DRB4 has a distinct motif and presents a proinsulin epitope that is recognized in subjects with type 1 diabetes

Abstract

DRB4*01:01 (DRB4) is a secondary HLA-DR product which is part of the high risk DR4/DQ8 haplotype that is associated with type 1 diabetes (T1D). DRB4 shares considerable homology with HLA-DR4 alleles that predispose to autoimmunity, including DRB1*04:01 and DRB1*04:04. However the DRB4 protein sequence includes distinct residues that would be expected to alter the characteristics of its binding pockets. To identify high affinity peptides that are recognized in the context of DRB4, we utilized an HLA class II tetramer based approach to identify epitopes within multiple viral antigens. We applied a similar approach to identify antigenic sequences within GAD65 and preproinsulin that are recognized in the context of DRB4. Seven sequences were immunogenic, eliciting high affinity T cell responses in DRB4⁺ subjects. DRB1*04:01-restricted responses toward many of these peptides have been previously described, but responses to a novel preproinsulin 9–28 peptide were commonly observed in subjects with T1D. Furthermore, T cells that recognized this peptide in the context of DRB4 were present at significantly higher frequencies in patients with type 1 diabetes than in healthy controls, implicating this as a disease relevant specificity that may contribute to the breakdown of beta cell tolerance in genetically susceptible individuals. We then deduced a DRB4 motif and confirmed its key features through structural modeling. This modeling suggested that the core epitope within the preproinsulin 9–28 peptide has a somewhat unusual binding motif, with tryptophan in the fourth binding pocket of DRB4, perhaps influencing the availability of this complex for T cell selection.

Introduction

HLA class II molecules are heterodimers that are expressed on thymic epithelial cells and the surface of antigen presenting cells and function to select, maintain, and activate a CD4⁺ T cell repertoire. In addition to the well-studied HLA-DRB1 genes, three distinct groups of haplotypes include a second functional DRB gene product (either DRB3, DRB4 or DRB5) that also combines with the common DRA chain to form a secondary HLA-DR molecule (1). Prior work demonstrates presentation of both shared and distinct peptides by DRB5*01:01 (DR2a) and by DRB1*15:01 (DR2b) and supports a role for both in presenting peptides to CD4⁺ T cells (2). Likewise, DRB3*01:01 and DRB3*02:02 were shown to present distinct epitopes that complement the peptide repertoire of their associated DRB1 proteins (3). DRB4*01:01 (DRB4) is serologically defined as the HLA-DR53 antigen and is ubiquitously found in association with HLA-DR4, -DR7, and -DR9 alleles (4). Consequently, DRB4 is part of the high risk DR4/DQ8 haplotype that is associated with T1D. DRB4 is reported to be transcribed at lower levels than its associated DRB1 alleles (5). However, its level of expression appears to be quite high on the cell surface of resting and LPS-stimulated B cells, and correspondingly up-regulated in activated monocytes (6). Therefore, DRB4 restricted T cells could play a role in protective and auto-reactive T cell responses.

DRB4 shares considerable homology with the HLA-DR4 alleles that predispose to autoimmunity, including DRB1*04:01 (DR0401) and DRB1*04:04 (DR0404). However, its sequence includes distinct residues, mainly found within and around its peptide binding pockets. As such, DRB4 might be expected to bind and present a distinct subset of peptides, thereby presenting unique self-peptides that contribute to formation of a potentially auto-reactive T cell repertoire. Texier et al. partially characterized a binding motif for DRB4 (7). However their study evaluated only a limited number of substitutions at some positions and, consequently, was unable to define a complete binding motif, leaving some question about the complementarity of its peptide repertoire with its associated HLA-DRB1 alleles. We further addressed this question by defining DRB4 restricted T cell epitopes within common vaccine antigens. We then comprehensively assessed the binding of GAD65 and PPI peptides to DRB4 and evaluated the immunogenicity of sequences with detectable binding by expanding T cells from the peripheral blood of subjects with T1D who had DRB4-associated haplotypes. For peptides that showed evidence of immunogenicity, we compared in vitro responses to these peptides in subjects with T1D and in healthy controls. For an epitope that elicited disease-associated responses we directly assessed T cell frequencies in peripheral blood. To deduce a more comprehensive peptide binding motif for DRB4, we selected a representative viral peptide sequence, constructed a library consisting of peptides with one or more amino acid substitutions, and assessed their binding to recombinant DRB4 protein. Finally, we confirmed key features of the DRB4 motif through structural modeling and evaluated the binding motif of a disease relevant DRB4/proinsulin epitope.

Materials and Methods

Peptides

Peptides representing the full protein sequences of Influenza A/Panama/2007/99 Hemagglutinin (HA-Pan), Influenza A/New Caledonia/20/99 Hemagglutinin (HA-NC), Influenza A/Puerto Rico/8/34 Nucleoprotein (NP-PR), Influenza B/Hong Kong/330/2001 Hemagglutinin (HA-HK), Anthrax protective antigen (PA), Tetanus Toxin heavy chain (TT), and Tetanus Toxin light chain (TTL), preproinsulin (PPI) and glutamic acid decarboxylase 65 (GAD65) were synthesized as sets of 20mer peptides with an overlap of 12 amino acids (Mimotopes, Clayton, Australia). In addition, GAD_557I (GAD _555–567 with a 557I substitution) was synthesized by Anaspec (San Jose, CA) as a biotinylated peptide with two Fmoc-6-aminohexanoic acid spacers between the N-terminal biotin label and the remainder of the peptide sequence. A library of arginine substituted peptides and of 52 singly substituted peptides with sequences based on the TT_803–815 sequence (Supplementary Table 1) were synthesized by MIMOTOPES (Clayton, Australia). Each peptide was dissolved in DMSO at 20 mg/ml and subsequently diluted as needed.

