Myelin Basic Protein Fragmentation by Engineered Human Proteasomes with Different Catalytic Phenotypes Revealed Direct Peptide Ligands of MS-Associated and Protective HLA Class I Molecules

Proteasomes exist in mammalian cells in multiple combinatorial variants due to the diverse regulatory particles and exchange of catalytic subunits. Here, using biotin carboxyl carrier domain of transcarboxylase from Propionibacterium shermanii fused with different proteasome subunits of catalytic and regulatory particles, we report comprehensive characterization of highly homogenous one-step purified human constitutive and immune 20S and 26S/30S proteasomes. Hydrolysis of a multiple sclerosis (MS) autoantigen, myelin basic protein (MBP), by engineered human proteasomes with different catalytic phenotypes, revealed that peptides which may be directly loaded on the HLA class I molecules are produced mainly by immunoproteasomes. We detected at least five MBP immunodominant core regions, namely, LPRHRDTGIL, SLPQKSHGR, QDENPVVHFF, KGRGLSLSRF and GYGGRASDY. All peptides, except QDENPVVHFF, which originates from the encephalitogenic MBP part, were associated with HLA I alleles considered to increase MS risk. Prediction of the affinity of HLA class I to this peptide demonstrated that MS-protective HLA-A*44 and -B*35 molecules are high-affinity binders, whereas MS-associated HLA-A*23, -A*24, -A*26 and -B*51 molecules tend to have moderate to low affinity. The HLA-A*44 molecules may bind QDENPVVHFF and its deamidated form in several registers with unprecedently high affinity, probably linking its distinct protective phenotype with thymic depletion of the repertoire of autoreactive cytotoxic T cells or induction of CD8+ regulatory T cells, specific to the encephalitogenic MBP peptide.


Introduction
The ubiquitin-proteasome system and the autophagy-lysosome system are essential intracellular machinery [1,2]. The proteasome is assembled from a catalytic core particle, also termed the 20S proteasome, which is capped by different regulatory particles. Three

