Instability of the HLA-E peptidome of HIV presents a major barrier to therapeutic targeting

Abstract

Naturally occurring T cells that recognize microbial peptides via HLA-E, a nonpolymorphic HLA class Ib molecule, could provide the foundation for new universal immunotherapeutics. However, confidence in the biological relevance of putative ligands is crucial, given that the mechanisms by which pathogen-derived peptides can access the HLA-E presentation pathway are poorly understood. We systematically interrogated the HIV proteome using immunopeptidomic and bioinformatic approaches, coupled with biochemical and cellular assays. No HIV HLA-E peptides were identified by tandem mass spectrometry analysis of HIV-infected cells. In addition, all bioinformatically predicted HIV peptide ligands (>80) were characterized by poor complex stability. Furthermore, infected cell elimination assays using an affinity-enhanced T cell receptor bispecific targeted to a previously reported HIV Gag HLA-E epitope demonstrated inconsistent presentation of the peptide, despite normal HLA-E expression on HIV-infected cells. This work highlights the instability of the HIV HLA-E peptidome as a major challenge for drug development.

Graphical abstract

Wallace and colleagues describe target discovery for a universal T cell receptor (TCR)–adapted therapy against HIV, exploiting the nonpolymorphic HLA class I molecule HLA-E. Their systematic interrogation of the HIV proteome revealed a distinct lack of stable HLA-E/peptide complexes, presenting a challenge for drug and vaccine development.

Introduction

T cell receptors (TCRs) recognize pathogen-derived antigens in the form of peptides displayed by major histocompatibility complex (MHC) class I molecules on the infected cell surface. The exquisite specificity of TCR-peptide–MHC interactions has been successfully exploited therapeutically by TCR-based biologics against cancer antigens and offers opportunities to treat chronic infectious diseases.^1,2,3 However, clinical application is limited by the diversity of MHC-encoded human leukocyte antigens (HLAs). A possible solution is to target antigens presented by the nonpolymorphic HLA class Ib molecule HLA-E, which has only two functional alleles. HLA-E∗01:01 and HLA-E∗01:03 differ by a single amino acid located outside the peptide-binding groove and offer the potential for universal therapy.^4,5 Although the principal function of HLA-E is to present HLA class Ia signal peptides to CD94/NKG2 receptors on natural killer (NK) cells, and thereby regulate NK activity, the detection of HLA-E-restricted CD8⁺ T cells in the context of diverse infections, including Mycobacterium tuberculosis (Mtb), cytomegalovirus (CMV), Salmonella typhi, and Epstein-Barr virus, suggests that HLA-E may play a role in adaptive immunity.^{6,7,8,9,10,11,12,13} These observations have spurred efforts to develop new biologics and vaccines directed to HLA-E.^14,15

The discovery of an unprecedented, crucial role for MHC-E-restricted CD8⁺ T cells in intercepting and eliminating a pathogenic simian immunodeficiency virus (SIV) challenge in a rhesus CMV-SIV (RhCMV-SIV) vaccination model provided a strong rationale to exploit HLA-E-mediated antigen presentation in vaccines and immunotherapeutics against HIV.¹⁶ Isolation of CD8⁺ T cells that recognize HLA-E-binding HIV Gag peptides in the naive T cell repertoire of HIV-seronegative donors and in people living with HIV (PLWH) lent support to the notion that the MHC-E-mediated protection observed in RhCMV-SIV vaccinated animals could be translated to humans.^17,18,19 However, the tractability of the HLA-E ligandome of HIV requires careful evaluation given the lack of clarity regarding the mechanisms by which HIV antigens may gain access to the HLA-E presentation pathway.²⁰

In this study, we performed a systematic search for HLA-E ligands in the HIV proteome using both unbiased and targeted approaches.²¹ Peptides from HIV proteins were positively identified in HLA class Ia molecules immunopurified from infected cells but not in purified HLA-E molecules. In addition, 80 bioinformatically predicted HIV peptides showed poor stability in screening assays, with the previously reported HLA-E ligand, Gag_275–283, confirmed as the top-scoring peptide. Following the successful engineering of a high-affinity soluble bispecific TCR (immune mobilizing monoclonal TCR against virus, ImmTAV) for HLA-E-HIV Gag_275–283 complexes, we observed inconsistent peptide presentation in killing assays using HIV-infected targets. These results were corroborated by observations that continual loading of HLA-E with exogenously supplied peptide was necessary to elicit consistent ImmTAV-mediated T cell responses. This comprehensive study underscores the distinct instability of HIV peptide-HLA-E complexes and provides a possible explanation for the rarity of HLA-E-restricted virus-specific T cell responses in PLWH.