DRB4 Protein and tetramers

Recombinant DRB4 protein was produced essentially as previously described (8). Briefly, DRB4 was purified from insect cell culture supernatants by affinity chromatography and dialyzed against phosphate storage buffer, pH 6.0. The protein was biotinylated at a sequence-specific site using biotin ligase (Avidity, Denver, CO) and dialyzed into phosphate storage buffer. The biotinylated monomer was loaded with 0.2 mg/ml of peptide by incubating at 37°C for 72 hours in the presence of 2.5 mg/ml n-octyl-β-D-glucopyranoside and 1 mM Pefabloc SC protease inhibitor mix (Sigma-Aldrich, St. Louis, MO). Peptide loaded monomers were subsequently conjugated as tetramers using R-PE streptavidin (Biosource International, Camarillo, CA) at a molar ratio of 8 to1.

Human Subjects

Healthy subjects and subjects with T1D with DRB4 haplotypes were recruited with written informed consent under studies approved by the Benaroya Research Institute IRB. Healthy subjects were previously vaccinated against influenza and tetanus and had no history of autoimmune disease. Subjects with T1D were autoantibody positive and insulin dependent.

Tetramer-guided Epitope Mapping

For vaccine antigens, the Tetramer-guided Epitope Mapping procedure was conducted essentially as previously described for each protein (HA-Pan, HA-NC, NP-PR, HA-HK, PA, TT, and TTL) (9). Briefly, PBMC were isolated from the blood of vaccinated healthy DRB4⁺ subjects by Ficoll® underlay and CD4⁺ T cells purified by negative selection using an isolation kit (Miltenyi, Auburn, CA). Adherent cells from the CD4^- fraction were used as antigen presenting cells by incubating 3×10⁶ cells per well (300 μL volume) in 48 well plates for 1 h and washing. Two million CD4⁺ T cells per well were stimulated with pools of five consecutive peptides and expanded by adding fresh medium and IL-2 starting on day 7. After 14 days, ~2×10⁵ cells (100 μL volume) were stained with pooled peptide PE-conjugated tetramers for 60 min at 37°C. Subsequently, cells were stained with CD4-PerCP (BD Biosciences, Mountainview CA), CD3-FITC and CD25-APC mAbs (eBioscience, San Diego CA) and analyzed by flow cytometry. Cells that gave positive staining were analyzed again using the corresponding individual peptide tetramers.

Peptide binding competition assay

Various concentrations of each test peptide were incubated in competition with 0.05 μM biotinylated GAD_557I peptide in wells coated with DRB4 protein, essentially as previously described (10). After washing, residual biotin-labeled peptide was detected using europium-conjugated streptavidin (Perkin Elmer, Waltham MA) and quantified using a Victor2D time resolved fluorometer (Perkin Elmer). Peptide binding curves were simulated using a sigmoidal dose-response curve with Prism software (Version 4.03, GraphPad Software Inc., La Jolla, CA). IC₅₀ binding values were calculated as the peptide concentration needed for 50% inhibition of reference peptide binding. Relative binding affinity (RBA) values were calculated as the IC₅₀ value of the un-substituted peptide divided by the IC₅₀ value of the substituted peptide. The ratio of two IC₅₀ values reflects the fold difference in peptide binding affinity.

In vitro Tetramer Assay for GAD65 and insulin Peptides

To investigate T cell responses to GAD65 and insulin peptides, CD4⁺ T cells were isolated from the PBMC of DRB4⁺ subjects with T1D by negative selection using a magnetic isolation kit (Miltenyi). Adherent cells from the CD4^- fraction were used as antigen presenting cells. Two million CD4⁺ T cells per well were stimulated with groups of peptides that had been shown to bind to DRB4. After 14 days, ~2×10⁵ cells were stained (100 μL volume) with PE-conjugated tetramers and analyzed as described in the previous section. To assess disease relevance of the resulting epitopes, assays were repeated using the PBMC of DRB4⁺ healthy subjects with no history of autoimmunity.

Ex vivo Tetramer Analysis

Direct HLA class II tetramer staining was performed essentially as previously described (11, 12), except that cells were labeled with PE and PE-CF594 labeled tetramers. Briefly, 30×10⁶ PBMCs were re-suspended in 200 µL of T cell media, incubated with 50 nM dasatinib for 10 minutes at 37°C, and stained with PE-DRB4/PPI 9–28 tetramer and PE-CF594-DR0401/PPI 76–90 (with an Lys Ser substitution at position 88, as described (13)) at room temperature for 100 minutes. Cells were washed, incubated with anti-PE magnetic beads (Miltenyi) for 20 minutes at 4°C and enriched with a magnetic column, retaining 1% of the cells as a non-enriched sample. The enriched and pre-column samples were labeled with anti-CD4 PerCP-Cy5.5 (Biolegend), anti-CD14 and anti-CD19 FITC (both eBioscience), ant-CD45RA AF700 (BD Biosciences), anti-CXCR3 APC (BD Biosciences), anti-CD38 APC-Cy7 (Biolegend), and SYTOX Green (ThermoFisher) as a viability indicator for 15 minutes at 4°C and analyzed on an LSR II (BD Biosciences), gating on Viable CD4⁺ cells and excluding events that were positive for more than one tetramer color. Frequencies of tetramer positive cells were calculated as previously described (12).

