Recruiting ESCRT to single-chain heterotrimer peptide MHCI releases antigen-presenting vesicles that stimulate T cells selectively Authors: Blade A. Olson, Kathryn E. Huey-Tubman, Zhiyuan Mao, Magnus A. G. Hoffmann, Pamela J. Bjorkman, Richard M. Murray, Stephen L. Mayo Published: 2025 ABSTRACT SignificanceImmune cells are known to naturally release pMHC-displaying extracellular vesicles (EVs), called antigen-presenting vesicles (APVs), which can orchestrate immune responses either directly or with the aid of antigen-presenting cells (APCs). For decades, researchers have pursued ways to replicate these APVs for immunotherapy by using chemically modified nanoparticles or by engineering the increased expression of APVs from immune cells which are typically low yield. Here, we present a broadly applicable platform for generating high concentrations of pMHCI-displaying APVs that can selectively modulate T cells, demonstrating a significant advance in the engineering of APVs for cell-free immunotherapy. Almost 30 y ago, Raposo et al. first reported on the release of B cell–derived extracellular vesicles (EVs) with functional peptide:MHC (pMHC) complexes on their surface (1). Since then, EVs that were previously labeled as simple cellular “debris” have been increasingly recognized for their immunoregulatory potential as a simpler, cell-free alternative to cell-based immunotherapies (2, 3). Naturally occurring pMHC-displaying antigen-presenting vesicles (APVs) in particular have been demonstrated to orchestrate immune responses either directly or with the aid of antigen-presenting cells (APCs) (4). Research teams have ventured to engineer more complex versions of these APVs: whole mammalian cells conjugated with immunostimulatory molecules (5–9), nanoparticles coated with APC-derived membranes via extrusion or sonication (10, 11), or engineered extracellular blebs from dendritic cells (12–16). These previous approaches demonstrated the therapeutic potential of APVs, but production is often low yield or requires the patient’s own cells to be harvested and processed before treatment can be administered. Significant attention has been given to manufacturing APVs via microfluidics (17), but existing microfluidic vesicles suffer from short half-lives and are susceptible to rupture due to osmotic changes and the turbulent flow of the lymphatic and cardiovascular systems. Polymer-coated iron-oxide nanoparticles conjugated with pMHCI have shown remarkable therapeutic potential for autoimmune diseases such as type 1 diabetes, but the stiff spherical structure of the nanoparticle scaffold reduces the potential surface area for ideal pMHC:T cell receptor (TCR) interactions (18–20); additionally, pMHC complexes are rigidly adhered to the nanoparticle surface, limiting pMHC movement and clustering considered important for the formation of a T cell stimulating immunological synapse (21). Optimal, engineered APVs would ideally recreate the endogenous form of APVs, while maximizing the concentration of proteins that are critical for immunoregulation. Here, we describe the engineered assembly and budding of peptide MHCI-displaying APVs from the cell surface of nonimmune cells by DNA transfection of an engineered single-chain heterotrimer (SCT) peptide MHCI (pMHCI). Building on a recently described method for generating eVLPs displaying the SARS-CoV-2 spike protein (22), we fused an endosomal sorting complex required for transport (ESCRT)- and ALG-2-interacting protein X (ALIX)-binding region (EABR) to the cytoplasmic tail of SCT pMHCI to directly recruit the ESCRT machinery and induce the release of pMHCI-displaying vesicles. Immuno-electron microscopy (immuno-EM) of the supernatant from transfected cells showed halos of black punctae surrounding exosome-like vesicles, confirming the presence of SCT pMHCI on APVs. ELISPOT assays show that SCT pMHCI/EABR APVs are capable of stimulating IFN-γ release from T cells at a level equivalent to refolded pMHCI tetramer, and that the APVs selectively stimulate T cells presenting their cognate TCR. Altogether, these results demonstrate a method for generating synthetic, native-like APVs in high copy number and valency with the therapeutic potential for improved cell-free immunoregulation. Results EABR Addition to the Cytoplasmic Tail of SCT pMHCI Promotes APV Release. To generate APVs from nonimmune cells, we attached an EABR sequence from the human CEP55 protein (23) to the cytoplasmic tail of MHCI (Fig. 1A). To ensure each APV presented the same, intended peptide on MHCI, the complete peptide:MHCI complex was formed into a SCT; the N terminus of the HLA-A*02:01 α chain was extended with a 4xG4S linker connecting to the C terminus of beta-2 microglobulin (β2 M), which was itself extended at its N terminus by a 3xG4S linker connecting to either the 10-mer MART-1 (MART1) or 9-mer NY-ESO-1(C165V)157-165 (NYESO1) cancer-related peptides (24). The EABR sequence used here was identical to the sequence used by Hoffmann et al. (22) to create enveloped virus-like particles (eVLPs) displaying the SARS-CoV-2 spike protein, and was attached to the cytoplasmic tail of the HLA-A*02:01 α chain by a G3S linker (Fig. 1B). Adopting the nomenclature of Chour et al.’s thermostability studies (25, 26) on previous SCT pMHCI designs, our first SCT pMHCI construct was named “D1/EABR” to denote a “D1” SCT design with an “EABR” sequence appended to the end of its cytoplasmic tail. A second “D9” variant of SCT pMHCI, containing Y84C and A139C substitutions, was also tested given its reported improvements in thermostability and TCR binding compared to the D1 design (SI Appendix, Table S1) (27). EABR addition to the cytoplasmic tail of SCT pMHCI promotes APV budding and release. (A) Schematic of membrane-bound SCT pMHCI proteins on a cell surface containing cytoplasmic tail EABR additions that induce budding of an APV comprising a lipid bilayer with embedded SCT pMHCI proteins. (B) SCT pMHCI/EABR construct. Top: the SCT pMHCI