TLR9 plus STING Agonist Adjuvant Combination Induces Potent Neopeptide T Cell Immunity and Improves Immune Checkpoint Blockade Efficacy in a Tumor Model Open Access

Immune checkpoint blockade (ICB) therapies have demonstrated a remarkable therapeutic efficacy against a variety of cancers in humans over the past decade (1–6). Despite their clinical success, however, only a subset of patients benefits from ICB therapies and only ∼30% of different solid tumors respond (7–10). One predictor of ICB therapy response in patients is the abundance of tumor mutational load that presumably leads to increased expression of neoantigens, which elicits more Ag-specific T cells that attack tumor cells (11–16). On the basis of the above information, amplifying antitumor neoantigen-specific T cell responses is expected to improve ICB therapy efficacy, and a number of clinical trials have attempted this with neoantigen vaccines (10, 17–19).

One such neoantigen vaccine approach is based on neopeptide vaccines that use synthetic peptides in combination with adjuvants to boost antitumor immune responses, specifically targeting mutation-derived neoantigen-specific CD4⁺ and CD8⁺ T cells with the goal to eliminate tumor cells. However, although such neopeptide vaccines have been tested in clinical trials alone or in combination with ICB, clinical benefit has yet to be demonstrated while elicited T cell responses remain low (18–21). Currently, the most frequently used adjuvant in clinical trials of neopeptide vaccines is the TLR3 agonist poly-ICLC, an attenuated human analog of polyinosinic:polycytidylic acid [poly(I:C)] mixed with the stabilizers carboxymethylcellulose and polylysine. Personalized neoantigen vaccines formulating synthetic long peptides (SLPs) with poly-ICLC have been tested in patients with melanoma and glioblastoma. These demonstrated the induction of Ag-specific CD4⁺ and CD8⁺ T cells, albeit in very low frequencies, especially for the CD8⁺ T cell compartment, and with no clinical benefit (20, 21). Therefore, there is still a need for new vaccine strategies that induce better cytotoxic CD8⁺ T cell responses. Although neopeptide selection is one aspect of the problem, a more important issue resides in the potency of vaccine adjuvants used. These adjuvants are especially needed in neopeptide vaccines, because neopeptides derived from somatic tumor mutations are often similar to self-peptides and are generally of low affinity, whereas peptides on their own generally lack sufficient immunogenicity (22, 23). Therefore, employing the appropriate adjuvants is crucial to induce strong CD8⁺ T cell responses and to produce an effective peptide-based vaccine. For these reasons, in this study, we aimed to develop a highly immunogenic vaccine using a combination of pattern recognition receptor agonists in combination with 20-aa-long SLPs to induce potent Ag-specific cytotoxic CD8⁺ T cell responses. Such SLPs are known to be efficiently cross-presented by dendritic cells, and, when combined with appropriate adjuvants, they can stimulate T cell immunity (19, 24).

A growing number of pattern recognition receptor agonists have been proposed and tested as adjuvants in vaccine design (25). CpG oligodeoxynucleotides (CpG-ODNs) are adjuvants that activate APCs through binding to TLR9 (26, 27). In particular, the K-type (also known as B) of CpG-ODN is capable of inducing both humoral and cellular immune responses (28, 29). Meanwhile, it has been shown that stimulator of IFN genes (STING) agonists, such as cyclic GMP-AMP, are also potent adjuvants that are able to induce the expansion of Ag-specific CD8⁺ T cells via the induction of type I IFN production (30). Importantly, K3 CpG-ODN and STING agonists can act synergistically to induce Th1-type immune responses and suppress tumor growth in mice in vivo (29, 31).

In the present study, we employed the combination of humanized K-type K3 CpG-ODN and c-di-AMP as an adjuvant, and we examined their immunogenicity in OVA and neoantigen-derived SLP vaccines. We compared this adjuvant combination to K3 CpG-ODN, c-di-AMP, poly(I:C), and squalene alone and examined their capacity to induce T cell immunity against neopeptides and low-affinity peptides. Finally, we tested their ability to provide tumor protection and synergize with ICB in a mouse tumor model.

