Safety and immunogenicity of a tri-antigen vaccine targeting IGFBP-2, HER2, and IGF-IR in participants with non-metastatic breast cancer

Abstract

Background

Ductal carcinoma in situ (DCIS) is a preinvasive form of breast cancer. Current treatment consists of surgery, radiation, and often systemic therapy exposing patients to unnecessary health risks. Vaccines targeting DCIS may be a way to intercept preinvasive lesions and prevent the development of invasive breast cancer.

Methods

We developed a Th1 selective multiantigen, polyepitope plasmid-DNA vaccine encoding segments of IGFBP-2, HER2, and IGF-IR, all antigens expressed in hormone receptor positive and negative DCIS. We then performed a Phase I study in participants with non-metastatic breast cancer with no evidence of disease. The primary objective was to assess the safety of 3 monthly intradermal doses (150, 300, or 600 µg) of the tri-antigen vaccine with granulocyte macrophage colony-stimulating factor as an adjuvant. 32 participants were enrolled, 10 per dose level. Toxicity evaluations occurred monthly with vaccination and at 1 and 6 months after the last vaccine. Blood was collected at baseline and at 1 and 6 months after the last immunization to assess cellular immune responses. Participants were followed annually for 5 years for long-term toxicity.

Results

There was no significant difference in adverse events (AEs) across dose levels and all related AEs were grades 1 or 2. All doses were immunogenic; responders included 70% of participants at the 150 µg dose level, 67% at the 300 µg dose, and 40% at the 600 µg dose level. All participants at the 300 µg dose retained significant Th1-antigen-specific immune response at 6 months after end of immunizations. T-cells derived from vaccine immunologic responders exhibited gene expression profiles that indicated an increased metabolic fitness as compared with immunologic non-responders.

Conclusions

The tri-antigen vaccine appears safe and immunogenic. The intermediate dose (300 µg) was chosen as the Phase II dose due to the long-term persistence of immunity after vaccination. The vaccine will be studied in Phase II trials for the treatment of DCIS.

Trial registration number

NCT02780401.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Single antigen cancer vaccines, targeting HER2, have been used for the interception of ductal carcinoma in situ (DCIS) and have been found to be safe and immunogenic. Early stage data is promising, with some resolution of lesions, but there is a suggestion of the development of antigen loss with immunization, and the vaccines are limited to lesions expressing HER2.

WHAT THIS STUDY ADDS

• This study describes a novel multiantigen vaccine that would potentially have immunologic efficacy against most DCIS lesions. The vaccine was designed using methods that would ensure Th1 selectivity, which was documented during the course of the study. Finally, T-cell fitness is suggested as a critical biomarker of vaccine responsiveness.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The tri-antigen vaccine appears safe and immunogenic and may offer a means of cancer interception for DCIS. The vaccine will advance to phase II studies.

Introduction

Cancer vaccines are increasingly being developed to treat lesions at the premalignant stage, a concept termed “cancer interception”.¹ One of the most common premalignant tumors diagnosed is ductal carcinoma in situ (DCIS). With the adoption of routine mammography, DCIS accounts for about a quarter of all new cases of breast cancer each year.² Only a minority of patients with DCIS will develop invasive breast cancer, and to date, there are no tests or tools that are effective in determining which patients are at the highest risk.³ For this reason, most patients are treated with surgery and often radiation and systemic therapy.⁴ There is a growing consensus that DCIS is overtreated, which exposes patients to unnecessary therapy-associated risks, and active surveillance is currently being studied as an option to avoid those risks in some patients.⁵ Low-risk, low-toxicity treatment approaches, such as vaccines, would be ideally suited to this population.

Clinical trials have evaluated active immunization as a treatment option for DCIS. Nearly half of DCIS overexpress the HER2 protein, and HER2-specific Th1 can be detected in the peripheral blood of these patients. In patients with invasive HER2-positive breast cancer, the loss of circulating HER2-specific Th1 is associated with suboptimal clinical response to adjuvant and neoadjuvant chemotherapy and can be restored by HER2 vaccination.⁶ Investigators initiated a clinical trial of a HER2-directed dendritic cell (DC)-based vaccine designed to elicit Th1 immune responses to treat DCIS lesions.⁷ Subjects with HER2+DCIS were given 4 weekly intranodal HER2 peptide-loaded DC vaccines and monitored for resolution of DCIS prior to their planned surgical resections. All patients developed increased type I peripheral blood immune responses to HER2. After immunization, 11 of 13 patients had persistent DCIS, but 7 of those had significantly reduced HER2 expression in lesions. Two of the patients had no DCIS or inadequate tissue for evaluation. The persistence of DCIS with a loss or decrease of HER2 expression indicates the potential for immune editing; a productive immune response against the immunizing antigen but an immune escape of the remaining antigen-negative cells.⁷

