E7_11-19 placement and orientation dictate CD8⁺ T cell response in structurally defined spherical nucleic acid vaccines

Abstract

To develop effective nanostructured immunotherapeutics, identifying structural parameters that maximize immune response is essential. Spherical nucleic acids (SNAs) provide a modular platform for coordinated antigen-adjuvant delivery, where subtle structural differences can markedly influence potency. Herein, three SNAs were designed with HLA-A2–restricted HPV16 E7_11-19 peptide and CpG adjuvant, nearly identical in composition but differing in antigen presentation. All enhanced dendritic cell activation and CD8⁺ T cell cytotoxicity in primary human cells compared to peptide-CpG admixture; however, one variant, N-HSNA, elicited the strongest response, inducing ~8-fold higher interferon-γ secretion and ~2.5-fold greater cytotoxicity. In tumor-bearing AAD mice, N-HSNA reduced tumor burden by ~3.5-fold, prolonged survival, and expanded CD8⁺ T cells. Transcriptomic profiling revealed up-regulation of activation genes and suppression of exhaustion markers. In patient-derived HPV⁺ head and neck cancer spheroids, N-HSNA enhanced cytotoxicity ~2.5-fold, establishing antigen placement and orientation as key parameters for translational cancer immunotherapy.

Antigen placement and orientation drive CD8⁺ T cell and antitumor responses in a clinically relevant HPV model.

INTRODUCTION

The incidence of human papillomavirus (HPV)–associated head and neck squamous cell carcinoma (HNSCC) is rising, with about 18,000 new cases reported annually in the United States (US) (1, 2). Most cases present with locally advanced disease, which is highly curable with surgery, radiation, and chemotherapy. However, treatment is very toxic, and a large number of patients suffer recurrence (3). Despite advances in treatment, particularly in the field of immunotherapy, the vast majority of patients with recurrent or metastatic HNSCC eventually succumb to their disease (4), highlighting the need for better treatments.

HPV-positive HNSCC expresses the viral oncoproteins E6 and E7, with the E7_11-19 epitope representing a particularly promising immunotherapy target (5). A phase 1 trial of T cell receptor–engineered therapy (TCR-T) targeting E7_11-19 showed tumor regression in 50% of patients with metastatic HPV-associated cancers (6), but was limited by off-target toxicity, immune escape, and scalability (7). Therapeutic vaccines targeting E7_11-19 show potent preclinical efficacy yet fail to elicit robust epitope-specific T cell responses in clinical trials (NCT02865135) (8). Therefore, scalable vaccine platforms capable of inducing high-avidity T cells comparable to those elicited by TCR-T are urgently needed.

Liposome-based spherical nucleic acids (L-SNAs) are composed of a liposome core densely functionalized with radially oriented oligonucleotides (9–17). In particular, L-SNAs that incorporate a Toll-like receptor 9 (TLR9) agonist CpG DNA and a tumor-specific peptide antigen have been explored. These L-SNAs are efficiently taken up by antigen-presenting cells (APCs), such as dendritic cells (DCs), and enhance cross-presentation and T cell priming. The immunostimulatory potency of L-SNAs can be tuned by adjusting key structural parameters, including liposome stability (18, 19), antigen attachment chemistry and localization (20–24), and oligonucleotide anchoring strength (25, 26). For example, when cholesterol anchors on oligonucleotides were replaced with dodecane oligomers [(C12)₉], the DNA anchoring strength of the L-SNAs increased as did their ability to spur enhanced DC activation and E7_11-19-specific CD8⁺ T cell responses in humanized mice and human primary immune cells (25, 26). Likewise, when the spatial placement of antigens was altered [i.e., encapsulated inside the liposomal core (E-SNA), anchored on the surface (A-SNA), or hybridized to the DNA shell via a complementary oligonucleotide (H-SNA)], cross-presentation kinetics and DC priming efficiency markedly differed (22). Notably, the L-SNA variant, H-SNA, improved murine major histocompatibility complex class 1 (MHC-I)–specific CD8⁺ T cell responses and tumor inhibition in an HPV-transgenic TC-1 model compared to the other nearly chemically identical forms of SNAs. However, whether these structural design principles extend to clinically relevant, human HLA-A2–restricted HPV epitopes has yet to be determined.

This study explores how antigen configuration influences HPV vaccine efficacy by incorporating E7_11-19 antigens (YMLDLQPET) into the (C12)₉-anchored L-SNA structure using two strategies: (i) encapsulation within the liposomal core (E-SNA) and (ii) hybridization to the DNA shell via complementary DNA-antigen conjugates (H-SNA). In H-SNA constructs, a cysteine residue was introduced at the N or C terminus of E7_11-19 peptide to enable site-specific DNA conjugation and preserve the core epitope sequence (5). N-HSNA (CYMLDLQPET) and C-HSNA (YMLDLQPETC) were compared for their ability to elicit HLA-A2–restricted CD8⁺ T cell responses and inhibit tumor growth in HLA-A2 transgenic mice bearing HPV⁺ tumors and patient-derived HPV⁺ HNSCC models. These findings demonstrate that antigen placement and orientation are key determinants of SNA vaccine performance, with the N-HSNA driving superior CD8⁺ T cell function, transcriptional activation, and antitumor efficacy in both humanized mouse and patient-derived tumor models.

RESULTS

Characterization of HPV SNAs

The oligonucleotide shell density on SNAs influences their biological properties, including their cellular uptake and immune activation (24, 27, 28). Leveraging the 2.33-fold greater DNA loading capacity, when the (C12)₉ DNA anchor is used compared to when the cholesterol DNA anchor is used (22, 25), the impact of DNA loading on SNA properties was assessed. Constructs with high (2.67 pmol/cm²), medium (1.34 pmol/cm²), and low (0.67 pmol/cm²) DNA surface densities were explored (figs. S1 and S2A). Consistent with previous findings (27), the SNAs with the high DNA loading showed a ~7-fold and ~4-fold increase in uptake in CD11c⁺ human peripheral blood mononuclear cells (hPBMCs) compared to those with low and medium loading, respectively (fig. S2B). This enhanced uptake correlated with a 1.8-fold elevation in CD83 expression in HLA-DR⁺ and CD123⁺ DCs for the high-loading SNAs relative to the medium-loading SNAs (fig. S2C). These results confirm that increasing DNA surface density enhances both SNA internalization and DC activation in human immune cells, consistent with prior studies showing that the increased DNA shell density of SNA strengthens avidity for class A scavenger receptors and drives increased caveolae-mediated endocytosis (24, 27).

Building on this finding, three (C12)₉-anchored SNAs were designed with high DNA loading and with different E7_11-19 antigen orientations and spatial presentations (Fig. 1A): E-SNA, N-HSNA, and C-HSNA. For N-HSNA and C-HSNA, the E7_11-19 peptides were conjugated to the CpG complementary strands (table S3) using a succinimidyl 3-(2-pyridyldithio)propionate (SPDP) linker and subsequently hybridized to the CpG DNA. To isolate the effects of antigen orientation and placement, all SNA formulations were synthesized with a 1:1 antigen:CpG ratio, ensuring equivalent antigen density across constructs. Remaining surface sites were backfilled with (C12)₉-T20 DNA to achieve ~175 total DNA strands per liposome (fig. S3). While modulation of antigen density could provide additional insight into how surface presentation influences immune activation, this variable was intentionally held constant here to ensure that the observed differences reflect structural configuration rather than antigen loading variability. The successful anchoring of the DNA to the liposome cores was confirmed using dynamic light scattering (DLS), zeta-potential measurements, and agarose gel electrophoresis (Fig. 1, B and C, and figs. S4 and S5). DLS number-, volume-, and intensity-weighted size distributions all showed an increase in hydrodynamic diameter following SNA assembly relative to the parent liposomal core, consistent with successful DNA loading and SNA formation (Fig. 1C and fig. S5). Furthermore, cryo–electron microscopy (cryo-EM) images did not reveal morphological differences between E-SNA, N-HSNA, and C-HSNA (Fig. 1D). The size distributions observed in the cryo-EM images were consistent with DLS measurements. N-HSNA and C-HSNA exhibited larger particle sizes than E-SNA, which can be attributed to peptide-encapsulated liposomes forming more compact vesicles than empty dioleoylphosphatidylcholine liposomes. This is consistent with previous findings that peptide encapsulation can influence lipid packing and vesicle size (29). Together, these results confirm the successful formulation of E-SNA, N-HSNA, and C-HSNA, enabling further investigation into how SNA antigen configuration influences E7_11-19-specific immune stimulation.