Male and female C57BL/6 mice (8–12 wk old) were used for bone marrow isolation. C57BL/6 female mice were used for immunizations (8–10 wk old) and to test the vaccine effect with or without anti-PD-1 treatment in tumor-bearing C57BL/6 female mice (6–8 wk old). Animal studies were carried out in accordance with the recommendations of the local authorities (Instantie voor Dierenwelzijn) that approved all protocols (license AVD1010020209604). Specific pathogen-free mice were purchased from Charles River and housed in the Erasmus Medical Center animal facility (Erasmus Dierenexperimenteel Center) in groups of two to four mice and kept in type IV cages. Food and water were administered ad libitum.

Melanoma cell lines B16-F10-OVA and B16-F10 (kindly provided by Dr. M. Wolkers, Sanquin, Amsterdam, the Netherlands) were cultured in RPMI media supplemented with 10% FBS, 1% l-glutamine, 50 µM 2-ME, and 100 U/ml penicillin/streptomycin. Mesothelioma cell line AE17 (kind gift from Dr. D. Nelson, Curtin University, Perth, Australia) were cultured in RPMI media supplemented with 10% FBS, 1% l-glutamine, 50 µM 2-ME, 48 µg/ml gentamicin, and 60 µg/ml benzylpenicillin. All cell lines were cultured in a 5% CO₂ incubator at 37°C.

Mouse BMDCs were generated in vitro from mouse bone marrow. Briefly, femurs and tibias from C57BL/6 mice were collected, cleaned with 70% ethanol, and kept in RPMI. Bones were crushed or flushed with a syringe, and bone marrow cells were collected and filtered using a 70-µm cell strainer. Cells were washed in RPMI media followed by centrifugation at 500 × g for 5 min at room temperature (RT) and resuspended in RPMI media supplemented with 10% FBS, 100 U/ml penicillin/streptomycin, 50 µM 2-ME (complete media), and 20 ng/ml GM-CSF and cultured in non–cell culture-coated 10-cm petri dishes. Culture media were refreshed after 4 d. On day 6, cells were harvested with Ca²⁺/Mg²⁺ free HBSS-EDTA, washed with complete media, and used for subsequent experiments.

The following OVA peptides were used for immunizations and in vitro restimulations: 20-mer OVA_(252-271): LEQLESIINFEKLTEWTSSN, 20-mer OVA E1_(252-271): LEQLEEIINFEKLTEWTSSN, 20-mer OVA R4_(252-271): LEQLESIIRFEKLTEWTSSN, 8-mer OVA E1_(257-264): EIINFEKL, 8-mer OVA R4_(257-264): SIIRFEKL (all from GenScript), and 8-mer OVA_(257-264) peptide SIINFEKL (Anaspec). TLR9 agonist K-type of CpG-ODN (K3 CpG) and STING agonist c-di-AMP adjuvants were used in vitro at 1 µM and in vivo at 10 µg per injection, respectively, in sterile saline. K3 CpG was synthesized by GeneDesign (Japan), and c-di-AMP was kindly provided by Yamasa (Japan). TLR3 agonist poly(I:C) HMW VacciGrade (InvivoGen) and TLR4 agonist LPS (Escherichia coli O111:B4) (Sigma-Aldrich) were commercially purchased. AddaVax (InvivoGen), a squalene-based adjuvant, was used by mixing it with an equal volume of sterile saline.

Neoantigens were identified from B16-F10 and AE17 based on the C57BL/6 spleen and tumor whole-exome DNA-sequencing results combined with RNA-sequencing expression data using personalized variant Ags by cancer sequencing (pVAC-seq) without applying the epitope predictor (32). The output list from pVAC-seq was manually checked, and the neoantigens were selected on the basis of high expression levels and having missense, frameshifts, or inframe indels in the protein sequence. Synthesized 20-mer neopeptides were screened for in vivo immunogenicity using C57BL/6 mice. Selected neopeptides exhibited no cross-reactivity to wild-type peptides. The selected neopeptides were used for immunizations and in vitro restimulations (Supplemental Table II).