Results

Immunopeptidome analysis and empirical testing of predicted HLA-E binders demonstrate instability of putative HIV HLA-E peptides

To explore the HIV peptidome in virus-infected cells in an unbiased manner, we isolated peptide-HLA complexes from HIV-infected C8166 cells and used liquid chromatography-tandem mass spectrometry (LC-MS/MS) to identify HLA class I–associated peptides according to established methods (Figure 1A).²¹ Infection was confirmed by intracellular staining, which showed that 19.4% cells (of a total of 1 billion) expressed Gag p24 (data not shown). C8166 cells were chosen because they are highly permissive to HIV, they express HLA-E at similar levels to primary CD4⁺ T cells (Figure 1B), and they are HLA-A∗02⁻, thus eliminating the possibility of low-affinity HLA-E peptides preferentially binding to HLA-A∗02 due to a shared binding motif.²² In total, we identified 43,741 unique peptides at a false discovery rate of 1% (peptide spectrum match), with 95% of peptides between 8 and 14 amino acids in length. The pan-HLA class I fraction (W6/32, anti-HLA class I A/B/C antibody) contained 57 unique HIV-derived peptides, the majority of which were 9-mers, as expected (Figure S1).²³ HLA class I leader peptides (LPs) were identified in the 3D12 fraction (anti-HLA-E antibody), confirming that we had successfully enriched for known HLA-E binders. We also identified 22 unique HIV peptides in the 3D12 fraction, 15 of which were between 8 and 14 amino acids in length. However, none of these peptides were predicted HLA-E binders,^7,24 nor did they stabilize HLA-E at the cell surface (Figure S1; Table S1).²⁵ The most likely explanation for the lack of HLA-E HIV peptides identified from this approach was poor stability and/or low abundance of the pHLA-E complexes.

Figure 1.

HIV peptides are not detected in the HLA-E immunopeptidome of virus-infected cells by highly sensitive LC-MS/MS

(A) Diagram showing workflow for immunoprecipitation of HLA-E- and HLA class Ia-peptide complexes from cell lysates using 3D12 and W6/32 antibodies, respectively, followed by identification of eluted peptides by LC-MS/MS using discovery (top) and targeted (bottom) proteomics approaches. (B) Cell surface HLA-E expression on uninfected C8166 (top, HLA-E in red, isotype control in gray) and primary CD3⁺CD8⁻ T cells from an HIV-naive donor (bottom, HLA-E in blue, isotype control in gray), indicating comparable expression (geometric mean fluorescence intensity, HLA-E/isotype control). (C) Peptide copies/cell for Gag_275–283 (solid circle, left), Mtb inhA_53–61 (solid diamond, right), and HLA-A1 or HLA-A2 LPs (open symbols) were quantified in peptide-pulsed C8166 cells (10 μM Gag_275–283 or Mtb inhA_53–61 peptides, 2 h), Gag_275–283 or Mtb inhA_53–61 minigene-transfected PANC1 cells, or HIV-infected C8166 cells (for Gag_275–283 only) by targeted MS with SIL peptide standards. (D) Comparison of stability of pHLA-E complexes containing either HIV Gag_275–283 or Mtb inhA_53–61 peptide as determined by refold yield, Tm,²⁶ ILT2 binding (BLI²⁶), or cell surface stabilization (flow cytometry). See materials and methods and Table S1 for details.

To address the possibility of a low abundance of pHLA-E complexes, in parallel we performed a targeted experiment using the stable isotope labeled (SIL) Gag_275–283 peptide, a previously reported HLA-E ligand, as an internal standard (Figure 1A).^17,26,27 Parallel reaction monitoring (PRM) is a highly sensitive approach to detect and quantify peptides of interest using high-resolution MS.^28,29 Positive controls included PANC1 cells (∼400 million) transfected with minigenes encoding either the Gag_275–283 peptide or a known HLA-E binding Mtb enoyl reductase-derived peptide, inhA_53–61 and C8166 cells (>200 million) pulsed with the above peptides. Immunopurification was again performed with 3D12 and W6/32 antibodies; in addition, an HLA-A∗02-specific antibody was used to enable isolation of HLA-A∗02 peptides from PANC1 cells.

No Gag_275–283 was detected in the 3D12 immunopurifications from either HIV-infected C8166 cells or the minigene-transfected PANC1 cells. Furthermore, peptide quantification of C8166 cells pulsed with Gag_275–283 peptide (10 μM) yielded only 0.5 copies/cell. By contrast, LPs from HLA-A∗01 (C8166 cells) or HLA-A∗02 (PANC1 cells) were detected at a high copy number (1254–1775 and 558–637 copies/cell, respectively), demonstrating consistent detection of peptides from stable pHLA-E complexes. In addition, C8166 pulsed with Mtb inhA_53-61 peptide or PANC1 cells endogenously expressing Mtb inhA_53–61 (minigene) had high levels of Mtb inhA_53–61 detected in the 3D12 immunopurifications (>3,900 and >1,600 copies/cell, respectively) (Figure 1C).