Fig. 1. Design and characterization of SNAs with distinct E7_11-19 antigen configurations.

(A) Schematics of three SNA designs presenting the HLA-A2–restricted HPV16 E7_11-19 peptide: E-SNA (peptide encapsulated in the liposomal core), N-HSNA (peptide displayed on the SNA surface via N-terminal DNA conjugation), and C-HSNA (peptide displayed via C-terminal conjugation). DNA-peptide conjugates were synthesized using a SPDP linker and a site-specific additional cysteine residue at either the N or C terminus of the peptide. (B) Zeta potential measurements comparing SNA formulations with liposomes. (C) Number mean size distribution of SNAs and liposomes measured by DLS. Data represent mean ± SD (n = 3). (D) Cryo–electron microscopy (cryo-EM) images of E-SNA, N-HSNA, and C-HSNA. Particle size quantification was performed on at least 50 particles per group. Scale bars, 50 nm.

SNA design influences DC activation and antigen-specific CD8⁺ T cell responses in hPBMCs

Given the critical role of DCs in T cell priming (30), the SNA constructs were evaluated for their ability to be taken up by and activate primary human DCs (hDCs) within a population of hPBMCs, leading to the induction of E7_11-19 antigen–specific cytotoxic T cells against the HPV⁺ cancer cell line, UM-SCC-104 (31). To assess cellular uptake, hPBMCs were incubated for 1 or 4 hours with E-SNAs, N-HSNAs, C-HSNAs, or bare liposomes (without peptide). Uptake was quantified using flow cytometry to determine the efficiency of SNA internalization by primary human immune cells. All of the SNA constructs showed significantly higher internalization into CD11c⁺ DCs within the hPBMC population as compared to the bare liposomes at both 1- and 4-hour time points. However, E-SNA [mean fluorescence intensity (MFI) = 1460 at 1 hours or 5621 at 4 hours] exhibited ~2-fold greater uptake compared to either N-HSNA (MFI = 327 at 1 hours or 2765 at 4 hours) or C-HSNA (MFI = 436 at 1 hours or 2956 at 4 hours). The N- and C-HSNAs displayed comparable levels of internalization at both time points (Fig. 2A). The enhanced uptake of E-SNAs is likely due to the higher affinity of the single-stranded phosphorothioate oligonucleotides for scavenger receptors (fig. S6), abundantly expressed on hDCs (32).

Fig. 2. In vitro immune stimulation by SNAs with distinct antigen configurations using human immune cells.

(A) Uptake of various (C12)₉-anchored SNAs constructs by CD11c⁺ cells in human PBMCs after 1 or 4 hours (h), quantified by flow cytometry. (B) Activation of plasmacytoid dendritic cells (pDCs; CD123⁺) after 24-hour incubation with admix control, E-SNA, N-HSNA, or C-HSNA, assessed by expression of the costimulatory marker CD83. (C) IFN-γ secretion from E7_11-19-specific CD8⁺ T cells measured by ELISpot 24 hours after coculture with monocyte-derived mature dendritic cells (mDCs) pretreated with admix, different SNA constructs, or controls (E6 peptide admix as negative control, ionomycin as positive control). (D) Proliferation of E7_11-19-specific CD8⁺ T cells normalized to untreated controls, after coculture with mDCs pretreated with each SNA construct. (E) Cytotoxic responses of E7_11-19-specific CD8⁺ T cells against HPV⁺ UM-SCC-104 cells. Data represent mean ± SD with n = 4 to 9; statistical significance was calculated by one-way analysis of variance (ANOVA) with Tukey’s post hoc test; *P < 0.05, **P < 0.01, ***P < 0.005, and ****P < 0.001.

To determine whether uptake correlates with stronger DC activation, hPBMCs were treated for 24 hours with phosphate-buffered saline (PBS), admix (free adjuvants and antigens), E-SNA, N-HSNA, or C-HSNA. The activation of the HLA-DR⁺ DCs was evaluated using flow cytometry. Specifically, the expression of costimulatory marker CD83 was measured in CD123⁺ plasmacytoid DCs (pDCs) (33–35). The E-SNA treatment exhibited a ~2-fold increase in the percentage of CD83⁺ pDCs (59%) compared to the N-HSNA (32%) and C-HSNA (30%) groups (Fig. 2B). These results suggest that the presence of single-stranded CpG DNA on the surface of E-SNAs, as opposed to duplexed CpG-antigen constructs on N-HSNAs and C-HSNAs, contribute to the enhanced cellular uptake, leading to more robust activation of hDCs (25).

Next, the SNA-mediated activation of hDCs and antigen-specific cytotoxic T cell responses was evaluated by measuring interferon-γ (IFN-γ) secretion and T cell proliferation (fig. S6). We hypothesized that CD83 expression on DCs correlates with the magnitude of IFN-γ secretion and E7_11-19-specific CD8⁺ T cell proliferation. Mature hDCs (mhDCs) derived from THP-1 cells were treated with PBS, admix, E-SNA, N-HSNA, C-HSNA, or a nonspecific E6 MHC class II antigen containing admix for 24 hours (36). To ensure that peptide modifications did not impair antigen presentation, transporter-associated with antigen processing–deficient T2 cells were used to test HLA-A2 surface expression following incubation with peptides: E7_11-19, N-terminal cysteine–modified E7_11-19 (CE7_11-19), C-terminal cysteine-modified E7_11-19 (E7_11-19C), or E6 MHC class II (37). While untreated or E6 antigen–treated cells showed negligible HLA-A2 expression, all of the E7_11-19 variants induced ~30-fold enhancements in HLA-A2 expression, confirming that the addition of the cysteine residue at either terminus does not interfere with HLA-A2 binding in the absence of intracellular antigen processing (fig. S9).

mhDCs were cocultured with E7_11-19-specific CD8⁺ T cells (see fig. S9 for optimized conditions), in the presence of the corresponding peptide antigen with ionomycin serving as a positive control. IFN-γ secretion was quantified after 24 hours using an IFN-γ ELISPOT assay. Unexpectedly, E-SNA [33 spot-forming cells (SFCs)] and C-HSNA (30 SFCs) did not elicit a significantly greater number of SFCs as compared to the specific E7_11-19 admix (55 SFCs), untreated (28 SFCs), or nonspecific E6 admix (37 SFCs) controls (fig. S10 and Fig. 2C). However, N-HSNA treatment resulted in an average of 261 SFCs, representing an eight- to nine-fold increase over E-SNA and C-HSNA, respectively, indicating that N-HSNA strongly promotes antigen-specific CD8⁺ T cell activation.