The immunostimulatory capacity of adjuvants was tested in BMDCs that were plated at 1 × 10⁵ cells/ml/well in 12-well plates and incubated overnight (ON) at 37°C and 5% CO₂. The following day, cells were stimulated ON with 1 µM K3 CpG and/or c-di-AMP, and, after harvesting, cells were stained by flow cytometry for costimulatory and MHC molecules.

To test the immunogenicity of the different vaccine formulations, mice were s.c. immunized with 100 µl of each formulation in sterile saline in the left flank three times, once every week. One week after the last immunization, mice were euthanized, and blood, spleen, and left axillary and inguinal lymph nodes were collected. For OVA peptide immunizations, the formulations contained 10 µg of 20-mer OVA_(252-271) peptide and 10 µg of each of the corresponding adjuvants. To examine the safety profile of the vaccine formulation, blood was collected 6 h and 24 h after each vaccination, and the body weight was monitored for 24 d. As a positive control, mice were i.p. immunized with 10 µg of 20-mer OVA_(252-271) peptide and 200 µg of LPS. For tumor neoantigen vaccinations, a pool of six or eleven 20-mer neopeptides were included per vaccine at 7 µg each peptide and 10 µg per adjuvant. Tumor neopeptides for B16-F10 and AE17 tumors were determined by an in-house proprietary bioinformatic neopeptide prediction pipeline using whole-exome sequencing and RNA-sequencing data of the above tumors and C57BL/6 spleens.

For the in vivo tumor model experiments, 6–8-wk-old mice were injected s.c. with 100 µl of tumor cell suspension mixed 1:1 with Matrigel (Corning Life Sciences) containing 0.5 × 10⁵ cells of B16-F10-OVA cell lines. Vaccinations started 7 d after tumor injection, when the average of tumor size is 54.3 ± 5.6 mm³ (mean ± SE), and were delivered as above in three doses once per week using 10 µg of 20-mer OVA peptide and 10 µg of individual adjuvants per vaccine. In some experiments, animals were vaccinated without peptide. When α-PD-1 ICB therapy was included, vaccinations were delayed and started 12 d after tumor injection, when the average tumor size was 112.9 ± 10.2 mm³ (mean ± SE), to minimize the protective effect of ICB. Anti-PD-1 Abs (clone RMP1-14, catalog no. 114119; BioLegend) or isotype control Abs (clone RTK2758, catalog no. 400565; BioLegend) were injected i.p. at 100 μg/mouse in a final volume of 200 µl sterile saline twice per week. Mice were monitored for tumor growth and survival. To assess tumor growth, tumor size was measured every other day using a digital Vernier caliper. Tumor volume was calculated with the formula V = L × W × H, where V is tumor volume, L is tumor length (longer diameter), W is tumor width (shorter diameter), and H is the tumor height. Mice were euthanized, and their organs were harvested when the tumor volume reached the humane endpoint of 1500 mm³ volume.

For single-cell suspensions of splenocytes and lymph nodes, tissues were mechanically disrupted and filtered through a 40-µm cell strainer (Falcon, San Jose, CA). Cells were washed in RPMI and counted, and 2 × 10⁶ cells were used for staining. For blood staining, 60 µl blood was used. After lysing erythrocytes, cells were washed and stained. In all stains, cells were first pretreated with anti-CD16/CD32 blocking Abs for 10 min. For surface stains, cells were stained for 20 min on ice with different panels of Abs (Supplemental Table I). In some stains, PE-labeled tetramers of H-2kb MHC class I (MHC-I) loaded with OVA_(257-264) (in-house preparation) were used. To exclude dead cells, cells were stained with PerCP-Cy 5.5–, BV421-, or V500-labeled annexin V (BD Biosciences, San Jose, CA). After staining, cells were washed with HBSS containing 3% FBS and 0.02% sodium azide and fixed with 1% PFA. When annexin V was used, all buffers contained 2.5 mM CaCl₂. For intracellular cytokine staining, cells were stimulated with 10 µg/ml OVA_(257-264) for 6 h at 37°C in 5% CO₂ in the presence of GolgiPlug (BD Biosciences) or stimulated with 2 µg/ml of B16-F10 20-mer neopeptide pool ON at 37°C in 5% CO₂ and GolgiPlug for the last 6 h. Cells were surface stained as above, fixed for 1 h with IC Fixation Buffer at 4°C, washed with Perm/Wash buffer (eBioscience), and stained for intracellular cytokines for 45 min at 4°C using fluorochrome-conjugated Abs (Supplemental Table I). Finally, cells were washed twice with Perm/Wash buffer and fixed with 1% PFA. All samples were acquired in a Fortessa flow cytometer (BD Biosciences) and analyzed with FlowJo version 9.9.6 software.