To further understand the lack of LC-MS/MS-identified HLA-E-binding HIV peptides from infected cells, we used bioinformatic predictions coupled with high-throughput biochemical and cellular assays to screen and triage the stability of putative HLA-E binding 9-mer and 10-mer peptides from across the HIV proteome. Due to the large number of possible peptides, we focused on Gag and Pol because these proteins have the highest conservation across viral clades. Over 80 peptides were screened, including Gag_275–283 and Gag_335–343 (KALGPAATL) (Table S1).^17,18,30 Peptides were ranked according to (1) pHLA-E thermostability (Tm), (2) pHLA-E complex stability by biolayer interferometry (BLI), (e) cell surface stabilization by flow cytometry, and (4) whether they could be produced as a soluble complex using refolding methods. Gag_275–283 ranked highest, with BLI (t_1/2) values of 11.4 and 21 min for the 6V “index” (RMYSPVSIL) and 6T variant (RMYSPTSIL), respectively, and quantifiable refolded complexes. The next highest ranked peptides passed only one of the four assays (Table S1). These observations are consistent with previous studies showing low stability of the Gag_275–283 peptide, which was most apparent when compared with the Mtb inhA_53–61 peptide (stability and Tm values: 11.4 min versus 231.4 min and “not possible to determine” versus 48.2°C; Figure 1D).^26,27

High-affinity TCRs distinguish stable and unstable pHLA-E complexes in functional cellular assays

We have previously developed picomolar affinity HLA-A∗02:01-Gag_77–85-specific TCRs that can eliminate HIV-infected cells expressing very low pHLA copy numbers.^31,32,33 Building on this, we successfully adapted our TCR discovery platform to enable the isolation of HLA-E-restricted TCRs from naive human libraries. A wild-type TCR was affinity enhanced by >10⁵-fold, achieving 90 pM affinity for Gag_275–283; an ImmTAV molecule was then generated by fusing the TCR to an anti-CD3 scFv domain (Figure 2A). Strong affinity to both the Gag_275–283 peptide (RMYSPVSIL) and the common variant 6T (RMYSPTSIL) was demonstrated by surface plasmon resonance (SPR) and specificity was confirmed by a lack of measurable off-target binding to HLA class Ia leader sequences (LP mixes; Figure 2A). Potency of the ImmTAV molecule was assessed in a dose-response T cell redirection assay involving coculture of peptide-pulsed C8166 cells with peripheral blood mononuclear cells (PBMCs) from five HIV-naive donors and titrated concentrations of the ImmTAV. A dose-dependent interferon-γ (IFN-γ) response was observed, with a mean half-maximal effective ImmTAV concentration of 267 pM (Figure 2B). In all of the aforementioned experiments, a consistent IFN-γ response was elicited in the presence of supraphysiological concentrations of the Gag_275–283 peptide (50–100 μM), which we predicted would stabilize HLA-E on the cell surface based on prior studies.^19,27,34

Figure 2.

Affinity-enhanced soluble TCR bispecific molecules confirm the poor stability of Gag_275–283 in HLA-E

(A) Binding of wild-type TCR (left) or affinity-enhanced TCR (center) to immobilized HLA-E∗01:03-Gag_275–283 was assessed by steady-state analysis using a serial dilution of TCR or single-cycle kinetics, respectively; affinity (K_D) and t_1/2 to both the 6V and 6T variants are indicated below. (Right) Binding kinetics of affinity-enhanced TCR specific for Gag_275–283 (6V and 6T variants) was assessed by single-cycle kinetics. Two panels of LPss (LP mix 1 and 2) were tested as controls for cross-reactivity. (B) T cell redirection by an affinity-enhanced TCR-anti-CD3 bispecific molecule (ImmTAV) was assessed in an IFN-γ ELISpot assay in which peptide-pulsed (50 μM) C8166 cells were cocultured with PBMCs from healthy donors (n = 5; average EC50 = 267 pM, r² = 0.974) together with titrated concentrations of Gag_275–283-specific ImmTAV (100 fM–10 nM). Data shown are representative of 2 independent experiments. (C) (Left) Impact of peptide wash step on ImmTAV-mediated T cell redirection was assessed by IFN-γ ELISpot assay: C8166 cells pulsed with titrated concentrations of Gag_275–283 peptide (5–50 μM) were either washed (dashed line) or not washed (solid line) after pulsing, and then cocultured with PBMCs (n = 5 donors) together with Gag_275–283-specific ImmTAV (10 nM). Data shown represent mean ± SD and are representative of 2 experiments. (Right) Impact of peptide wash step on ImmTAV versus ImmTAB-mediated T cell redirection was assessed by IFN-γ ELISpot assay: C8166 cells pulsed with 10 μM peptide were either washed (unfilled bars) or left unwashed (filled bars) after pulsing, and then cocultured with PBMCs (n = 5 donors) and 10 nM ImmTAX. Gag_275–283 peptide and specific ImmTAV shown in black versus Mtb inhA_53–61 peptide and specific ImmTAB in gray. (D) To confirm HLA-E presentation of Gag_275–283 peptide, peptide-pulsed (10 μM) C8166 cells were washed 2 h postpulsing and cocultured with PBMCs (n = 3 donors), ImmTAV (10 nM), and blocking antibodies against HLA-E (3D12) and HLA-A2 (BB7.2) (25 μg/mL each). Responses were assessed by IFN-γ ELISpot assay. In parallel, Mtb inhA_53–61 peptide-pulsed (10 μM) C8166 cells were cocultured with PBMCs (n = 3 donors) plus the corresponding ImmTAB and the same blocking antibodies. Data shown are representative of 2 independent experiments.

