocates some of these degradation products to the endoplasmic reticulum lumen (1). The peptides with the correct length (8 to 11 residues) and interactions with specific pockets of the antigen recognition site of the HLA class I molecule (2), usually at position 2 (P2) as anchor residue and auxiliary residues at the C-terminus and/or other positions of the peptide (3,4) can stabilize their direct binding to HLA class I molecules. The fraction of amino-terminally extended precursors can be also utilized for antigen presentation after precursor editing and customization by endoplasmic reticulum (ER)-resident aminopeptidase activities. In humans, two related aminopeptidases, ERAP1 (5) and ERAP2 (6,7), with nonredundant specificities and different substrate preferences trims N-terminally extended residues of these peptide precursors to their final length. Of both, ERAP1 has a wider specificity since, with the exception of Pro, cleaves virtually all N-terminal residues (8).
In contrast, basic residues are the preferential target of ERAP2 activity (6,(9)(10)(11). The binding of peptide to HLA class I molecule in the ER stabilizes the nascent trimolecular peptide-HLA-β 2 -microglobulin complexes and allows for their subsequent transport to the cell membrane, where they are exposed to CD8 + cytotoxic T lymphocyte (CTL) activity (12). The recognition of foreign or self-peptide ligands can lead to the beneficial killing of pathogen-infected cells or, instead, to initiate an autoimmune damage, respectively.
HLA-B*27 is an interesting HLA class I allele, whose prevalence varies significantly among different subpopulations, and it is more common in Caucasoids (http://www.allelefrequencies.net/). HLA-B*27 is strongly associated with ankylosing spondylitis (AS) (13), a chronic arthritis that cause inflammation, primarily of the spinal by guest on  https://www.mcponline.org Downloaded from joints although hip or shoulder joints can also be involved (14). Even though most of the HLA-B*27 subtypes are strongly associated with this chronic inflammatory spondyloarthropathy, HLA-B*27:06 and -B*27:09 subtypes are either not or perhaps only very weakly associated with this autoimmune disease, respectively (reviewed in (15)). This fact suggests that the polymorphism of HLA-B*27 subtypes modulates disease susceptibility. Previous, unambiguous functional distinction between the closely related AS-associated B*2704 and non-AS-associated B*2706 subtypes was found (16). However, after forty years of scientific effort, the bases for this association remain largely unknown. Several hypotheses, based on different features of HLA-B*27, have been previously proposed to explain this intriguing association. However, to this day, none of them has reasonably explained the mechanism and the differential association between HLA-B*27 subtypes and the AS disease.
The classical arthritogenic peptide hypothesis (17) is focused in the antigen presenting properties of HLA-B*27. In a bacterial infection, microbial peptide epitopes bound to HLA-B*27 elicit a normal CTL response against the pathogen. This hypothesis assumes that some of these antimicrobial effector T cells would cross-react with autologous self-ligands, which are also presented by this HLA class I molecule, peptides showing a molecular mimicry with the primary bacterial epitope. Thus, this undesired cross-reaction triggers both autoimmune tissue injury and inflammation.
Under this assumption, variations in HLA binding or immune recognition between mimetic self-peptides and bacterial epitopes with the HLA-B*27 subtypes would explain their differential association with AS. In the last years, the two endoplasmic reticulum aminopeptidases involved in the trimming of peptides for HLA class I antigen presentation, ERAP1 and ERAP2, have been described as the main non-HLA susceptibility genes for AS (18)(19)(20)(21) and thus, the interest on the arthritogenic peptide hypothesis has been highly renewed. ERAP1 is expressed in all individuals and shows a significant degree of polymorphism, which affects the activity and / or level of enzyme by guest on May 4, 2020 expression (22). Although the polymorphism of ERAP2 seems to be very limited, a differential splicing promotes nonsense-mediated RNA decay and thus, about 25% of individuals fail to express this aminopeptidase (23). The highly active ERAP1 variants and the ERAP2 expression favor AS, while the less active variants of ERAP1 and the absence of ERAP2 are protective (24). Similarly to AS, the expression of functional ERAP2 with other different HLA class I molecule: HLA-A*29 is also associated to the birdshot chorioretinopathy, an eye-specific autoinflammatory disorder (25). Moreover, the role of ERAP2 in association with HLA-C*06 is more controversial in the triggering of the psoriasis, other immune-mediated disease (22).
Previously, only two studies using unrelated ERAP2-positive and -negative lymphoblastoid cell lines (LCLs) tried to determine the effect of ERAP2 protease on the generation of HLA-B*27 peptide ligands (26,27). Moreover, although a rigorous pairwise comparison between both types of cell lines were carried (26,27), these LCLs may vary in the expression of many proteins, including the related with antigen processing and thus, the possibility of background differences cannot be formally ruled out. Thus, to elucidate conclusively the role of this aminopeptidase in the generation of HLA-B*27 ligands, high throughput mass spectrometry analyses isolated from ERAP2edited cell clones were carried out in this study. In the unedited and edited-ERAP2 isogenic cell clones, all the antigen presentation machinery and the rest of cellular proteins are common with the exception of the deleted ERAP2 protein. Our results determine that ERAP2 select HLA-B*27:05 ligands with a higher contribution of P2 anchor motif and lower influence of auxiliary residues. and Int_Rv (ACTTGTTACGTGCCTAGACCT). Transfected cells were cloned and the ERAP2 expression was analyzed by PCR and Western Blot as previously described (31). Two unedited (UN17, and UN20 with the complete ERAP2 sequence) and two homozygously edited (KO9, and KO19 without the ERAP2 exons 5 and 6) cell clones were selected.
by guest on May 4, 2020