Molecular modeling

Models of DRB4 in complex with the TT_803–818 sequence and selected variants of interest were prepared on a Silicon Graphics Fuel work station using the program Insight II, version 2005 (Accelrys, San Diego CA), essentially as previously described (14). Energy minimization was performed at pH 5.4, the experimental pH used for binding studies. The crystal structure of HLA-DRB1*0401 in complex with the Collagen II peptide (15) was used as the base molecule for all simulation studies. Figures were drawn with the aid of WebLabViewer version 3.5 and DSViewer Pro version 6.0, of Accelrys using previously published formatting and color conventions (16). The .pdb coordinates of these complexes will be provided by GKP or AKM to interested researchers upon request.

Epitope Prediction

To predict the binding affinity to DRB4 for peptides of interest we utilized an approach similar to that in our published work (16). The RBA of any peptide was calculated as the product of the observed RBA values for the substituted TT_803–818 peptide with the equivalent residue in that position within that pocket or an interpolated value (for amino acids not measured) which was calculated based on the observed values for chemically similar amino acids. This can be expressed using the following formula:

$${\text{RBA}}_{\text{C}}={\text{RBA}}_{\text{P1}}\times {\text{RBA}}_{\text{p4}}\times {\text{RBA}}_{\text{p6}}\times {\text{RBA}}_{\text{p7}}\times {\text{RBA}}_{\text{p9}}$$

To predict minimal epitopes within DRB4-restricted epitopes the RBA_C was compared for all possible registers within the peptide.

T cell cloning and assays

T cell lines were generated by staining in vitro expanded T cells with tetramer and then sorting gated tetramer-positive CD4⁺ cells using a FACSAria (at single-cell purity) and expanding in a 96-well plate in the presence of 1.0 × 10⁵ irradiated PBMCs and 2 μg/ml PHA (Remel, Lenexa, KS). To confirm their specificity, T cells were re-stained using tetramer that had been loaded with the same peptide as the tetramer used for sorting or with various truncated versions of the peptide. In parallel to assess proliferation, clones were stimulated with 10 μg/ml of peptides, adding HLA-DR–matched irradiated PBMCs as APCs, pulsed after 48 h with 1 μCi [3H] thymidine (Amersham Biosciences, Piscataway, NJ), and harvested after an additional 16 h. Thymidine uptake was measured with a scintillation counter to assess proliferation.

Results

Tetramer-based identification of DRB4 restricted viral epitopes

DRB4-restricted epitopes within HA-Pan, HA-NC, NP-PR, HA-HK, PA, TT, and TTL proteins were identified using a tetramer-based approach (17). The sequences of the immunogenic peptides identified through these experiments are summarized in the top portion of Table 1. Interestingly, although high affinity DRB4 restricted epitopes were present within each of these antigens, the proportion of antigenic peptides was markedly lower than we had previously observed for HLA-DRB1 proteins. For example, within the tetanus toxoid heavy chain we previously reported 15 peptides that contain DR0401-restricted epitopes and 7 peptides that contain DR0404-restricted epitopes, but observed only a single DRB4 restricted epitope (9).

Table 1:

Novel and Previously Published DRB4 Epitopes

Peptide	Sequence*	RBA^
HA-PAN5 65–84	TLIDALLGDPHCDGFQNKEW	0.20
HA-NC6 504–520	KYSEESKLNREKIDGVK	0.03&
HA-NC6 510–526	KLNREKIDGVKLESMGV	0.08&
NP-PR7 193–212	LVRMIKRGINDRNFWRGENG	0.04&
HA-HK8 265–284	RGILLPQKVWCASGRSKVIK	0.88
PA9 137–156	LEKGRLYQIKIQYQRENPTE	0.25&
TT10 802–821	RSFLVNQMINEAKKQLLEFD	1.0
TTL11 257–276	QEIYMQHTYPISAEELFTFG	0.46&
NY-ESO-1 115–132	PLPVPGVLLKEFTVSGNI	0.14
NY-ESO-1 121–138	VLLKEFTVSGNILTIRLT	0.14&
NY-ESO-1 139–156	AADHRQLQLSISSCLQQL	0.03&
Vacc IMVMP 127–137	KIQNVIIDECY	0.05
VZV env GI 144–155	YVLLVRLDHSRS	0.32&
HSV5 pMP 285–299	GKISHIMLDVAFTSH	0.02&

^*Primary binding registers are indicated in boldface

^{^}Predicted Relative Binding Affinity based on observed preferences

^&Secondary registers or weak primary binding registers are underlined

T cell responses to GAD65 and PPI Peptides

Given that there is an association between DRB4-containing haplotypes and T1D, we next sought to identify sequences from GAD65 and PPI that contain DRB4-restricted epitopes. As an initial screening tool, we first comprehensively assessed the binding of peptides spanning GAD65 and pre-proinsulin to DRB4. The sequences and measured affinities for peptides with positive binding results are shown in Table 2. All remaining peptides had no detectable binding to DRB4 and therefore were not included in subsequent T cell assays. We evaluated the immunogenicity of all DRB4-binding GAD65 and PPI peptides by expanding and tetramer staining peripheral blood samples from subjects with T1D, identifying a total of 6 GAD65 peptides and 1 proinsulin peptide that elicited expansion of tetramer positive T cells (Figure 1A). Three of these specificities were further confirmed through the isolation of tetramer positive T cell clones (Figure 1B). We had previously identified 17 GAD65 peptides that contain DR0401-restricted epitopes, some of which were subsequently confirmed as DR0404-restricted epitopes (18). Comparing our newly reported DRB4-restricted epitopes with previously reported DR0401-restricted epitopes (last column, Table 2) within these beta cell antigens, epitopes were present within several “shared” peptides, suggesting that 4 out of 6 DRB4-restricted GAD65 epitopes can be analogously presented by DR0401. However, the epitope within the proinsulin leader sequence was unique to DRB4

