Introduction

The potential of therapeutic cancer vaccines has motivated extensive research and development efforts. Although the majority of this effort has led to disappointing results, much has been learned in the last years that will enable the generation of successful cancer vaccines. First, several advances have enabled better cancer antigen selection. For example, while B cells and innate immunity responses contribute to the efficacy of immune-mediated cancer therapies, research has shown that cytotoxic T cells are the dominant effector and thus T cell antigens should be the basis of therapeutic cancer vaccines. Improved techniques to profile protein abundance in normal and cancer tissue as well as to empirically assess the MHC presentation and T cell reactivity of various epitopes have been employed to identify suitable target antigens. Observations of antigen down-regulation by cancer cells have also prompted the inclusion of multiple vaccine antigens to prevent escape from immunosurveillance. Second, highly potent vaccine vectors appear to be required to stimulate immunity strong enough to yield meaningful clinical efficacy. Many cancer antigens are native self-proteins for which high affinity T cells have been eliminated by central and peripheral tolerance mechanisms. Even in the case of neoantigens, which result from cancer-specific mutations, high affinity T cells may be lacking because either the neoantigens still closely resemble the wild type antigen or because the T cells have been depleted or exhausted by cancer cell defense mechanisms. Several different therapeutic cancer vaccine platforms have been developed and tested in recent years. These have incorporated antigens delivered directly as peptides, encoded by nucleic acids, or by viral, bacterial, and cellular vectors [1]. In addition, mixtures of different types of antigens have been used for heterologous prime-boost strategies, and combinations with potent adjuvants have been tested to increase immune responses [1]. Which platforms will sufficiently activate and expand T cells to effect cancer regression is an active area of research and will probably vary depending on antigen selection and tumor type. Third, progress in understanding the interaction of cancer cells with the immune system and the development of immune-mediated cancer therapies has illuminated new paths to therapeutic cancer vaccine development. Cancer cells exploit natural immune checkpoint mechanisms that normally function to prevent hyperactivation and maintain peripheral tolerance by limiting T cell activity [2]. Antibodies that block the PD-1 and CTLA-4 immune checkpoints have demonstrated efficacy in some patients by unleashing suppressed anti-cancer T cells [2, 3]. Combinations with such immune checkpoint inhibitors (ICIs) may also augment the activation of T cells by cancer vaccines. Multiple preclinical studies have shown synergy between therapeutic vaccination and ICIs, and numerous clinical studies are now evaluating these combinations [4, 5].

We have applied the lessons of previous cancer vaccine studies and newly acquired knowledge of immunosuppressive mechanisms used by cancer cells to develop a potent cancer vaccine plus ICI combination regimen we call VBIR, for vaccine-based immunotherapy regimen. VBIR employs T cell-reactive tumor antigens delivered via a priming vaccination with a chimpanzee adenovirus and booster vaccinations with electroporated DNA. The vaccine is also combined with tremelimumab, an anti-CTLA-4 antagonist antibody, and sasanlimab, an anti-PD-1 antagonist antibody [6]. We describe here the development of VBIR, which involved testing and optimizing various vaccine platforms, ICI combinations, and dosing approaches to achieve maximal antigen-specific T cell stimulation and cancer killing activity. VBIR stimulates strong CD8 and CD4 T cell responses in mice and monkeys, and anti-tumor responses in murine cancer models. These results have prompted the initiation of clinical trials to evaluate VBIR in cancer patients.

Materials and methods

More information in Supplemental Materials and Methods.

Antibodies and vaccines

Anti-CTLA-4 mAb, tremelimumab, and anti-PD-1 mAb, sasanlimab, were produced in Pfizer labs. Anti-mouse CTLA-4 mAb 9D9, anti-mouse PD-1 mAb 29F.1A12, and matched isotype control mAbs were obtained from BioXCell. DNA plasmids and adenoviruses encoding the following antigens were constructed: human prostate specific membrane antigen (PSMA), rhesus macaque PSMA (rhPSMA), mouse Trp2, and rat Her2. Vectors encoding no antigens or encoding a luciferase-GFP fusion were constructed and used as controls and for expression profiling. Adenoviruses used were human serotype 5 (AdHu5) and chimpanzee serotype 68 (AdC68) and were administered by intramuscular (IM) injection. DNA vaccines were delivered intradermally using particle-mediated epidermal delivery (PMED, Pfizer) [7]or intramuscularly (IM) using the TriGrid electroporation (EP) device (Ichor Medical Systems) [8].