protein is composed of a peptide fused to a 3x(Gly4Ser) linker, β2 M, a 4x(Gly4Ser) linker, HLA-A*02:01 α chain, a Gly3Ser spacer, and an EABR sequence. Bottom: EABR sequence. (C) Schematic showing production and purification of APVs. (D) Chromatogram representative of the typical yield of SCT pMHCI APVs from a 100 mL transfection of Expi293F cells. Inset highlights void volume peak. (E) Western blot analysis detecting SCT pMHCI protein from APVs purified by ultracentrifugation on a 20% sucrose cushion from transfected HEK293T cell culture supernatants. Cells were transfected with SCT MART1/D1, SCT MART1/D1/EABR, SCT MART1/D9, SCT MART1/D9/EABR, and SARS-CoV-2 spike(S)/EPM/EABR (S/EPM/EABR) constructs. Rabbit anti-HLA-A primary antibody and Alexa488-conjugated anti-rabbit secondary antibody were used to detect SCT pMHCI protein. (F) Immuno-EM images of SCT MART1/D1/EABR APVs purified from transfected Expi293F cell culture supernatants by ultracentrifugation and SEC. Primary rabbit anti-HLA-A antibody identical to primary antibody used in (E). Secondary 6 nm gold-conjugated anti-rabbit antibody appears as black punctate in image. Left: representative APVs are highlighted in boxes. (Scale bar, 100 nm.) Right: close up of black-boxed APV. (Scale bar, 20 nm.) APVs were expressed by transfecting HEK293T or Expi293F cells with DNA plasmids encoding the SCT pMHCI constructs. Vesicles were purified by ultracentrifugation of cell culture supernatant on a 20% sucrose gradient as is typical for the purification of viral particles and exosomes (28) (Fig. 1C). Ultracentrifuged product was resuspended in pH 7.4 phosphate-buffered saline (PBS) and optionally further purified by size exclusion chromatography (SEC) to yield a high concentration of product in the void volume (Fig. 1D). Western blot analysis of the purified product showed that the addition of the EABR sequence to the cytoplasmic tails of both the MART1/D1/EABR and MART1/D9/EABR constructs generated higher levels of MHCI compared to their non-EABR counterparts (Fig. 1E). Because pMHCI is difficult to resolve with cryo-EM or cryo-ET, we prepared immuno-electron microscopy grids of the purified product to confirm the presence of pMHCI-displaying APVs, which presented as cup-like, collapsed spheres resembling the canonical shape of exosomes that partially collapse upon paraformaldehyde fixation and negative staining (29) (Fig. 1F and SI Appendix, Fig. S1). Sandwich ELISAs of these pMHCI vesicles showed that pMHCI copresented with the hallmark exosomal tetraspanin proteins CD81 and CD63 (SI Appendix, Fig. S2), which was further confirmed by mass spectrometry showing enrichment of pMHCI and exosomal tetraspanin proteins in the purified product (Document S1). As a result, the vesicles have been denoted as pMHCI-displaying “APVs” resembling natural immunoregulatory EVs or “exosomes.” EABR Addition to SCT pMHCI Promotes Budding of the Complete, Intended pMHCI Complex on APVs. To better quantify the relative improvement in APV formation with EABR addition, we performed ELISA of SCT MART1 and SCT NYESO1 constructs. Both showed over 100-fold improvements in APV formation with the addition of the EABR sequence (Fig. 2A). Subsequent ELISAs showed that appending EABR to MHCI heavy chain (HC) alone (i.e., without covalently linked β2 M and presenting peptide) still boosted MHCI yield in the purified transfection product relative to HC without an EABR tag, and an even greater concentration of MHCI resulted from transfection with EABR-tagged single-chain heterodimer (SCD) constructs composed of MHCI HC covalently linked to β2 M without a presenting peptide (SI Appendix, Table S1 andFig. S3). Consequently, we next investigated whether the SCT form of pMHCI was necessary to assemble the intended pMHCI complex on released APVs. EABR addition to SCT pMHCI promotes budding of the complete, intended pMHCI complex on APVs (A) Indirect ELISA of APVs purified by ultracentrifugation from transfected HEK293T cells with a primary rabbit anti-HLA-A antibody and secondary HRP-conjugated anti-rabbit antibody. Cells were transfected with SCT NYESO1/D1, SCT NYESO1/D1/EABR, SCT NYESO1/D9, SCT NYESO1/D9/EABR, SCT MART1/D1, and SCT MART1/D1/EABR constructs (2 replicates per dilution). (B) Sandwich ELISA of APVs purified by ultracentrifugation from transfected HEK293T cells using anti-NYESO1:HLA-A*02:01 (3M4E5) for capture, biotinylated 3M4E5 for detection, and HRP-conjugated streptavidin secondary. Cells were transfected with SCT NYESO1/D9/EABR, SCD D9/EABR and NYESO1 peptide (SCD/EABR + NYESO1)), HC D9/EABR and NYESO1/β2 M (HC/EABR + NYESO1/β2 M), SCT NYESO1/D9 (SCT), SCD D9 and NYESO1 peptide (SCD + NYESO1), HC D9 and NYESO1/β2 M (HC + NYESO1/β2 M) constructs (2 replicates per dilution). (C) Schematic showing full SCT pMHCI/EABR construct (SCT/EABR) and various alternative constructs used for the ELISA in (B): NYESO1/β2 M, HC, and SCD. (D) Dynamic light scattering (DLS) measurement of SCT MART1/D1 and SCT MART1/D1/EABR APVs represented as a probability mass function of the expected diameter values for each sample. Points are connected by a simple interpolating spline. (E–G) Immuno-EM micrographs of different EABR vesicles purified from transfected Expi293F cell culture supernatants by ultracentrifugation and SEC. Samples were stained with a primary rabbit anti-HLA-A antibody and a secondary 6 nm gold-conjugated anti-rabbit antibody that appears as black punctate in image. (Scale bars, 100 nm.) (E) SARS-CoV2-S/EPM/EABR. Spike protein coronas are visible around the edges of the spherical vesicles. A single gold nanoparticle from a secondary antibody is visible in the middle of the image. (F) SCT MART1/D1/EABR gold particles from secondary antibody can be seen as halos of black punctate around deflated vesicles. (G) SCT MART1/D1 scattered gold particles from secondary antibody can be seen as nonspecific staining without the presence of vesicles. To compare the impact of different NYESO1/MHCI construct