We questioned whether a multiple antigen vaccine targeting DCIS-associated proteins, designed to selectively elicit Th1, could be effective in eliminating these lesions and preventing antigen loss variants. We have demonstrated that it is possible to edit the amino acid sequence of nonmutated tumor-associated antigens to specifically identify class II interacting epitopes that selectively elicit a type I T-cell response as evidenced by secretion of interferon-gamma (IFN-γ) but not type II cytokines by antigen-specific T-cells.^{8 9} Using a combination of multi-algorithm class II epitope prediction methods and functional phenotypic screening to prioritize epitopes that interact across multiple HLA-DR alleles which could be part of a universal vaccine approach, we developed a tri-antigen epitope-based vaccine consisting of Th1-selective epitopes derived from IGFBP-2, HER2, and IGF-IR.^{8 10 11} All three proteins have been shown to be overexpressed in both hormone receptor-positive and -negative DCIS and atypical ductal hyperplasia.12,14 We tested the tri-antigen Th1 selective vaccine in the TgMMTV-neu transgenic mouse model which develops DCIS progressing to invasive mammary cancers.¹¹ The tri-antigen vaccine was more effective in preventing the development of invasive breast cancer and DCIS than each single antigen used alone and could prevent the development of mammary cancer in 65% of mice. The protective effects of vaccination were mediated by CD4+T cells.¹¹ Immunized animals did not develop evidence of any autoimmune toxicities.

The ability to prevent DCIS and invasive breast cancer and the lack of toxicity in transgenic mouse models supported the clinical development of the tri-antigen vaccine. As a first-in-human study, the trial was performed in subjects with non-metastatic HER2-negative breast cancer in prelude to advancing the vaccine to DCIS interception. We specifically chose HER2-negative breast cancer patients as HER2-positive breast cancer patients often have high levels of HER2-specific T-cells induced via the use of trastuzumab, a HER2-specific monoclonal antibody used for primary treatment of the disease.^{15 16} High levels of pre-existing immunity would obscure the ability to assess the immunogenicity of HER2, one of the component antigens of this vaccine. Here we report the results of the Phase I dose escalation study of the tri-antigen vaccine targeting IGFBP-2, HER2, and IGF-IR.

Methods

Participants

32 participants with non-metastatic, node-positive, HER2-negative breast cancer who were in remission and defined as having no evidence of disease were enrolled between September 2016 and December 2019. Written informed consent was obtained for all participants (NCT02780401). The study was approved by the National Cancer Institute Central IRB. Eligible subjects had to be at least 28 days from cytotoxic, steroid, or immune therapy, have no contraindication to receiving granulocyte macrophage colony-stimulating factor (GM-CSF), and have no history of uncontrolled autoimmunity, diabetes, or significant heart disease. Concurrent bisphosphonate, denosumab, and anti-estrogen therapy was allowed.¹⁷ Of the 32 participants enrolled, 30 completed all 3 vaccinations and 27 completed all follow-up visits (online supplemental Figure 1).

Study design

The study was conducted in accordance with the principles outlined in the US Common Rule.¹⁸ This multisite phase I study, Universities of Washington and Wisconsin, had a primary objective to assess the safety of 3 escalating doses of a Th1 selective multiantigen, polyepitope plasmid-based vaccine encoding segments of IGFBP-2, HER2, and IGF-IR.¹¹ Secondary objectives reported: (1) determining the level of antigen-specific Th1 immunity generated with vaccination, (2) whether antigen-specific Th1 and/or Th2 persist in the peripheral blood after immunizations had ended, (3) whether vaccination modulated peripheral blood T-regulatory cells (Treg) and myeloid-derived suppressor cells (MDSC), (4) whether there were any long-term adverse events (AEs) associated with vaccination, and (5) determining a phase II dose.

10 participants were sequentially assigned to one of three dose levels (150, 300, and 600 µg) of the tri-antigen plasmid DNA vaccine. Participants who left the study prior to analysis of an immune response were replaced so that complete immunologic data could be obtained 1 month after the last vaccination; hence, 12 individuals were enrolled on the 300 µg dose level (online supplemental Figure 1). All enrolled patients (n=32) were included in safety assessments. Vaccines were administered intradermally with 100 µg of GM-CSF admixed with the plasmid DNA as an adjuvant monthly for 3 months. Participants without a recent tetanus toxoid vaccine received one prior to the first study immunization as a positive control. Toxicity evaluations and safety assessment occurred immediately prior to each of the monthly study vaccinations and at 1 and 6 months after the last vaccine (including physical exam, cardiac evaluation by ECHO or MUGA, autoimmune serologies, CBC, HbA1c, GAD65, and comprehensive metabolic panel). Toxicity was graded per Common Terminology Criteria for Adverse Events (CTCAE) V.4.0. Peripheral blood mononuclear cells (PBMCs) were collected at baseline, 1, and 6 months after the last immunization to assess cellular immune responses to the vaccinating antigens. Dose level 1 was enrolled first, and after dose safety was determined, dose level 2 participants were enrolled. Once the second dose was deemed safe and immunologically more efficacious than the first, participants were enrolled at the highest dose. Participants were then followed annually for 5 years through records collected from their oncologist or primary care physician. The clinical protocol can be found in online supplemental materials.