Table 2:

Observed Binding and Immunogenicity of GAD65 and Insulin Peptides

Peptide	Sequence*	IC50^	DR0401&	DRB4%
PPI 9–28	PLLALLALWGPDPAAAFVNQ	1.2	No	Yes@
GAD65 113–132	DVMNILLQYVVKSFDRSTKV	1.4	Yes	Yes@
GAD65 201–220	NTNMFTYEIAPVFVLLEYVT	1.2	Yes	No
GAD65 209–228	IAPVFVLLEYVTLKKMREII	21	No	Yes
GAD65 241–260	PGGAISNMYAMMIARFKMFP	0.6	No	No
GAD65 249–268	YAMMIARFKMFPEVKEKGMA	0.5	No	No
GAD65 265–284	KGMAALPRLIAFTSEHSHFS	10	Yes	No
GAD65 305–324	DERGKMIPSDLERRILEAKQ	29	Yes	No
GAD65 313–332	SDLERRILEAKQKGFVPFLV	22	No	No
GAD65 353–372	ICKKYKIWMHVDAAWGGGLL	0.9	Yes	Yes@
GAD65 449–468	HVDVFKLWLMWRAKGTTGFE	32	No	No
GAD65 513–532	LRTLEDNEERMSRLSKVAPV	5.9	No	No
GAD65 529–548	VAPVIKARMMEYGTTMVSYQ	2.8	No	Yes
GAD65 553–572	KVNFFRMVISNPAATHQDID	7.8	Yes	Yes
GAD65 566–585	ATHQDIDFLIEEIERLGQDL	12	No	Yes

^*Predicted binding registers are indicated in boldface

^{^}IC₅₀ for binding to DRB4, values are expressed in μM

^&Previously reported as a DR0401 restricted epitope

^%Confirmed as a DRB4 restricted epitope through in vitro responses

^@Confirmed as a DRB4 restricted epitope through T cell cloning

Figure 1. Immunogenicity of Gad65 and PPI peptides.

All peptides with positive binding to DRB4 were evaluated for immunogenicity by tetramer staining of PBMC after 14 days of peptide stimulation. (A) As demonstrated by the representative FACS plots shown, seven peptides elicited levels of T cell expansion that were more than twofold above background staining. (B) For three of the candidate epitopes of interest (GAD65 113, GAD65 353, and PPI 9–28) tetramer positive T cell clones were isolated by single cell sorting.

To assess the potential relevance of these epitopes, we first compared in vitro responses to these peptides in 16 DRB4 positive subjects with established T1D and 12 HLA-matched healthy controls. As summarized in Table 3, the PPI 9–28 and GAD65 209–228 peptides elicited positive responses in more than half of the patients tested. Notably, only the PPI 9–28 peptide elicited a significantly higher proportion of positive responses in patients than controls. Consequently, this DRB4 restricted epitope within PPI appeared to have the highest potential relevance. To further confirm the relevance of this epitope, we directly examined the frequency of DRB4/PPI 9–28 reactive T cells and, as a comparator, DR0401/PPI 76–90 reactive T cells in the peripheral blood of 6 DRB4 and DR0401 positive subjects with established T1D and 6 HLA matched healthy controls through direct ex vivo tetramer staining. The observed frequencies of DRB4/PPI 9–28 reactive T cells were significantly higher in subjects with established T1D than in HLA-matched controls (Figure 2A). As expected, observed frequencies of DR0401/PPI 76–90 reactive T cells (an epitope with established disease relevance) were also significantly higher in subjects with established T1D than in HLA matched controls (Figure 2B). Unexpectedly, a paired analysis indicated that frequencies of DRB4/PPI 9–28 reactive T cells were significantly higher than DR0401/PPI 76–90 reactive T cells both in subjects with T1D and controls (Figure 2C). Examining this relationship further, we observed that the frequencies of DRB4/PPI 9–28 specific T cells and DRB1*0401/PPI 76–90 reactive T cells were significantly correlated in subjects with T1D but not HLA matched controls (Figure 2D). We observed no significant differences between subjects with T1D and controls with regard to common cell surface markers (Supplementary Figure 1). Cumulatively, these findings appear to support a role for DRB4/PPI 9–28 specific T cells in the diabetogenic response. Specifically, T cell receptors which recognize this peptide are present within the T cell repertoire of all DRB4 positive subjects but appear to expand to higher frequencies in subjects with T1D.

Table 3:

Comparing in vitro responses to GAD65 and Insulin Peptides in DRB4+ Subjects

	PPI 9–28	GAD65 113	GAD65 209	GAD65 353	GAD65 529	GAD65 553	GAD65 566
T1D	13/16	5/16	11/16	5/16	4/16	4/16	4/16
Control	2/12	4/12	5/12	4/12	5/12	6/12	4/12
P-value*	0.0016	1	0.24	1	0.43	0.24	0.69

^*P-value calculated using 2-tailed Fischer’s Exact Test

Figure 2. Direct analysis of DRB4/proinsulin and DR4/prosulin reactive T cell frequencies.

PBMCs from for 6 subjects with diabetes and 6 HLA-matched controls were thawed and immediately co-stained with PE labeled DRB4/PPI 9–28 (DRB4 Ins) and PE-CF594 labeled DR0401/PPI 76–90 (DR4 Ins) tetramers to quantify antigen-specific CD4+ T cells. (A) DRB4 Ins frequencies were significantly higher in subjects with type 1 diabetes (filled circles) than in controls (open circles). (B) DR4 Ins frequencies were also significantly higher in subjects with type 1 diabetes (filled squares) than in controls (open squares). (C) DRB4 Ins frequencies (circles) were significantly higher than DR4 Ins frequencies (squares) both in subjects with T1D (filled symbols) and HLA-matched controls (open symbols). (D) There was a significant linear relationship between the frequencies of DRB4 Ins and DR4 Ins reactive T cells in subjects with T1D but no significant correlation in controls. In each panel one, two, or three stars indicates p-values less than 0.05, 0.01, and 0.001 respectively.