designs on APV assembly and release, we selectively removed the Gly4Ser linkers from the SCT NYESO1/D9 construct—either removing the 3xG4S linker between the peptide and β2 M yielding SCD, or the 4xG4S linker between β2 M and the MHCI HC yielding NYESO1/β2 M—and performed ELISAs with a TCR-like antibody specific for the fully formed NYESO1:HLA-A*02:01 complex (Fig. 2 B and C) (30). Decoupling the NYESO1 peptide from the EABR-tagged SCD MHCI complex led to an over 10-fold reduction in the detected level of expression of NYESO1/MHCI APVs in HEK293T cells to the extent that it was nearly undetectable by sandwich ELISA (Fig. 2B). Even in the higher-expressing Expi293F cell line, an exhaustive search of every possible combination of MHCI α HC, β2 M, and either the full-length NY-ESO-11-165 or 9-mer NYESO1 peptide did not produce measurable pMHCI product in the purified APVs when these constructs were assembled without glycine-serine linkers between the 9-mer NYESO1 peptide and β2 M (SI Appendix, Fig. S4). Cotransfecting two separate constructs of HC/EABR and NYESO1/β2 M, however, generated APVs detectable by sandwich ELISA, and the fully connected SCT form of NYESO1/D9/EABR further improved the expression of pMHCI-displaying APVs out of all constructs tested (Fig. 2B); in addition, because EABR-tagged MHCI HC is capable of generating APVs without transfection of a presenting peptide such as NYESO1 (SI Appendix, Fig. S3), the SCT form of pMHCI is the only design that ensures the APVs are presenting the transfection constructs’ intended peptide. Interestingly, we still recovered low levels of MHCI protein after purification of non-EABR SCT pMHCI from transfected HEK293T supernatant, an effect that was amplified by expressing the same constructs in the higher-expressing Expi293F cell line that has been used previously for exosome production (31) (Fig. 2A and SI Appendix, Fig. S4). DLS of both the purified SCT MART1/D1 and SCT MART1/D1/EABR constructs showed similar median particle diameters of 60 to 80 nm (range 30 to 250 nm) with the SCT MART1/D1/EABR construct exhibiting a slightly narrower distribution compared to the non-EABR SCT MART1/D1 construct (Fig. 2D). This particle size distribution is consistent with the vesicle sizes seen in immuno-EM images of the EABR constructs. We were unable, however, to observe pMHCI vesicles when imaging immuno-EM grid preparations of the non-EABR SCT MART1/D1 construct purified from Expi293F supernatant. Instead, the non-EABR SCT MART1/D1 grids showed what appeared to be pMHC aggregates or blebbing of sparsely populated membrane, which was less frequently observed in the grids prepared for the EABR-tagged vesicles (Fig. 2 F and G and SI Appendix, Fig. S1). As a direct contrast, identically prepared EM grids for both SARS-CoV-2 S/EPM/EABR and SCT MART1/D1/EABR constructs were densely populated with vesicles displaying their respective proteins, and only the MART1/D1/EABR vesicles stained positively with a halo of anti-HLA-A antibody around each vesicle (Fig. 2 E and F). Altogether, our results show that SCT pMHCI promotes the display of complete pMHCI complexes on APVs, and that addition of the EABR sequence increases APV release. EABR-Mediated APV Release Generalizes to mRNA Transfection and Across pMHCI Variants. To further explore the generalizability of EABR sequence addition and heterotrimerization of pMHCI on APV release, we tested whether EABR-tagged SCT constructs of murine MHCI and chimeric human/mouse MHCI would also generate APVs displaying pMHCI relevant for in vivo studies of immunoregulation. Chimeric human/mouse SCT MHCI HLA-A2/H-2Db and HLA-A2/H-2 Kb were designed according to the D9 SCT design, with an NYESO1 presenting peptide fused to a 3x(Gly4Ser) linker, human β2 M, a 4x(Gly4Ser) linker, and a chimeric MHCI α HC composed of the extracellular α1 and α2 domains of human Y84A,A139C HLA-A*02:01 HC and the α3, transmembrane, and cytoplasmic domains of murine H-2 Kb or H-2Db HC, i.e., NYESO1/D9/H-2Db and NYESO1/D9/H-2 Kb, respectively (Fig. 3A and SI Appendix, Table S1). D1 SCT constructs of murine MHCI H-2 Kd were prepared with either an NRP-V7 or IGRP206-214 type 1 diabetes-relevant presenting peptide fused to a 3x(Gly4Ser) linker, murine β2 M, a 4x(Gly4Ser) linker, and the H-2 Kd α chain, i.e., NRP-V7/H-2 Kd and IGRP/H-2 Kd, respectively; D9 designs were further generated by introducing Y84A and A139C mutations in the H-2 Kd α HC to make NRP-V7/D9_H-2 Kd and IGRP/D9_H-2 Kd (Fig. 3A and SI Appendix, Table S1). EABR-mediated APV release generalizes to mRNA transfection and across pMHCI variants. (A) Schematic of pMHCI constructs used in (B–D) showing which domains are sequences of human or murine origin. Further descriptions of constructs and domains can be found in SI Appendix, Table S1. (B) Sandwich ELISA of chimeric human/mouse SCT pMHCI protein on APVs purified by ultracentrifugation from transfected HEK293T cell culture supernatants (2 replicates per dilution). Mouse anti-NYESO1:HLA-A*02:01 capture antibody, biotinylated mouse anti-NYESO1:HLA-A*02:01 primary antibody, and HRP-conjugated streptavidin secondary antibody. (C) Sandwich ELISA of murine SCT pMHCI of APVs purified by ultracentrifugation from DNA transfected HEK293T cells (2 replicates per dilution). Mouse anti-H-2 Kd capture antibody, biotinylated mouse anti-H-2 Kd primary antibody, and HRP-conjugated streptavidin secondary antibody. (D) Sandwich ELISA of murine SCT pMHCI of APVs purified by ultracentrifugation from mRNA transfected NMuMG mammary epithelial cells (1 replicate per dilution). Mouse anti-H-2 Kd capture antibody, biotinylated mouse anti-H-2 Kd primary antibody, and HRP-conjugated streptavidin secondary antibody. For the chimeric human/mouse SCT NYESO1/D9/H-2Db and NYESO1/D9/H-2 Kb, addition of an EABR sequence to the cytoplasmic tail of MHCI α HC amplified the production of APVs as detected by sandwich ELISA (Fig. 3B). Similar to the previous SCT MART1/D1, NYESO1/D1, and NYESO1/D9 pMHCI