Tri-antigen vaccine

The pUMVC3-IGFBP2-HER2-IGF-IR plasmid contains 5947 base pairs. A DNA insert encoding the N-terminal 163 amino acids of IGFBP2,¹⁹ HER2 amino acids 686-994/1036-1051/1156-1197,²⁰ and IGF-IR amino acids 1196-1261/1323-1360¹⁰ was inserted between the Kpn1 and Not1 restriction sites. Along with binding promiscuity across multiple HLA-DR alleles, another selection criterion for inclusion of epitopes in the vaccine formulation was that epitope-specific T-cells respond to recombinant protein presented by autologous antigen-presenting cells²⁰ (and representative examples in online supplemental Figure 2). The vaccine is supplied in single-use vials as a sterile, frozen solution. Each single-use vial had a final concentration of 1 mg of DNA/1 mL tromethamine/EDTA (TE). Reconstituted recombinant human GM-CSF (rhuGM-CSF, Sargramostim) was admixed with the plasmid-based vaccine prior to administration.

Antigen-specific immune responses

All samples for each patient were cryopreserved, thawed, and analyzed simultaneously to ensure comparability.²¹ There was no significant difference in PBMC viability (p=0.94) or yield/ml blood drawn (p=0.11) between samples collected at the Universities of Washington and Wisconsin. Samples from all patients enrolled (n=10 each) were analyzed in the 150 µg and 600 µg dose levels. However, due to an insufficient cell yield in one patient, samples only from n=9 patients were analyzed in dose level 300 µg. IFN-γ ELISPOT on 8-day antigen-specific cultured T-cells and IL-10 ELISPOT on PBMC was performed as previously described.^{8 10 22} We chose IL-10 as the Type II cytokine to screen as it is secreted by both type II T-cells (Th2) and T-regulatory cells, although cytokines such as IL-5 are more commonly produced by Th2. One µg/mL of overlapping peptide pools for HER2 amino acids 686-1197 (512aa), IGFBP2 amino acids 1-163 (163aa), and IGF-IR amino acids 1196-1261 (66aa) and 1323-1360 (38aa) (JPT) were used as stimulating antigen as a surrogate to assess immune response to protein.²³ Tetanus toxoid (0.5 U/mL) and phytohemagglutinin (2.5 µg/mL) were positive controls. Each antigen was assessed in four replicates of 2×10⁵/ well. Data are presented as IFN-γ or IL-10 spots per well corrected for background in 10⁶ cultured PBMC (corrected spots per well, cSPW) for each individual antigen or as the sum of the magnitude for all antigens. No ex vivo assays were performed due to lack of remaining cells after per-protocol assays were completed.

Participants were considered to have pre-existing immunity to the immunizing antigens if, at baseline, the mean antigen-specific cSPW were statistically different from the no antigen wells (p<0.05) for at least one antigen. Participants were considered to be a “responder” if the postvaccination cSPW at any time point evaluated after immunization was significantly (p<0.05) above the prevaccination value for at least one antigen. For a “non-responder” designation, the postvaccination cSPW after immunization was not statistically different from the prevaccination value for any antigen at any time point.²⁴ Participants had a persistent immune response if the postvaccination cSPW at 6 months after the last vaccine was greater than two SD above the prevaccination value for at least one antigen (p<0.05). Fold change for responders was calculated from the sum baseline and sum maximum response for only those peptides that induced an IFN-γ response as defined above.

Flow cytometry

Cryopreserved PBMC collected at baseline and 1 month after the last vaccine were thawed and washed according to our published methods,²⁵ then incubated with 100 µg of 10% normal mouse serum in PBS at room temperature for 30 min to block non-specific binding. After washing, the cells were stained with Live/Dead stain (Aqua, Molecular Probes). PD-1 expressing T-cell subsets were enumerated with anti-human-CD3 FITC (clone SK7); anti-CD4 APC-Cy7 (clone RPAT4); anti-CD8 PerCP-Cy5.5 (clone SK-1), anti-CD279 PC-Cy7 (clone EH12.1). T-regulatory cells were enumerated according to eBioscience FOXP3 staining protocol with anti-human-FOXP3 PE (clone 236A/E7), anti-CD3 Pac Blue (clone: UCHT), anti-CD4 APC (RAPT4), anti-CD25 APC Cy7 (clone: BC96), anti-CD127 FITC (clone A019D5). m-MDSC were enumerated with human lineage cocktail-FITC of anti-CD3 (clone UCHT1), anti-CD14 (clone: HCD14), anti-CD16 (clone: 3G8), anti-CD19 (clone: HIB19), anti-CD20 (clone: 2H7), anti-CD56 (clone: HCD56) and anti-human-HLA-DR PE-Cy7 (clone: L243), anti-CD14 BV421 (clone: MOP9), anti-CD33 PercP (clone: WM53) and anti-CD11b (clone: P67.6) and defined by %HLADR-CD14+CD33+CD11b+in live cells. All of the antibodies were purchased from BioLegend or BD Biosciences. After washing the cells with PBS/1% FBS, data acquisition was performed on a FACS Canto flow cytometer (BD Biosciences) and analyzed using FlowJo software.