Determining Binding Preferences for DRB4

Given the apparent relevance of DRB4-restricted T cell responses, we next sought to more fully characterize its peptide binding preferences. Among the peptides identified as DRB4-restricted epitopes, TT_802–821 was selected as a reference peptide to evaluate its binding preferences, as it showed the highest binding affinity. To determine the most likely binding register for this peptide, a library of substituted peptides with successive arginine substitutions was synthesized and the binding of each peptide assessed using an in vitro competition assay (Supplementary Table 1). Arginine substitutions at residues 805, 808, 810, 811, and 813 markedly reduced the binding of the peptide (Supplementary Table 1), suggesting that these represent anchor positions. The spacing of these residues establishes 805L as the P1 anchor, 808Q as the P4 anchor, 810I as the p6 anchor, 811N as the P7 anchor, and 813A as the P9 anchor. Based on this information we designed a panel of TT_803–815 derived peptides with single amino acid substitutions (including obligatory anchor residues for pocket 1 and general classes of amino acids for pockets 4, 6, 7, and 9) and utilized these to determine preferences for the binding pockets of DRB4. The sequences of these peptides and their relative binding affinities (RBA) are summarized in Supplementary Table 1. Figure 3 summarizes our overall findings, which are somewhat consistent with the findings of Texier et al. (7), who inferred a binding motif for DRB4 using a more limited set of substituted peptides. Pocket 1 accommodated Val, Leu, Met, and Ile, followed by Phe, while tolerating Lys and Ala. Pocket 4 preferred Gln and Glu; tolerating Met and Ala; and barely accepting Ser, Leu and Phe. Pocket 6 preferred Pro, Ile, and Asn; also accepted Met, His, Ala; barely tolerated Ser; but excluded various aromatic and charged residues. Pocket 7 was more permissive, preferring Ala, Asn, and Me; also accepting Tyr, Ser, Phe, Ile, and His; but excluding charged residues and proline. Pocket 9 was the most permissive, preferring Ser, His, Met, Ala, Asn, and Asp, but also accommodating Ile, Tyr, Pro, and barely accepting Phe, while excluding only basic residues (Arg and Lys).

Figure 3. Summary of the DRB4 motif.

Amino acid preferences for A) pocket 1, B) pocket 4, C) pocket 6, and D) pocket 9 of the DRB4 motif. Black bars represent values within 2.5-fold of the reference peptide (RBA ≥ 0.4, “preferred”), hatched bars represent values within 10-fold of the reference peptide (RBA ≥ 0.1, “tolerated”), and open bars represent values within less than 10-fold of the reference peptide (RBA ≤ 0.1, “excluded”).

Motif Analysis of Peptides Containing DRB4 epitopes

We evaluated the experimentally determined DRB4 motif by utilizing an algorithm (implemented as described in Materials and Methods) to predict minimal epitopes within the 8 viral peptides we had identified that contain DRB4 restricted epitopes (Table 1, upper rows). For 6 of these peptides, the algorithm identified at least one adequate DRB4 motif. A few peptides contained a second distinct register that could be expected to bind. However, two peptides (HA-NC_504–520 and NP_193–212) only contained a weak DRB4 motif. The key unfavorable residue in the best predicted motifs of both of these peptides is an Asp in pocket 7. We then utilized the same strategy to predict minimal epitopes within 6 previously published DRB4 restricted epitopes (Table 1, bottom rows). For 5 of these antigenic peptides, the prediction algorithm identified a strong DRB4 motif. However, NY-ESO-1_139–156 and the HSV5 phosphorylated matrix protein peptide only contained a weak motif. The key unfavorable residue the best predicted motifs, respectively, are an Arg in pocket 4 and an Asp in pocket 4. We conclude that, in general, the sequences of newly identified and previously published DRB4 restricted epitopes are consistent with the DRB4 motifs that we defined.

Defining and modeling a minimal DRB4 motif within PPI 9–28

T cell responses to epitopes within the proinsulin leader sequence have been reported in the context of other HLA types, including both Class I as well as Class II (19, 20). Therefore, we next sought to deduce the minimal epitope and binding register of the DRB4 restricted epitope within PPI 9–28. To accomplish this, we assessed the binding of truncated peptides and arginine substituted versions of the peptide. Among the truncated peptides tested, only PPI 12–23 was able to bind, indicating that the minimal DRB4 epitope must be within this interval. Arginine replacement of L14 blocked binding, suggesting that this residue serves as the P1 anchor (Figure 4Α). Consequently, the minimal motif for this epitope appears to be LALWGPDPA (anchors underlined)

Figure 4. DRB4 binding register of the PPI peptide.