To evaluate antigen-specific T cell responses, the proliferation of E7_11-19-specific CD8⁺ T cells was assessed using flow cytometry. Immature hDCs, generated from primary hCD14⁺ monocytes, were treated with PBS, admix, or the various SNA constructs to induce maturation (mhDCs), and then cocultured with E7_11-19-specific CD8⁺ T cells to measure their proliferation. N-HSNA–treated mhDCs induced a 1.4-fold increase in E7_11-19 antigen–specific CD8⁺ T cells as compared to the untreated control (Fig. 2D) and exceeded that elicited by E-SNA (0.9-fold) and C-HSNA (1.1-fold) treatment. Notably, the increased proliferation was observed despite N-HSNA exhibiting lower DC uptake and activation than E-SNA (Fig. 2, A and B). Overall, these results indicate that the arrangement of antigens on SNAs can affect the quality of E7_11-19-specific CD8⁺ T cell responses and that uptake efficiency alone does not fully account for the observed functional differences (38).

To evaluate the correlation between the enhanced stimulation of E7_11-19-specific CD8⁺ T cells and increased cytotoxicity against HPV⁺ cancer cells, CD8⁺ T cells isolated from hPBMCs were treated with PBS, admix, E-SNAs, N-HSNAs, and C-HSNAs, and cocultured the T cells with HPV-positive UM-SCC-104 cells or HPV negative UM-SCC-01 cells (39) for 24 hours. Flow cytometry analysis of early apoptotic (Annexin V⁺) and necrotic (7-aminoactinomycin D⁺) cells revealed that the CD8⁺ T cells derived from hPBMCs treated with N-HSNA (19%) or E-SNA (17%) exhibited significantly enhanced cytotoxicity against UM-SCC-104 cancer cells representing a 2-fold and 1.8-fold increase compared to admix (10%) treated groups (Fig. 2E). Cytotoxicity against HPV-negative UM-SCC-01 cells remained similar across all groups (~30 to 38%) (fig. S11), indicating that SNA treatments do not induce off-target CD8⁺ T cell killing. These results show that while all SNA constructs enhance the immunogenicity of the HLA-A*02 restricted E7_11-19 antigen, the N-HSNA is most effective in inducing CD8⁺ T cells capable of mounting robust, antigen-specific cytotoxic responses, highlighting that antigen configuration of on SNAs is a critical driver of functional and antigen-specific cytotoxic T cell immunity.

E7_11-19-CD8⁺ T cell infiltration and tumor suppression are maximized by N-HSNA in HLA-A2⁺ tumor model

To evaluate whether the in vitro findings translated to in vivo antitumor efficacy, transgenic AAD mice expressing the human MHC-I HLA-A*02 allele and bearing TC-1/A2 tumor were used for a clinically relevant model (40). Mice were subcutaneously inoculated with TC-1/A2 cells and vaccinated every 5 days with SNA formulations or admix controls (Fig. 3A). No abnormal clinical signs were observed in any treatment group, and serum Aspartate Aminotransferase (AST) and Alanine Aminotransferase (ALT) levels remained unchanged 48 hours after administration of (C12)₉-anchored CpG DNA (fig. S12), consistent with prior reports showing reduced systemic cytokine release and a favorable safety profile for (C12)₉-anchored SNA formulations (26). By day 19 postinoculation, N-HSNA–treatment reduced tumor growth by ~2.5-fold compared to PBS and admix groups (Fig. 3B), accompanied by a modest increase in survival. The median survival was 28 days for PBS- and admix-treated mice, 30 days for E-SNA and C-HSNA, and 33 days for N-HSNA–treated mice (Fig. 3C). Only N-HSNA delayed tumor growth compared to the admix group, although no statistically significant differences were observed between the SNA variants themselves.

Fig. 3. In vivo antitumor efficacy and immune activation of SNAs in the TC-1/A2 model.

(A) Vaccination protocol of C57BL/6-AAD mice subcutaneously inoculated with 2 × 10⁵ TC-1/A2 cells. (B) Tumor growth curves and (C) survival plots are shown. Data are mean ± SEM from two independent experiments (n = 7 mice per group). P values are shown for Admix versus N-HSNA (orange) at days 19. (D) Vaccination protocol of tumor-bearing mice intraperitoneally treated with anti–PD-1 antibody following SNA immunization. (E) Tumor growth and (F) survival data are shown. P values compare Admix + anti–PD-1 versus N-HSNA + anti–PD-1 (orange) or C-HSNA + anti–PD-1 (yellow) at the indicated time points. Data represent mean ± SEM from two independent experiments (n = 8 mice/group). [(G) to (L)] Immune profiling on day 20 using flow cytometry of splenocytes, PBMCs, and tumor-infiltrating lymphocytes. (G) Frequency of total CD8⁺ T cells in spleen. (H) Effector memory CD8⁺ T cells (CD44⁺/CD62L⁻) in spleen. (I) IFN-γ and CD107a expression in splenic CD8⁺ T cells after 4-hour E7_11-19 peptide restimulation. (J) Circulating E7_11-19-specific CD8⁺CD19⁻ T cells in PBMCs. (K) Total intratumoral CD8⁺ T cells. (L) Intratumoral E7_11-19-specific CD8⁺CD19⁻ T cells. Data represent mean ± SD (n = 6 to 7 mice per group). Significance was determined by one-way ANOVA with Tukey’s post hoc test for all panels except (C) and (F), where survival was assessed by log-rank (Mantel-Cox) test.

To evaluate whether the inclusion of immune checkpoint inhibitor, anti–PD-1 antibody, could enhance the efficacy (23), AAD mice were inoculated with TC-1/A2 cells and treated every 5 days with SNAs or admix formulations. Anti–PD-1 antibody was administered every 3 days after SNA treatment (Fig. 3D). By day 19 postinoculation, N-HSNA treatment reduced tumor growth by 2.5- and 2-fold compared to PBS and admix groups, respectively, with continued suppression until day 23 (~3-fold compared to PBS and ~2-fold compared to admix) (Fig. 3E). Although no statistical difference was observed between the various SNA formulations, N-HSNA delayed tumor growth and led to a significant extension in survival (median survival: PBS = 26 days, admix = 26 days, E-SNA = 29 days, C-HSNA = 30 days, and N-HSNA = 33 days) (Fig. 3F).

To evaluate whether SNA formulations differentially induced E7_11-19-specific CD8⁺ T cell responses, spleen, blood, and tumor tissues (41) were analyzed on day 20 post-tumor inoculation, when N-HSNA treatment showed superior tumor control over admix. Although total splenic CD8⁺ T cell frequencies decreased in SNA-treated mice (Fig. 3G and fig. S13), this reduction likely reflects antigen-driven migration of effector T cells from lymphoid tissues into circulation and tumor sites (42) rather than systemic suppression. Therefore, the effector memory CD8⁺ T cell subset (CD44⁺CD62L⁻) increased, with N-HSNA treatment yielding ~16% compared to ~11% in admix-treated mice (Fig. 3H). Upon E7_11-19 peptide restimulation, SNA-treated mice exhibited elevated intracellular IFN-γ and CD107a expression, indicative of cytolytic antigen-specific responses (Fig. 3I and fig. S14), with N-HSNA treatment resulting in a 2.3-fold increase in IFN-γ⁺CD107a⁺ CD8⁺ T cells compared to admix. IFN-γ and CD107a were selected as the primary functional readouts of antigen-specific CD8⁺ T cell activation, as these markers directly reflect cytolytic function and correlate with antitumor efficacy. This focused analytical approach aligns with established immunological standards and is consistent with prior SNA vaccine studies (22, 23, 25, 43). Consistent with in vitro results, N-HSNA significantly expanded antigen-specific CD8⁺ T cells in both peripheral blood and tumors, inducing a threefold increase in E7_11-19-specific CD8⁺ T cells (0.6%) in PBMCs compared to E-SNA (0.2%) and a twofold increase compared to admix (Fig. 3J). Tumor analysis further revealed a threefold increase in E7_11-19-specific CD8⁺ T cells with N-HSNA (2.4%) versus admix (0.9%) (Fig. 3, K and L, and fig. S15), corroborated by CD8⁺ T cell infiltration observed via immunohistochemistry (fig. S16). These results demonstrate that N-HSNA most effectively drives systemic and intratumoral expansion of functional, antigen-specific CD8⁺ T cells.