Mouse tumor models and non-human primates

All animal care, handling, and study-related procedures were reviewed and approved by the Pfizer IACUC and conducted in accordance with the USDA Animal Welfare Act, the PHS policy on Humane Care and Use of Laboratory Animals, and the Guide for the Care and Use of Laboratory Animals from the National Research Council. The TUBO tumor model, in which BALB/c mice transgenic for the transforming rat Her-2/neu oncogene are implanted with the murine TUBO cell line that expresses this oncogene, were generated and used as described previously [9]. The TUBO cell line was kindly provided by Dr. Guido Forni (University of Turin, Orbassano, Italy). The B16F10 tumor model was generated and used as described previously [10]. TUBO and B16F10 murine tumor models were used for both anti-tumor efficacy and immunogenicity studies with an adenovirus prime followed by bi-weekly DNA boosts, encoding self-antigens of rHer2 and TRP2, respectively, in combination of anti-CTLA4 and/or anti-PD1. Male Indian rhesus macaque or Chinese cynomolgus macaques monkeys were used for vaccine immunogenicity studies and therapeutic antibody pharmacokinetic studies. Vaccination involved various combinations of adenovirus and plasmid DNA encoding antigens. Tremelimumab and sasanlimab were co-administered via intravenous (IV) or (SC) injections. Doses and schedules are described in Figure Legends and Supplemental Materials and Methods.

Immunological assays

ELISpot was performed using 96-well human IFNγ ELISpot plates (MabTech). These were washed and blocked with AIM-V media prior to adding 100 μL AIM-V media only or media containing pools of peptides derived from either rhPSMA or hPSMA proteins. Peptides used are described in Supplemental Materials and Methods. 100 μL aliquots of peripheral blood mononuclear cell (PBMC) samples were then added to designated wells (4 × 105 cells/well) and plates were incubated at 37 °C and 5% CO2 overnight. Wells were then washed with water and PBS to remove cells and developed as per manufacturer’s directions using a biotinylated anti-human IFNγ detection antibody, streptavidin-HRP, and TMB substrate.

Intracellular cytokine staining was performed using freshly isolated PBMCs (1.5–2 × 106) in media AIM-V media containing 20 µg/mL DNase (Sigma), 1 µg/mL BFA (BD Biosciences), and a 1:20 dilution of anti-CD107α-PE antibody (H4A3, BD Biosciences) in U-bottom 96-well tissue culture plates. Pools of antigen-specific peptides (see Supplemental Materials and Methods, each peptide at 2 µg/mL) were added, plates were incubated at 37 °C and 5% CO2 overnight, and then wells were stained to detect intracellular IFNγ expression from CD8+ and CD4+ T cells. PBMCs cultured with medium plus DMSO served as negative control.

Tumor infiltrating lymphocytes (TILs) were assessed in tumors dissected from mice and dissociated mechanically using a GentleMACS Dissociator (Miltenyi Biotec) in media containing 1 mg/mL collagenase IV (Roche) and treated with 50 μg/mL DNase I (Roche) for 10 min at 37 °C. After centrifugation, single cells were counted using a ViCell cell counter (Beckman) and 1 million viable cells were stained with LIVE⁄DEAD® Fixable Aqua Dead Cell Stain Kit (Invitrogen), anti-CD45 (APC-Cy7, clone 30-F11), anti-CD3 (Pacific Blue, clone 17A2), anti-CD45R/B220 (PerCp-Cy5.5, clone RA3-6B2), anti-CD11b (PerCp-Cy5.5, clone M1/70), anti-MHC II (PerCp-Cy5.5, clone M5/114.15.2), anti-CD8 (FITC, clone 53–6.7), anti-PD-1 (PE-Cy7, clone RMP1-30), anti-PD-L1 (PE, clone MIH5). Samples were assessed using a Canto II flow cytometer (BD Biosciences) and data files were analyzed using FlowJo software (Tree Star, Inc.).

Pharmacokinetics of checkpoint inhibitor antibodies

Bilateral inguinal and axillary lymph nodes were removed from 10 monkeys at 8 h and 10 monkeys at 48 h after tremelimumab injection and snap frozen (− 80 °C). Individual lymph nodes were weighed and lysed with a polytron (model PT 10–35) in a buffer containing 70 mMNaCl, 1% Triton-X-100, 50 mM b-glycerophosphate, 10 mM HEPES, and a protease inhibitor cocktail (Roche, 058929700001, one tab/50 mL) for 90 s. Sample was kept on ice, and an aliquot centrifuged at 10,000×g for 15 min at 4 °C. Samples was transferred to two duplicate 96-well plates and immediately frozen (− 80 °C). A protein concentration was determined for each sample using a BCA kit (Pierce). Prepared lymph node samples and serum samples were assessed for the presence of tremelimumab by ELISA.