Plates precoated with anti-murine IFN-γ Ab or anti-murine IL-5 Ab were used (ImmunoSpot – CTL). The protocol followed was as the manufacturer recommended. Briefly, 1 × 10⁵ cells for IFN-γ plates and 4 × 10⁵ cells for IL-5 plates were seeded in 100 µl per well, and 100 µl of media, 8-mer OVA peptide at 10 µg/ml, or neopeptide pools at 1 µg/ml was used to restimulate cells. As a positive control, PMA and ionomycin were used at 20 ng/ml and 500 ng/ml final concentrations, respectively. Cells were incubated ON at 37°C in 7% CO₂. The next day, the wells were washed twice with PBS and twice with PBS Tween 0.05%. Anti-murine detection Ab was added and incubated at RT for 2 h. After washing the plates three times, with PBS Tween 0.05%, streptavidin-AP solution was added and incubated at RT for 30 min. The plates were washed two more times with PBS Tween 0.05% and two times with deionized H₂O, and the developer solution was added and incubated at RT for 15 min. The reaction was stopped by washing the plates with water; afterward, they were allowed to air dry for at least 24 h before reading. The plates were read and analyzed in a CTL counter with ImmunoSpot Software.

ELLA Simple Plex Cartridge kits for IFN-γ, IL-1β/IL-1F2, IL-6, and TNF-α were used to measure cytokines (Bio-Techne). The protocol followed was as the manufacturer recommended. Briefly, the plasma was diluted four times with sample diluent and applied to wells. The cartridges were immediately measured by ELLA Simple Plex system.

Data, except the survival curve, the tumor growth curve, and the body weight curve, were first evaluated for normal distribution by Shapiro–Wilk normality test. Groups in the tumor growth curve and the body weight curve were compared by two-way ANOVA followed by Dunnett or Tukey posttests, respectively. The survival curves were compered by log-rank (Mantel–Cox) test. In other group comparisons, one-way ANOVA followed by Tukey tests or unpaired t tests were used if the data were normally distributed or Friedman or Kruskal–Wallis followed by Dunn test or Mann–Whitney U test if not. A p value ≤0.05 was considered significant. Statistical analysis was performed using GraphPad Prism version 9 software. The figure legends indicate the test used.

We first investigated the in vivo ability of different adjuvants to stimulate T cell immunity against 20-mer OVA_(252-271). We used adjuvants K3 CpG, c-di-AMP, and their combination. As a control, we used AddaVax, a squalene-based mimetic of vaccine adjuvant MF59 that is used in approved influenza virus vaccines. In addition, both AddaVax and MF59 are known to induce robust CD8⁺ T cell responses against whole proteins (33). We immunized mice three times with different formulations containing 20-mer peptide and adjuvants. One week after the last immunization, the frequency of OVA_(257-264)-specific CD8⁺ T cells was analyzed in the blood, spleen, and lymph nodes using peptide-loaded MHC-I tetramers. Adjuvants K3 CpG, c-di-AMP, and their combination with 20-mer peptide were significantly better at inducing OVA_(257-264)-CD8⁺ T cells than 20-mer peptide alone (Fig. 1A). In contrast, peptide vaccination with AddaVax failed to induce significant responses (Fig. 1A). Furthermore, we measured OVA_(257-264)-specific CD8⁺ T cells producing IFN-γ and TNF-α after OVA_(257-264) SIINFEKL peptide restimulation of splenocytes and found, in concordance with tetramer stains, a high frequency of cytokine-producing CD8⁺ T cells when animals were vaccinated with the combination of K3/c-di-AMP (Fig. 1B). Animals immunized with K3 CpG or c-di-AMP alone tended to have lower responses than their combination. Consistent with this, the adjuvants showed high immunostimulatory capacity on mouse primary BMDCs in vitro. We found that K3 CpG alone did not induce any of these markers, whereas c-di-AMP increased only CD86 and MHC-I above background (Supplemental Fig. 1). In contrast, combining K3 CpG and c-di-AMP together significantly increased the expression of all these molecules on the surface of BMDCs (Supplemental Fig. 1). These results show that the combination of K3/c-di-AMP is a promising vaccine formulation that elicits potent CD8⁺ T cell immunity against 20-mer SLPs.