Distinct CD8⁺ T cells transcriptional profiles driven by antigen configuration on SNAs

To examine how different antigen configurations on SNAs influence CD8⁺ T cell transcriptional programming, bulk RNA sequencing was performed on splenic CD8⁺ T cells from tumor-bearing mice. Despite expressing similar surface markers across SNA-treated groups (Fig. 3, H and I), transcriptional analysis revealed significant differences in gene expression that correlated with functional outcomes. Principal components analysis (PCA) showed distinct clustering of CD8⁺ T cells by different treatment groups, indicating transcriptional divergence across different SNA forms (Fig. 4A). N-HSNA–treated samples clustered more tightly than E-SNA– and C-HSNA–treated samples, reflecting consistent transcriptomic profiling. Among the different SNA formulations, N-HSNA induced the most pronounced transcriptional shift, aligning with its superior in vivo efficacy (Figs. 2 and 3).

Fig. 4. Transcriptomic profiling of CD8⁺ T cells reveals enhanced immune activation by N-HSNA.

(A) PCA plot of transcriptomes from sorted CD8⁺ T cells isolated from TC-1/A2 tumor–bearing mice immunized with different SNA constructs, based on 4148 DEGs. (B) Heatmap of z-scored gene expression for selected immune-relevant genes in CD8⁺ T cells between a pairwise comparison of E-SNA versus C-HSNA, followed by N-HSNA versus E-SNA and C-HSNA, and lastly, Admix versus N-HSNA after comparison with E-SNA and C-SNA. (C to E) Volcano plots showing differential gene expression in CD8⁺ T cells from pairwise comparisons: (C) N-HSNA versus Admix, (D) N-HSNA versus E-SNA, and (E) N-HSNA versus C-HSNA. Colored dots indicate significantly expressed genes; a positive LFC indicates an up-regulation for N-HSNA with respect to admix, E-SNA, and C-HSNA (red) whereas a negative LFC indicates a down-regulation for N-HSNA with respect to admix, E-SNA, and C-HSNA (blue).

Hierarchical clustering revealed distinct gene expression signatures across different groups (Fig. 4B). CD8⁺ T cells from N-HSNA-treated mice up-regulated genes associated with T cell activation, survival, and cytotoxicity, including Traj4, 11, 26, 29, 35, and 39, and Cxcr4, Bcl2, Nkg7, and Cd6, compared to the admix or other SNA groups (Fig. 4B). Volcano plot analysis further showed that N-HSNA–treated T cells down-regulated immunosuppressive or stress-related genes (Cd74, Hes1, Nfkbib, and Tnfaip8l2) and up-regulated genes involved in trafficking (Ccr5 and Cxcr4), effector function (Nkg7), survival (Cd27), and immune regulation (Ikzf2) as compared to the admix (Fig. 4C). Compared to E-SNA, N-HSNA down-regulated exhaustion- and suppression-associated genes (Lag3, Pirb, Ptpn6, Cd244a, Serpina3g, and Atg16l2), while up-regulating activation and cytokine signaling genes (Trim39, Il18r1, Txk, and Rarg), indicating a more functional CD8⁺ T cell phenotype (Fig. 4D). Compared to C-HSNA, N-HSNA induced expression of cytotoxicity- and migration-related genes (Gzmb, Dtx1, Ccr2, Tgfbi, Cdkn2b, and Serpinb9b), while suppressing genes associated with lymphoid tissue homing, costimulation, and tissue residency (S1pr1, Cd28, Cd96, and Itgae), suggesting a shift away from a regulatory-like phenotype (Fig. 4E).

Gene ontology (GO) enrichment analysis (44) reinforced these findings, showing that N-HSNA treatment up-regulated pathways involved in T cell differentiation, lymphocyte activation, and αβ T cell commitment relative to E-SNA treatment (fig. S17A). In contrast, E-SNA–treated T cells showed enrichment in pathways associated with cytoskeletal organization and reduced motility. Compared to C-HSNA enrichment, N-HSNA enriched programs linked to T cell activation, lipid metabolism, and nuclear division, suggesting enhanced functional readiness (fig. S17B). Together, these transcriptional data highlight the ability of N-HSNA to reprogram CD8⁺ T cells toward an activated, cytotoxic phenotype aligned with effective antitumor responses. These results mechanistically support the superior systemic and intratumoral CD8⁺ T cell expansion and effector function induced by N-HSNA observed in vivo.

Ex vivo evaluation of SNA efficacy in patient-derived HPV-HNSCC tumor models

To determine whether the enhanced E7_11-19-specific CD8⁺ T cell responses induced by the N-HSNA structure translated into cytotoxic efficacy against patient-derived HPV-HNSCC tumors, an ex vivo 3D culture system was established (38, 39). Patient-derived tumor spheroids were maintained within a microfluidic device, and viability and multicellular structure were monitored over time. Confocal imaging revealed that the spheroids remained ~89% viable (Calcein-AM⁺; green) for at least 1 week and that the multicellular architecture of the HPV-infected tumor cells (coexpressing red and green fluorescence) and CD45⁺ immune cells (cyan) was maintained (Fig. 5A).

Fig. 5. Evaluation of SNA-induced cytotoxicity in ex vivo HPV-HNSCC patient–derived tumor spheroid models.

(A) Bright-field image of patient-derived tumor spheroids cultured in a microfluidic platform for 7 days. Representative 3D confocal image shows live (green) and dead (red) cells, and immunofluorescence staining for CD45⁺ (cyan), EGFR⁺ (red), and HPV-E7⁺ (green) cells. Scale bar, 40 μm. (B) Frequency of CD3⁺, CD11c⁺, and CD19⁺ immune cell populations in tumor specimens from patients with HPV-HNSCC analyzed by flow cytometry. (C) Quantification of live and dead cells per spheroid based on Imaris analysis of ≥10 spheroids per group. Data are presented as mean ± SD (n = 4 to 9). Statistical significance was assessed by one-way ANOVA with Tukey’s post hoc test; *P < 0.05, **P < 0.01, and ***P < 0.005.

SNA uptake was measured within the tumor spheroids using fluorophore-labeled formulations. The results indicate that compared to liposome-only or CpG-only controls, SNAs functionalized with single-stranded DNA (S-SNA, analogous to E-SNA) or double-stranded DNA (D-SNA, analogous to C-HSNA and N-HSNA) lacking antigens showed substantially enhanced colocalization of liposome and CpG signals within the tumor spheroids (fig. S18). Furthermore, SNA-treated patient-derived tumor cells exhibited enhanced uptake by CD11c⁺ DCs, consistent with in vitro trends (Fig. 2A), with S-SNA achieving ~2-fold greater DC uptake than D-SNA (fig. S19). These findings underscore the structural advantage of SNAs in improving delivery and cellular engagement within the tumor microenvironment.

Tumor spheroids were derived from three patients with HPV-HNSCC with distinct immune profiles: Tumors from the neck of patient 1 exhibited low immune cell infiltration (~2%) and was HLA-A2 negative, and tumors from the tonsils of patients 2 and 3 exhibited high immune cell composition (~85%), including CD3⁺ T cells, CD11c⁺ DCs, and CD19⁺ B cells, and they were HLA-A2 positive (Fig. 5B and fig. S20).