Neutralizing antibody levels to hAd5 and AdC68 in prostate cancer patients

Plasma samples from 66 prostate cancer patients residing in the United States were obtained from BioServe and Asterand Bioscience. Neutralizing antibodies levels to AdHu5 and AdC68 in each plasma sample were measured as described in Supplemental Materials and Methods. Briefly, plasma samples were incubated in 96-well tissue culture plates for 1 h at 37 °C along with 8 × 107 virus particles of either AdC68 or hAd5, each engineered to express firefly luciferase upon infection. A549 cells were added to the plasma-virus mixture at a multiplicity of infection of 8000 viral particles per cell, and after 24 h luciferase activity was measured by adding Bright-Glo Luciferase assay buffer to each well and reading plates on a Wallac 1420 fluorimeter.

Results

Selection of adenovirus as a cancer vaccine priming vector

Potent T cell priming and expansion are essential for efficacy of a therapeutic cancer vaccine [1]. This is especially true for vaccines targeting tumor-associated self-antigens, since the vaccine must break immune tolerance to activate suppressed or low affinity autoreactive T cells. We considered several vaccine vectors, including peptides, nucleic acids, and several viruses for potent and durable priming of T cells against a self-antigen. We identified DNA and adenoviruses as the most promising options. DNA vectors are a versatile vaccine platform that have proven effective in delivering bacterial, viral, and tumor-associated antigens in preclinical studies and in early clinical trials [11, 12]. We constructed a DNA vaccine plasmid expressing PSMA and then compared two delivery methods, particle-mediated epidermal delivery (PMED) and intramuscular electroporation, to activate T cells in rhesus macaque monkeys. We first used human PSMA (hPSMA) as the encoded antigen, which we assumed would be recognized by the rhesus immune system as a foreign antigen. Antigen-specific IFNγ T cell responses were measured by an ELISpot assay of PBMCs taken before and then 9–10 days and 15–17 days after a single priming vaccination. Antigen-specific T cells were undetectable prior to vaccination. Small but significant T cell responses to hPMSA above pre-vaccination levels were observed from several animals using both delivery methods on day 9–10 post-vaccination (Fig. 1A, B). After 15–17 days, higher responses (> 300 SFC/106 PBMCs) were observed only in PBMCs from animals that were vaccinated by electroporation. However, electroporation delivery of a DNA vaccine expressing the self-antigen, rhesus PSMA (rhPSMA), failed to break tolerance and activate T cells (Fig. 1C), emphasizing the higher barrier to immune responsiveness presented by a self-antigen.

Fig. 1
figure 1

Antigen-specific IFNγ T-cell responses induced by DNA plasmids expressing hPSMA or rhPSMA delivered by PMED or electroporation. Groups of rhesus monkeys (N = 6/group) were vaccinated with DNA plasmid expressing hPSMA by PMED (A) or electroporation (B) or with DNA plasmid expressing rhPSMA by electroporation (C). Animals in group A first received two SC immunizations with huPSMA long peptides in Montanide ISA-51 (days 1 and 22) prior to receiving a hPSMA PMED boost (day 57). Human PSMA-specific or rhPSMA-specific IFNγ-producing cellular responses were measured on freshly isolated PBMC samples before (pre), 9–10, and 15–17 days after vaccination using an IFNγ ELISpot assay. PSMA-derived peptide pools are described in Supplemental Materials and Methods. Responses for individual animals in each group are displayed as the sum of averaged replicate responses to each peptide pool that had been subtracted from the average background responses recorded for matched unstimulated samples

Full size image

We next tested a replication-defective human adenovirus vaccine delivered by a single intramuscular injection. Adenovirus vector vaccines have been shown to potently induce multi-functional immune responses to transgene-encoded antigens [13]. Human adenovirus serotype 5 (AdHu5) bearing a deletion of the E1 gene has been studied extensively in vaccine development, so we employed this vector to construct vaccines for immunogenicity testing in monkeys. In contrast to DNA electroporation, priming with an AdHu5 vaccine was able to elicit antigen-specific IFNγ-producing T cell responses to the self-antigen, rhPSMA, at a similar magnitude achieved using the foreign hPSMA antigen in most vaccinated monkeys (Fig. 2). Additionally, rhPSMA-specific T cell responses induced by AdHu5 priming were readily detectable at an earlier time point (D9 post-vaccination) than observed for hPSMA-specific T cell responses induced by DNA electroporation. We thus selected adenovirus as our preferred priming vector for use in a therapeutic cancer vaccine.