Purification and Characterization of the HTBH-Engineered Proteasomes
The DNA constructs coding for HTBH-tagged [18][19][20] proteasome subunits β5 (PSMB5), b5i (PSMB8), β7 (PSMB4) and Rpn11 (PSMD14) (Figure 1a) were integrated into the genomes of HEK293T and HeLa cells using the Sleeping Beauty transposon system [21]. The β5 and β5i subunits are responsible for the chymotrypsin-like activity of the constitutive proteasome and immunoproteasome, respectively. Proteasome purification from HEK293T and HeLa cells with overexpressed β5i and β5, respectively, resulted in extremely low yields of the complex, and these variants were excluded from further experiments (data not shown). β7 is a non-catalytic subunit of the 20S inner β-ring, and Rpn11 is a subunit of the 19S regulatory particle with deubiquitinase activity. The HTBH tag consists of the tobacco etch virus nuclear-inclusion-a endopeptidase (TEV) site and the biotin carboxyl carrier domain of transcarboxylase from Propionibacterium shermanii, flanked with two 6-histidine clusters.
Mammalian cells stably expressing HTBH-tagged proteasome subunits (Figure 1b) were lysed in the presence or absence of ATP, and further protein lysates were incubated with streptavidin beads (Figure 1c). Washing buffer for samples without ATP was supplemented with 0.75 M NaCl to dissociate 20S and 19S particles. Purified proteasomes were analyzed by gel electrophoresis utilizing Coomassie and Sypro Orange staining, in-gel fluorescence with Me 4 BodipyFL-Ahx 3 Leu 3 VS fluorescent proteasome probe (UbiQ18) [22] and native gel (Figure 2a). Mammalian cells stably expressing HTBH-tagged proteasome subunits ( Figure 1b) were lysed in the presence or absence of ATP, and further protein lysates were incubated with streptavidin beads (Figure 1c). Washing buffer for samples without ATP was supplemented with 0.75 M NaCl to dissociate 20S and 19S particles. Purified proteasomes were analyzed by gel electrophoresis utilizing Coomassie and Sypro Orange staining, ingel fluorescence with Me4BodipyFL-Ahx3Leu3VS fluorescent proteasome probe (UbiQ18) [22] and native gel (Figure 2a).  The 20S and 26S proteasomes (Figure 2a,b) purified from HEK293T cells were analyzed in terms of activation by sodium dodecyl sulfate (SDS) in the presence or absence of the 100 mM KCl (Figure 2c). The 20S proteasomes isolated via HTBH-tagged β5 and β7 subunits had similar activation profiles with maxima at SDS concentrations of 0.005-0.01%. The presence of the 100 mM KCl in the reaction buffer resulted in an unsaturated activation curve. Moreover, in the presence of 100 mM KCl, dilution of 20S samples resulted in an enhanced activation rate.
The HeLa cells were exposed to 400 U/mL of IFNγ for 72 h to induce expression of the proteasome catalytic immunosubunits (Figure 3a,b). Ratio of chymotryptic to caspase activity of 26S proteasomes, isolated from cells exposed to IFNγ, increased approximately three times in comparison with non-treated cells (Figure 3c). Despite a similar immunophenotype of the 20S and 26S proteasomes from IFNγ-treated HeLa cells according to The HeLa cells were exposed to 400 U/mL of IFNγ for 72 h to induce expression of the proteasome catalytic immunosubunits (Figure 3a,b). Ratio of chymotryptic to caspase activity of 26S proteasomes, isolated from cells exposed to IFNγ, increased approximately three times in comparison with non-treated cells (Figure 3c). Despite a similar immunophenotype of the 20S and 26S proteasomes from IFNγ-treated HeLa cells according to fluorescent proteasome probe (Figure 3b), the chymotryptic to caspase activity ratio of the 20S proteasomes increased only 20-30% upon IFNγ addition (Figure 3d). fluorescent proteasome probe (Figure 3b), the chymotryptic to caspase activity ratio of the 20S proteasomes increased only 20-30% upon IFNγ addition (Figure 3d). Analysis of the proteasomes purified with or without ATP from cells exposed or not exposed to IFNγ. From top to bottom: SDS-PAGE stained Coomassie and Sypro Orange, in-gel fluorescence staining using Me4BodipyFL-Ahx3Leu3VS fluorescent proteasome probe (UbiQ18). (c) Analysis of chymotryptic (blue), tryptic (deep blue) and caspase-like (violet) activities of proteasomes isolated from HeLa cells. (d) Ratio of chymotryptic to caspase activity of proteasomes, isolated from HeLa cells and HEK293T cells in the absence of ATP not exposed (green) and exposed to IFNγ (aquamarine) or in the presence of ATP not exposed (blue) and exposed to IFNγ (deep blue). Data are shown as averages and standard deviations.

Fragmentation of Myelin Basic Protein by Proteasomes with Different Catalytic Phenotypes
In order to obtain mostly homogenous immunoproteasomes, we increased exposure of HeLa cells to IFNγ to up to 96 h and further isolated proteasomes with different catalytic phenotypes from HEK293T, HeLa and HeLa cells subjected to IFNγ treatment (Figure 4a). As HeLa cells exposed to IFNγ still contain constitutive proteasomes, this sample represented a mixture of immuno-and constitutive proteasomes, predominantly containing immunoproteasomes. Further, myelin basic protein (MBP) was incubated with purified proteasomes, and hydrolyzates were next analyzed using LS-MS/MS (Figure 4b-e). As we previously showed, MBP can be hydrolyzed by both 26S and 20S proteasomes, at least in vitro [12,14]. Although the 20S subparticle is an integral part of the 26S holoenzyme, it is also quite abundant as a free complex in many cell types [23], especially under stress conditions [24]. Thus, we analyzed equimolar mixture of MBP hydrolysates generated by 20S and 26S proteasomes from HEK293T cells, HeLa cells and HeLa cells exposed Analysis of the proteasomes purified with or without ATP from cells exposed or not exposed to IFNγ. From top to bottom: SDS-PAGE stained Coomassie and Sypro Orange, in-gel fluorescence staining using Me 4 BodipyFL-Ahx 3 Leu 3 VS fluorescent proteasome probe (UbiQ18). (c) Analysis of chymotryptic (blue), tryptic (deep blue) and caspase-like (violet) activities of proteasomes isolated from HeLa cells. (d) Ratio of chymotryptic to caspase activity of proteasomes, isolated from HeLa cells and HEK293T cells in the absence of ATP not exposed (green) and exposed to IFNγ (aquamarine) or in the presence of ATP not exposed (blue) and exposed to IFNγ (deep blue). Data are shown as averages and standard deviations.