To probe the sensitivity of the ImmTAV to cell surface pHLA-E levels, T cell redirection assays were subsequently performed with peptide titrations, with or without a wash step 2 h after peptide pulsing. An IFN-γ response was maintained over peptide concentrations of 10–50 μM when excess peptide was present in the culture medium (no wash; Figure 2C). When the pulsed cells were washed to remove excess peptide, a lower magnitude of IFN-γ response was elicited and, furthermore, declined steeply at peptide concentrations ≤25 μM (Figure 2C). By contrast, the peptide wash had no impact on the IFN-γ response of T cells redirected to targets pulsed with the Mtb inhA_53–61 peptide in a comparable assay using an immune mobilizing monoclonal TCR against bacteria (ImmTAB) (Figure 2C).

As a final step, we verified that Gag_275–283 and Mtb inhA_53–61 peptide presentation were both solely HLA-E mediated using blocking antibodies to HLA-E (3D12) and HLA class I A/B/C (W6/32) in T cell redirection assays with the HIV ImmTAV and the Mtb ImmTAB (Figure 2D). Of note, the Gag_275–283 response was completely inhibited by 25 μg/mL 3D12 antibody, whereas the Mtb inhA_53–61 response was only partially blocked, further highlighting the robustness of the HLA-E-restricted Mtb response; neither was blocked by the pan-class I antibody (Figure 2D).

Taken together, these observations confirmed the instability of HLA-E/Gag_275–283 complexes on cells, and furthermore, showed a continual supply of exogenous Gag_275–283 peptide for the duration of the assay (“no wash”) is able to stabilize HLA-E on the cell surface at levels required for T cell recognition in vitro.

Endogenously expressed HIV Gag_275–283 peptide is not presented at sufficient levels to trigger T cell activation or cytolysis

To determine whether endogenously derived Gag_275–283 peptide could be presented by HLA-E for T cell recognition, we explored two methods of intracellular antigen presentation: a target cell line expressing Gag_275–283 as a minigene and HIV infection of primary CD4⁺ T cells.

Minigenes encoded ubiquitin-fused peptide sequences based on the 6V and the 6T variants of Gag_275–283. PANC1 cells transfected with each of the Gag_275–283 variant minigenes were cocultured with polyclonal T cells, together with the HIV ImmTAV, in a T cell redirection assay; HIV peptide-pulsed targets were included as a positive control. In a parallel assay, PANC1 cells transfected with an Mtb inhA_53–61 minigene or pulsed with the corresponding peptide were tested with the Mtb ImmTAB molecule. The Gag_275–283 minigene-transfected cells failed to elicit an IFN-γ response, in contrast to the Gag_275-283 peptide-pulsed cells (6V variant, 3/3 donors; 6T variant, 1/3 donors) and Mtb inhA_53–61 minigene-transfected cells (Figure 3A).

Figure 3.

The Gag_275–283 peptide does not elicit a T cell response via HLA-E when generated from an endogenous source

(A) IFN-γ responses were assessed in a T cell redirection assay with minigene-transfected PANC1 cells cocultured with PBMCs (n = 3 donors) and ImmTAV (HIV) or ImmTAB (Mtb) molecules (10 nM). The minigenes encoded Gag_275–283 (6V) (left), Gag_275–283 (6T) (center), and inhA_53–61 (right). PANC1 cells pulsed with peptide were included as a positive control. (B) Representative flow cytometry plots showing T cell activation (top) indicated by CD25 and CD137 coexpression in CD8⁺ T cell population and percentage of activated CD8⁺ T cells (bottom) after coculture of HIV-infected primary CD4⁺ cells with autologous CD8⁺ T cells plus ImmTAV for 6 h (n = 5 donors). Gag_275–283 peptide-pulsed cells were included as a positive control. Data shown are representative of 2 independent experiments. (C) Representative flow cytometry plots showing Gag p24 expression in CD3⁺CD8⁻ T cells (left) and percentage of elimination of p24⁺ cells after coculture of HIV-infected CD4⁺ cells with autologous CD8⁺ T cells (n = 5 donors) plus ImmTAV in 3 independent experiments (right; experiment “a” only includes 2 donors). (D) Cell surface HLA-E expression (geometric mean fluorescence intensity, HLA-E/isotype control) in primary CD4⁺ T cells of 5 donors after in vitro infection with HIV for 7 days (uninfected control, total population from infected cell culture, and p24⁺ subpopulation of infected cell culture) representative of 2 time points.

The in vitro infection model consisted of primary CD4⁺ T cells from HIV-naive donors (n = 5) infected with HIV pNL4.3, which encodes the 6T variant, cocultured with autologous CD8⁺ T cells in the presence or absence of the ImmTAV molecule at an effector:target ratio of 1:1. HLA-A∗02:01⁻ donors were selected to avoid the possibility of Gag_275–283 presentation via HLA-A∗02:01 molecules.²² T cell activation, indicated by the upregulation of CD25 and CD137, was assessed after 24 h of coculture, and infected cell elimination was assessed on day 7, as described previously.²⁰ The addition of ImmTAV led to T cell activation above background levels only in the presence of exogenous peptide; no response to HIV-infected cells was seen with any donor CD8⁺ T cells (Figure 3B). Conversely, infected cell elimination was measurable in three separate experiments, albeit with intradonor variation. Cytolytic responses of 40%–45% (normalized to no-ImmTAV conditions) were observed in two donors in different experiments; all of the other responses were <20% (Figure 3C). Differences in cell viability or infection level did not account for interexperimental variation. The lack of T cell activation or consistent killing of infected cells was not due to the downregulation of HLA-E on the cell surface, since HLA-E was maintained on Gag p24⁺ cells, consistent with previous reports (Figure 3D).^30,35 These data collectively suggested that the presentation of endogenously derived Gag_275–283 by HLA-E was transient at best and was insufficient to trigger rapid T cell activation or a reproducible effector response.