Analysis of gene expression in purified T-cells

PBMC derived from eight responding and eight non-responding participants randomly selected across all dose groups were stimulated with a pool of all seven peptides in the tri-antigen vaccine and a median of 2X10⁶ T-cells were negatively selected using the Pan T-cell Isolation Kit according to manufacturer’s instructions (Miltenyi Biotec). RNA was extracted and gene expression was assessed using the CAR-T Gene Expression Panel on the nCounter MAX Analysis System (NanoString Technologies). This panel includes a curated list of genes associated with T-cell persistence, metabolism, activation, and exhaustion. Samples were analyzed using the nSolver software (NanoString Technologies). All samples were normalized against positive controls and selected housekeeping genes using the geometric mean. Differential expression to identify specific targets was performed and p values were adjusted using the Benjamini-Hochberg procedure. Data are presented as differential expression (log2 fold change) of T-cells at baseline and 1 month after the last vaccine for only genes that were significantly upregulated or downregulated (p<0.05).

Statistical analysis

Per protocol, analyses of secondary endpoints were exploratory and descriptive in nature. With 10 participants per dose (n=30 total), the study had approximately 80% power at a two-sided α=0.05 F-test to detect a linear dose-response effect corresponding to an anticipated effect size (defined as R₂/(1−R₂)) of 0.281, indicating adequate power only for relatively large dose-response relationships.

Descriptive statistics were used for patient demographics at enrollment and for AEs. The differences across dose levels were analyzed by the analysis of variance for continuous variables and the χ² test for categorical variables. We used the Cochran-Armitage test to assess the dose-response relationship between the number of participants experiencing specific AEs attributable to the vaccine and dose level.

For comparisons of differences in the magnitude of cellular immunity to antigens across the three dose levels, normality was first checked using the Shapiro-Wilk test, and a non-parametric test was used due to detected skewness of the data. Kruskal-Wallis and Mann-Whitney U tests were conducted accounting for multiple comparisons. For comparisons within each dose level, data collected at different time points were paired. The incidence of cellular immunity across dose levels was analyzed using the χ² test. For responders, total vaccine response between baseline and two postvaccine magnitudes was evaluated using the Friedman test with Dunn’s test for multiple comparison.

Kaplan-Meier curves, generated to show the probabilities of overall survival (OS) and progression-free survival (PFS) from the time of study enrollment, were compared across dose levels using a log-rank test for trend, where OS and PFS were defined as the time elapsed between enrollment and death or earliest disease progression. All tests are two-tailed, and a p<0.05 was considered statistically significant. Statistical analysis was performed using GraphPad Prism V.10 (GraphPad Software) and R V.4.2.1.

Results

Immunization with the tri-antigen vaccine was well tolerated and associated with minimal toxicity

There were no significant differences in subject characteristics across dose levels (table 1). A total of 32 participants (median (range) age, 52 [31–79] years) were enrolled, including 31 female and 1 male subject (table 1). 15 (47%) were individuals with stage III breast cancer, with the remainder having earlier stage disease. The majority of the participants enrolled had hormone receptor-positive breast cancers and were postmenopausal.

Table 1. Baseline participant demographic characteristics (n=32).

	150 µg (n=10)		300 µg (n=12)		600 µg (n=10)
	No.	%	No.	%	No.
Age, years (median, range)	51 (37–64)		53 (31–64)		45 (36–79)
Breast cancer diagnosis
Stage IB	0	0	1	8	0
Stage IIA	2	20	2	17	3
Stage IIB	3	30	5	42	1
Stage IIIA	3	30	2	17	3
Stage IIIB	1	10	0	0	1
Stage IIIC	1	10	2	17	2
Menopausal status at enrollment
Premenopausal	2	20	0	0	4
Postmenopausal	8	80	12	100	5
Not applicable	0	0	0	0	1
Estrogen receptor status
Positive	8	80	10	83	8
Negative	2	20	2	17	2
Progesterone receptor status
Positive	8	80	9	75	8
Negative	2	20	3	25	2
Ethnicity
Non-Hispanic or Latino	10	100	10	83	8
Hispanic or Latino	0	0	1	8	0
Unknown/not reported	0	0	1	8	2
Race
White	10	100	12	100	8
Asian	0	0	0	0	1
Unknown/not reported	0	0	0	0	1
Gender
Female	10	100	12	100	9
Male	0	0	0	0	1