Fragmentation of Myelin Basic Protein by Proteasomes with Different Catalytic Phenotypes
In order to obtain mostly homogenous immunoproteasomes, we increased exposure of HeLa cells to IFNγ to up to 96 h and further isolated proteasomes with different catalytic phenotypes from HEK293T, HeLa and HeLa cells subjected to IFNγ treatment ( Figure 4a). As HeLa cells exposed to IFNγ still contain constitutive proteasomes, this sample represented a mixture of immuno-and constitutive proteasomes, predominantly containing immunoproteasomes. Further, myelin basic protein (MBP) was incubated with purified proteasomes, and hydrolyzates were next analyzed using LS-MS/MS (Figure 4b-e). As we previously showed, MBP can be hydrolyzed by both 26S and 20S proteasomes, at least in vitro [12,14]. Although the 20S subparticle is an integral part of the 26S holoenzyme, it is also quite abundant as a free complex in many cell types [23], especially under stress conditions [24]. Thus, we analyzed equimolar mixture of MBP hydrolysates generated by 20S and 26S proteasomes from HEK293T cells, HeLa cells and HeLa cells exposed to IFNγ. Proteasomes purified from HeLa cells exposed to IFNγ, as anticipated, demonstrated more chymotryptic and less caspase and tryptic phenotype (Figure 4b). All proteasomes revealed more sites in the middle part of the MBP sequence (Figure 4c), as judged by the absolute peptide count. Pairwise comparison showed similar patterns for proteasomes from HEK293T and HeLa cells, whereas proteasomes isolated from HeLa cells exposed to IFNγ demonstrated a significant change in spectrum of the generated MBP peptides (Figure 4d). Detailed analysis of 9-10-amino-acid peptides generated by proteasomes purified from HeLa cells exposed to IFNγ-and HEK293T proteasomes revealed a significant increase in the amount of immunodominant peptides upon MBP hydrolysis by proteasomes purified from HeLa cells exposed to IFNγ (Figure 4e). judged by the absolute peptide count. Pairwise comparison showed similar patterns for proteasomes from HEK293T and HeLa cells, whereas proteasomes isolated from HeLa cells exposed to IFNγ demonstrated a significant change in spectrum of the generated MBP peptides (Figure 4d). Detailed analysis of 9-10-amino-acid peptides generated by proteasomes purified from HeLa cells exposed to IFNγ-and HEK293T proteasomes revealed a significant increase in the amount of immunodominant peptides upon MBP hydrolysis by proteasomes purified from HeLa cells exposed to IFNγ (Figure 4e). (a) Analysis of the proteasomes purified with or without ATP from cells exposed or not exposed to IFNγ, in-gel fluorescence staining using Me4BodipyFL-Ahx3Leu3VS (UbiQ18) fluorescent proteasome probe. (b) N-and C-terminal amino acids of peptides generated by proteasomes purified from HEK293T cells and HeLa cells exposed to IFNγ during MBP hydrolysis. The numbers of peptides with the corresponding combinations of first and last amino acids were counted and normalized to the total number of MBP peptides. (c) Fragmentation of MBP (P02686-5 isomer) by proteasomes from HEK293T cells and HeLa cells exposed to IFNγ. The hydrolysis sites were counted and normalized to the total number of hydrolysis events. (d) All MBP peptides generated by proteasomes purified from HeLa cells exposed to IFNγ versus HEK293T proteasomes (left plot) and proteasomes isolated from HeLa cells not exposed to IFNγ and HEK293T cells (middle plot); overlap (right). (e) The 9-10-amino-acid MBP peptides generated by proteasomes purified from cells exposed to IFNγ versus HEK293T proteasomes (left plot), proteasomes isolated from HeLa cells not exposed to IFNγ and HEK293T cells (middle plot) and overlap (right). Inserts represent the axis scale in the 0-0.5% range. Underlined amino acid represents Q to E deamidation. Each dot represents an individual peptide. Relative peptide count was normalized to total ion current (TIC).