Discussion

This report describes an exhaustive search for HLA-E ligands in the HIV proteome that may be exploited as therapeutic targets. No HLA-E-bound HIV peptides were identified in an exploratory immunopeptidomic analysis of HIV-infected cells, and the previously reported Gag_275–283 epitope was not detected using a targeted approach with stable isotope-labeled peptides.^17,19,26,27 However, the peptides that were bound to classical HLA class I molecules on infected cells included known HIV epitopes, and the well-described Mtb inhA_53–61 peptide was detected in HLA-E purified from minigene-transfected cells, confirming the validity of our experimental approach.^23,36,37 Empirical testing of the half-life, thermal stability, and cell surface expression of bioinformatically predicted peptides, including the two most common variants of Gag_275–283, showed that their stability in HLA-E was universally low, providing an explanation for their lack of detection in HIV-infected cells by MS. Although the use of an HLA-E∗01:01-expressing cell line for the discovery MS experiment may have limited the possibility of detecting HIV peptides due to differences in the peptide repertoires presented by these two alleles,^4,17,38 this is unlikely to be the sole explanation since no Gag_275–283 peptide was eluted from the HLA-E∗01:03-expressing minigene-transfected PANC1 cells.

In T cell redirection assays with an affinity-enhanced soluble bispecific TCR, we showed that reactivity to HLA-E-Gag_275–283 was dependent on a continual supply of peptide in the culture medium. Although the cell surface stability of HLA-E is low compared with HLA class Ia molecules in general, our observations likely reflect the particularly short half-life of Gag_275–283 pHLA-E complexes, which was reported by Barber et al.²⁶ Together, these observations provide a mechanistic basis for the infrequent detection of HIV-specific HLA-E-restricted T cell responses in PLWH.^17,18,19 Although this contrasts sharply with HLA class Ia-restricted CD8⁺ T cells, which are measurable in the majority of PLWH, sometimes at frequencies of up to 20% total CD8⁺ T cells,^39,40,41 our data align with prior studies that have identified a correlation between HLA class Ia-peptide complex stability and CD8⁺ T cell immunodominance hierarchy.^42,43,44

Assessment of endogenous processing and presentation of Gag_275–283 via HLA-E was a necessary step in the validation of this peptide as a potential biologic or vaccine target. The lack of response to minigene-transfected cells and the inconsistent responses to HIV-infected primary CD4⁺ T cells suggested that Gag_275–283 was not generated in sufficient quantity to compete with HLA class I LPs for HLA-E binding. Alternatively, HLA-E-Gag_275–283 complexes may have been generated under the experimental conditions used, but due to their short half-life did not attain measurable levels on the cell surface. This is supported by a recent study showing that HLA-E is typically more abundant in the endoplasmic reticulum and endosomes than on the cell surface, largely as a consequence of the limited availability of high-affinity peptides.⁴⁵ The sporadic responses we did observe against HIV-infected cells in this study (2/12 data points) could reflect stochastic interactions between the ImmTAV and HLA-E-Gag_275–283 complexes on infected cells.⁴⁵ We excluded the possibility of HLA-E downregulation in the present study, because cell surface HLA-E expression was maintained in the targets (Gag⁺ cells) after in vitro infection, consistent with previous reports^30,35,46 This, together with the positive signal obtained with the Mtb inhA_53–61 minigene in a T cell redirection assay, provides confidence that peptide instability was the sole explanation for the lack of a reproducible HLA-E-Gag_275–283-specific T cell response in the infected cell elimination assay. In addition, in a parallel study, we used a soluble bispecific TCR molecule targeted to HLA-E-inhA_53–61 (ImmTAB) to demonstrate killing of Mtb-infected cells; this showed that HLA-E-inhA_53–61 complexes were generated by the processing of naturally expressed inhA and were present on the cell surface, confirming that TCR bispecific molecules can detect stable HLA-E peptide complexes in a relevant model system (data not shown).

The mechanisms underpinning the loading of pathogen-derived peptides on HLA-E are best defined for Mtb and modified rhesus CMV strains that have been successfully used as an SIV vaccine vector.^16,47,48 It has been proposed that peptide loading may occur via transporter associated with antigen processing–independent mechanisms in intracellular compartments such as endosomes, especially considering the rapid internalization of HLA-E from the cell surface, although the conditions that enable this have yet to be defined.^20,45 This highlights the need for a better understanding of the circumstances that could lead to the priming of HLA-E-restricted T cell responses to HIV antigens in humans. It is notable that the previously reported naturally occurring T cell responses were typically detected in assays using very high peptide concentrations and/or peptide trapping methods to overcome the poor stability of Gag peptides in HLA-E.^26,27 In vitro studies demonstrate that the formation of fully folded and stable HLA-E-Gag_275–283 complex requires excess peptide³⁴; therefore, we speculate that the reported T cell responses are a consequence of low-affinity interactions with transitory and heterogeneous pHLA-E complexes in vivo.