constructs, the SCT NYESO1/D9/H-2Db pMHCI construct showed evidence of substantial vesicle production in HEK293T cells regardless of the addition of the EABR sequence. Swapping the α3, transmembrane, and cytoplasmic domains to a different MHC haplotype, NYESO1/D9/H-2 Kb, led to comparatively reduced APV expression (Fig. 3B), a phenomenon that could be attributed to increased cell surface presentation of the H-2Db haplotype due to differences in the MHCI α3 domain which has been observed previously (32). In a similar fashion, the fully murine SCT NRP-V7/H-2 Kd and SCT IGRP/H-2 Kd constructs showed amplified APV formation upon addition of an EABR sequence to the cytoplasmic tail of MHCI α HC (Fig. 3C). While NRP-V7/H-2 Kd was capable of measurable vesicle formation without the addition of EABR, IGRP/H-2 Kd vesicle production was comparatively undetectable despite the only sequence difference being a change in the presenting peptide, likely reflecting differences in pMHCI stability based on changes in the presenting peptide (27, 33, 34); mRNA transfection with murine-optimized sequences of the murine NRP-V7/H-2 Kd and IGRP/H-2 Kd constructs using the lower-expressing murine NMuMG mammary epithelial cell line generated a much more pronounced improvement in vesicle formation, as high as 50-fold, upon addition of the EABR sequence, highlighting the vesicle-forming potential of the EABR sequence in a more translationally relevant context involving mRNA delivery. Collectively, these results suggest that MHC haplotype and the loaded presenting peptide can influence vesicle production for SCT constructs. As shown previously by immuno-EM of the MART1/D9 and MART1/D9/EABR constructs (Fig. 2 F and G), the increased yield of pMHCI product observed by ELISA in the non-EABR NRP-V7/H-2 Kd and NYESO1/D9/H-2Db constructs may be an artifact of pMHCI aggregation or indiscriminate membrane blebbing rather than intact APV formation. Further, vesicle formation by SCT pMHCI without EABR may be a result of improved protein stability or protein overexpression in the relatively higher-expressing HEK293T and Expi293F cell lines (35). Regardless of this intrinsic vesicle-forming potential, addition of the EABR sequence consistently increased APV release across these different pMHCI variants, and was shown to be further amplified in the context of mRNA transfection, demonstrating the generalizability of the EABR sequence for inducing APV production. pMHCI/EABR APVs Stimulate Their Cognate T cells Selectively In Vitro. Having confirmed that the EABR sequence reliably produces pMHCI-displaying APVs, we next asked whether EABR-induced pMHCI-displaying APVs were functional as T cell stimulators. SCT NYESO1/D9/EABR APVs were purified from transfected cell supernatants by ultracentrifugation and SEC, and the MHCI protein concentrations in purified APVs were determined by quantitative western blot analysis (SI Appendix, Fig. S5). An equivalent concentration of MHCI for MART1/D9/EABR, NYESO1/D9/EABR, and NYESO1:HLA-A*02:01 tetramers was mixed with human donor T cells that were transduced with the 1G4 TCR specific for the NYESO1:HLA-A*02:01 epitope (SI Appendix, Fig. S6), and T cell stimulation was evaluated using ELISPOT assay (Fig. 4A). pMHCI/EABR APVs stimulate their cognate T cells selectively in vitro. (A) Top: Scans of ELISPOT wells detecting IFN-γ release from one replicate test of Donor 3’s T cells after being dosed with serial dilutions of NYESO1:HLA-A*02:01 tetramers (NYESO1 Tetramer), SCT NYESO1/D9/EABR APVs, and SCT MART1/D9/EABR APVs. Bottom: Scans of ELISPOT wells detecting IFN-γ release from T cells dosed with 1.5% PHA as a positive control, wells withouT cells as a negative control (No cells), and T cells plated without antigen challenge (No antigen). (B) ELISPOT spot forming units (SFUs) for IFN-γ from 1G4 T cells of three different donors dosed with SCT NYESO1/D9/EABR APVs, SCT MART1/D9/EABR APVs, NYESO1:HLA-A*02:01 tetramers (NYESO1 Tetramer), or no antigen challenge. Each data point is the response per 1 × 105 cells. Baseline stimulation (i.e., “No antigen”—plated T cells without stimulation by antigen) was measured in triplicate for each donor and presented as a single mean in each plot with SD. Because the NYESO1:HLA-A*02:01 tetramers generated larger spot sizes on average compared to the APVs, Donor 3’s SFUs were too numerous to count when dosed with more than 10 ng of NYESO1:HLA-A*02:01 tetramer, preventing direct comparison between antigen conditions at higher concentrations for this donor. (C) ELISPOT SFUs for IL-2 from 1G4 T cells of “Donor 3” in (B). T cells were dosed with SCT NYESO1/D9/EABR APVs, NYESO1:HLA-A*02:01 tetramers (NYESO1 Tetramer), NYESO1:HLA-A*02:01 monomers (NYESO1 Monomer), or no antigen challenge (3 replicates per dilution). Significant differences between conditions linked by vertical lines are indicated by asterisks: ns > 0.05; *P < 0.05; **P < 0.01; ***P < 0.001; ****P < 0.0001. Analysis of ELISPOT assay SFUs from three different human T cell donors showed that both refolded NYESO1:HLA-A*02:01 tetramers and NYESO1/D9/EABR APVs significantly stimulated 1G4 T cells above background (Fig. 4B). This level of activation was specific to the NYESO1 peptide, as 1G4 T cells dosed with MART1/D9/EABR APVs demonstrated a statistically similar response to 1G4 T cells without antigen challenge. Interestingly, the elevated baseline of T cell activation without antigen challenge in Donor 3 coincided with an increase in the potency of NYESO1/D9/EABR APVs relative to NYESO1:HLA-A*02:01 tetramers for this donor. The possible association between the baseline activation level of a donor’s T cells and the potency of the NYESO1/D9/EABR APVs is also seen in the relatively weakly activated “Donor 1” and moderately activated “Donor 2” T cells, which demonstrate correspondingly weak and moderate receptiveness, respectively, to NYESO1/D9/EABR APV stimulation as compared to NYESO1:HLA-A*02:01 tetramers (SI Appendix, Fig. S7). Notably, soluble