Because many neopeptides are low-affinity peptides (22), we examined whether the combination of K3/c-di-AMP enhances the immunogenicity of low-affinity peptides. To test this, we used 20-mer peptides of OVA E1 or R4 peptides, two variants of SIINFEKL peptide known for their lower affinity to MHC-I (34, 35). We immunized mice with the same schedule as before, using 20-mer OVA E1 or R4 peptides with or without K3/c-di-AMP combination or AddaVax. Ag-specific CD8⁺ T cell responses against 8-mer OVA E1 or R4 peptides were determined by measuring intracellular cytokine production upon in vitro restimulation with these peptides. Vaccination with 20-mer OVA E1 or R4 peptides when combined with K3/c-di-AMP adjuvant but not the peptide alone or with AddaVax induced a potent Ag-specific CD8⁺ T cell response against 8-mer OVA E1 or R4 peptides (Fig. 2). The above findings support the combination of K3 CpG and c-di-AMP as being able to induce potent T cell responses against both high- and low-affinity 20-mer SLP.

We next investigated the immunogenicity of the K3 CpG plus c-di-AMP vaccine formulations with neopeptides obtained from the two different tumor models. We first compared the K3/c-di-AMP combination with c-di-AMP alone and AddaVax. We vaccinated mice with neopeptides harboring missense or frameshift mutations from each of melanoma B16-F10 and mesothelioma AE17 tumors. Mice were immunized as above with the 20-mer neopeptide pools consisting of 11 or 6 peptides from each tumor combined with adjuvants (Supplemental Table II). T cell immunity was assessed 1 wk after the last vaccination in spleens. Both Th1 and Th2 immunity was evaluated by performing ELISPOT assays for IFN-γ– and IL-5–producing cells after in vitro restimulation with the corresponding pool of peptides used in the immunization. We found that the adjuvant combination K3/c-di-AMP resulted in significantly more IFN-γ–producing neopeptide-specific T cells than the response induced by AddaVax (Fig. 3, left panels). Although c-di-AMP alone as an adjuvant induced levels of IFN-γ–producing T cells similar to those of K3/c-di-AMP, c-di-AMP alone induced much higher levels of IL-5–producing neopeptide-specific T cells (Fig. 3, right panels). This Th2 skewing induced by c-di-AMP alone was also evident with nonspecific PMA/ionomycin stimulation of T cells and suggested that c-di-AMP alone shifts the overall balance of the immune system in favor of Th2 immunity in a non–Ag-specific and systemic manner (Supplemental Fig. 2). These PMA/ionomycin Th2 responses were 10 times higher than those measured by neopeptide restimulation and suggest a more general Th2 skewing (Fig. 3 and Supplemental Fig. 2). Importantly, K3/c-di-AMP induced significantly lower Th2 skewing than c-di-AMP alone in this setting.

We next compared the K3/c-di-AMP adjuvant combination with poly(I:C) because its attenuated derivative poly-ICLC is currently a favorite in cancer neoantigen vaccines. We immunized mice with the 20-mer neopeptide pool from the B16-F10 tumor three times, and, 1 wk after the last boost, we analyzed T cell responses. In terms of IFN-γ–producing T cells, K3/c-di-AMP adjuvant combination induced ∼10-fold higher T cell immunity than poly(I:C) (Fig. 4A). K3/c-di-AMP adjuvant combination and poly(I:C) did not differ in IL-5–producing T cells (Fig. 4A). This increased Th1-driven immunity by K3/c-di-AMP was observed for both CD4⁺ and CD8⁺ T cells (Fig. 4B). Overall, these results demonstrate that the K3/c-di-AMP adjuvant combination preferentially induced potent neopeptide-specific Th1-type immune response while eliciting low Th2 immunity.