Fig. 2
figure 2

Antigen-specific IFNγ T cell responses induced by adenoviral vectors expressing hPSMA or rhPSMA. Indian Rhesus macaques (N = 6/group) were vaccinated with AdHu5 adenoviral vector expressing rhPSMA (A) or hPSMA (B). IFNγ T-cell responses specific to the respective vaccine antigens were measured on freshly isolated PBMC samples by an IFNγ ELISpot assay before vaccination and nine days after vaccination. PSMA-derived peptide pools are described in Supplemental Materials and Methods. Responses for individual animals in each group are displayed as the sum of averaged replicate responses to each peptide pool that had been subtracted from the average background responses recorded for matched unstimulated samples

Full size image

Local co-administration of anti-CTLA-4 antibody enhances the magnitude of T cell responses induced by an adenoviral vaccine

Cytotoxic T lymphocyte-associated antigen (CTLA-4) is a co-inhibitory molecule that represses T cell activation. CTLA-4 is upregulated on both CD4 and CD8 T cells subsequent to TCR-MHC engagement and competes with the co-stimulatory receptor CD28 for binding to CD80 and CD86 on antigen-presenting cells to attenuate T cell activation, thus serving as a checkpoint for T cell activation [2]. Antagonism of the CTLA-4 checkpoint enhances T cell activation, and anti-CTLA-4 antibodies are efficacious cancer therapies in some patients [14,15,16,17,18]. We tested whether an anti-CTLA-4 antibody, tremelimumab, would augment antigen-specific T cell activation when combined with AdHu5 vaccination in rhesus macaques. Tremelimumab, which binds with subnanomolar affinity to both human and monkey CTLA-4, is typically administered to patients by intravenous (IV) infusion, resulting in systemic exposure [19]. We compared IV infusion to subcutaneous (SC) injection in close proximity to the vaccine injection site, hypothesizing that the latter approach might increase exposure in the lymph nodes where vaccination was initiated as well as reduce toxicities associated with systemic tremelimumab treatment. Circulating levels of tremelimumab were lower upon SC injection as compared to IV infusion for more than two days after administration (Fig. 3A). However, higher concentrations of tremelimumab were detected in lymph nodes proximal to the AdHu5 vaccination site at both 8 and 48 h after local SC injections compared the tremelimumab concentrations in monkeys that received IV infusion (Fig. 3B). In contrast, SC administration resulted in a lower concentration of tremelimumab in lymph nodes distal to the vaccination site compared to that observed with an IV route. Importantly, concurrent SC administration of tremelimumab with the rhPSMA AdHu5 vaccine markedly enhanced the frequency and magnitude of antigen-specific IFNγ CD8+ and CD4+ T-cell responses (Fig. 3C), demonstrating optimal characteristics for a priming regimen. Tremelimumab use with concurrent administration of DNA plasmid vaccination did not increase antigen-specific CD8 T cell responses but instead only increased CD4 T cells in half of the animals (Fig. 3C). These results supported the rationale of local anti-CTLA4 monoclonal antibody delivery to amplify the adenovirus vaccine-induced T cell responses and achieve lower systemic exposure and possibly toxicity.

Fig. 3
figure 3

Local subcutaneous delivery of tremelimumab increases proximal exposure and activation of vaccine-specific T cells. A total dose of 33 mg tremelimumab was injected intravenously (IV) or subcutaneously (SC) bilaterally into the upper thighs of male rhesus monkeys that had also received bilateral quadriceps injections of 1 × 1011 virus particles of AdHu5 encoding luciferase and GFP reporter proteins. A Circulating levels of tremelimumab antibody in sera were collected at various time points post-administration and measured by ELISA. B Tremelimumab and total protein levels were measured in lysates prepared from inguinal (proximal to SC injections) and axillary lymph nodes excised at 8 and 48 h post-administration. C PSMA-specific, IFNγ producing CD4+ and CD8+ T cell responses were measured by ICS on freshly isolated PBMC collected day 15 after priming immunization. Immunization consisted of either 1e11 vp of AdHu5 or 5 mg of plasmid DNA (delivered by EP) encoding rhPSMA in the upper left quadriceps muscle with or without a concurrent and adjacent 10 mg/kg SC injection of tremelimumab. Symbols represent responses from individual animals and horizontal bar represent geometric mean averages for each group

Full size image

Use of a chimpanzee adenovirus vector, AdC68, circumvents pre-existing antivirus immunity

Because infection by naturally occurring human adenoviruses is common, most people carry antibodies to common serotypes, including AdHu5, that can quickly neutralize human adenovirus vaccines [20,21,22,23]. To circumvent this issue, we switched to the use of chimpanzee adenovirus serotype 68, AdC68, for which pre-existing immunity is low to non-existent in humans, including cancer patients [24,25,26,27,28]. Similar to the AdHu5 vector, the AdC68 virus was also rendered replication-defective by deletion of the E1 gene and has been shown to work well as a vaccine vector due to its high transduction efficiency, high expression of transgenes, and good genetic stability [24, 25, 29]. To further support use of AdC68 in humans, we extended the work of previous reports describing differential immunity to AdHu5 and AdC68 by screening plasma samples from 66 prostate cancer patients for neutralizing antibodies to each virus vector (Fig. 4). Nearly half of the plasma samples screened had detectable neutralizing antibodies to AdHu5, and more than one quarter displayed moderate to high levels of antibodies that would be expected to reduce AdHu5 infectivity following intramuscular administration. In contrast, only 18% of plasma samples carried AdC68 neutralizing antibodies, and in all cases, titer levels were lower than those that would be expected to blunt AdC68 infectivity. The non-human primate models used in other experiments described below have similarly low pre-existing antibodies to AdC68 (data not shown).