tetramers and monomers of natural, refolded pMHCI can activate pools of T cells by the release and subsequent representation of the pMHCI presenting peptide on T cell surfaces containing costimulatory signals, which then act as APCs for their T cell neighbors (36). Thus, T cell stimulation by these SCT pMHCI APVs may be altogether different than stimulation by refolded tetramer or monomer, which is further supported by our subsequent assay of IL-2 release from the strongly responding Donor 3 (Fig. 4C); at an equivalent mass of pMHCI, both monomer and tetramer forms of pMHCI induced significantly more IL-2 release than the EABR-mediated pMHCI. A dampened IL-2 release despite the substantial release of IFN-γ would coincide with in vitro studies of pMHCI-loaded EVs in the context of Type 1 diabetes (3), which showed limited IL-2 release when cocultured alone with human T1D-relevant T cells, and pMHCI nanoparticles, which showed a substantial release of IFN-γ when cocultured with murine T1D-relevant T cells and were therapeutic when administered in vivo (18, 37). Overall, these data confirm that the presence of pMHCI/EABR APVs is sufficient to stimulate IFN-γ release from T cells harboring their cognate TCR, and that their stimulation is specific to those T cells that feature their cognate TCR. Future studies will explore whether the observed difference in IL-2 release induced by APVs, and their seemingly dynamic potency for inducing IFN-γ release, is consistent across in vitro T cell clonal types and in vivo models of disease. Discussion Here, we present an approach for generating immunomodulatory pMHCI-displaying APVs from nonimmune cells. SCT pMHCI/EABR constructs containing a cytoplasmic EABR sequence (22) from the human centrosomal protein CEP55 recruit ESCRT proteins involved in cell division and viral budding to drive efficient assembly and release of pMHCI-displaying APVs. The lipid bilayer of the APV is expected to allow for pMHCI reorganization in the membrane and pMHCI clustering, both of which are considered important for the formation of a T cell stimulating immunological synapse. Purified pMHCI/EABR APVs elicited IFN-γ release from T cells, and the response was specific to T cells harboring the cognate TCR. Unlike previous methods that have used viral coat proteins to induce budding (38), the pMHCI/EABR construct does not require accessory proteins for vesicle release. Consequently, these pMHCI/EABR APVs could prove to be less immunogenic compared to other APV production methods as the pMHCI/EABR vesicles are composed of protein sequences already found in the human proteome. Thus, these T cell–targeting APVs, with putatively reduced immunogenicity and reduced potential for inducing release of IL-2 known for T cell proliferation and survival, could be applicable as a targeted immunosuppressive therapeutic for autoimmune diseases. As an example, the presenting peptides that were tested for SCT murine pMHCI H-2Kd in this study—NRP-V7 and IGRP206-214—are two therapeutically relevant epitopes for a murine type 1 diabetes (T1D) model that showed promise as a T1D therapeutic in vitro when formulated as SCT IGRP/H-2 Kd tetramers (39); these pMHCI complexes have also been conjugated to iron-oxide polymer nanoparticles and demonstrated in vivo efficacy as a targeted T1D therapy by selectively expanding the pool of antidiabetogenic memory-like autoregulatory CD8+ T cells (18), suggesting a potential clinical application for the APVs described here. As we demonstrated in this study, another advantage of using the EABR sequence for APV budding is that pMHCI APV assembly requires expression of only a single pMHCI/EABR protein and the corresponding gene construct could be delivered as an mRNA therapeutic. Previous work has explored the use of mRNA therapeutics for multiple sclerosis by delivering a mRNA-LNP therapeutic coding for an autogenic peptide with the goal of loading and presenting that peptide on the endogenous MHC that resides on, and is often limited to, the surface of the cell (40). An mRNA therapeutic coding for pMHCI/EABR may result in increased presentation of the fully formed pMHCI complex on the cell surface in addition to the release of pMHCI-loaded APVs that could distribute throughout the body, mirroring the natural release of endogenous APVs seen during an immune response and potentially increasing their therapeutic effect (2). The increase in purified pMHCI product resulting from the single-chain construct may be due to the SCT’s increased stability (24). The D9 variant in particular has previously been shown to have increased traffic to the cell surface and maintains a steady-state presence there by bypassing cellular quality control steps (27). This increased cell surface expression and stability may be acting synergistically with the EABR sequence to increase APV budding and release in the comparatively lower-expressing HEK293T and NMuMG cell lines, which would make the pMHCI/EABR APV approach particularly powerful as an injectable mRNA therapeutic that would likely be translated in lower-expressing primary cells in vivo. Future studies will investigate whether pMHCI/EABR APVs, or lipid nanoparticles encapsulating mRNA encoding a pMHCI/EABR construct, are effective as an autoimmune therapy. In summary, we present a method for efficiently generating APVs from nonimmune cells. We demonstrate that these APVs display complete pMHCI complexes and are capable of specific stimulation of cognate T cells, warranting further investigation as a potential cell-free therapeutic strategy for targeted immunoregulation. Materials Availability All expression plasmids generated in this study are available upon request through a Materials Transfer Agreement. Experimental Model and Subject Details Bacteria. Escherichia coli DH5 Alpha cells (Zymo Research) used for expression plasmid productions were cultured in LB broth (Sigma-Aldrich) with shaking at 225 rpm at 37 °C. Plasmids were purified for transfection