Treatment parameters were optimized in tumor spheroids isolated from TC-1 tumor–bearing AAD mice (figs. S21 and S22), including SNA concentration (CpG and antigen) and incubation duration. A concentration of 50 nM with a 24-hour incubation was identified as the optimal condition. This dose elicited maximal SNA-mediated cytotoxicity and produced consistent responses across replicates. These optimized parameters were subsequently applied to patient-derived tumor spheroids. Therapeutic efficacy was assessed by quantifying live (Calcein-AM⁺; green) and dead (BOBO-3 Iodide⁺; red) cells using confocal imaging. Direct measurement of IFN-γ or CD8⁺ T cell activation within patient-derived spheroids was not performed because of (1) the limited number of immune cells (40 spheroids per microfluidic chip) and (2) cytokine sequestration within the hydrogel matrix (45) make accurate quantification impractical. However, spheroids lacking immune cells and treated under identical conditions did not exhibit increased cytotoxicity, confirming that the observed effects are immune dependent rather than nonspecific. The spheroids derived from patient 1 exhibited minimal cytotoxicity (<10% dead cells) across all treatment groups, likely due to the absence of HLA-A2 expression and limited immune infiltration (fig. S20 and Fig. 5C). In contrast, spheroids derived from patient 2 and patient 3 exhibited enhanced cytotoxicity following SNA treatment, particularly with N-HSNA, resulting in ~20% and ~30% dead cells, respectively, compared to the admix control (fig. S23 and Fig. 5C). In summary, these results show that N-HSNA effectively penetrates tumor spheroids and triggers cytotoxicity in an immune-dependent manner, highlighting the translational potential of SNA as a promising strategy for personalized cancer immunotherapy.

DISCUSSION

Using a clinically relevant human tumor antigen (E7_11-19), this study demonstrates that antigen configuration, specifically epitope location and orientation within SNAs, shapes CD8⁺ T cell activation and antitumor immunity, establishing it as a key parameter in vaccine designs based on nanomedicines. The results show that both antigen placement and orientation on the SNA dictates the magnitude and quality of human antigen–specific cytotoxic T cell responses. While E-SNA displayed robust uptake and activation in hDCs, the N-HSNA induced the most potent systemic and intra tumoral expansion of E7_11-19-specific CD8⁺ T cells, exhibiting a transcriptional and phenotypic profile that correlates with strong antitumor response. Consistent with our previous findings linking SNA-mediated T cell activation to tumor regression in murine models (22, 23, 26, 43), the data presented here collectively support a CD8⁺ T cell–mediated mechanism underlying the antitumor efficacy of N-HSNA.

Although the overall magnitude of in vivo immune activation among constructs was modest, the consistent trends across multiple readouts, including T cell functionality, cytokine secretion, and tumor cytotoxicity, support the conclusion that the N-HSNA confers a measurable functional advantage. The more pronounced differences observed in the patient-derived tumor spheroid model likely reflect the higher sensitivity of the ex vivo human tumor microenvironment where antigen processing, MHC-I presentation, and immune cell interactions occur with higher spatial and temporal resolution than whole animal systems. In vivo immune responses may be attenuated by systemic variables such as lymphatic drainage, biodistribution, and inter animal heterogeneity. Furthermore, although the AAD mouse model expresses human HLA-A2, it still relies on a murine immunobiology, including species specific DC subsets, costimulatory signaling networks, and cytokine milieu. These differences can limit the capacity to resolve human antigen–specific immune responses when compared with human-derived ex vivo models.

The observed differences in immune potency likely arise from differences in formulation architecture and antigen stability. Although not directly measured in this study, H-SNAs are expected to exhibit more stable architectures than E-SNAs due to the interfacial reinforcement conferred by DNA-peptide hybridization at the particle surface. This configuration may reduce antigen diffusion and premature release compared to encapsulation within the aqueous liposomal core. These interpretations align with prior reports (22, 23), which demonstrate that H-SNAs exhibit slower antigen release and improved intracellular delivery of peptide and CpG compared to encapsulated constructs, supporting the enhanced formulation stability of the hybridized design.

Mechanistically, the superior performance of N-HSNA over C-HSNA and E-SNA likely reflects enhanced epitope accessibility and more efficient MHC-I presentation. N-terminal conjugation preserves the native C terminus of the E7_11-19 peptide, which is critical for proteasomal cleavage and MHC-I binding, whereas C-terminal modification may disrupt proteasome-defined termini and hinder efficient antigen processing (5, 46). These findings highlight that epitope integrity (47–49) and anchor residue preservation are essential considerations for vaccine formulation.

In contrast, previous work with murine MHC-I epitopes (e.g., OVA1) and cholesterol-anchored CpG DNA showed improved antigen retention, endosomal colocalization, and synchronization of antigen and adjuvant processing in murine DCs (22), promoting superior cross-presentation and T cell priming. However, the translation relevance was limited by the use of model antigens and murine MHC-I epitope, since murine and human MHC-I processing differ substantially with regard to peptide binding (50). Furthermore, DC subsets across species diverge in receptor expression and subset composition (51), altering how SNAs are internalized and processed, increasing translational complexity.

Leveraging high-fidelity preclinical models, including AAD mice expressing human HLA-A2 and ex vivo patient-derived tumor spheroids from HPV⁺ HNSCC, we showed that even subtle changes in antigen configuration profoundly influence nanovaccine efficacy. Shifting the conjugation site from the N terminus (as in N-HSNA) to the C terminus (as in C-HSNA), markedly reduced CD8⁺ T cell responses (5), reinforcing the critical role of structural precision in epitope presentation and immune activation and underscoring the importance of evaluating human-restricted epitopes to guide clinical translation.

Despite these advances, the current study does not provide mechanistic insight into antigen presentation due to the technical challenges in detecting E7_11-19 HLA-A2 complexes (52). Future work using advanced antigen-tracking tools or high-sensitivity mass spectrometry may address this gap and further refine our understanding of how human antigens within SNAs modulate intracellular processing pathways. In addition, while HLA-A2 is one of the most prevalent MHC-I alleles among patients with HPV⁺ HNSCC (53), our therapeutic evaluation remains restricted to this single allele. This represents a study limitation, as the reliance on one HLA type constrains generalizability. Broader clinical translation will require identifying immunodominant epitopes for other HLA types (54) and validating them in appropriate humanized models (55) to ensure population-wide applicability. Our findings using AAD mice and HLA-A2⁺ patient-derived organotypic tumor cultures establish a strong foundation, but further collaborations with clinicians and immunogenomics experts will be necessary to enable personalized vaccine strategies across diverse patient populations.

In summary, this work adds to the growing body of evidence that structure may be as important as choice of adjuvant and antigen in designing the most effective vaccine architectures (15) Moreover, an understanding of structure-function correlations may bridge the gap between murine-based mechanistic insights and clinically relevant human HLA-restricted immunotherapies. By integrating human tumor antigens, (C12)₉ anchoring, and humanized models, this study provides a framework for the design of structural nanomedicines that will account for the immunological nuances of the human immune system.

MATERIALS AND METHODS

Materials and cell lines

The phosphoramidites and oligonucleotide synthesis reagents were purchased from Glen Research. The peptides, used in this study, HPV E7_11-19 (YMLDLQPET), CE7_11-19 (CYMLDLQPET), and E7_11-19C (YMLDLQPETC) were purchased from Genscript, with a guaranteed purity >95%. Human PBMCs and CD14⁺ monocytes were obtained from Cytologics, thawed and used immediately for experiments as indicated. Human anti-HPV E7_11-19 CD8⁺ T cells were acquired from Charles River Labs Cell Solutions, thawed and used immediately for experiments. T2 cells were purchased from American Type Culuture Collection and cultured according to their recommendations in Iscove’s modified Dulbecco’s medium supplemented with 20% heat-inactivated fetal bovine serum (FBS) and 1% penicillin-streptomycin. UM-SCC-01 (HPV16⁻ and HLA-A*02⁻) and UM-SCC-104 (HLA-A*02⁺) cells were acquired from Millipore Sigma and cultured in Dulbecco’s modified Eagle’s medium containing 10% FBS, 1% penicillin-streptomycin, and 1% nonessential amino acids. TC-1 cells were provided by B. Zhang at Northwestern University. TC-1/A2 was received from T. C. Wu at Johns Hopkins University and cultured in Roswell Park Memorial Institute (RPMI) 1640 medium supplemented with 2 mM glutamine, 1 mM sodium pyruvate, 20 mM HEPES, 50 μM β-mercaptoethanol, 1% penicillin-streptomycin, and 10% FBS.