Fig. 4
figure 4

Seroprevalence of adenovirus neutralizing antibodies in prostate cancer patients. Plasma samples from 66 prostate cancer patients were screened for neutralizing antibodies to human adenovirus serotype 5 (AdHu5) and chimpanzee adenovirus serotype 68 (AdC68). Symbols represent the average titer measured for each plasma sample and horizontal bars represent the geometric mean average for the test group. Dashed lines indicate the upper (8000) and lower (33) titer limits of quantitation for the assay. Statistical comparison between test groups was performed using a paired T test

Full size image

Heterologous prime-boost vaccination enhances T cell activation

To extend the magnitude and duration of antigen-specific T cell activation, we further investigated the use of additional booster vaccinations. Previous studies have shown that repetitive administration of the same virus vector is ineffective due to induction of neutralizing anti-viral antibodies following the first vaccination [30, 31]. We thus explored the use of DNA encoding the same rhPSMA self-antigen, delivered by electroporation and in conjunction with local tremelimumab, as our booster vaccination. We evaluated several different dosing schedules using adenovirus and DNA vaccines to optimize the heterologous prime-boost regimen. rhPMSA-specific IFNγ CD4 and CD8 T cell titers induced by AdC68 vaccination plus tremelimumab contracted by seven weeks after treatment, whereas adding a booster vaccination with a DNA plasmid plus tremelimumab ten weeks after priming sustained high levels of activated T cell titers for at least 18 weeks (Fig. 5A). We also tested boosting at either four or eight weeks after priming. In animals that were boosted with the DNA vaccine at week eight, the T cell titers were restored by week 10 back to levels seen after priming (Fig. 5B). A similar increase in the magnitude and frequency of antigen-specific T cell responses was observed in animals boosted at week four, and a second boost at week eight further increased responses (Fig. 5B). This response was maintained after a third DNA vaccine plus tremelimumab boost (data not shown). Based on these collective data, we chose to move forward with a vaccine schedule in which one adenovirus prime and three DNA boosts were given at one-month intervals.

Fig. 5
figure 5

Heterologous adenovirus prime plus DNA boost increases magnitude and duration of vaccine responses in monkeys. A Rhesus macaques were primed with 1011 virus particles of adenovirus encoding rhPSMA given IM and 10 mg/kg of tremelimumab given ID. The animals were subsequently boosted on week 10 with 5 mg of DNA encoding rhPSMA by electroporation and 10 mg/kg of tremelimumab. rhPSMA-specific IFNγ+ CD8 T cell responses were measured on weeks 2, 7, 12, and 18 after adenovirus prime for each animal. B Animals were primed with bilateral IM injections of AdC68 encoding rhPSMA and boosted with DNA expressing rhPSMA at week 4 or week 8. Tremelimumab was administered with each vaccine. rhPSMA-specific IFNγ+ CD8 and CD4 T-cell responses were measured on weeks 2, 6 and 10 for each animal. C Animals were primed with an IM injection of 1011 VP of adenovirus encoding rhPSMA and 33 mg of tremelimumab given SC. They were then boosted at week 8 with 5 mg of DNA and 50 mg of tremelimumab SC and at week 16 with an IM injection of 2 × 1011 VP adenovirus encoding rhPSMA with 75 mg of tremelimumab. rhPSMA-specific IFNγ+ CD8 T cell were measured two weeks after the last vaccination for each animal. The geometric means of the responses and responder frequencies are indicated in the table. D Animals were treated with two cycles of a regimen consisting of adenovirus priming followed by three monthly DNA boosts, each expressing rhPSMA. Tremelimumab was co-administered SC with each vaccination. One additional DNA plus tremelimumab treatment was also given 6.5 months after the two-cycle regimen was completed. For each animal, the numbers of rhPSMA-specific polyfunctional (IFNγ+ and IL-2+ or TNFα+) CD8 T cells (top), cytolytic (CD107α+) IFNγ+ CD8 T cells (middle), and polyfunctional CD4 T cells (bottom) were measured during and after treatment. Lines indicate the geometric mean of the group. Response to DMSO was subtracted as background from all T cell measurements described above