To conclude, stable peptide presentation is a prerequisite for consistent on-target activity and clinical efficacy of TCR-based immunotherapeutics and is likely to be crucial to the efficacy of T cell–based vaccines.^2,49 The strong biochemical and biological evidence for the unstable interaction of Gag_275–283 and other HIV peptides with HLA-E presented here and by others highlight the challenges that will need to be addressed in the pursuit of HLA-E-targeted drug and vaccine development.⁴⁵

Materials and methods

Cell lines and primary cells

Cell lines were obtained from the following suppliers: C8166 cells (E∗01:01 homozygous) (European Collection of Authenticated Cell Cultures, Salisbury, UK; catalog no. 88051601) and K562 cells DSMZ (Braunschweig, Germany; catalog no. ACC10), transduced with single-chain HLA-E∗01:01 or 01:03-β2m. These cell lines were cultured in RPMI medium. PANC1 cells (E∗01:01/E∗01:03) were supplied by American Type Culture Collection (catalog no. CRL-1469) and were cultured in DMEM medium. All of the media were supplemented with 1% (v/v) penicillin/streptomycin, 2 mM l-glutamine, and 10% fetal calf serum (R10 or D10). Cell line authentication and mycoplasma testing were routinely carried out by the LGC Standards Cell Line Authentication Service (Teddington, UK) and Mycoplasma Experience Ltd. (Redhill, UK), respectively. Healthy donor (presumed HIV naive) PBMCs were obtained from StemCell Technologies (Cambridge, UK), Cellular Technology Ltd. (Rutesheim, Germany), or Tissue Solutions (Glasgow, UK).

Preparation of HIV-infected cell pellets

C8166 cells were infected with HIV-1 IIIB (MOI 0.01; NIH AIDS Reagent Program, Rockville, MD) by spinoculation for 2 h at 2,000 rpm and cultured for 2 days. Infection was confirmed by intracellular staining for Gag p24 as previously described.⁵⁰

Generation of minigene-expressing cell lines

PANC1 cells were transfected with a linearized minigene construct containing blasticidin selection marker and HIV or Mtb peptides (Gag_275–283 RMYSPTSIL or RMYSPVSIL and inhA_53–61 RLPAKAPLL) fused to ubiquitin. At 24 h after transfection, cells were grown in cell culture media supplemented with 20 μg/mL blasticidin to select for minigene stable cell clones.

Immunopurification of peptide-HLA complexes

HIV-infected C8166 cell pellets (>1 × 10⁹ cells) were suspended in ice-cold lysis buffer (50 mM Tris, pH 8.0, 150 mM NaCl, 0.5% NP-40) containing 1× HALT Protease and Phosphatase Inhibitor Cocktail (Thermo Fisher Scientific, Waltham, MA) and incubated for 1 h at 4°C with agitation. Iodoacetamide was added to the lysate at a final concentration of 25 mM to alkylate cysteine residues during centrifugation at 15,000 × g for 45 min at 4°C. Lysates were subsequently applied to 3D12 (αHLA-E, minimally cross-reactive with HLA class Ia⁵¹) antibody-crosslinked protein L agarose resin (Agarose Bead Technologies, Madrid, Spain) to enrich for peptide-HLA-E complexes, and the flow-through was applied to W6/32 (αHLA class I) antibody-crosslinked protein A agarose resin (Sigma, Gillingham, UK). Columns were generated as previously described, with minor modifications.²¹ Briefly, columns were packed with protein L agarose resin (3D12) or protein A agarose (W6/32). The 3D12 or W6/32 antibody was crosslinked for 1 h at room temperature using 30 mM dimethyl pimelimidate dihydrochloride in 200 mM triethanolamine, pH 8.3. The crosslinking reaction was terminated using 100 mM Tris-HCl, pH 8.0. After incubation with the lysate, the columns were flushed with wash buffer 1 (50 mM Tris-HCl, pH 8.0, 150 mM NaCl, 5 mM EDTA, 0.05% NP-40), wash buffer 2 (50 mM Tris-HCl, pH 8.0, 150 mM NaCl), wash buffer 3 (50 mM Tris-HCl, pH 8.0, 450 mM NaCl), and wash buffer 4 (50 mM Tris-HCl, pH 8.0). Peptides were eluted using 5% acetonitrile (ACN) and 0.5% trifluoroacetic acid (TFA) and desalted using C18 SepPak cartridges (Waters, Wilmslow, UK).

PANC1 cell pellets (0.5 × 10⁹ cells) were lysed and processed as described above, but with additional immunopurifications with BB7.2 (αHLA-A2) antibody-crosslinked protein A agarose resin (Agarose Bead Technologies) between 3D12 and W6/32 immunopurifications.