(A) The binding of the full length PPI 9–28 peptide, various truncations of the peptide, and two arginine substituted versions of the peptide were assessed using an in vitro competition assay. Among the peptides tested, only PPI 12–23 and the full length peptide had detectable binding. Replacement of L14 with arginine disrupted binding, implicating this as a primary anchor residue. (B) Structural modeling confirms LALWGPDPA as a possible core epitope (anchors in bold). TCR view of the modeled complex of HLA-DRB4*01:01 with the human PPI 12–24 peptide. The remarkable anchoring of p4Trp is achieved because of several factors at play: i) the presence of p5Gly/p6Pro allow for sufficient maneuvering room for p4Trp, as Gly, the smallest of amino acids, has more flexibility than other residues; also, proline is the most preferred residue at pocket 6. ii) Aromatic-aromatic interactions between β78Tyr and p4Trp; iii) Cation-pi (π) interaction between β70Arg and p4Trp. iv) Charge-charge interaction between β70Arg-β66Asp-β71Arg-p7Asp. Surface water molecules may modify the latter interactions but their importance is apparent. Color conventions for atoms: oxygen, red; nitrogen, blue; carbon, orange in DRB4 and green in the antigenic peptide; hydrogen, white; sulfur, yellow. Transparent surfaces are colored as follows: positively charged, blue; negatively charged, red; partially charged, scaled shades of the previous two colors; neutral, grey. Note that from this perspective β71Arg is “under” β70Arg, and thus not labeled for clarity.

This experimental finding is at odds with the DRB4 motif that we determined, which would predict LALLALWGP as the most likely motif. Therefore, we performed structural modeling as described in Materials and Methods to verify the plausibility of this experimentally determined motif (Figure 4B). This modeling confirmed that placing Trp in pocket 4, which would normally be unfavorable, is facilitated by the rotational freedom provided by the Gly residue at position 5, as glycine residues possess much greater ranges of Ramachandran angles allowing the amino- and carboxy-bonded neighboring residues greater flexibility of movement (21); glycine thus allows P4Trp enough freedom to fit snugly into pocket 4 (Figure 4B). Furthermore, proline at pocket 6 is the most preferred residue in DRB4*01:01, while P8Pro does not result in any missed hydrogen bonds due to its imine group (22). The result is a stable but non-standard anchoring of this peptide within the cleft of DRB4.

To further refine our understanding of the DRB4 motif for more conventional peptides, additional models were prepared. As shown in Figure 5, pocket 4 accommodates Glu through favorable interactions with β70Arg. In contrast, the more compact Asp residue (Supplementary Figure 2A) has a much higher calculated total energy when docked within this pocket and is predicted to rapidly dissociate rather than being accommodated. As shown in Figure 6, the combination of α65Val/β11Ala and nearby β13Cys/β28Ile enables proline to fit snugly in pocket 6 (even at the expense of a lost hydrogen bond due to its imine group), and would likewise allow small/medium aliphatic residues to remain in the pocket; furthermore, the charged residues α11Glu/α66Asp and the polar β30Tyr allow small/medium sized polar residues and the positively charged histidine to anchor in the pocket (Supplementary Figure 2B). As shown in Figure 7, histidine is accommodated in pocket 9 through favorable interactions with negatively charged residues (β9Glu and β57Asp) and with β37Tyr. The presence of β57Asp (which forms a salt bridge to α76Arg) greatly enhances the stability of this pMHCII complex in comparison those which lack β57Asp. In contrast, the less intuitive anchoring of aspartic acid in this pocket (Supplementary Figure 2C), is aided by favorable interactions with α76Arg and β37Tyr. Notably, in this instance the p9Asp anchor (rather than β57Asp) preferentially forms a salt bridge with α76Arg.

Figure 5.

TCR view of pocket 4 with anchor residue Glu, with DRB4 residues in stick mode with transparent surfaces colored according to atom charge and the anchor residue in space-filling mode. This pocket is restrictive, favoring Gln/Glu but not Asp/Met/Ala and barely Ser/Leu. Residues β13Cys (behind p4Glu) and β26Asn are unique to DRB4 alleles (except DRB4*02:01), while the combination of β70Arg/β71Arg/β74Glu/β78Tyr (with its aromatic ring presented here perpendicular to the plane of the paper/screen) is unique to some 15 DRB1* alleles out of over 1000 known sequences to date. Note how β26Asn and β74Glu interact with each other and p4Glu points towards β70Arg minimizing thus the p4Glu—β74Glu and β70Arg—β71Arg electrostatic repulsions. The figure is rotated by 30^o around the x-axis (bottom of the page towards the reader, top of the page away from the reader), and 10^o around the y-axis (left-hand side towards the reader, right-hand side away from the reader) with respect to the standard depiction mode of MHC II molecules with their β-sheet floor parallel to the plane of the paper/screen. Depiction and color conventions are as in Figure 4B.

Figure 6.

TCR view of pocket 6 with Pro in space-filling mode and the pocket residues in stick form with transparent surfaces colored according to atom charges. The figure is rotated by 30° around the x-axis (bottom of the page towards the reader, top of the page away from the reader), in comparison to the conventional depiction of MHCII molecules as shown in Figure 4B. Color conventions are as in Fig. 4B.

Figure 7.

TCR view of pocket 9 with anchor residue His in space filling mode and the pocket residues in stick form with transparent surfaces colored according to atom charges. P9His (ambient pH = 5.4, so His is protonated) interacts with the two negatively charged residues (β9Glu and β57Asp) as well as the aromatic β37Tyr, to insure good fitting into the pocket, and to overcome the repulsion from positively charged α76Arg. The figure is rotated by 30° around the x-axis (bottom of the page towards the reader, top of the page away from the reader), in comparison to the conventional depiction of MHCII molecules shown in Fig 4B. Color conventions as in Fig. 4B.