using Zymogen miniprep kits (Zymo Research) or maxiprep kits (Macherey-Nagel). Cell Lines. HEK293T cells plated in 6-well plates (Falcon) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS, R&D Systems) and 1 U/mL penicillin-streptomycin (Gibco) at 37 °C and 5% CO2. Transfections were carried out with FuGENE transfection reagent (Promega) diluted in Opti-MEM (Gibco). Expi293F cells (Gibco) for APV expression and antibody expression were maintained at 37 °C and 8% CO2 in Expi293 expression medium (Gibco). Transfections were carried out with an Expi293 Expression System Kit (Gibco). Falcon tubes sealed with AirOtops (VWR) or culture flasks (Corning) containing Expi293F cells were maintained under shaking at 470 rpm for 10 to 30 mL transfections and 130 rpm for transfections larger than 30 mL. Expi293F and HEK293T cell lines were derived from female donors and were not specially authenticated. NMuMG cells plated in 6-well plates (Falcon) were cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS, R&D Systems) and 10 μg/mL insulin (Gibco) at 37 °C and 5% CO2. Transfections were carried out with Invitrogen Lipofectamine MessengerMAX Transfection Reagent (Invitrogen) diluted in Opti-MEM (Gibco). Transfection mRNA was synthesized using the mMessage mMachine T7 Transcription Kit (Invitrogen) and purified with the MEGAclear™ Transcription Clean-Up Kit (Invitrogen) according to the manufacturer’s instructions. T cells. T cells were isolated from human donor whole blood samples by negative selection using the RosetteSep Human T cell Enrichment Cocktail Kit (STEMCELL Technologies). Isolated T cells were seeded in a fresh complete ImmunoCult™-XF T cell expansion medium (STEMCELL Technologies) at 1 × 106 cells/mL with 2 μL/mL ImmunoCult™ human CD3/CD28 T cell activator. T cells were activated for 3 d and expanded for up to 12 d by changing into fresh expansion medium every 2 to 3 d. All cells were cultured in incubators at 37 °C and 5% CO2. Methods Design of EABR Constructs. The EABR sequence was identical to the sequence used by Hoffmann et al. to create SARS-CoV2-S eVLPs (22). Briefly, the EABR domain (residues 160 to 217) of the human CEP55 protein was fused to the C terminus of the SCT MART1/HLA-A*02:01 α chain separated by a (Gly)3Ser (GS) linker to generate D1 and D9 SCT pMHCI constructs inserted in the p3BNC expression plasmid. To generate the D1 SCT construct, the N terminus of the HLA-A*02:01 α chain was extended with a 4x(G4S) linker connecting to the C terminus of β2 M, which was itself extended at its N terminus by a 3x(G4S) linker connecting to either the 10-mer MART1 (ELAGIGILTV) or 9-mer NYESO1 (SLLMWITQV) cancer-related peptides to make a SCT (22) (Fig. 1B). To generate the SCT D9 construct, the D1 SCT construct was edited with Y84C and A139C substitutions (27). The “single-chain dimer” (SCD) construct removed both the peptide and the 3x(G4S) linker between the peptide and β2 M of the D9 SCT construct. The HC construct removed the peptide, the 3x(G4S) linker between the peptide and β2 M, β2 M, and the 4x(G4S) linker between β2 M and the MHCI HC of the D9 SCT construct (Fig. 2C). Both the SCD and HC constructs were inserted in p3BNC expression plasmids. Murine SCT MHCI H-2Kd was constructed similar to the human D1 SCT HLA-A*02:01 design, with either an NRP-V7 or IGRP206-214 presenting peptide fused to a 3x(Gly4Ser) linker, murine β2 M, a 4x(Gly4Ser) linker, and the H-2Kd HC, i.e., NRP-V7/H-2Kd and IGRP206-214/H-2Kd, respectively. The chimeric human/mouse MHCI HLA-A2/H-2Db and HLA-A2/H-2 Kb were designed according to the D9 SCT design, with an NYESO1 presenting peptide fused to a 3x(Gly4Ser) linker, human β2 M, a 4x(Gly4Ser) linker, and a chimeric MHCI α chain composed of the extracellular α1 and α2 domains of human Y84A,A139C HLA-A*02:01 HC and the α3, transmembrane, and cytoplasmic domains of murine H-2 Kb or H-2Db HC, i.e., NYESO1/A2_D9/H-2Db and NYESO1/A2_D9/H-2 Kb, respectively. Murine and chimeric human/murine SCT pMHCI constructs were inserted into the p3BNC expression plasmid. Production of EABR APVs. EABR APVs were generated by one of three means: 1) transfecting Expi293F cells (Gibco) cultured in Expi293F expression media (Gibco) on an orbital shaker at 37 °C and 8% CO2 with plasmid DNA prefiltered through 0.22 µm Spin-X filters (Corning); 2) transfecting HEK293T cells cultured in Dulbecco’s modified Eagle’s medium (DMEM, Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Sigma-Aldrich) and 1 U/mL penicillin-streptomycin (Gibco) at 37 °C and 5% CO2. 3) transfecting NMuMG cells cultured in Dulbecco’s modified Eagle’s medium with 4.5 g/L glucose (DMEM, Gibco) supplemented with 10% heat-inactivated fetal bovine serum (FBS, Sigma-Aldrich) and 10 μg/mL insulin at 37 °C and 5% CO2. 72 h posttransfection, cells were centrifuged at 1,000×g for 10 min, supernatants were passed through a 0.45 μm filter (Corning) with Luer-Lok syringes (BD) or a 0.45 μm vacuum filter (Thermo Fisher Scientific), and concentrated using Amicon Ultra-15 centrifugal filters with 100 kDa molecular weight cut-off (Millipore). APVs were purified by ultracentrifugation at 50,000 rpm (135,000×g) for 2 h at 4 °C using a TLA100.3 rotor and an OptimaTM TLX ultracentrifuge (Beckman Coulter) on a 20% w/v sucrose cushion in polycarbonate centrifuge tubes (Beckman Coulter). Supernatants and the sucrose cushion were removed, and pellets were resuspended in 200 μL sterile pH 7.4 PBS at 4 °C overnight. To remove residual cell debris, resuspended samples were transferred to microcentrifuge tubes and centrifuged at 10,000×g for 10 min; clarified supernatants were collected for subsequent ELISA, DLS, and western blot assays. For immuno-EM grid preparations, quantitative western blots, and ELISPOT assays, APVs were further purified by SEC using a Superose 6 Increase 10/300 column (Cytiva) equilibrated with pH 7.4 PBS. Peak fractions in the void volume corresponding to pMHCI APVs were combined and concentrated to 250 to 500 μL in Amicon Ultra-4 