Once we confirmed the immunogenicity of the formulations containing 20-mer peptides combined with our adjuvants, we examined the effect of those formulations in controlling tumor growth. To this end, mice were first inoculated with B16-F10-OVA tumor cells, and, after 7 d, the mice were immunized with 20-mer OVA_(252-271) peptide and K3 CpG and/or c-di-AMP three times as indicated in Fig. 5A. We found that 20-mer OVA_(252-271) peptide combined with K3/c-di-AMP was the most effective formulation at reducing tumor growth and increasing survival of B16-F10-OVA tumor-bearing mice (Fig. 5B, 5C). When c-di-AMP alone and AddaVax were used as adjuvants of the 20-mer OVA_(252-271) peptide, we did not observe control of tumor growth (Fig. 5B, 5C). The protection conferred by K3/c-di-AMP was accompanied by higher frequencies of OVA_(257-264)-specific CD8⁺ T cells in the blood, spleen, and lymph nodes as determined by flow cytometry and peptide-loaded MHC-I tetramers (Fig. 5D). Finally, we excluded that the protective effect of vaccinating with the K3/c-di-AMP combination was due to a non–Ag-specific effect of the adjuvants because we found no such protection when we vaccinated animals with the adjuvants but without OVA_(252-271) peptide, thus confirming that the tumor control was not due to non–Ag-specific effects of the adjuvants (Fig. 6).

Once we confirmed that the tested peptide vaccine formulations were able to control tumor growth in immunized mice, we tested their ability to enhance α-PD-1 Ab treatment. In order to do this, we injected the partially refractory to anti–PD-1 therapy B16-F10-OVA tumor cells (36) and waited for 12 d to establish large enough tumors that do not respond to α-PD-1 Ab therapy. Starting at 12 d, we next vaccinated animals three times as above with 20-mer OVA_(252-271) peptide and K3/c-di-AMP combination as adjuvants. In addition, animals received either α-PD-1 Ab or isotype control Ab twice weekly as indicated in Fig. 7A. Treating mice with α-PD-1 Ab alone had no effect on tumor growth and survival of mice (Fig. 7B, 7C). When 20-mer peptide plus AddaVax vaccination was combined with α-PD-1 Ab treatment, we observed some protection in the tumor model (Fig. 7B, 7C), indicating that vaccination can improve the response to α-PD-1 Ab. However, the most potent antitumor effect, in terms of both tumor regression and survival, was seen when mice were vaccinated with 20-mer peptide combined with K3/c-di-AMP and treated with α-PD-1 Ab (Fig. 7B, 7C). Interestingly, even in this 12-d delayed vaccination scheme, vaccinating mice with 20-mer peptide combined with K3/c-di-AMP alone, without α-PD-1 Ab treatment, still conferred protection, albeit not as well as when α-PD-1 Ab was coadministered (Fig. 7B, 7C). The above information suggests that 20-mer peptide vaccination with K3/c-di-AMP combination adjuvant synergizes with ICB to provide better antitumor protection.

Finally, we examined the safety profile of our vaccine formulation containing K3/c-di-AMP. To this end, we immunized mice with 20-mer OVA_(252-271) peptide plus K3/c-di-AMP three times once per week and collected blood 6 h and 24 h after each vaccination. We analyzed inflammatory cytokines in plasma and used LPS as a positive control. Compared with LPS, K3/c-di-AMP induced much lower levels of IL-1β and IL-6. Some TNF-α and IFN-γ was detected in the K3/c-di-AMP group (Fig. 8A). Consistently, the K3/c-di-AMP group did not induce excess CD69⁺ activated T cells and MHC-II⁺ macrophages in blood and body weight loss upon vaccinations (Fig. 8B, 8C). These results indicate that the vaccine formulation containing K3/c-di-AMP induces only low levels of systemic inflammation.