Animal use and care

All animal protocols (IS00010970) were approved by the institutional animal care and use committee at Northwestern and were conducted in accordance with national and local guidelines and regulations. Female B6.Cg-Immp2lTg(HLA-A/H2-D)2Enge/J (AAD) mice (6 to 9 weeks old) were purchased from Jackson Laboratory and used immediately for in vivo studies.

SNA synthesis and characterization

DNA oligonucleotide synthesis, DNA-antigen conjugation, and liposome nanoparticle functionalization and characterization are discussed in the Supplementary Materials.

Human PBMC cell uptake

Rhodamine dye-labeled liposomes containing 1% (molar) 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (ammonium salt) (Avanti) were synthesized and used to assemble SNAs. Human HLA-A*02⁺ PBMCs (Zenbio or Cytologics) were thawed from −80°C storage quickly in a 37°C water bath and diluted in 10 ml of RPMI (+/+). The PBMCs were then counted, pelleted, aspirated, and resuspended at 2 × 10⁶ cells ml⁻¹. PBMCs (100 μl, 200,000 cells) in RPMI (+/+) were treated with either liposomes, or (C12)₉-anchored SNA (liposome concentration of 1.4 nM) for 1 or 4 hours. Following this incubation, the PBMCs were washed with 600 μl of PBS, pelleted, and resuspended in 100 μl of Dulbecco’s PBS (DPBS) containing 0.5 μl of the following staining antibodies: fixable ultraviolet (UV) live/dead, CD11c [fluorescein isothiocyanate (FITC)]. The PBMCs were then incubated at 4°C for 15 min then washed with 600 μl of DPBS, pelleted, and resuspended in 100 μl of DPBS for flow cytometry analysis.

Human pDC activation

PBMCs (100 μl, 200,000 cells) were treated with an additional 100 μl of media containing human CpG (500 nM) targeting human TLR9 with peptide or E7_11-19 (C12)₉–anchored E-SNA, N-HSNA, or C-HSNA was added and incubated with the PBMCs at a final treatment concentration of 250 nM for 24 hours. Afterward, the PBMCs were washed with an additional 600 μl of PBS, then pelleted, aspirated, and stained for 15 min at 4°C in 100 μl of DPBS containing 0.5 μl of the following antihuman antibodies: fixable UV live/dead, HLA-DR (APC), CD123 (PE), and CD83 (FITC). The PBMCs were then washed with 600 μl of DPBS, pelleted, aspirated, and fixed in 100 μl of fixation buffer then stored in 100 μl of DPBS at 4°C before flow cytometry analysis.

Human IFN-γ ELISpot assay

IFN-γ secretion was measured with a human IFN-γ ELISpot kit (BD) according to the manufacturer’s protocol. Before performing the IFN-γ ELISpot assay, human HLA-A*02⁺ THP-1 were cultured with hGM-CSF (100 ng/ml) and hIL-4 (100 ng/ml) for 5 days to differentiate them into immature DCs (iDCs), followed by 1 day of culturing with ionomycin (200 ng/ml) and hTNF-α (20 ng/ml) to generate mature DCs (mDCs) (36). mDCs were treated with SNAs (250 nM, CpG and antigen) for 24 hours. The day before coculture of mDCs and E7_11-19 CD8⁺ T cells (Charles River Laboratories), a 96-well ELISpot plate was coated with a 100-μl solution of IFN-γ capture antibody (1:200) in DPBS and incubated at 4°C for 16 hours. The plate was then washed and blocked for 2 hours with 200 μl of RPMI (+/+). The blocking solution was next removed and replaced 100 μl RPMI (+/+) containing the E7_11-19, CE7_11-19, or E7_11-19C peptides (100 μg/ml), a control E6MHCII peptide (100 μg/ml), or a positive control (ionomycin at 200 ng/ml). An additional 100 μl of RPMI (+/+) containing 1 × 10⁵ mDCs and 2 × 10⁵ E7_11-19 CD8⁺ T cells was added to the plate and incubated for 24 hours at 37°C in a 5% CO₂ incubator. The plate was then washed twice with deionized water and three times with wash buffer prepared according to the manufacturer’s protocol. Washes were followed by the addition of the biotinylated IFN-γ detection antibody (1:250) and a 2-hour incubation at room temperature (RT). The plate was washed three times with wash buffer I, and the streptavidin–horseradish peroxidase enzyme conjugate (1:250) was added. Spot formation was monitored and quenched with deionized water ~20 min following the addition of substrate. The plate was dried overnight, and spot-forming cells automatically were counted on a CTL Immunospot Analyzer.

Human E7_11-19 CD8⁺ T cell proliferation

Human HLA-A*02⁺ CD14⁺ cells were cultured with hGM-CSF (50 ng/ml) and hIL-4 (100 ng/ml) for 7 days to differentiate into iDCs, followed by 1 day of treatment with SNAs (100 nM, CpG and E7_11-19) to form mDCs. Before the coculture of mDCs (20,000 cells) and E7_11-19 CD8⁺ T cells (50,000 cells) with the ratio of 1:2.5 for 48 hours, mDCs were stained with Cell Proliferation Dye eFluor 450. After a 48-hour incubation on the round bottom 96-well plate, all cells transferred to flow tubes were washed with an additional 600 μl of PBS, then pelleted, aspirated, and stained for 15 min at 4°C in 100 μl DPBS containing 0.5 μl of the following antihuman antibodies: fixable UV live/dead and CD8 (FITC). The cells were then washed with 600 μl of DPBS, pelleted, aspirated, and resuspended in 100 μl of DPBS for flow cytometry analysis, which was performed immediately.

Human CD8⁺ T cell killing assay

Before a CD8⁺ T cell killing assay, CD8⁺ T cells in human PBMCs were expanded and reactivated following this process: human HLA-A*02⁺ PBMCs [1 × 10⁶ cells in 450 μl of RPMI (+/+)] were placed to each well of a 24-well plate and incubated with either an admixture of linear CpG and E7_11-19 peptide, cholesterol SNAs, or (C12)₉ SNAs at a final concentration of 250 nM in 900 μl of RPMI (+/+) for 48 hours at 37°C in a 5% CO₂ incubator. CD8⁺ T cells from human PBMCs following expansion with SNAs were isolated using a human CD8 Positive Selection Kit II (Stem Cell Technologies), according to the manufacturer’s directions. Following isolation, which was performed in 1X MOJOSort Buffer (BioLegend), the cells were counted and resuspended at 5 × 10⁶ cells ml⁻¹ in RPMI (+/+). Concurrently, the target cells (either UM-SCC-01 or UM-SCC-104 cells) were washed with DPBS and stained for 10 min with a 10 μM solution of eFluor 450 cell proliferation dye (Thermo Fisher Scientific) in DPBS with a cell concentration that did not exceed 1 × 10⁷ cells ml⁻¹). The target cells were then diluted with two volumes of RPMI (+/+) and incubated at 4°C for 5 min. The target cells were washed with RPMI (+/+) and resuspended at 5 × 10⁴ cells ml⁻¹ in RPMI (+/+) and 100 μl plated in a 96-well round bottom plate. The isolated CD8⁺ T cells were then added to a 96-well plate at ratios of 25:1 (CD8⁺ T cells:target cells) in a total volume of 200 μl and incubated for 24 hours at 37°C in a 5% CO₂ incubator. Following this incubation, the cells were trypsinized and transferred to flow tubes, washed with 1× DPBS, and stained in 100 μl of 1× Annexin V binding buffer (BioLegend) containing 0.5 μl of Annexin V (BioLegend) and 7-aminoactinomycin D (Thermo Fisher Scientific) before flow cytometry analysis, which was performed immediately.