Full size image

To ensure maintenance of strong T cell responses after a first cycle of one adenovirus prime and three DNA boost vaccinations, we tested the effectiveness of a second heterologous prime-boost cycle. Although the initial adenovirus vaccination activates anti-viral immunity, we found that this immunity waned after several months (data not shown) and that a second adenovirus vaccination given 16 weeks after the first dose resulted in a further stimulation of antigen-specific T cells compared to the levels induced using only a single adenovirus priming vaccination (Fig. 5C). In animals treated with two prime-boost cycles, we observed that titers of active (IFNγ-producing) rhPSMA-specific CD4 and CD8 T cells remained high (between 500 and 5000 SPC/106 PBMC using ELISpot) throughout the 32-week treatment and stayed elevated for several months after discontinuation of the regimen (Fig. 5D). An apparent decline detected starting at about five months after the last boost could be reversed by another DNA plus tremelimumab boost given at week 57 (26 weeks after completing the second cycle of the regimen), and high (> 500 SPC/106 PBMC) cell levels persisted for at least another four weeks. We have not measured activated T cell titers beyond this time, nor whether additional boost vaccinations would further extend the duration of the response. These results establish that this heterologous prime-boost vaccine regimen can break tolerance to a self-antigen and maintain antigen-specific T cell responses in primates for more than a year.

Addition of anti- PD-1 antibody to counteract immune checkpoint inhibition

The PD-1/PD-L1 checkpoint functions to suppress anti-tumor T cell responses, and several antibodies that antagonize PD-1 and PD-L1 have been approved and are efficacious in a variety of cancers [2]. We thus investigated whether addition of a PD-1 antagonist antibody would enhance the activity of our vaccination regimen. Sasanlimab is a monoclonal antibody that binds with high affinity to both human and monkey PD-1 and blocks interaction with its ligands, PD-L1 and PD-L2, resulting in T cell activation and proliferation [6]. In addition, sasanlimab has demonstrated activity using a monthly SC administration, which fits well with the delivery route and schedule established for our vaccine plus tremelimumab combination regimen. We first evaluated the potential contribution of sasanlimab on the level of antigen-specific T cell responses when administered either by SC injection, near the sites of adenovirus and tremelimumab injection, or by IV infusion in rhesus macaques. As in previous experiments, the addition of SC tremelimumab to our rhPSMA adenovirus vaccine increased CD8 T cell activation by about tenfold (Fig. 6A). In contrast, neither a 1 mg/kg dose of sasanlimab delivered SC nor a 5 mg/kg dose delivered IV augmented the T cell response induced by the vaccine alone. However, addition of 5 mg/kg SC sasanlimab to the vaccine increased T cell activation to a level similar to that induced by the vaccine plus tremelimumab. Moreover, the combination of 1 mg/kg sasanlimab SC with the vaccine and tremelimumab further stimulated T cell responses to levels about 30 times higher on average than that of vaccine alone. These results indicate that the PD-1 checkpoint is activated in monkeys by our vaccine, and that blocking this checkpoint enables induction of higher tumor antigen-directed immune responses.

Fig. 6
figure 6

Proximal subcutaneous administration of anti-PD-1 antibody further enhances T cell stimulation in combination with vaccine and tremelimumab in monkeys. Rhesus macaques were vaccinated with a total of 2e11 vp AdC68 encoding rhPSMA given IM with or without 33 mg tremelimumab given SC. Four groups of monkeys (N = 8/group) were also treated with the anti-PD-1 antibody, sasanlimab (PF-06801591), as indicated. SC antibody injections were given in the vicinity of the vaccination injections. rhPSMA-specific T cell responses two weeks after treatment were measured using frozen/thawed PBMC samples by intracellular cytokine staining. Symbols represent the responses of measured for individual animals while horizontal lines indicate the geometric mean for each group. Responses to DMSO only (no peptide) for each matched sample were subtracted as background

Full size image

VBIR has potent anti-tumor activity in a mouse model of cancer

In order to evaluate the ability of VBIR to inhibit tumor growth, we constructed vaccines expressing rodent versions of tumor-associated antigens and used these in combinations with anti-murine CTLA-4 and PD-1 antibodies to treat mouse tumor models. In the first experiment, we used a TRP2-encoding adenovirus (prime) and DNA plasmid (boost) to immunize C57BL/6 mice that had been implanted SC with B16F10 tumor cells, which express the syngeneic mouse TRP2 antigen [32]. Several days after tumor implantation, one prime and four boost vaccinations were given at biweekly intervals with or without the ICI antibodies. More tumor-bearing mice that were administered the TRP2 vaccine alone initially survived longer than those treated with negative control agents, but all mice in both groups ultimately died within 46 days tumor cell implantation (Fig. 7A). Addition of either anti-CTLA-4 or anti-PD1 antibody to the cancer vaccine increased survival, with the former providing enhanced benefit that closely matched the anti-tumor efficacy achieved with the anti-CTLA-4 plus anti-PD1 antibody combination without vaccine. The VBIR combination regimen, consisting of vaccine, anti-CTLA-4 and anti-PD1 antibody, more than doubled survival compared to the negative control treatment. We also assessed the numbers of TRP2-specific, activated T cells in these mice and found that the VBIR combination elicited the highest levels (Fig. 7B).