We tested the well-established adjuvants AddaVax, a squalene analog of MF59, and the TLR3 ligand poly(I:C) together with K3 CpG ODN, a TLR9 agonist, and c-di-AMP, a STING agonist, and examined their effect on SLP immunogenicity in vitro and in vivo. We found that the combination of K3/c-di-AMP together with SLP derived from model Ag OVA_(257-264) peptide (SIINFEKL), low-affinity peptides OVA E1_(257-264) and OVA R4_(257-264) (EIINFEKL and SIIRFEKL, respectively), and neopeptides derived from two different tumor models stimulated potent T cell immunity in vivo. OVA SLP plus K3/c-di-AMP vaccination induced OVA-specific CD8⁺ T cells, which, on day 7 after the third boost in spleens, were 91.9 ± 2.5% (mean ± SE) CD44⁺ CD62L⁻ effector (memory) cells (data not shown). These cells were low for PD-1, TIGIT, and Lag-3, but 62.3 ± 1.4% (mean ± SE) of these OVA-specific CD8⁺ T cells expressed TIM3 (data not shown). Vaccination with OVA SLP protected mice from B16-F10-OVA tumors. Despite inducing neopeptide-specific T cell responses, however, s.c. neopeptide vaccination failed to protect from tumor growth of B16-F10 melanoma and AE17 mesothelioma tumors (data not shown). This might be because vaccine-induced T cells are not recruited into tumors due to the cold tumor microenvironment or because T cells with low-affinity TCR against neopeptides are not triggered efficiently in the suppressive microenvironment of these tumors. Importantly, however, this combination induced potent Th1 and CD8⁺ T cell immunity in vivo, but only low Th2 T cell responses against neopeptides, something critical because single-nucleotide variant–derived neopeptides tend to be low-affinity peptides, and it is well established that low-affinity peptides predispose to induction of Th2-type immune responses rather than Th1-type immune responses (37, 38). K3 CpG alone proved a weaker Th1 adjuvant than the K3/c-di-AMP combination. In contrast, STING agonist c-di-AMP, when used alone, consistently showed increased Th2 immunity against neopeptides, and this is in agreement with previous reports showing Th2 immunity elicited when STING agonists were used as adjuvants (29). Even more concerning was that STING agonist alone induced a non–Ag-specific Th2-prone immune state that could affect more general immunity. Our studies therefore raise the concern that some adjuvants may promote Th2 responses to neoantigens that could be deleterious in some cancers by promoting fibrosis and alternatively activated macrophages (M2 macrophages) (39). Therefore, Th2 immunity should always be evaluated in vaccine trials, something that to date most trials have not performed. This is also important for the use of adjuvants alone as cancer immunotherapeutics. At least one completed clinical trial tested the efficacy of STING agonist treatments of patients with cancer showing no clinical benefit, despite showing signs of systemic immune activation (40). Evaluation of the Th1–Th2 balance in these patients could provide insight into the reasons for trial failure.

In our study, we used 20-mer synthetic long neopeptides because SLPs have been known to induce stronger CD8⁺ T cell responses in vivo than short peptides. Short peptides can induce tolerance, whereas SLPs do not, owing to their efficient and prolonged cross-presentation on MHC-I by professional APCs and their capacity to induce CD4⁺ T cells that help CD8⁺ T cell induction (19, 24, 41, 42). Although SLPs as a vaccine platform for personalized neoantigen vaccines have advantages such as ease of production and cost-effectiveness (19), a critical component of such peptide-based vaccines is the selection of adjuvant that is necessary for potent immunogenicity, because peptides alone are generally of low affinity and their immunogenicity relies on adjuvants (22, 23). Currently, more than 50% of all clinical trials in ClinicalTrials.gov test neoantigen vaccines that are peptide-based vaccines. Of these peptide vaccines, >60% use adjuvant poly-ICLC, an attenuated for human use analog of TLR3 ligand poly(I:C). Results from such clinical trials have shown that poly-ICLC induces low but detectable ex vivo T cell immunity against neopeptides (18, 20, 21, 43). Unfortunately, these clinical trials have not demonstrated clinical efficacy, suggesting that more potent adjuvant formulations need to be developed. One such adjuvant formulation could be the K3/c-di-AMP combination that we found to be much more immunostimulatory than poly(I:C) and elicited ∼10-fold higher neopeptide-specific T cell immunity. This increased immunostimulatory activity was not accompanied by adverse effects, and only rarely did we observe transient skin irritation at the injection site, which was always less than 1 mm in diameter and transient. We also found that K3/c-di-AMP induced only low levels of transient inflammation in vivo, suggesting that this adjuvant combination is tolerable.