In vivo tumor inhibition studies

Female AAD mice, which contain the α-1 and α-2 domains of the human HLA-A2.1gene and the α-3 transmembrane and cytoplasmic domains of the mouse H-2Dd gene were inoculated with subcutaneous injections of 2 × 10⁵/2.8 × 10⁵ TC-1/A2 cells in their right flanks. Mice were treated with 3 nmol of CpG/peptide as either a simple mixture or an SNA formulation via subcutaneous injection in the abdomen. Immunizations were administered according to the treatment schedule (see Fig. 3). For combination therapy with the immune checkpoint inhibitor anti–PD-1, mice received 20 μg of anti-mouse PD-1 (Clone RMP1-14, BioXCell) via intraperitoneal injection. Starting 5 days after tumor inoculation, tumor length and width were measured every 2 to 3 days in a partially blinded manner. Tumor volume was calculated using the following equation: tumor volume = length × width² × 0.5. Mice were sacrificed after their tumors reached 1500 mm³.

Collection procedure

At day 20 of tumor inhibition studies, spleen, tumor, and blood were collected from TC-1/A2 tumor–bearing AAD mice after three rounds of SNA vaccination and anti–PD-1 antibody treatment. The removed spleens or tumors were collected and temporarily stored in 5 ml of RPMI (+/+) until the physical dissociation process. In brief, either spleen or tumor was gently mashed with a sterile syringe plunger and passed through a 70-μm cell strainer with a constant flow of DPBS. The tumor cells were then counted, pelleted, and resuspended in DPBS at 1 × 10⁸ cells/ml. The splenocytes were pelleted, and red blood cells were lysed with 3 ml of ACK lysing buffer (Gibco) for 4 min. After lysis, the cells were diluted to 30 ml in DPBS, counted, pelleted, and resuspended in RPMI (+/+) at 1 × 10⁸ cells/ml. Approximately 250 μl of blood per mouse was collected via cardiac puncture using an EDTA-coated syringe and transferred to an EDTA-lined collection tube (BD). Blood samples were diluted to 1 ml with DPBS, and PBMCs were isolated using SepMate PBMC Isolation Tubes and Lymphoprep (STEM Cell Technologies) according to the manufacturer’s protocol. In brief, Lymphoprep was added through a SepMate insert, followed by the addition of diluted blood samples onto the insert. After centrifugation at 1200g for 10 min, the second layer from the top was collected and washed twice with DPBS. The cells were counted, pelleted, and resuspended in DPBS at 1 × 10⁸ cells/ml.

T cell memory phenotype profiling

Splenocytes (3 × 10⁵) in RPMI (+/+) were washed with 600 μl of DPBS, aspirated, then blocked with 0.5 μl Trustain murine FcX blocker in 50 μl of DPBS. A staining solution containing 0.5 μl of the following antibodies in 50 μl of DPBS was then prepared: CD8 (APC, BD), CD44 (FITC, BD), CD62L (BV421, BD), and a fixable blue live/dead stain (Thermo Fisher Scientific). This solution was added to the splenocytes, vortexed briefly and incubated at 4°C for 15 min. The splenocytes were then washed with 600 μl of DPBS, pelleted, and fixed in 100 μl of fixation buffer (BD) at 4°C for 15 min. After washing with 600 μl of DPBS, the cells were pelleted and stored in 100 μl of DPBS at 4°C before flow cytometry analysis.

Intracellular IFN-γ staining

Splenocytes (4 × 10⁶) were restimulated for 4 hours at 37°C in a 5% CO₂ incubator with 500 μl of RPMI (+/+) containing monensin (2 μM, BioLegend), brefeldin A (5 μg/ml, BioLegend), FITC-CD107a surface antibody, and the E7_11-19 peptide (10 μg/ml). Following this incubation, splenocytes were washed with 500 μl of DPBS, then resuspended in 50 μl of DPBS with 0.5 μl of Trustain murine FcX Blocker (BioLegend). An additional 50 μl of DPBS containing 0.5 of μl of the following surface antibodies: CD8 (APC, BD), and a blue fixable live/dead cell stain (Thermo Fisher Scientific) was added followed by a brief vortex and a 15-min incubation at 4°C. The splenocytes were then washed with 600 μl of DPBS, pelleted and resuspended in 100 μl of fixation and permeabilization solution (BD), vortexed, and incubated for 20 min at 4°C. The splenocytes were then washed with 300 μl of 1× permeabilization wash buffer (BD), pelleted, and resuspended in 100 μl of permeabilization wash buffer containing 0.5 μl of IFN-γ antibody (PE-Cy7, BD) and stored at 4°C before flow cytometry analysis to quantify the amount of intracellular IFN-γ production.

E7_11-19 CD8⁺ T cells assessment in PBMCs and tumors

PBMCs or tumor cells (2 × 10⁵) in DPBS were washed with 600 μl of DPBS, aspirated, then resuspended in 50 μl of DPBS containing 10 μl of APC-Pro5 MHC Pentamer for E7_11-19 (PROIMMUNE) followed by a brief vortex and 10-min incubation at RT. The cells were then washed with 600 μl of DPBS, pelleted, and resuspended in 100 μl of DPBS containing 0.5 μl of the following surface antibodies: CD8 (PE, BD), CD19 (FITC, BD), and a blue fixable live/dead cell stain (Thermo Fisher Scientific) was added followed by a brief vortex and a 15-min incubation at 4°C. The cells were then washed with 600 μl of DPBS, pelleted, and resuspended in 100 μl of fixation buffer and stored at 4°C before flow cytometry analysis.

Bulk RNA sequencing

CD8⁺ T cells were isolated from whole splenocytes collected using magnetic mouse CD8⁺ positive selection kits (Stem Cell Technologies) from 20 days of TC-1/A2 tumor inoculation with three rounds of combined SNA and anti–PD-1 treatments. From these isolated CD8⁺ T cells (1 × 10⁶), RNA was extracted, and total RNA sequencing were conducted using DV200 (RNA fragments >200 bp) at the Northwestern University NUSeq Core Facility. In brief, total RNA samples were checked for quality using RNA integrity numbers generated from an Agilent Bioanalyzer 2100. RNA quantity was confirmed using a Qubit fluorometer. The Illumina Total RNA-Seq with Ribo-Zero Plus Library Preparation kit was used to prepare sequencing libraries from 20-ng RNA input. The kit procedure, including rRNA depletion, fragmentation, cDNA synthesis, Illumina adapter ligation, library PCR amplification and validation, was performed without modifications. Libraries were sequenced using an Illumina NovaSeqX sequencer to generate 50-bp single reads at a depth of 20 to 25 million reads per sample.

RNA sequencing analysis

The quality of reads, in FASTQ format, was evaluated using FastQC. Reads were trimmed to remove Illumina adapters from the 3′ ends using cutadapt (56). Trimmed reads were aligned to the Mus musculus genome (mm39) using STAR (57). Read counts for each gene were calculated using htseq-count (58) in conjunction with a gene annotation file for mm10 obtained from Ensembl. Normalization and differential expression were calculated using DESeq2 that uses the Wald test (59). The cutoff for determining significantly differentially expressed genes (DEGs) was a false discovery rate–adjusted P value less than 0.05 using the Benjamini-Hochberg method.