Discussion
We obtained highly homogenous human 20S and 26S constitutive-and immunoproteasomes from HEK293T and HeLa cells constitutively expressing various proteasome subunits with the HTBH tag. Our data suggest that proteasomes should be maintained in buffers of low anionic strength, as salt likely leads to aggregation of the 20S particles and partial dissociation of the regulatory particles from the 26S proteasomes. The proteasome samples were used for fragmentation of MBP, the MS autoantigen, to elucidate what MBP peptides may be presented by molecules encoded by human HLA class I alleles.
The product of HLA-A*30:02 allele, which is more frequent in Bahrain MS patients compared with in control subjects [25], is capable of binding the MBP peptide GYGGRASDY, which is released by proteasomes purified from cells exposed to IFNγ two orders of magnitude more efficiently than HEK293T proteasomes (Figure 5b). MBP peptide KGRGLSLSR, generated solely by proteasomes purified from HeLa cells exposed to IFNγ (Figure 5b), is a potential high-affinity ligand for molecules encoded by HLA-A*30:01 and HLA-A31:01 alleles (Figure 5a). The same peptide may be bound by a product of the HLA-A*03:01 allele (Figure 5a), which is more frequent in Iranian patients [26]. The HLA-A*03:01 molecule also binds the KLIETYFSK peptide-a naturally processed epitope of the proteolipid protein (PLP) [27], which is similar to MBP peptide in terms of the flanking basic residues. Thus, one distinct HLA class I molecule may present peptides from both major MS autoantigens. Interestingly, Rajaei et al. showed that in MS patients, the HLA-A*31 allele was often in combination with HLA-A*03 and HLA-A*24 [28]. We thus suggest that synergetic presentation of myelin peptides by heterozygote HLA-A-encoded alleles may significantly increase the risk of MS development.
The HLA-B*07:02, -B*08:01 and -B*51:01 molecules may bind the MBP peptide LPRHRDTGIL with moderately high affinity (Figure 5a). The amount of this peptide was increased 50-fold in MBP hydrolysates mediated by proteasomes purified from HeLa cells exposed to IFNγ than with HEK293T proteasomes (Figure 5b). Jilek et al. showed that the HLA-B*07:02-restricted EBV-specific (epitope spanning residues 379-387 of EBNA-3, EBVRPP, RPPIFIRRL) CD8+ T cell response is dysregulated in MS patients [29]. One may suggest that the ability of molecules encoded by the HLA-B*07:02 allele to present MBP and EBNA-3 epitopes may be additional evidence confirming theory of molecular mimicry, linking EBV and MS [30,31]. These findings are especially important in terms of the recently reported association of the EBV and MS [32]. The product of HLA-A*30:02 allele, which is more frequent in Bahrain MS patients compared with in control subjects [25], is capable of binding the MBP peptide GYG-GRASDY, which is released by proteasomes purified from cells exposed to IFNγ two orders of magnitude more efficiently than HEK293T proteasomes (Figure 5b). MBP peptide KGRGLSLSR, generated solely by proteasomes purified from HeLa cells exposed to IFNγ (Figure 5b), is a potential high-affinity ligand for molecules encoded by HLA-A*30:01 and HLA-A31:01 alleles (Figure 5a). The same peptide may be bound by a product of the HLA-A*03:01 allele (Figure 5a), which is more frequent in Iranian patients [26]. The HLA-A*03:01 molecule also binds the KLIETYFSK peptide-a naturally processed epitope of the proteolipid protein (PLP) [27], which is similar to MBP peptide in terms of the flanking basic residues. Thus, one distinct HLA class I molecule may present peptides from both major MS autoantigens. Interestingly, Rajaei et al. showed that in MS patients, the HLA-A*31 allele was often in combination with HLA-A*03 and HLA-A*24 [28]. We thus suggest that synergetic presentation of myelin peptides by heterozygote HLA-A-encoded alleles may significantly increase the risk of MS development.
The HLA-B*07:02, -B*08:01 and -B*51:01 molecules may bind the MBP peptide LPRHRDTGIL with moderately high affinity (Figure 5a). The amount of this peptide was increased 50-fold in MBP hydrolysates mediated by proteasomes purified from HeLa cells exposed to IFNγ than with HEK293T proteasomes (Figure 5b). Jilek et al. showed that the HLA-B*07:02-restricted EBV-specific (epitope spanning residues 379-387 of EBNA-3, EBVRPP, RPPIFIRRL) CD8+ T cell response is dysregulated in MS patients [29]. One may suggest that the ability of molecules encoded by the HLA-B*07:02 allele to present MBP Figure 5. (a) MBP (P02686-5 isomer) sequence with marked peptides (bold), which may be presented by MS-associated (bold red) and protective (bold black) HLA class I molecules. Calculated percentile rank is shown by a gradient from white to blue (0.01-1). (b) Relative peptide count, normalized to total ion current, observed in the MBP hydrolyzates by HEK293T proteasomes (blue bars) and proteasomes purified from HeLa cells exposed to IFNγ (red bars). Ratio between values is shown, ND-not detected. (c) The MS2 spectra of analyzed MBP peptides.
Among MS-protective HLA alleles, we found that HLA-B*44 molecules may bind MBP peptide QDENPVVHFF, whereas HLA-A*02:06-encoded molecule may bind RTQDEN-PVV (Figure 5a). Interestingly, both peptides span the classical MBP immunodominant epitope [33]. Healy et al. reported that HLA-A*02 and HLA B*44 alleles both reduce susceptibility to MS; however, only HLA-B*44 seems to influence disease course, preserving brain volume and reducing the burden of T2 hyperintense lesions in subjects with MS [34]. On the other hand, we showed that core peptide DENPVVHFF also may be a potential ligand of MS-associated HLA-A*23:01, HLA-A*24:02, HLA-A*26:01 and HLA-B*51:01-encoded molecules (Figure 5a). The frequencies of HLA-A*23 and -A*26 alleles were significantly higher in MS patients than in controls in Italian population [35]. The HLA-B*51 was more frequent in patients with pars planitis [36], often accompanying MS [37].
Here we demonstrated that the predicted affinity of HLA class I to the core MBP peptide QDENPVVHFF of MS-associated HLA-A*23, -A*24, -A*26 and -B*51 molecules tends to be moderate to low (Figure 5a). In contrast, molecules encoded by protective HLA-A*44 alleles may bind this peptide, and its deamidated form, in several registers with unprecedently high affinity (percentile rank = 0.01) (Figure 5a). Further studies should elucidate if the ability of HLA-B*44 to efficiently present an encephalitogenic MBP peptide may affect MS's course, probably due to the exhaustive thymic negative selection of myelin-reactive-or induction of regulatory MBP-specific-CD8+ T cells [38,39].
Proteasome inhibitors may be of interest as therapeutic agents for MS treatment. Some inhibitors, such as bortezomib (PS-341), have been tested as a treatment for MS in experimental autoimmune encephalomyelitis (EAE), which serves as the animal model of MS. These experiments have resulted in a decrease in the number of T cells that secrete pro-inflammatory cytokines [11]. Neurodegeneration is also characterized by activation of the immunoproteasome to the detriment of the constitutive proteasome, 202 [17,40,41]. Therefore, the development of inhibitors specific to proteasome immunosubunits may represent new therapeutic approaches to various forms of MS and other neurodegenerative and neuro-immunological disorders, especially for individuals carrying HLA class I MS-risk alleles.