149 AEs that were possibly, probably, or definitely related to study vaccination were reported while participants were on study up to the 6-month follow-up visit. These potentially related AEs were all grades 1 or 2 in severity (table 2; online supplemental Tables 1 and 3). Online supplemental Table 2 depicts the number of participants who reported each AE, regardless of attribution, by the highest grade experienced. There were no significant differences in AEs across dose levels (table 2). The most common AEs were consistent with those observed after most infectious disease-related immunizations: injection site reaction (81%), influenza-like symptoms (44%), and fatigue (25%) temporally related to vaccination. While no patient reported symptoms concerning for autoimmune disease, one participant on the 150 µg dose level experienced a new transient grade 1 elevation of ANA during the course of immunization, which resolved spontaneously. As the tri-antigen vaccine targets two members of the insulin-like growth factor family, we assessed both HbA1c and GAD65 throughout the trial. There was no evidence of clinically significant hyperglycemia induced by vaccination (table 2).

Table 2. Adverse events (AEs) (possibly, probably, or definitely related).

	Total		150 µg (n=10)		300 µg (n=12)		600 µg (n=10)
	No.	%*	No.	%*	No.	%*	No.	%*
Most common related adverse events
Injection site reaction	26	81	8	80	10	83	8	80
Influenza-like symptoms	14	44	5	50	7	58	2	20
Fatigue	8	25	4	40	3	25	1	10
Nausea	6	19	4	40	1	8	1	10
Chills	6	19	3	30	3	25	0	0
Myalgia	6	19	3	30	3	25	0	0
Headache	5	16	1	10	3	25	1	10
Bone pain	5	16	0	0	3	25	2	20
All related abnormal autoimmune serologies
ANA	1	3	1	10	0	0	0	0
Anti-DS DNA	0	0	0	0	0	0	0	0
C3	0	0	0	0	0	0	0	0
TSH	0	0	0	0	0	0	0	0
GAD65	0	0	0	0	0	0	0	0
Glucose	0	0	0	0	0	0	0	0
Adverse events during long-term follow-up†
Psoriatic arthritis	1	3	0	0	0	0	1	10
Subclinical hypothyroidism	2	6	1	10	1	8	0	0
All related AE gradings (n=149)‡
1	144	97	55	37	58	39	31	21
2	5	3	2	1	0	0	3	2
3	0	0	0	0	0	0	0	0
4	0	0	0	0	0	0	0	0
5	0	0	0	0	0	0	0	0

^*Percentage of patients with related AEs.

^†Observed more than 6 months after trial completion.

^‡Of all related AE (n=149), long term toxicity not included. Source of terms CTCAE V.4.0.

CTCAE, Common Terminology Criteria for Adverse Events; TSH, thryoid stimulating hormone.

Vaccination-induced type I cellular immune responses to all three antigens in the vaccine

We evaluated the maximum immune response generated to each antigen individually. Based on comparisons of maximal IFN-γ response magnitude, no consistent pattern suggesting preferential dominance of one antigen over the others was observed. There was no difference in the magnitude of the maximal IFN-γ response to HER2 and IGFBP-2 between the dose levels. There was a significant difference at baseline for the level of immunity to IGF-IR between the 150 mg vs the 600 mg dose level (p=0.007), but no difference for the maximal response for that antigen between any dose level (figure 1A–C). The vaccine was indeed Th1-selective as there was no antigen-specific IL-10-secreting T cells generated at any dose level or time point evaluated (figure 1D–F). Participants generated a variable incidence of IFN-γ secreting T-cell immune responses among the 150 µg dose level (60% to HER2, 0% to IGF-IR, 40% to IGFBP-2; figure 1A), the 300 µg dose level (33% to HER2, 22% to IGF-IR and 67% to IGFBP-2; figure 1C) and the 600 µg dose level (10% to both HER2 and IGF-IR and 20% to IGFBP-2; figure 1G–I). These responder proportions are presented descriptively; the study was not powered for formal comparisons of incidence within dose level or for pooled comparisons across dose levels, and such analyses were not prespecified in the protocol.

Figure 1. Immunization induced type I cellular immune responses to all three antigens in the vaccine. Corrected IFN-γ (A–C) or IL-10 (D–F) corrected spots per well (cSPW) in 10⁶ PBMC or incidence of IFN-γ response (G–I) for the indicated individual antigen after immunization at (A, D, G) 150 µg; (gray bars), n=10, (B, E, H) 300 µg (blue bars), n=9 and (C, F, I) 600 µg (black bars), n=10. Connecting lines denote matching baseline (left symbol) and maximum (right symbol) response for each subject. PBMC, peripheral blood mononuclear cell.