Electrospray-ion trap mass spectrometry analysis
Peptide mixtures were desalted using OMIX Tips (C 18 , Agilent Technologies, Santa Clara, CA) (33,34), and were analyzed by nanoLC-MS/MS using a Q-Exactive-Plus mass spectrometer fitted with an Ultimate 3000 RSLC nanocapillary UHPLC (Thermo Fisher Scientific), using the same parameters previously described (35). The The effect of ERAP2 depletion on the HLA-B*27:05 peptidome was analyzed by quantitative label-free MS from two unedited (with the complete ERAP2 sequence) and two homozygously edited (without the ERAP2 exons 5 and 6) cell clones. Three independent preparations of the HLA-B*27:05-bound peptides were obtained from each of the four cell lines in this study, and used as biological replicates, whose reproducibility was assessed by Pearson correlation analyses ( Figure S1). A precolumn was utilized to remove non-specific binding proteins and peptides.
Additionally, a 0.01 false discovery rate (FDR) was utilized.
Since a majority of the identified peptides were found in both the UN and KO cells ( Figure S2), quantitative differences in peptide amounts between cell lines were assigned in pairwise comparisons as follows (38) the HLA-B*27:05 ligands were assessed by multiple t tests. P values < 0.05 were considered to be statistically significant.

RESULTS
Homozygous ERAP2-deficient cell clones were generated by deleting a 4741bp fragment of the ERAP2 gene, including the sequence encoding the active site of ERAP2 from B27-C1R cell line (Fig. 1A). The presence or absence of ERAP2 exons 5 and 6 and protein expression in unedited control (UN) or KO cells was confirmed by PCR analysis (Fig. 1B) and immunoblotting (Fig. 1C) peptides were identified in UN17 and UN20 cell clones, respectively (Table S1).
Similarly, 6077 peptides, and 4170 peptides were identified bound to HLA-B*27:05 in KO9 and KO19 cell clones, respectively (Table S1). The anchor motifs for HLA-B*27:05 binding: Arg, Gln or Lys at position 2 (P2) were detected for peptides from both UN and KO cell clones identified by mass spectrometry (Table S1). In silico binding prediction differences between the HLA-B*27:05 ligands from UN and KO were not found ( Figure S3).

Global effect of ERAP2 on the HLA-B*27:05 ligandome
To analyze the influence of ERAP2 on the relative amounts of B  (38). In addition, the specific peptides only identified in UN or KO clones were included in their respective IR > 1.5 subgroup. No differences in MW were found when 1.0 < IR ≤ 1.5 subgroups from UN and KO cell clones were compared (Fig. 2, left panel). In  (Fig. 3). In addition, a simultaneous and statistically significant decrease in the frequencies of Ile, Ser, and Val was also detected in absence of ERAP2 (Fig. 3). peptide position were found in absence of ERAP2 (Fig. 3B).
In addition, LC-MS signal intensities of shared ligands with Gln and Lys at anchor motif P2 peptide position were statistically stronger in absence of ERAP2 relative to the same peptides identified in presence of this aminopeptidase when compared with shared ligands with Arg at P2 position (Fig. 4). No differences were found between shared ligands with Gln and Lys at the anchor motif position when peptide LC-MS signal intensities from UN and KO cell clones were compared (Fig. 4). frequencies at P7, Phe was increased whereas Asp, Gly, and Glu were decreased in the ERAP2-deficient cell clones (Fig. 3D). Finally, Leu was increased and concomitant

Qualitative differences in residue frequencies
Arg was decreased at Pposition when ERAP2 is absent (Fig. 3E). Ppeptide position were found in absence of ERAP2 versus ERAP2-positive cell clones (Fig. S4).

