HLA-B and cysteinylated ligands distinguish the antigen presentation landscape of extracellular vesicles

Abstract

Extracellular vesicles can modulate diverse processes ranging from proliferation and tissue repair, to chemo-resistance and cellular differentiation. With the advent of tissue and immunological targeting, extracellular vesicles are also increasingly viewed as promising vectors to deliver peptide-based cancer antigens to the human immune system. Despite the clinical relevance and therapeutic potential of such 'cell-free' approaches, the natural antigen presentation landscape exported in extracellular vesicles is still largely uncharted, due to the challenging nature of such preparations and analyses. In the context of therapeutic vesicle production, a critical evaluation of the similarity in vesicular antigen presentation is also urgently needed. In this work, we compared the HLA-I peptide ligandomes of extracellular vesicles against that of whole-cells of the same cell line. We found that extracellular vesicles not only over-represent HLA-B complexes and peptide ligands, but also cysteinylated peptides that may modulate immune responses. Collectively, these findings describe the pre-existing provision of vesicular HLA complexes that may be utilized to carry peptide vaccines, as well as the propensity for different peptide and post-translationally modified ligands to be presented, and will outline critical considerations in devising novel EV vaccination strategies.

Fig. 1. Extracellular vesicle isolation and characterization.

a Schematic workflow of extracellular vesicle (EV) isolation. Lysates made from JY WC (WCL, green) and JY-derived EVs (EVL, purple) were used subsequently as input for immunoaffinity purification of HLA complexes (Figs. 2–4). b Nanoparticle tracking analysis (NTA). Size and concentration of isolated EVs were estimated. Purity (particle:protein ratio) of EVs was 3e9, with mean particle size of 121 nm. Representative trace. c Negative stain transmission electron microscopy (NS-TEM) analysis. Morphology of the isolated EVs were cup-shaped. Representative images. Scale bar: 200 nm. Full images provided in Supplementary Fig. 6. d SDS-PAGE of protein band patterns in WCL and EVL. Dominant protein bands in WCL and EVL were mutually exclusive. Full image provided in Supplementary Fig. 7. e HLA-I and cd81 western blot. From 5 µg load of EVL and WCL, prominent enrichment of HLA-I and CD81 were observed in EVL. Full images provided in Supplementary Fig. 8. f Quantitative proteomics analysis of WCL (n = 3) and EVL (n = 3). Numerous classical EV markers were enriched in EVL. g Quantitative proteomics comparison of EVL and WCL. HLA proteins, numerous cell surface CD proteins and tetraspanins were significantly enriched in EVL. Dashed lines indicate proteins with a 2-fold change in abundance, while dotted lines indicate proteins with p value ≤ 0.05.

Fig. 2. Global characterization of the EV HLA-I peptide ligandome.

a Preparative steps in the down-scaled HLA-I complex retrieval workflow. An input load of 120 μg was used for immunoaffinity purification of HLA-I complexes from both WCL and EVL. The flowthrough from 10 kDa filtration contains HLA peptide ligands that were subjected to LC-MS/MS analysis. The retentate contains HLA-I and HLA-I interacting proteins, which were digested, for the analysis of HLA-I abundance and interactome. b HLA-I peptide ligand specificity. From the 10 kDa flowthrough fraction, more than 80% of peptides detected were predicted to bind to the JY HLA type (HLA-A*02:01; HLA-B*07:02; HLA-C*07:02). c Western blot (WB) detection of HLA-I. HLA-I was detected in the eluate of WCL HLA-I IP, and more strongly in the eluate of EVL HLA-I IP. Full image provided in Supplementary Fig. 9. d Mass spectrometry analysis of immuno-purified HLA-I proteins and HLA-I peptide ligands. From EVs, more HLA-I proteins (cumulative MS intensities of HLA-A, HLA-B and HLA-C) and HLA-I peptide ligand species were detected. e Peptide length distribution. HLA-I peptide ligands from both whole-cells (WC) and EVs distributed similarly, and expectedly, between 8 to 12 amino acids. f Predicted binding affinity. Marginal differences in predicted binding affinity were observed in HLA-I peptides retrieved from WC and EVs. Box plots represent n peptides (where n has been annotated under each box). The 25%, 50% (median), and 75% quantiles are represented in each box, and the whiskers represent the ±1.5*IQR from the closest quantile. Bar plots represent the total pool of eluted ligands detected in three technical replicates from either WC (green) or EVs (purple).

Fig. 3. Unique characteristics of the EV HLA-I peptide ligandome.

a HLA-I peptide species overlap. The JY whole-cell HLA-I peptide ligandome is almost completely sampled in EVs. b Predicted peptide binding affinity to JY HLA type. More HLA-B binding peptides were detected from EVs. Distribution of strong (≤50 nM affinity) and weak (≤500 nM affinity) peptide binders respectively. Bar plots represent the total pool of eluted ligands detected in three technical replicates from either WC (green) or EVs (purple). c Proportion of HLA-A (blue), HLA-B (orange), HLA-C (yellow) binders. HLA-A binders were over-represented in the JY whole-cell HLA-I peptide ligandome in both peptide species count and summed abundance, which was corroborated by the relative abundance of captured HLA-A and HLA-B proteins. A more balanced proportion of HLA-A to HLA-B was observed in EVs. d HLA-I complexes purified from the plasma membrane show a similar abundance distribution to what observed on whole-cells. e Spectral count of CD20 co-immunoprecipitated with HLA-I. CD20 (MS4A1) was confidently detected exclusively in the EV HLA-I interactome.

Fig. 4. Post-translational modification (PTMs) analysis of EV and whole-cell HLA-I peptide ligands.

a PTM distribution. EV HLA-I peptide ligands are more frequently cysteinylated than HLA-I peptides from WC. Cysteinylation (yellow) was more abundantly detected on HLA-A binders (blue) than in HLA-B binders (orange). The number of cysteinylated ligands (n) derived from each allele is shown in parentheses. b Length distribution of the cysteinylated ligands. Y-axis was scaled to total number of peptides identified in three technical replicates from either WC or EVs. c Predicted binding affinity (nM) of all cysteine-containing peptides. Cysteinylated cysteine-containing peptides (yellow) were predicted to bind more strongly than non-cysteinylated cysteine-containing peptides.