The tri-antigen vaccine is immunogenic in the majority of patients in the low-dose and mid-dose groups

Next, we considered the total vaccine response as the sum IFN-γ magnitude from all antigens for each dose level. At the 150 µg dose level, 70% (7/10) of participants were responders. For the responders, responses were heterogeneous; thus, as a group, there was no significant difference between the baseline magnitude (median 300, range 6–1816, cSPW/10⁶ cultured PBMC) and the postvaccine magnitude at 1 month (median 611, range 238–1928 cSPW/10⁶ cultured PBMC) or at 6 months after the final vaccine (median 444, range 20–1073 cSPW/10⁶ cultured PBMC; figure 2A). At the 300 µg dose level, 67% of participants (6/9) were responders. For the responders, as a group, there was also no significant difference between the sum antigen-specific IFN-γ response before vaccination (median 453, range 0–1386 cSPW/10⁶ cultured PBMC) and 1 month or 6 months following the last vaccine (median, 456, range, 24–1682 cSPW/10⁶ cultured PBMC and median, 548, range, 78–2439 cSPW/10⁶ cultured PBMC, respectively; figure 2B). At the 600 µg dose level, 40% (4/10) of participants were responders. Similar to the lower dose levels, as a group, there was no difference in the magnitude of the immune response across all responders between baseline sum IFN-γ response (median, 834, range 323–2420 cSPW/10⁶ cultured PBMC) 1 month (median, 412, range 0–1288 cSPW/10⁶ cultured PBMC) or 6 months after vaccination (median, 335, range 26–1169 cSPW/10⁶ cultured PBMC; figure 2C).

Figure 2. The tri-antigen vaccine is immunogenic in the majority of patients in the low and mid-dose groups. Sum of all antigens corrected spots per well (cSPW) in 1×10⁶ PBMC for dose levels (A) 150 µg (n=10), (B) 300 µg (n=9) and (C) 600 µg (n=10 for the 0 and 1-month timepoint and n=8 for the 6-month time point). Lines connect values from each subject determined pre-vaccine (0) and at 1 and 6 months after the last vaccine. Responding participants are in blue lines, non-responders are in gray lines. (D) Mean (±SEM) fold change corrected spots per well from pre-vaccine to maximum response summed from only the responding antigens. PBMC, peripheral blood mononuclear cell.

At the maximum response observed at either the 1-month or 6-month time point, at the 150 µg dose level, the responders augmented immunity by a median 3.5-fold (range, 1.3–118 fold; figure 2D). Nine of 10 participants (90%) demonstrated pre-existing immunity to at least one antigen, and 67% (6/9) of those with pre-existing immunity were able to augment immunity after vaccination. Persistence of immune response was observed in 40% of participants (4/10) measured at the 6-month time point. At the 300 µg dose level, immunity was increased by a median 4.6-fold (range, 1.2–1377 fold) in responders (figure 2D). Pre-existing immunity to at least one antigen was observed in seven of nine (78%) participants, and antigen-specific immunity could be boosted in four of those participants. 56% (5/9) of participants retained an immune response at the 6-month time point. For those patients responding to the 600 µg dose level, immunity was augmented by a median twofold (range, 1.4–39.5 fold; figure 2D). Six of 10 (60%) participants exhibited pre-existing immunity to at least one antigen, with three (50%) of those participants increasing the immune response postvaccine. Persistence of the immune response evaluated at the 6-month time point was observed for only 13% (1/8) of participants.

We evaluated whether the incidence of pre-existent immune responses at any dose levels was associated with loss of immunity over time, but there was no significant association.

We sequentially evaluated m-MDSC and Treg cells as cell populations that may contribute to immune suppression, and there were no significant differences between prevaccine and postvaccine levels at any dose (online supplemental Figure 3). Additionally, we monitored for changes in PD-1 expression on circulating CD4 and CD8 T-cells. Again, there was no significant difference in expression at any dose level postimmunization as compared with prevaccination (online supplemental Figure 3).

We questioned whether we could determine any characteristics in the T-cells of responder and non-responder populations which could be associated with the generation of an antigen-specific immune response. Differential expression analyses in gene signatures of purified CD3⁺ T-cells derived from matched responder samples at baseline and 1 month postvaccine revealed a significant upregulation in the expression of 17 genes: UBE2I, PGK1, TGIF2, COX16, MPC2, MAX, COX4I1, GLUD1/2, HMGCR, CD247, HSPA9, NAA20, SLC25A6, ALDOA, PSMA3, CCNC, SDHB (figure 3A). The majority of these genes encode proteins associated with energy metabolism (figure 3B) and mitochondrial biogenesis (figure 3C). CD247 is T-cell receptor zeta, which is required for an active immune response. There were no genes that showed significant downregulation or loss of expression in immune responders. Conversely, no genes were differentially upregulated in the non-responder samples. Five genes were downregulated: PTGER4, GLS, IL23A, AhR, MMP9 (figure 3A). These five genes either play a role in T-cell migration (MMP9) or activation/exhaustion (PTGER4), are markers of the Th17 phenotype (IL23A and AhR), or are involved in energy metabolism (GLS; figure 3D).