Gene set enrichment analysis

DEGs obtained from DESeq2 were stratified into up-regulated (log2 fold change >0) and down-regulated (log2 fold change <0) subsets. Then, GO enrichment analysis was performed separately for each subset using the enrichGO() function from the clusterProfiler R package (version 4.6.2). Gene annotations were retrieved from the org.Mm.eg.db database (version 3.17.0), with gene symbols used as input keys (keyType = “SYMBOL”). All GO domains (“Biological Process,” “Molecular Function,” and “Cellular Component”) were included (ont = “ALL”). The Benjamini-Hochberg method was used for P value adjustment (pAdjustMethod = “BH”), and GO terms with adjusted P values <0.05 and q values <0.2 were considered significant. The top GO terms with the lowest adjusted P values were selected from both the up-regulated and down-regulated gene sets. Enrichment scores were calculated as the ratio of observed gene counts (Count) to the number of genes in the GO category, extracted from the leading fraction of the GeneRatio string.

Gene expression profiles

The overlapping DEGs were extracted for further visualization (Control = “Admix”; Experimental = “N_SNA”, “C_SNA”, “E_SNA”). The “counts” for the overlapping DEGs were averaged and stratified, followed by the z-score normalization of the gene expression matrix. The heatmap was generated using the pheatmap package (version 1.0.12). Both rows and columns were clustered using hierarchical clustering with Euclidean distance and complete linkage, allowing unsupervised pattern discovery.

Preparation of MDOTS/PDOTS and ex vivo microfluidic 3D culture

Tumor samples from patients with HPV-HNSCC were acquired from the Robert H. Lurie Comprehensive Cancer Center at Northwestern under IRB-approved protocols. Informed consent was obtained from all patients. For MDOTS preparation, female C57BL/6 or AAD mice were inoculated with subcutaneous injections of 5 × 10⁵ E.G7 OVA or 2 × 10⁵ TC-1 cells, respectively in their right flanks. Once the tumor volumes exceeded 1000 mm³, the tumor was collected for MDOTS preparation. MDOTS/PDOTS (40 to 100 μm) were prepared as described in previous reports (45). In brief, fresh tumor specimens either from patients or mice were physically minced using sterile forceps and scalpels until the samples could be pipetted using 1000-μl pipette tips. The minced samples were sequentially passed over 100- and 40-μm filters in RPMI (+/+) to obtain spheroid (40 to 100 μm) and single-cell (<40 μm) fractions. The spheroids were pelleted and resuspended in the neutralized type I bovine collagen (4 mg/ml). Approximately 40 spheroids in collagen pre-gel (10 μl) were loaded into the center gel region of a 3D microfluidic chamber (DAX-1, AIM BIOTECH). The 3D microfluidic chambers were placed in sterile humidity chambers. After collagen gelation at 37°C for 30 min in a 5% CO₂ incubator, RPMI (+/+) was added to the media port and cultures were maintained at 37°C in a 5% CO₂ incubator. RPMI (+/+) was replaced every day up to 7 days of culture.

SNA localization within PDOTS culture

Rhodamine dye–labeled liposomes containing 1% (molar) 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine rhodamine B sulfonyl) (ammonium salt) (Avanti) were synthesized and used with Cy5-labeled (C12)₉-hCpG to assemble SNAs. On day 5 of culturing ~200 PDOTS in 3D microfluidic chambers, rhodamine-labeled liposomes, Cy5-CpG, S-SNA [containing Cy5-labeled (C12)₉-hCpG, mimicking E-SNA], or D-SNA [containing Cy5-labeled (C12)₉-hCpG duplex, mimicking N-HSNA and C-HSNA] were added at the final concentration of 1.42 nM liposomes and 250 nM CpG, followed by overnight incubation. After three DPBS washes, the samples were visualized using Z-stack imaging on a Leica SP8 confocal microscope, and 3D reconstructions were generated using Imaris software.

Live/dead and immunofluorescence staining of MDOTS/PDOTS

Live/dead fluorescence staining was performed using calcein-AM (live)/ethidium homodimer-1 (dead) staining solution in RPMI (+/+) (Thermo Fisher Scientific). Following the treatment (either simple mixture or SNAs with 25 nM CpG/E7_11-19) for 2 days, at day 5 of PDOTS or MDOTS culture, PDOTS or MDOTS were incubated with calcein-AM/ethidium homodimer-1 staining solution at 37°C for 30 min in a 5% CO₂ incubator. Live and dead cells within PDOTS or MDOTS were visualized with Z-stack imaging using a Zeiss LSM800 confocal microscope. Live and dead cell quantitation was performed using Imaris software (Oxfold Instruments) with the automated methods applying defined intensity thresholds for the live and dead fluorescence channels. Cell viability was calculated by dividing the integrated fluorescence intensity for the live channel by the sum of the integrated fluorescence intensities for both the live and dead channels. For immunofluorescence staining, PDOTS was fixed with 4% PFA at 4°C for 2 hours, and washed three times with DPBS. They were then permeabilized with 0.2% Triton X-100 at 4°C for 2 hours. After three additional DPBS washes, PDOTS was blocked with 5% bovine serum albumin (BSA) overnight at 4°C. The directly conjugated antibodies used for PDOTS staining were CD45-AlexaFluor-647 (BioLegend, HI30), EGFR-AbBy Fluor-594 (BioLegend, AY13), and HPV16 E7-AlexaFluor-488 (Bioss). Antibodies were diluted to the following concentrations: CD45 (0.1 mg/ml), EGFR (0.005 mg/ml), and E7 (0.025 mg/ml) in 10 μg/ml solution of Hoechst 33342 (prepared in 1% BSA in DPBS). The antibody mixture was loaded into microfluidic chambers and incubated for 48 hours at 4°C in the dark. The spheroids were then washed three times with DPBS containing 0.02% Tween-20 for 24 hours, followed by three additional DPBS washes. PDOTS was visualized with Z-stack imaging on a Leica SP8 confocal microscope, and 3D reconstructions were generated using Imaris software.

Cellular phenotypic profiles of tumors from patients with HPV-HNSCC

Single cells (5 × 10⁵; <40 μm) in DPBS from each of the patients with HPV-HNSCC were washed with 600 μl of DPBS, aspirated, then resuspended in 100 μl of 1% BSA DPBS containing 0.5 μl of PE-CD3 (BioLegend, UCHT1), APC-CD19 (BioLegend, HIB19), FITC-CD11c (BioLegend, Bu15), PE-HLA-A2 (BD, BB7.2), or their corresponding isotype control antibodies. A blue fixable live/dead cell stain (Thermo Fisher Scientific) was added followed by a brief vortex and a 15-min incubation at 4°C. The cells were then washed with 600 μl of DPBS, pelleted, and resuspended in 100 μl of fixation buffer and stored at 4°C before flow cytometry analysis. Samples were analyzed on a Symphony A3 flow cytometer (BD Biosciences). Debris, multiplets, and nonviable cells were gated out, and marker positivity was determined relative to the isotype controls.

Software and statistical analysis

Flow cytometry data were analyzed using FlowJo (version #10.10.0) for Windows 10 (BD, www.flowjo.com). Statistical analyses indicated (as in the figure captions) were performed using GraphPad Prism version 9.2.0 for Windows 10 (www.graphpadprism.com). Outliers in the in vivo dataset were removed using the ROUT method in Prism with a false discovery rate (Q) set to 1%.

Supplementary Materials

The PDF file includes:

Supplementary Materials and Methods

Figs. S1 to S23

Tables S1 to S3

Legends for datasets S1 and S2

Other Supplementary Material for this manuscript includes the following:

Datasets S1 and S2