Mammalian Cells Constitutively Expressing HTBH-Tagged Proteasomes
HEK293T (human embryonic kidney) and HeLa (human, cervical adenocarcinoma) cells were obtained from shared research facility "Vertebrate cell culture collection" of the Institute of Cytology Russian Academy of Sciences. All mammalian cell lines were cultured in DMEM medium (GIBCO, Thermo Fisher Scientific, Waltham, MA, USA). The media was supplemented with 10% fetal bovine serum (GIBCO, Thermo Fisher Scientific, Waltham, MA, USA) and 1% antibiotic-antimycotic (GIBCO, Thermo Fisher Scientific, Waltham, MA, USA). The cell lines were incubated in 37 • C humidified incubator with 5% CO 2 . All cells were routinely tested for Mycoplasma contaminations.

Purification of the HTBH-Tagged Proteasomes
Stable HeLa or HEK293T cell lines expressing HTBH-tagged proteasomes were grown to confluence and then washed with PBS buffer. The cell pellets were lysed in buffer (30 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 0.5% NP-40, 1 mM TCEP, 1 mM PMSF, +/−100mM KCl and 5 mM ATP (ATP was added only to purify 26S/30S proteasomes)). The lysates were centrifuged at 20,000× g for 15 min to remove cell debris, and the supernatant was incubated with Streptavidin-agarose (Thermo Fisher Scientific, Waltham, MA, USA) resin overnight at 4 • C with constant rotation. To purify the 20S proteasome, the streptavidin beads were washed 2 times with 10 volumes of the buffer (30 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 1 mM TCEP, +/−100 mM KCl, 750 mM NaCl), followed by another 2 washes with buffer without 750 mM NaCl. To purify the 26S proteasome, the streptavidin beads were washed 4 times with 10 volumes of the buffer (30 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 1 mM TCEP, +/−100 mM KCl and 1 mM ATP). Next, the beads were resuspended in the required volume of wash buffer with TEV protease, His (Genscript Biotech, Piscataway, NJ, USA) and were incubated at 30 • C for 2 h.