Figure 3. T-cells derived from participants responding to the tri-antigen vaccine exhibit increases in energy metabolism and mitochondrial biogenesis. (A) Volcano plot of fold change (FC) in gene expression levels prevaccine to 1 month postvaccine from subjects responding to the tri-antigen vaccine (blue circles) or non-responders (gray circles). A dotted line denotes the cut-off for significance (p<0.05). Violin plots showing FC levels from prevaccine to postvaccine for significantly upregulated genes in responders associated with (B) energy metabolism and (C) mitochondrial biogenesis or (D) for significantly downregulated genes in non-responders associated with T-cell activation and energy metabolism. n=8 responders and eight non-responders. AHR, Aryl Hydrocarbon Receptor.

Long-term toxicity and clinical outcomes

The median (range) follow-up was 70.8 (33.9–84.5) months, 57.8 (28.2–69.0) months, and 45.7 (12.0–67.7) months for the 150 µg, 300 µg, and 600 µg dose levels, respectively. During the long-term follow-up period, from the 6-month follow-up, annually, until the subjects’ last status report, three participants reported a worsening of pre-existing autoimmune disease (table 2). One had a history of symptomatically controlled psoriatic arthritis at the time of study entry. During active immunization, at the 600 µg dose level, she had only faint psoriasis plaques on her face, legs, and back. After vaccinations had ended, she developed joint pain that required treatment with sulfasalazine 500 mg and calcipotriene 0.005% solution for adequate symptom control. Two participants had a history of hypothyroidism. One individual, at the 150 µg dose, had a low-level elevation of thyroid stimulating hormone (TSH) (never more than 2 points above the upper range of normal) that persisted into long-term follow-up. She was managed with oral levothyroxine by her primary care provider. The second participant, who had received the 300 µg vaccine dose, developed a transient elevation of her TSH within the year of completing vaccinations, which resolved without treatment and was not associated with symptoms.

At the time of this report, the median OS (online supplemental Figure 4A) has not been reached at the 300 µg dose level, while the median OS for 150 µg and 600 µg dose level was 77.4 months and 92.8 months, respectively. The median PFS (online supplemental Figure 4B) was 52.2 months and 47.7 months for doses 150 and 300 µg, respectively, and the median PFS had not been reached for the 600 µg dose. There were no statistically significant differences in OS and PFS across arms.

Discussion

Data presented here suggest that the tri-antigen vaccine is safe and well tolerated, including in long-term follow-up. Immunization generated Type I antigen-specific T-cells without augmenting antigen-specific type II T-cells or immune suppressive cells. While participants receiving the 150 and 300 µg doses developed similar incidence and magnitude of immune responses 1 month after vaccines had ended, antigen-specific Th1 immunity persisted for months in the majority of those participants’ who received the 300 µg dose; thus, this dose will be the recommended Phase II dose. Vaccine-induced T-cell responses were not persistent in the highest (600 µg) dose group. Exploratory studies suggest the ability to immunologically respond to immunization is associated with optimal T-cell metabolism and energy expenditure, that is, T-cell fitness.

The nonmutated tumor antigens in the tri-antigen vaccine are all markedly overexpressed in DCIS lesions. Marked overexpression of nonmutated proteins is a mechanism by which immunity is enhanced.²⁶ Overexpression unmasks subdominant epitopes that have not been available in great enough concentrations to the MHC for recognition due to more effective competition by tolerizing epitopes.²⁷ Overexpression may allow specificity of the vaccine-induced immune response for the lesion without the induction of autoimmunity. We did not observe the development of significant toxicity. Our findings are similar to a number of other studies of nonmutated tumor antigen vaccines.²⁸ A systemic analysis of the toxicities of 239 phase I cancer vaccine trials, the majority targeting nonmutated proteins, showed the rate of grade 3 or 4 vaccine-related AEs was 2 per 1000 vaccinations, many of which were attributable to the vector used for immunization.²⁸ Studies in mice have shown that DNA vaccines do not induce long-lasting anti-DNA antibodies, which is consistent with data presented here.²⁹ No long-lasting or significant changes in serologies were observed in this study. The encouraging safety profile of the tri-antigen vaccine supports continued clinical development.