centrifugal filters with 100 kDa molecular weight cut-off (Millipore) or Nanosep with 100 K Omega centrifugal filters (Pall). Samples were stored at 4 °C and imaged directly after purification. Transduction of TCRs in PBMC. Peripheral blood mononuclear cells (PBMCs) were isolated and processed following a previously published protocol with some modifications (41). The MSGV retroviral backbone was utilized to overexpress the target TCRs with a truncated human nerve growth factor receptor (tNGFR) serving as the transduction marker. TCR alpha and beta chains were expressed with a F2A linker to maintain approximately equal copy numbers. To reduce mispairing with endogenous human TCRs, the human TCR constant regions were replaced with murine constant regions. Commercially obtained, cryopreserved PBMCs from healthy donors (AllCells) were revived in T cell media (TCM) comprising AIM V media (Thermo Fisher, #12055-091), 5% Human AB serum (Omega Scientific), 50 U/mL human recombinant IL-2 (Peprotech, #200-02), 1 ng/mL human recombinant IL-15 (Peprotech, #200-15), 1X Glutamax (Thermo Fisher, 35050-061), and 50 μM β-mercaptoethanol. These cells were mixed with CD3/CD28 dynabeads (Thermo Fisher, #11132D) at a concentration of 25 µL beads per million PBMCs and incubated overnight at 37 °C and 5% CO2 (day 0). On days 1 and 2, TCR-expressing retroviral supernatant was added at a ratio of 2 mL per two million PBMCs, followed by a 90-min spin infection at 1,350×g and 30 °C. Excess retrovirus was removed on day 3 with a PBS wash, and cells were resuspended in TCM. CD3/CD28 dynabeads were removed using magnets on day 5, and PBMCs were maintained in TCM at a concentration of one million cells per mL. Cell phenotypes (CD3, CD8), transduction efficiency (NGFR), and TCR membrane trafficking (murine TCRβ or mTCRb) were assessed using flow cytometry on day 7. For long-term storage, transduced PBMCs were preserved in CryoStor CS10 serum-free cryopreservation medium (STEMCELL Technologies #07930) at −80 °C overnight before transferring to liquid nitrogen. Antibody Expression. Anti-HLA-A2-peptide (SLLMWITQV) Complex (3M4E5) antibody was made in-house by expressing the HC and light chain of the 3M4E5 antibody at a 1.5:1 plasmid ratio in Expi293F cells (30). Supernatant from the transiently transfected Expi293F cells (Gibco) was separated from the cells by pelleting at 3,500×g for 15 min and filtering the clarified supernatant through a 0.2 μm syringe filter (VWR) or 0.45 μm vacuum filter unit (Thermo Fisher Scientific). The filtered supernatant was further purified using Fc-affinity chromatography (HiTrap MabSelect, Cytiva) and SEC (HiLoad 16/600 Superdex 200 column, Cytiva). Peak fractions corresponding to purified antibody proteins were pooled, concentrated, and stored at 4 °C. Biotinylated antibodies for ELISAs were generated and purified using an EZ-Link biotinylation kit (Thermo Fisher Scientific). The in-house generated antibody was compared to a commercially available 3M4E5 antibody (RHM03002-100; ProteoGenix) with ELISA and showed identical performance. Western Blot Analysis. The presence of HLA-A on purified APVs was detected by Western blot analysis. Samples were diluted in NuPage SDS-PAGE loading buffer (Thermo Fisher Scientific) under reducing conditions, separated on Bolt™ Bis-Tris Plus Mini Protein Gels, 4 to 12% (Thermo Fisher Scientific), and transferred to nitrocellulose membranes in the form of an iBlot2 NC Mini Stack (Thermo Fisher Scientific) using the iBlot2 Gel Transfer Device (Thermo Fisher Scientific). Nitrocellulose membranes from the stack were transferred to water for 1 to 5 min, then transferred to Flex FD solution (Thermo Fisher Scientific), before placing on an iBind card (Thermo Fisher Scientific) for staining with antibodies per the iBlot2 protocol. The following antibodies were used for detecting pMHCI: rabbit anti-HLA-A protein (PA5-29911; Thermo Fisher Scientific) at 1:1,000, mouse anti-β2 M protein (MA5-36022; Thermo Fisher Scientific) at 1:1,000, Alex488-conjugated goat anti-rabbit IgG (A32731; Thermo Fisher Scientific) at 1:1,000, and Alex488-conjugated goat anti-mouse IgG (A55058; Thermo Fisher Scientific) at 1:1,000. Protein was compared to a SeeBlue Plus2 Prestained protein ladder (Thermo Fisher Scientific). For T cell studies, the amount of MHCI on APVs was determined by quantitative Western blot analysis. Various dilutions of SEC-purified pMHCI APV samples and known amounts of soluble, biotinylated pMHCI protein (NIH Tetramer Core Facility at Emory) were separated by SDS-PAGE and transferred to nitrocellulose membranes as described above. Band intensities of the pMHCI standards and pMHCI APV sample dilutions were measured using ImageJ to determine MHCI concentrations. Mass Spectrometry. Expi293F cells were DNA transfected with NYESO1/D9/EABR or NYESO1/D9 pMHCI constructs, and supernatant was concentrated using 100 kDa MWCO (EMD Millipore; #UFC910096) to be ultracentrifuged on a 20% w/v sucrose cushion before SEC purification. Elution volumes corresponding with EVs (Fig. 1D) were collected, digested with trypsin, and run on an Orbitrap Eclipse Tribrid Mass Spectrometer. Proteins were searched and identified against the human proteome using Proteome Discoverer 3.2. Immuno-EM of pMHCI/EABR APVs. SEC-purified MART1/MHCI/EABR APVs were prepared on grids for negative stain transmission electron microscopy at room temperature. Formvar/Carbon 300 Mesh grids (Electron Microscopy Sciences) were first glow discharged. 