Analysis of the Proteasome Activity
The peptidase activity was determined with 0.125 µg proteasome incubated with 20 µM of the fluorogenic substrate Suc-LLVY-AMC, Boc-LRR-AMC, or Z-LLE-AMC (excitation wavelength of 380 nm and an emission wavelength of 440 nm) in a volume of 25 µL by a microplate reader (CLARIOstar plus, BMG Labtech, Ortenberg, Germany) at 37 • C. The buffer used for measurement of the activity of the proteasomes contained 30 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 1 mM TCEP and 1 mM ATP with various SDS concentrations. Data related to analysis of inhibition of purified proteasome samples by specific inhibitor MG132 are presented in Figure S1.

Native PAGE with LLVY-AMC
All samples were analyzed on mini gels using a Mini-Protean gel apparatus (Bio-Rad, Hercules, CA, USA). Proteasome samples were resolved by nondenaturing PAGE by a modified version of the method described in [42]. We used a gel consisting of 0.09 M Tris (pH 8.3), 0.09 M H 3 BO 3 , 5 mM MgCl 2 , 1 mM ATP, 1 mM DTT and acrylamide-bisacrylamide at a ratio of 37.5:1 (4% for resolving gel and 2.5% for stacking gel) and polymerized with 0.1% N,N,N ,N -tetramethylethylenediamine (TEMED) and 0.1% ammonium persulfate. The running buffer was the same as the gel buffer but without acrylamide. Nondenaturing minigels were run at 150 V for approximately 2 h at +4 • C. The gels were then incubated in 10 mL of 0.1 mM Suc-LLVY-AMC in activity assay buffer (20 mM Tris-HCl pH 7.5, 5 mM MgCl 2 , 1 mM ATP, 1 mM DTT) for 30 min at 37 • C. Proteasome bands were visualized upon exposure to UV light (360 nm) and photographed using the imaging system (ChemiDoc, Bio-Rad, Hercules, CA, USA).

Negative Staining Electron Microscopy
For negative staining, 3 µL of the purified proteasome samples was applied to glowdischarged EM grids covered by a thin layer of continuous carbon film (0183-F, Ted Pella Inc., Redding, CA, USA). After a 1 min incubation, the grid was "washed off" at an angle with 25 µL of 2% uranyl-acetate solution. After 1 min, excess uranyl-acetate was blotted, and the grid was dried. Observations of the sample were performed with a Titan 80-300 transmition electron microscope (Thermo Fisher Scientific, Waltham, MA, USA).
MS raw files were analyzed by Peaks studio 10.0 (Bioinformatics Solutions Inc., Waterloo, ON, Canada). Identification of proteins was performed by searching against the Homo sapiens Uniprot FASTA database, 9 July 2021 version, with carbamidomethyl Cys as a fixed modification and deamidation of Asn/Gln and Met oxidation as variable modifications. The false-discovery rate for peptide-spectrum matches was determined by searching a reverse database and was set to 0.01. Enzyme specificity was set as cleavage after arginine and lysine amino acid residues, and a maximum of two missed cleavages were allowed in the database search. Peptide identification was performed with an allowed initial precursor mass deviation up to 10 ppm and an allowed fragment mass deviation 0.05 Da.

Bioinformatics
The proteome data were analyzed within the in-house algorithm. Briefly, specific MBP peptides were separated out from other irrelevant peptides. The specific MBP peptide count was normalized to total MBP ion current (TIC). Overall, we identified roughly 18,000 specific MBP peptides (18,216), which included both modified and non-modified peptides. HLA affinity binding of the peptides was predicted by IEDB API Tools Python SDK (application iedb 0.0.2, https://pypi.org/project/iedb/, accessed on 11 May 2021), with recommended default settings [44]. The artificial neural network model returns a likelihood of a peptide being a natural ligand of the selected MHC(s). With this approach, we estimated binding affinity for peptides 9-10 amino acids in length versus 27 HLA clas I alleles.