LC-MS/MS

Lyophilized peptides were reconstituted in 5% ACN and 0.1% TFA. Stable isotope-labeled peptides (100 fmol) were spiked into each sample analyzed by PRM. MS analysis was performed on an Orbitrap (OT) Fusion Tribrid mass spectrometer (Thermo Fisher Scientific) connected to an UltiMate 3000 RSLCnano System (Thermo Fisher Scientific). Peptides were separated on an EASY-Spray nanoLC C18 column (75 μm inner diameter × 500 mm, 2 μm, 100 Å) using a linear gradient from 8% to 30% solvent B (0.1% formic acid in acetonitrile) for 80 min, followed by 30%–50% for 10 min at a flow rate of 250 nL/min. The OT Fusion Lumos was operated in positive-ion mode using a data-dependent acquisition (DDA) or a hybrid DDA and PRM method. Precursor ion (MS1) scans were performed in the OT mass analyzer in the range of 300–1,200 m/z, at a resolution of 120,000 (DDA) or 60,000 (hybrid) at 200 m/z. Precursor ions were isolated using a quadrupole mass filter (1.2 m/z isolation width) and fragmented using high-energy collisional dissociation set to 28%. MS2 spectra were acquired in the OT at a resolution of 30,000 at 200 m/z. The number of MS2 scans between full scans were determined on the fly to maintain a 3-s (DDA) or 1-s (hybrid) fixed duty cycle. Dynamic exclusion of ions was implemented using a 45-s exclusion duration. For the PRM method, the OT Fusion Lumos was operated in tMS2 mode at a resolution of 60,000 at 200 m/z. Peptides were selected for MS2 data acquisition using predetermined isolation windows.

Raw mass spectrometry data was analyzed using PEAKS Online X build 1.8 (Bioinformatics Solutions, Waterloo, Canada) and a proteome database consisting of the SwissProt Human reference proteome with isoforms and Uniprot HIV reference proteome sequences (42,866 database entries). A precursor mass tolerance of 10 ppm and a fragment mass tolerance of 0.02 Da was used during the database search. The enzyme specificity was set to unspecific. Carbamidomethylation was set as fixed modification, whereas protein N-terminal acetylation, deamidation of N and Q, and methionine oxidation were considered variable modifications. All PRM data were analyzed using Skyline (version 21.1.0.278).⁵² Fragment ions were extracted and light:heavy ratios of peptide pairs were automatically determined in Skyline. All of the extracted ion chromatograms were manually inspected before copy-number estimation. HLA motif clustering was performed using GibbsCluster version 2.0, and HLA-binding affinity predictions were performed using NetMHCpan-4.1 (DTU Health Tech, Kongens Lyngby, Denmark).^53,54 The data file can be found in Table S2.

Computational prediction of HLA-E-binding peptides

HIV per-protein amino acid sequences from clades B and C obtained from GenBank were aligned using MUSCLE and searched for 9-mer or 10-mer peptides within Gag and Pol using netMHCpan4.0.⁵⁵ Those with rank NetMHCpan4.0 affinity <1%, as well as peptides with rank NetMHCpan4.0 affinity <10% and GRAVY solubility score >-0.5, were selected for further analysis, together with associated variants present in >50% of clade B and/or C sequences across the Gag and Pol genes.⁵⁶ Approximately 80 peptides (including some viral variants and previously reported epitopes) were selected for assessment with stability assays.^17,18

pHLA-E stability assays

Putative HLA-E ligands were assessed for pHLA-E Tm (thermal shift), pHLA-E complex stability (by BLI), ability to refold pHLA-E complexes, and pHLA-E cell surface stabilization as previously described, with modifications to increase throughput.²⁶ Thermal shift assays were performed using the Quantstudio 6 (Applied Biosystems, Waltham, MA). Peptides were obtained by chemical synthesis (Peptide Protein Research, Fareham, UK) and solubilized in DMSO. Peptides of interest were added to refolded and purified HIV Gag_275–283-HLA-E∗01:03 complexes (in PBS at 0.25 mg/mL) at a 60:1 M ratio, combined with SYPRO Orange protein gel stain. This mixture was heated from 22°C to 95°C at 1 °C/min while detecting fluorescence using the FAM (fluorescein) filter set, with excitation and emission wavelengths at 495 and 518 nm, respectively. Positive hits (indicated by a typical melting curve) were analyzed using Protein Thermal Shift Software version 1.4 (Thermo Fisher Scientific) to determine Tm. pHLA-E complex stability was assessed using HBV Env_371–379 refolded with HLA-E as the starting complex, followed by incubation with peptides of interest (at 750 nM) to enable peptide loading by exchange. pHLA-E monomers were immobilized on a streptavidin-coupled biosensor tip and assessed for binding of soluble ILT2 (Ig-like transcript 2 receptor) using BLI (Octet RED96, ForteBio, Fremont, CA); positive hits showed an increase in ILT2 binding relative to the no-peptide control; complex stability (t_1/2) was determined by fitting the ILT2 binding curve plotted against time using nonlinear regression (GraphPad Prism version 8.3 or later).To determine pHLA-E refold yields, HLA-E heavy chain and β2M were refolded with the peptide of interest.⁵⁷ Yield was calculated by determining the amounts of the purified soluble complex relative to the refold volume. Cell surface stabilization of HLA-E on K562 cells transduced with single-chain HLA-E-β2m was assessed as described by Barber et al. with minor modifications.²⁶ Cells were cultured with peptide overnight at 26°C followed by a 2-h culture at 37°C. Following staining with phycoerythrin (PE)-conjugated anti-HLA-E antibody (3D12), samples were acquired on a MACSQuantX (Miltenyi, Bisley, UK) flow cytometer and analyzed with FlowJo software (version 10.6.2; Tree Star, Ashland, OR). The cell surface stabilization of HLA-E by the peptide was assessed using the following formula: (<5% shift = nonbinding peptide): % shift of surface HLA-E × PE = [(geoMeanFI × PE(peptide − unpulsed control))/geoMeanFI × PE(inhA_53–61 – unpulsed control)] × 100.