Fig. 7
figure 7

VBIR increases survival of tumor mouse models and activates self-antigen-specific CD8 T cells. Various combinations of vaccine (prime + boost), anti-CTLA-4 antibody, and anti-PD-1 antibody, as well as control vaccine and antibodies, were used to treat either the B16F10 mice three days after tumor implantation (A, B) or transgenic rHER2 mice 10 days after implantation with TUBO (rHER2) tumor cells (C, D). For the B16F10 study, three groups of mice were also evaluated for antigen-specific T cell activation one week after the last treatment (B). For the TUBO study, tumor infiltrating leukocytes were analyzed for proportion of CD3+ CD8+ cells, PD1+ cells among CD8+ cells, and the abundance of PD-1 in CD8+ cells (D)

Full size image

In a second experiment, we also tested VBIR in a mouse model in which tumorigenesis is driven by a syngeneic HER2 oncogene. This model involves SC implantation of TUBO cells, a mouse mammary carcinoma cell line expressing the rat HER2 gene, into rat HER2 transgenic mice [9]. TUBO tumor-implanted mice were vaccinated biweekly with a rat HER2-encoding adenovirus (prime) and DNA plasmid (boost) in various combinations with the anti-CTLA-4 and anti-PD-1 antibodies. As in the B16F10 model, the vaccine alone provided some survival benefit, and the addition of one or the other ICI antibodies extended this benefit, with the vaccine plus anti-CTLA-4 combination working better than the vaccine plus anti-PD-1 combination (Fig. 7C). The strongest efficacy was seen in mice given the full VBIR combination. We also assessed the numbers of tumor infiltrating lymphocytes (TILs) in these mice (Fig. 7D). The vaccine, with or without ICIs, induced a two-fold increase in TILs, and that these were activated T cells was confirmed by showing that they expressed high PD-1 levels.

Discussion

In developing VBIR we aimed to apply lessons from cancer vaccine and immuno-oncology research to construct a combination therapy with three key features: tumor antigens that could elicit specific T cell responses, a highly potent, “off-the-shelf” vaccine technology, and ICI antibodies to overcome both tolerance and tumor immunosuppressive mechanisms. Although other therapeutic cancer vaccines have demonstrated some clinical activity, this activity has been limited and some of these vaccines require cumbersome, personalized manufacturing for each patient. For example, sipuleucel-T (Provenge) affected a four-month increase in median overall survival compared to placebo in men with castrate-resistant prostate cancer in a pivotal phase III trial [33], but involves custom ex vivo engineering of each patient’s dendritic cells. Another prostate cancer vaccine in development, PROSTVAC-VF, uses a heterologous prime-boost strategy with two different poxviruses (vaccinia and fowlpox) engineered to encode immunostimulatory transgenes in addition to the PSA antigen [34]. This vaccine in combination with GM-CSF yielded a ten-month longer median overall survival compared to the empty vector control group (26.2 vs. 16.3 months, respectively) in a phase II trial [35], but induced antigen-specific T cell expansion to only about 0.03% of the total CD8 T cell population and was stopped in phase III due to futility [36, 37]. Thus, it is likely that this vaccine is not sufficiently potent and/or the vaccine alone is incapable of overcoming the tumor immunosuppressive microenvironment to achieve significant efficacy in patients. Consequently, PROSTVAC is now being tested in combination with ICIs in clinical trials [NCT02933255, NCT02506114].

Checkpoint inhibitor antibodies that antagonize PD-1, PD-L1, and CTLA-4 have shown activity in multiple tumor types and in patients whose tumors are unresponsive to other drugs [2, 3]. However, only a minority of patients respond to ICIs, and although combinations increase response rates, most patients still do not experience significant benefit [38]. A likely reason for this poor response is the absence of pre-existing tumor-specific T cells, which may fail to arise due to insufficient activation or a lack of appropriate target antigen expression by the tumor. In addition, recent research has shown that therapeutic PD-1 blockade in suboptimal T cell priming conditions, which are known to occur in the tumor microenvironment, results in generation of dysfunctional CD8 T cells and tumor resistance, but that this can be reversed by proper priming with a cancer vaccine [39]. Multiple preclinical studies have shown synergy between therapeutic vaccination and ICIs [40, 41], and several clinical studies are now evaluating this strategy. A recent phase II trial testing the addition of the PD-1 inhibitor, nivolumab, to an HPV E6 and E7 vaccine yielded overall survival nearly double that of nivolumab alone in incurable HPV-positive oropharyngeal cancer patients [42]. These results indicate that the combination of a potent cancer vaccine to stimulate robust T cell activation with ICIs to disable checkpoint-mediated immunosuppression would extend therapeutic benefit to more patients than either approach alone.