20 μL of purified sample was pipetted onto paraffin, glow discharged formvar/carbon 300 mesh grids were placed on the droplet of sample for 5 min, then gently wicked away by dabbing the edge of the grid on filter paper. The grids were then placed on a 20 μL droplet of 1% paraformaldehyde (Electron Microscopy Sciences) in pH 7.4 PBS for 10 min, and subsequently wicked away. Grids were washed three times by placing the grids on droplets of pH 7.4 PBS for 5 min, and wicking away the PBS with filter paper after each wash. Grids were then blocked by placing the grids on droplets of 5% FBS diluted in pH 7.4 PBS for 1 h. Blocking solution was wicked away from the grids, and the grids were then placed on droplets of rabbit anti-HLA-A antibody (PA5-29911; Thermo Fisher Scientific) diluted 1:100 dilution in 5% sucrose, 5% FBS pH 7.4 PBS for 2 h; to prevent drying of the primary antibody solution, the grids were placed in a humidified chamber made from wet paper towels folded under a pyrex glass dish. Grids were again washed three times by placing the grids on droplets of pH 7.4 PBS for 10 min, and wicking away the PBS with filter paper after each wash. The grids were then placed on droplets of 6 nm colloidal gold goat anti-rabbit antibody (PA5-29911; Thermo Fisher Scientific) diluted 1:20 dilution in 5% sucrose, 5% FBS pH 7.4 PBS for 1 h. Grids were then washed three times, 10 min per wash with pH 7.4 PBS, wicking away the PBS after the 10 min wash each time. Finally, the grid was placed on a droplet of 1.5% uranyl formate (Electron Microscopy Sciences), prepared fresh within 1 wk of use, for 1 min. The excess uranyl formate was wicked away and the grid was left to air dry for 30 min before storing at room temperature away from light. Imaging of the grids was performed on a FEI Tecnai T12 Transmission Electron Microscope within 2 wk after the grids were prepared. Indirect ELISA. 96-well plates (Corning) were coated with APV samples diluted 1:10 or 1:4 or 1:1, fourfold serially diluted in sterile 0.1 M NaHCO3 pH 9.6, and sealed with TempPlate sealing foil (USA Scientific) overnight at 4 °C. Plates were emptied and blocked with TBS-T/3% BSA (Sigma-Aldrich) for at least 30 min. Plates were again emptied, and 5 μg/mL of primary antibody diluted in TBS-T/3% BSA was added to the plates. The following were used as primary antibodies for detecting pMHCI: rabbit anti-HLA-A antibody (PA5-29911; Thermo Fisher Scientific), mouse anti-HLA-A2-peptide (SLLMWITQV) Complex (3M4E5) antibody (RHM03002-100; ProteoGenix), mouse anti-H-2Kd antibody (14-5998-82; Thermo Fisher Scientific). After a 2-h incubation at room temperature, plates were emptied, washed three times with TBS-T, and 5 μg/mL of HRP-conjugated secondary antibody diluted in TBS-T/3% BSA was added to the plates for 30 min at room temperature. The following were used as secondary antibodies: HRP-conjugated goat anti-rabbit IgG (4058-05; Southern Biotech) at 1:2,000, HRP-conjugated goat anti-mouse IgG (1013-05; Southern Biotech) at 1:2,000. After washing three times with TBS-T, plates were developed using 1-Step™ Ultra TMB-ELISA Substrate Solution (Thermo Fisher Scientific) and absorbance was measured at 450 nm. Standard 4-parameter sigmoidal binding curves were calculated using GraphPad Prism 10.3.1 without any further editing. Sandwich ELISA. 96-well plates (Corning) were coated with 5 μg/mL of mouse anti-HLA-A2-peptide (SLLMWITQV) Complex (3M4E5) antibody (RHM03002-100; ProteoGenix) or in-house-made anti-HLA-A2-peptide (SLLMWITQV) Complex (3M4E5) antibody diluted in sterile 0.1 M NaHCO3 pH 9.6, and sealed with TempPlate sealing foil (USA Scientific) overnight at 4 °C. Plates were emptied and blocked with TBS-T/3% BSA for at least 30 min. Plates were again emptied, and APV samples diluted 1:10 or 1:4, fourfold serially diluted in TBS-T/3% BSA were added to the plates. After a 2-h incubation at room temperature, plates were emptied, and 5 μg/mL of biotinylated detection antibody diluted in TBS-T/3% BSA was added to the plates. 5 μg/mL of biotinylated (Thermo Fisher Scientific) mouse anti-HLA-A2-peptide (SLLMWITQV) Complex (3M4E5) antibody (RHM03002-100; ProteoGenix) or in-house-generated human anti-HLA-A2-peptide (SLLMWITQV) Complex (3M4E5) antibody was used as a capture antibody. After another 2-h incubation at room temperature, plates were washed three times with TBS-T. HRP-conjugated streptavidin (Abcam) was diluted to manufacturer’s recommendations in TBS-T/3% BSA and added to plates for 30 min at room temperature. After washing three times with TBS-T, plates were developed using 1-Step™ Ultra TMB-ELISA Substrate Solution (Thermo Fisher Scientific) and absorbance was measured at 450 nm. Standard 4-parameter sigmoidal binding curves were calculated using GraphPad Prism 10.3.1 without any further editing. ELISPOT. 1G4 T cells were centrifuged at 300×g for 10 min, and resuspended in CTL-TestTM media (ImmunoSpot) containing 1% L-glutamine (Gibco) for ELISPOT analysis to evaluate T cell responses. APVs were added to human IFN-g single-color ELISPOT plates (ImmunoSpot) and 10-fold serially diluted. Phytohemagglutinin (Gibco) was added in a separate well at a final concentration of 1.5% as a positive control. 100,000 cells were added per well, and plates were incubated at 37 °C in 8% CO2 for 24 h. Biotinylated detection, streptavidin-alkaline phosphatase (AP), and substrate solutions were added according to the manufacturer’s guidelines. Plates were gently rinsed with water three times to stop the reactions. Plates were air-dried for 2 h in a running laminar flow hood. The number of spots and the mean spot sizes were quantified using a CTL ImmunoSpot S6 Universal-V Analyzer (Immunospot). Quantification and Statistical Analysis. ELISA serial dilutions were plotted by fitting a standard sigmoidal four parameter logistic curve without weighting or special handling of outliers using Graphpad Prism 10.3.1. Statistical significance in IFN-γ ELISPOT data was calculated using multiple unpaired t tests with Welch correction and Holm–Šídák’s multiple comparison tests to obtain adjusted P-values in Graphpad Prism 10.3.1; adjusted P-values are reported for comparisons of the maximum concentrations. Statistical significance in IL-2 ELISPOT data was calculated using two-tailed unpaired t tests with Welch’s correction to obtain P-values in Graphpad Prism 10.3.1. Key Resources. SI Appendix, Table S2 lists key resources used in this study. Supplementary Material Appendix 01 (PDF) Dataset S01 (XLSX)