Discussion

The amino acid sequences of HLA class II proteins dictate their binding specificity, thereby defining the peptide motifs that are presented to CD4+ T cells as the T cell repertoire is shaped in the thymus and activated in the periphery. DRB4 is present in all DR4, DR7 and DR9 haplotypes and therefore is part of the high risk DR4/DQ8 haplotype that is associated with T1D. Although DRB4 shares structural homology with DR0401 and DR0404, we hypothesized that key polymorphic residues within DRB4 facilitate the presentation of unique peptides (sequences which are not presented by homologous DRB1* alleles) from self-antigens that can be recognized by potentially auto-reactive T cells. To address this hypothesis, we performed a comprehensive screening of peptides derived from PPI and GAD65 to identify those that are able to bind to DRB4. We then utilized tetramer based assays to identify a total of six GAD65 peptides and one peptide from the proinsulin leader sequence (residues 9–28) that were recognized in the context of DRB4. Four of the GAD65 peptides were previously shown to be immunogenic in the context of DR0401 (18). Recognition of the proinsulin leader peptide was unique to DRB4. Among these immunogenic peptides, only PPI 9–28 elicited detectable in vitro T cell responses that were significantly more frequent in subjects with T1D than in HLA matched controls. Notably, observed frequencies of DRB4/PPI 9–28 reactive T cells were significantly higher in subjects with established T1D than in HLA matched controls and were also consistently higher than those of DR0401/PPI 76–90 reactive T cells (an epitope with previously reported disease relevance). Furthermore, frequencies of DRB4/PPI 9–28 specific T cells and DRB1*0401/PPI 76–90 reactive T cells were significantly correlated in subjects with T1D but not HLA-matched controls, perhaps suggesting that T cells with both specificities have coordinated expansion during disease progression.

Our findings support a role for DRB4/PPI 9–28 specific T cells in the diabetogenic response but leave open the question of why T cells with this specificity are able to escape negative selection and expand to relatively high numbers in subjects with diabetes. Surprisingly, further study revealed that proinsulin residues 14–22 comprised the minimal binding element of the DRB4-restricted epitope rather than the best predicted core epitope of PPI 11–19. Consequently, the core epitope is predicted to have an unfavorable Trp residue (based on the deduced DRB4 motif) as its p4 anchor. Modeling analysis indicated that this atypical anchor is accommodated because the sequence WGP fits well with its Pro as the preferred p6 anchor and p5Gly with its much wider spectrum of permitted Ramachandran angles allows sufficient flexibility to allow Trp to anchor within pocket 4 (21). This is a case of a non-anchor residue substantially influencing the binding of an otherwise non-allowed residue in an adjacent pocket. It seems plausible that the atypical binding motif of this epitope could influence the availability of this complex for T cell selection in the thymus.

It was of particular interest that we observed DRB4-restricted responses toward the PPI leader peptide. As summarized in Table 4, several studies have demonstrated T cell recognition of epitopes within (or overlapping with) the proinsulin leader sequence. Distinct epitopes derived from the proinsulin leader peptide which are recognized by CD8⁺ T cells in the context of HLA-A2, HLA-A24, HLA-B38, and HLA-B39, were recently shown to be processed for immune recognition through a non-canonical process that requires signal peptide peptidase (23). Concerning HLA class II presentation, the mechanism through which the signal peptide could reach the endosomes of professional antigen presenting cells, facilitating loading HLA class II molecules such as DRB4 remains unclear. Given recent reports suggesting HLA class II expression by human β-cells in islets exposed to interferon gamma, cross-presentation would be one possibility (24). Alternatively, it could be that the insulin fragments which are passed to islet resident macrophages via crinophagic granules from β-cells (as documented by Vomund et al. (25)) include derivatives of the signal peptide which can be processed and presented to T cells. We have taken steps here to define the minimal binding register of this DRB4 restricted PPI epitope, which overlaps with an antigenic region that was previously reported to elicit cytokine responses in subjects with HLA-DRB1*04/DQ8 haplotypes ((20) and Table 4). However, since the methodology of that study was not HLA-specific, the responses observed could have been DRB4 restricted.

Table 4.

HUMAN HLA-I/II epitopes within the PPI Leader Peptide

1	10	20	30	40	50
|	|	|	|	|	|
|	Signal	|	| B chain	|	|
MALWMRLLPLLALLALWGPDPAAAFVNQHLCGSHLVEALYLVCGERGFFYTPKT
LWMRLLPLL
MRLLPLLA*
MRLLPLLALL
ALWGPDPAAA
MALWMRLLPLLALLAL
MRLLPLLALLALWGPD
PLLALLALWGPDPAAA
LLALWGPDPAAAFVNQ
WGPDPAAAFVNQHLCG
PAAAFVNQHLCGSHLV
LALWGPDPA
HLA-A2
HLA-A24
HLA-B38
HLA-B39
HLA-DRB1*04
HLA-DRB4*01:01:01

Notes:

HLA-A2, HLA-A24, HLA-B38, HLA-B39

as in ref. 22.

HLA-DRB1*04

, ref. 19 (Durinovic-Bello et al, Diabetologia, 2004). Precise core nonamers not reported.

DRB4*01:01

, this work

^*According to the published motif this is a weak ligand, but the experiments of ref. 22 report isolating this peptide, not a longer one. Extending by one residue (Leu) would make for a higher affinity peptide, but antigenicity of this version has not yet been documented.

Beyond these insights with regard to self-reactive T cell responses against a proinsulin signal peptide epitope, our work reveals some general insights about the DRB4 peptide binding motif. This motif exhibited a few notable characteristics, including its accommodation of alanine and lysine residues in pocket 1, a specific subset of favored residues in pocket 4 (polar and small/medium-sized aliphatics, glutamate but not aspartate and no basic amino acids or histidine), a strong preference for proline at pocket 6 (plus a preference for polar uncharged residues and small/medium-sized aliphatics and an accommodation of histidine), and accommodation of a wide range of residues within pocket 9 (with preference for serine and histidine, accommodation of small/medium-sized aliphatics and polar, uncharged, and acidic residues, as well as a bare accommodation of tyrosine). A comparison of the residues lining the various anchoring pockets of DRB4 compared to various associated HLA-DR alleles and the somewhat homologous DR10 allele (Supplementary Table 2) prompts the following observations:

1. The presence of β81Tyr (in nearly all DRB4 alleles) in place of near invariant His accounts for the bare acceptance of Ala and Lys in pocket 1 [(22) and Supplementary Table 2]. The effect is greater at pH 5.4 (endosomal) where His is positively charged.