One reason that ICB is less effective in some patients and cancer types is the lack of preexisting tumor-infiltrating lymphocytes, which presumably are tumor specific (10). Therefore, personalized cancer vaccines that elicit neoantigen-specific antitumor T cell immunity can be one way to augment and improve the efficacy of ICB. This has yet to be proved clinically, however. One phase I clinical trial did use a vaccine of neopeptides formulated with poly-ICLC and combined it with anti–PD-1 Ab treatment, but without an arm with anti–PD-1 Ab treatment alone, it is impossible to confirm added clinical benefit of vaccination (18). In our present study, we clearly show that peptide vaccine formulated with the K3/c-di-AMP adjuvant combination conferred antitumor protection to mice, and this protection was further increased when vaccine was combined with anti–PD-1 Ab treatment in the ICB-resistant B16-F10-OVA tumor. Therefore, our vaccination strategy can convert a non–ICB-responsive tumor to being ICB responsive. It is possible that blocking other inhibitory receptors such as TIM3 may further improve vaccine efficacy. Importantly, we show that vaccine protection when using the K3/c-di-AMP adjuvant combination relied on the presence of Ag and was not due to a systemic nonspecific immunostimulatory activity of adjuvants alone.

Strikingly, the difference between K3/c-di-AMP combination and c-di-AMP alone in in vivo protection and synergy with ICB in the tumor models was very pronounced and did not reflect a major difference in Th1 immunity. A number of reasons may explain why the K3/c-di-AMP combination is a more potent adjuvant than c-di-AMP alone. First, the systemic Th2 priming effect that c-di-AMP induced could impair immunity against tumor. Although K3 CpG is a Th1 inducer in general, it is a weak inducer of IFN-γ but a potent IL-6 inducer (44). As we show, adding a STING agonist to K3 CpG strongly boosts Th1 and CTL immunity. Conversely, adding K3 CpG to the STING agonist inhibited Th2 immunity development by c-di-AMP. STING agonists can be Th2 inducers, and this is IRF3 mediated (45). The combination of K3 CpG plus STING agonists acts synergistically to induce large quantities of IFN-α and IL-12 by dendritic cells (29). This high induction of IL-12 would hinder Th2 development and explain the lack of Th2 immunity with this combination. More recently, it has been shown that STING agonists can induce IL-35 by B cells in an IRF-3–dependent manner. This IL-35 in turn can promote tumor growth in mice because it hinders immunity (46). It would be interesting to investigate whether adding K3 CpG to STING agonists can prevent IL-35 induction and offer an additional explanation for the improved antitumor efficacy of the K3/c-di-AMP combination.

In summary, we demonstrate that the K3/c-di-AMP adjuvant combination is able to induce potent Ag-specific Th1-type and CD8⁺ T cell immune responses against SLP and neopeptide pools derived from melanoma and mesothelioma tumors. This vaccine formulation induced low Th2 responses. Importantly, we show potent antitumor activity in vivo when SLPs are combined with the K3/c-di-AMP adjuvant combination, and this antitumor activity is synergistically increased by ICB treatment in the ICB-resistant melanoma model B16-F10-OVA. Thus, our peptide vaccine formulation confers ICB responsiveness to an unresponsive tumor. Taken together, our findings indicate that our vaccine formulation containing SLPs and K3/c-di-AMP adjuvant combination is a promising candidate for the development of efficient vaccine platforms that can be used in personalized cancer vaccines that potentiate the effect of ICB therapy in patients with cancer.

The authors have no financial conflicts of interest.

We thank Kevin de Vos, Marlous Wildemans, Rik Ruijten, Tessa Alofs, Tamara van Wees, and Danique Laport for assistance.