We developed VBIR, a vaccine-based immunotherapy regimen, for this purpose. We found that delivery of tumor antigens by first priming with a replication-defective adenovirus and then boosting with several rounds of DNA plasmid led to activation and expansion of T cells, and that this was significantly enhanced by concurrent administration of anti-CTLA-4 and anti-PD-1 antibodies. Vaccination with both self- and non-self-antigens yielded T cell responses, indicating that this regimen works even when self-reactive T cells would have been culled by immune tolerance mechanisms. The use of a chimpanzee rather than a human adenovirus for priming reduces the chance of rapid virus neutralization by pre-existing immunity. Additionally, we found that SC delivery of an anti-CTLA-4 antibody near the sites of vaccination improved its concentration in lymph nodes and T cell activation in monkeys, while reducing systemic exposure. CTLA-4 inhibition is known to promote priming of T cells in tumor-draining lymph nodes, but also has been found to elicit autoimmune toxicities when delivered IV in some patients [43]. Similarly, current PD-1/PD-L1 therapies are delivered IV and associated with adverse events in some patients [44], so we co-administered the PD-1 antibody by SC injection to reduce systemic exposure in monkeys as well. SC rather than IV delivery of the ICI antibodies should also improve the convenience of the VBIR regimen and potentially could reduce toxicities in patients.

A key challenge for developing a combination therapy like VBIR is determining the optimal dose levels and scheduling of its components. For the vaccine components, we found that our adenovirus vector led to the strongest T cell priming, and that we could sustain the high levels of activated, antigen-specific T cells with three monthly boosts of DNA plasmid encoding the same antigen. Simultaneous injection and in situ electroporation of the DNA yielded the best results, in accord with previous studies showing that this method dramatically amplifies vaccine responses compared DNA injection alone [45]. Because the ICI antibodies we tested have sufficiently long half-lives, we also delivered them concurrent with the monthly vaccinations, reasoning that this would enhance the T cell response and durability and optimize convenience to patients. To prolong the vaccine response, we also tested two cycles of vaccination, which in monkeys generated a robust T cell activation lasting more than a year. Several dose levels of each component were tested to find the optimal combination regimen (data not shown).

Anti-tumor activity of VBIRs expressing self-antigens was demonstrated in two syngeneic mouse models that respond poorly to individual ICIs. The B16F10 model is immunologically “cold,” with little immune cell infiltration into tumors, and does not respond to either anti-murine CTLA-4 or PD-L1 antibodies alone [46], although we saw anti-tumor activity when these two antibodies were combined. We also found that the HER2 transgenic/TUBO model is unresponsive to inhibition of CTLA-4 or PD-1 used either alone or in combination. However, the addition of either antibody to our prime-boost vaccine resulted in significant tumor growth inhibition and improved mouse survival, and the combination of both antibodies with the cancer vaccine (the VBIR combination) yielded the best responses. These responses also correlated with increased numbers of activated T cells in the tumors. We are currently testing VBIR in other murine models representing a broad range of tumor types and tumor immune infiltrate phenotypes.

These encouraging preclinical results with VBIR have motivated testing this regimen in the clinic. Two phase I trials are underway in different cancer patients: VBIR-1 combines our prime-boost vaccine expressing prostate specific antigen (PSA), prostate stem cell antigen (PSCA), and prostate specific membrane antigen (PSMA) with tremelimumab and sasanlimab, an anti-PD-1 antibody and is being studied in prostate cancer patients (NCT02616185); VBIR-2 combines a vaccine expressing antigens derived from telomerase reverse transcriptase (TERT), mucin-1 (MUC-1) and mesothelin (MSLN) with the same ICIs and is being tested in patients for activity against multiple solid tumor types (NCT03674827). For both VBIRs, we employed three antigens in order to reduce the risk of immune escape by tumor antigen down-regulation and then carefully designed and tested optimal target antigen constructs. These are among the first clinical trials to test a potent, multi-antigen, prime-boost vaccine with two complementary ICIs, a combination which may extend the benefit of immunotherapies to more patients. Results will be used to further advance our application of therapeutic cancer vaccine combination regimens.