Generation of ImmTAX molecules

TCRs specific for HLA-E-HIV Gag_275–283 or Mtb inhA_53–61 complexes were isolated from naive TCR libraries and affinity enhanced by phage display as previously described.^33,58,59 Negative selections were performed using pHLA-E complexes containing HLA class Ia leader sequences. An anti-CD3 scFv was fused to the TCR-β chain to generate a bispecific retargeting molecule (ImmTAX).

TCR binding affinity and cross-reactivity assessment by SPR

Biotinylated peptide-HLA-E complexes were immobilized on streptavidin-coated sensor chips for assessment of TCR affinity by SPR on a Biacore 8K system (GE Healthcare, Chicago, IL) with either steady-state affinity or single-cycle kinetics analysis, as previously described.^26,31,58,60

To assess potential cross-reactivity, two pools, each comprising 4 biotinylated LP-HLA-E complexes, were immobilized on individual flow cells of a streptavidin-coated sensor chip, following which the TCR was injected at a concentration of 320 nM (∼10,000× K_D) and responses recorded.

IFN-γ ELISpot

IFN-γ enzyme-linked immunospot (ELISpot) assays were performed in accordance with the manufacturer’s instructions (BD Bioscience, Wokingham, UK). In brief, C8166 cells (5 × 10⁴ cells/well) were pulsed with peptide (10–50 μM) for 2 h, washed or left unwashed, and then cocultured with cryopreserved HIV-naive donor PBMCs (1:1 ratio) and ImmTAX molecules overnight. IFN-γ release was quantified using an automated ELISpot reader (Immunospot Series 5 Analyzer, CTL, Cleveland, OH) and analyzed using GraphPad Prism version 9.0 (GraphPad Software, San Diego, CA). In selected experiments, cell surface HLA-E was blocked using anti-HLA-E (3D12) or anti-HLA-A2 (BB7.2) (25 μg/mL, InVivo BioTech Services, Hennigsdorf, Germany).

HIV-infected cell activation and elimination assays

PBMCs from HIV-naive donors were stimulated with anti-CD3 (100 ng/mL, BioLegend, San Diego, CA) and IL-2 (100 IU/mL; Peprotech, Hamburg, Germany) for 5 days, after which CD8⁺ cells were positively selected using anti-CD8 magnetic beads (Miltenyi). The CD8⁻ fraction was infected with HIV pNL4.3 by spinoculation for 2 h, then cocultured with autologous CD8⁺ T cells at a 1:1 ratio, with or without the addition of ImmTAV molecules. Peptide-pulsed cells were included as a positive control. T cell activation was measured after 24 h coculture by cell surface staining with anti-CD3 (allophycocyanin [APC]-Cy7), anti-CD8 (APC), anti-CD25 (PE), anti-CD137 (BV421) (all BioLegend), plus a viability stain (LiveDead Aqua) and analysis by flow cytometry. HIV-infected cells were quantified by intracellular Gag p24 staining on day 7 of coculture, gating on CD3⁺CD8⁻ cells to account for Nef-mediated CD4 downregulation, as previously described.^31,50 The percentage of elimination was calculated using the following equation: [(fraction of p24⁺ cells in CD8⁻ T cells cultured with CD8⁺ T cells (no ImmTAV) – fraction of p24⁺ in CD8⁻ T cells cultured with CD8⁺ T cells plus ImmTAV)/fraction of p24⁺ cells in CD8⁻ T cells cultured with CD8⁺ T-cells (no ImmTAV)] × 100%. HLA-E expression was assessed by cell surface staining with anti-HLA-E (3D12; PE) or isotype control (mouse immunoglobulin G1, κ; PE) and normalizing the geometric mean (GeoMean) of the fluorescence intensities (HLA-E/isotype). Samples were acquired on a MACSQuant X flow cytometer (Miltenyi) and analyzed using FlowJo (version 10). Gating strategies are shown in Figure S2.

Data and code availability

The datasets generated and/or analyzed during the present study are available from the corresponding author on reasonable request.

Supplemental information

Table S2. Data files from immunopeptidomics

(Figure 1) including (A) statistics of PEAKS database search results from 3D12 (anti-HLA-E) immunopurifications, (B) peptides identified at 1% peptide spectrum match false discovery rate in 3D12 immunopurifications, (C) statistics of PEAKS database search results from W6/32 immunopurifications, and (D) peptides identified at 1% peptide spectrum match false discovery rate in W6/32 immunopurifications, all performed on HIV-infected C8166 cells