The vaccine was designed to generate Th1-selective T-cell responses, and the majority of subjects at all three dose levels developed antigen-specific Th1 without augmenting antigen-specific IL-10 secreting T-cells.^{8 11 30} There was no augmentation in the level of immune suppressive cells such as m-MDSC or Treg in any of the dose levels to account for this difference. Of the patients who completed all 3 vaccines, 21 entered the study with pre-existing immunity to at least one of the immunizing antigens. This level of pre-existing immunity is consistent with other vaccine studies. A recent study of individualized mRNA vaccines in patients with triple-negative breast cancer indicated 8/14 patients (57%) had evidence of pre-existent immunity directed against epitopes that they were immunized against.³¹ Another neo-antigen vaccine trial demonstrated pre-existing immune responses in one third of the patients immunized.³² Indeed, immune monitoring in clinical trials has shown that using ex vivo assays to monitor the vaccinated response may result in underestimation of the pre-existent immune response and inflate the number of patients with a de novo vaccinated response.³³ Immunity to specific mutations could not be detected in ex vivo analysis in the blood of melanoma patients but could be readily identified by ELISPOT after an in vitro stimulation in the same patients.

Of those patients with pre-existing immunity, 21% decreased antigen-specific immune responses over time, while 24% boosted antigen-specific Th1 with vaccination. Fourteen percent had pre-existing immunity to two antigens and boosted immunity to one but decreased immunity to the other, while the remaining 14% did not have any change in immune response after vaccination. Does pre-existing immunity assist in the boosting of the cancer-specific immune response or impede the development of immunity? Studies of a live attenuated influenza vaccine demonstrated pre-existent antibody immunity interfered with effective immunization and inhibited the development of cellular immunity.³⁴ We evaluated the interaction of preexistent immunity to the immunizing antigens and the ability to boost the response and found no statistically significant association. However, we have observed blunted immune responses at the highest dose levels in a previous trial of a plasmid DNA vaccine targeting HER2.²⁰ In this study, three doses (10, 100, and 500 µg) were evaluated in 66 patients. While all doses were immunogenic, participants in the 500 µg dose level did not develop long-term HER2-specific T-cell responses. The vaccine sites were biopsied to assess for the persistence of DNA at the site of vaccination, which was most prevalent at the highest dose and was associated with loss of vaccine-induced immunity over time.²⁰ While we did not perform skin biopsies in the current trial, we assume a similar mechanism is limiting persistent immunity at the highest dose. The magnitude of Th1 immunity achieved and the persistence of immunity in the majority of patients after immunizations had ended suggest the 300 µg dose as the optimal dose for further study.

There were non-responders at each dose level. We explored potential differences in the genetic profiles of T-cells derived from participants who developed or did not develop antigen-specific immune responses after vaccination. Several of the most highly upregulated genes we identified in T-cells from responding subjects are associated with the metabolic fitness of lymphocytes and their long-term survival. As examples, phosphoglycerate kinase 1 (PGK1) and aldolase A (ALDOA) are glycolytic enzymes that are metabolic regulators of CD4+T cell differentiation, especially for Th17 cells.³⁵ Mitochondrial pyruvate carrier (MCP2) supports long-term survival of antibody-secreting B-cells.³⁶ 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase (HMGCR), an endoplasmic reticulum-residing enzyme, is essential for the survival of both T and B-cells.³⁷ In contrast, those individuals who did not respond to vaccination displayed downregulated genes associated with T-cell activation and metabolism. Aryl Hydrocarbon Receptor (AHR) is a transcription factor that has multiple roles in immune activation via either ligand-specific interactions or cell-specific effects.³⁸ Glutaminase (GLS) facilitates glutamine uptake, the differentiation of T-cells into Th1, and tumor lymphocytic infiltration.³⁹ IL-23 is a pro-inflammatory cytokine that induces the expansion of Th17.⁴⁰ Proteins encoded by these genes are all necessary for a tumor destructive immune response; thus, they may not be present in the T-cells of non-responders. The expression of many of these genes is associated with T-cell fitness, which defines the ability of the T-cell to function effectively, especially in creating lasting immune responses.

There are limitations to this study, which include the small number of subjects in each cohort. Due to the concerns of potentially generating autoimmunity to the non-mutated self-antigens, patients were enrolled sequentially rather than randomized, and the immunogenicity of the dose was validated prior to exposing the next cohort to the intervention. Ideally, cohorts would have been randomized to reduce any bias or confounding factors in outcome assessments. We did not have enough samples to separate out T-cell subsets for individual analysis. Only a limited number of samples were available for the gene analysis studies due to the large amounts of cells consumed for the T-cell assays. Ideally, we would have liked to evaluate all patients on trial. Despite these limitations, the tri-antigen vaccine directed against IGFBP-2, HER2, and IGF-IR appears safe and immunogenic and will be studied further in phase II trials for the treatment of DCIS.

Data availability statement

Data are available on reasonable request.