Similarity of ERAP2-dependent HLA-B*27:05 ligands with pathogenic bacterial sequences
In addition, in the search of homology between the HLA-B*27:05 ligands identified in the current study with the bacteria associated with HLA-B*27-dependent reactive arthritis, other multiple bacteria peptides were highly homologs (with only one or two amino acid changes) with several ERAP2-dependent cellular peptides ( Table 2).
The higher and lower conservation rates between HLA human ligands and bacteria sequences were found with Yersinia enterocolitica and Chlamydia trachomatis, respectively ( Table 2) (26,27). Also, an enrichment of peptides with N-terminal basic residues in absence of ERAP2 was detected in the mass spectrometry analyses using ERAP2-edited cell clones as was previously described using ERAP2-positive andnegative LCLs (26,27). by guest on  The interaction between the B pocket of HLA-B*27:05 and the P2 peptide position is the major contribution to B*27:05 binding and thus, this position was defined as anchor motif (SYFPEITHI database: http://www.syfpeithi.de (4)). Arg is overwhelmingly the amino acid most utilized at this P2 peptide position, with residual presence of Gln (42)(43)(44), and Lys (45,46) at this anchor position. Previously, computational simulations suggested a similar binding conformation between P2-Arg and P2-Lys peptides to the B pocket of the HLA-B*27:05 molecules, but slightly different from the P2-Gln peptides (46). In this modeling, the hydrogen bond profiles were energetically and geometrically enhanced for Arg at P2 position relative to Gln or Lys amino acids (46). In addition, binding analysis with the same peptide sequence and different changes at P2 position concluded that the HLA affinity was substantially lower for peptides with Lys or Gln as anchor motif than the respective P2-Arg ligands (46). In the present study, we found that the absence of ERAP2 increases, in both quantity and peptide hypothesis. The cellular peptides specifically generated by ERAP2, identified herein, may be tested as candidates to be the arthritogenic peptides cross-recognized by antibacterial T cells. Another possibility will be that quantitative differences by the ERAP2 influence detected by us in multiple HLA-B*27 ligands may alter the equilibrium between immunogenicity and tolerance. Such misbalance would trigger the autoimmune T cell response, expanding the universe of peptides to be tested as possible arthritogenic peptides.
In this context, since no arthritogenic peptides have been previously described, the human HLA-B*27:05 ligands akin to bacterial protein segments reported in the current study are interesting, and especially those fully conserved with arthritogenic bacteria as the IRPAINVGL ligand since the leakage of T cell tolerance after infection was previously described (48)(49)(50). In addition, it is also striking that the bacteria protein pathways activation, and later the inflammatory response (51), and the presence of HLA-B*27 heavy chain homodimers at the cell surface (52), which can be recognized by NK cells or a subset of Th17 cells (53). In these last two hypothesis no contribution of ERAP1 and/or ERAP2 aminopeptidases is necessary to trigger the AS disease. In addition, recently similar levels of HLA-B*27 heavy chain homodimers for ASassociated and non-AS-associated subtypes were described (54). In contrast, the role of ERAP2 in AS would be more compatible with the hypothesis of the cross-reactive peptide epitopes, involving this enzyme with either the generation of pathogenic peptide(s) or the destruction of putative protective epitope one(s). Then, on the basis of our observations, the significant destruction of peptides with basic residues in P1 peptide position could have a direct contribution to the risk of AS, activating T cells or breaking tolerance. This mechanism, unlike the other two proposed, would also explain that other HLA-B alleles have been linked to AS (55).
Finally, the complexity of ERAP1 haplotypes, with influence of individual mutations (56) and the combined effects of co-occurring polymorphisms (57) is also relevant in the risk of AS. Thereby, Hap1 to Hap3 (but not Hap8 or Hap10) are ASassociated haplotypes (24). Therefore, future studies analyzing the combined effect of ERAP2 and the ERAP1 haplotypes associated with AS must be carried out to target more specifically the possible set of arthritogenic peptides.  The peptides from UN or KO cell clones were subdivided based on their ion peak intensity relative (IR) to the following conditions: 1.0 < IR ≤ 1.5 (panel A) and IR > 1.5 (panel B) and the equivalent subgroups were compared as previously described (38).
The intensity assigned to each peptide in a cell line was normalized by dividing its      by guest on May 4, 2020