2. The combination of β13Cys/β26Asn in pocket 4 is again unique to DRB4 alleles, while the combination of β70Arg/β71Arg/β74Glu in the same pocket is to be found in most DR9 and DR14 alleles, very few scattered alleles of various other designations, as well as in all known DRB4 alleles.

3. The remarkably simple β11Ala residue (unique to DRB4 alleles) is key to the essential preference for proline and small/medium-sized aliphatic anchors at pocket 6. Also unique to DRB4 alleles is nearby β28Ile in pocket 7.

4. The preferences for pocket 7 are wide, as in most other MHC II alleles; the exception concerns long basic residues (Arg/Lys) because of the presence of two basic residues in the pocket β70Arg/β71Arg, which modeling shows as forming the border between pockets 4 and 7, thereby influencing the anchoring residue selection in both pockets; this is similar to the situation in HLA-DQ2 alleles, where these two positions are occupied by basic residues and thus basic residues are excluded from pockets 4 and 7 (23). The exclusion of p7Asp might in part arise from the electrostatic repulsion with p8Glu of the TT test peptide.

5. The combination of β37Tyr/β38Ala, also found in DR10 as well as DRB3 alleles is responsible for the accommodation of a wide variety of anchor residues in pocket 9. Alanine at β38 allows greater freedom of movement for β37Tyr and at the same time makes possible this wide spectrum of accommodated anchor residues from polar to small/medium-sized aliphatics, to both acidic residues, to histidine and even tolerably to tyrosine. Remarkably, all this takes place in a β57Asp⁺ MHC II allele, where the strong salt bridge between β57Asp and α76Arg greatly adds to the stability of the pMHCII complex and limits somewhat the capacity of pocket 9 to accept bulky residues, a situation very similar to what is observed with the DR10 allele (16).

The differences in the binding preferences of DRB4 and its DRB1 counterparts (such as DRB1*04:01) apparently cause their epitope repertoires to diverge to some degree. Comparison of the DRB4- and DRB1-restricted T cell epitopes within influenza and tetanus antigens identified some “shared” epitopes that were predicted to have identical binding registers. However, in agreement with its binding motif, DRB4-restricted epitopes tended to contain smaller residues at the P1 anchor position because of β81Tyr that limits entrance of bulky residues into this pocket and of β86Val that diminishes the pocket volume, so that the aromatic residues Phe/Tyr/Trp cannot anchor into this pocket. The capacity of DRB4 to present unique peptides such as PPI 9–28 is a consequence of various residues in key positions in the five pockets as recounted above (plus the influence of a very flexible glycine at P5, between pockets 4 and 6, facilitating the binding of what would otherwise be a poorly accommodated anchor). Supplementary Table 2 lists all the β-chain residue differences among alleles HLA-DRB1*04:01, 04:02, 04:03, 04:04, 04:05, 07:01, 09:01, and 10:01, along with DRB4*01:01 and DRB4*01:03. The DRB4 alleles have several differences in invariant residues located outside the anchoring pockets (e.g. β4, β18, β25, β41, β44, β48) and a few residues in the β2 domain, and one each in the transmembrane and intracellular short domains. The precise role of these substituted residues remains to be clarified.

It bears mentioning that a recent study reported an association between DRB4*01:03:01 and diabetes risk but associated DRB4*01:01:01 with protection (26). It is interesting to note that these two DRB4 alleles differ only by a single amino acid (DRbeta135, which is Gly and Ser respectively) that is located in the CD4-binding region of the protein. Consequently, DRB4*01:03 should have a peptide binding motif that is identical to the DRB4*01:01 motif defined here and would be expected to present the exact same epitopes. The DRB4*01:03 allele is plentiful in the Scandinavian population studied in that work, whereas DRB4*01:01 is relatively infrequent. Given that the reported differences in risk are not explainable by differences in epitope presentation, there is at least some possibility that the biochemical stability or regulation of expression of these DRB4 alleles may differ or that this subtle change in the CD4 binding domain has a functional consequence.

Cumulatively, our findings suggest that DRB4 has the capacity to present both shared and unique self-peptides derived from GAD65, PPI, and perhaps other beta cell antigens, adding to the repertoire of potentially auto-reactive T cells that could play a crucial role in the breakdown of self-tolerance in the islets. The set of peptides that is presented by DRB4 includes unique epitopes, leading to the selection of a distinct repertoire of T cells that more readily expand in the peripheral blood of subjects with T1D. In particular, our findings demonstrate that the PPI 9–28 peptide presented by DRB4 has likely disease relevance, as demonstrated by elevated T cell frequencies in subjects with T1D as compared with controls. In addition to its binding preferences, DRB4 has a different pattern of expression than its DRB1 counterparts, in that it is expressed at lower levels in non-inflamed tissues but is markedly upregulated in settings of inflammation (5, 6). Conceptually, this difference in expression could alter T cell selection and the subsequent induction of peripheral tolerance to DRB4 restricted T cells, perhaps leading to different ratios of regulatory versus effector T cells for these epitopes. As such, future studies that seek to monitor beta cell specific CD4⁺ T cells or that seek to implement strategies for the induction of antigen specific tolerance could benefit from the inclusion of these epitopes.