Immunization with a Plasmid DNA Vaccine Encoding the N-Terminus of Insulin-like Growth Factor Binding Protein-2 in Advanced Ovarian Cancer Leads to High-level Type I Immune Responses

Type I T cells are needed for tumor eradiation (1). Both type I cytokine secreting CD4 T cells as well as cytolytic CD8 T cells are required for a sustained antitumor immune response. Most patients with ovarian cancer do not have high levels of type I T cells in their tumors. Indeed, tumor-infiltrating CD4 and CD8 T cells are found at much lower levels in ovarian cancer than in other solid tumors (2). There are several mechanisms involved in limiting type I T cells in ovarian cancer. The immune microenvironment is dominated by M2 macrophage and regulatory T cells secreting cytokines, such as IL10, which prevent the recruitment and proliferation of type I T cells (3). Ovarian cancer generally has a low mutational burden and few neoantigens that might stimulate a CD8 T-cell response (4). Finally, most identified tumor antigens in ovarian cancer are nonmutated and associated with the development of type II T cells and humoral immune responses (5).

Insulin-like growth factor binding protein-2 (IGFBP-2) is a protein that is overexpressed in most high-grade ovarian cancers and is associated with increased metastases and cell invasion (6). One study of over 400 ovarian cancers demonstrated that IGFBP-2 was overexpressed (2+,3+) in 74% of high-grade tumors, 73% of samples derived from stage III, and 82% of samples from stage IV disease (7). IGFBP-2 is also a human tumor antigen as evidenced by significant IGFBP-2–specific antibody levels in the serum of patients with epithelial cancers (8–10). We found, within the native protein sequence of IGFBP-2, class II–interacting T-cell epitopes that preferentially elicit either type I selective (IFNγ secreting) or type II selective (IL10 secreting) antigen-specific immune responses in both patients with cancer and volunteer donors (11). The type II selective epitopes have a higher functional avidity for antigen than the type I epitopes, are more common, and cluster in the C-terminus of the protein (11). In contrast, the type I selective epitopes cluster in the N-terminus of the protein. We constructed plasmid-based vaccines composed of the N-terminus and the C-terminus domains of IGFBP-2. In mice, the type I and type II cytokine selectivity of the constructs was preserved with N-terminus immunized mice generating high levels of IGFBP-2–specific type I T cells with little to no type II cytokine secretion and C-terminus vaccinated mice generating type II selective IGFBP-2–specific responses. In a tumor challenge, vaccination with the N-terminus significantly inhibited tumor growth, the C-terminus vaccine had no impact on tumor growth, but admixing the N- and C-terminus vaccines abrogated the antitumor effect of the type I selective N-terminus immunization (11). Vaccines targeting nonmutated proteins have been limited by the generation of vaccine induced regulatory T cells and type II T cells preventing the development of high levels of tumor antigen-specific type I T cells (12, 13). We hypothesize constructing vaccines directed against nonmutated tumor antigens by including only type I selective epitopes will allow the generation of high levels of type I T cells with no induction of regulatory responses.

We constructed an IGFBP-2 N-terminus plasmid-based vaccine for clinical use and conducted a clinical trial of immunization in patients with advanced-stage ovarian cancer. We questioned whether the vaccine was safe, immunogenic, and generated type I selective immunity without the elaboration of IL10 secreting T cells or regulatory T cells.

Twenty-five patients with advanced-stage or recurrent ovarian cancer were enrolled between March 2012 and January 2015 after written informed consent was obtained (NCT01322802). The University of Washington Cancer Consortium granted institutional review board approval. Eligible patients had been treated to complete remission with standard therapies including primary debulking surgery. A CA-125 level within normal limits for the testing laboratory had to have been documented 90 days prior to enrollment. Eligible patients were at least 28 days from cytotoxic or steroid therapy, had no contraindication to receiving sargramostim, and had no history of uncontrolled autoimmunity, diabetes, or significant heart disease.

The study was conducted in accordance with the ethical guidelines outlined in the US Common Rule. This single-arm phase I study had a primary objective to assess the safety of a T-helper 1 (Th1) selective polyepitope plasmid-based vaccine, pUMVC3-hIGFBP-2 (1–163), encoding the N-terminus of IGFBP-2 including an evaluation of long-term toxicity. Toxicity was evaluated by Common Terminology Criteria for Adverse Events (CTCAE) v3.0. Long-term toxicity data were collected via yearly review of the medical record. Secondary objectives included: (i) the immunogenicity of vaccination, (ii) the development of epitope spreading or a broadening of the immune response to other antigens, and (iii) the level of regulatory T cells generated during vaccination. Patients received pUMVC3-hIGFBP-2 (1–163; 100 mcg) vaccine admixed with rhuGM-CSF (100 mcg) intradermally monthly for 3 months. A 100-mcg dose was chosen based on a dose-escalation study with a HER2 intracellular domain vaccine encoded in the same plasmid as we used with IGFBP-2 N-terminus (14). Blood was drawn at baseline, prior to each vaccine for toxicity monitoring, and at months 4 (1 month after vaccination) and 9 (6 months after vaccination) from enrollment for toxicity and immunologic evaluation. Targeted enrollment was 22 patients with up to 5 additional replacement patients to complete the month 4 toxicity analysis. A sample size of 22 patients would ensure that if no toxicities occur the probability of such an occurrence is at least 90% if the true toxicity rate, that is, any grade 3 or 4 toxicity, is 10% or more. For the immunologic response rate, 22 would allow 80% confidence that the estimated immune response rate is within at least 0.14 of the true immunologic response rate. Twenty-three patients completed the vaccine regimen and toxicity evaluation, 21 completed all blood collection time points, and 20 patients completed immunologic analyses (Supplementary Fig. S1).

All samples for each patient were cryopreserved under GLP conditions, thawed, and analyzed simultaneously to ensure comparability (15). IFNγ and IL10 ELISPOT assays were performed as described previously (11, 16, 17). Ten mcg/mL of IGFBP-2 class II binding peptides contained in the vaccine were used in a pool: p8–22 (PALPLPPPLLPLLP), p17–31 (LLPLLPLLLLLLGAS), p67–81 (VAAVAGGARMPCAEL), p99–113 (EACGVYTPRCQGLR), p109–123 (GQGLRCYPHPGSELP), and p121–135 (ELPLQALVMGEGTCE; ref. 11). Tetanus toxoid (0.5 U/mL) and phytohemaglutinin (2.5 mcg/mL) were positive controls. Each antigen was assessed in four replicates of 2 × 10⁵/well. Data are presented as IFNγ or IL10 spots per well corrected for background (cSPW). Data are reported as the maximum cSPW determined either at 4 or 9 months after enrollment. Patients were considered to be a “responder” if the post-vaccination cSPW was greater than 2 SD above the pre-vaccination value and a “nonresponder” if the post-vaccination cSPW was within 2 SD above the pre-vaccination value. Two SD is equivalent to a P value of 0.05 in that there is a 95% probability that the values are statistically significant (18). Patients were considered to have preexistent immunity to IGFBP-2 if, at baseline, the mean antigen-specific SPWs were statistically different from no-antigen wells (16).

Intra- and intermolecular epitope spreading were determined via IFNγ ELISPOT using 10 mcg/mL of IGFBP-2 C-terminal epitopes in a pool: p121–135 (ELPLQALVMGEGTCE), p164–178 (NHVDSTMNMLGGGGS), p190–204 (ELAVFREKVTEQHRQ), p213–227 (LGLEEPKKLRPPPAR), p235–249 (DQVLERISTMRLPDE), p251–265 (GPLEHLYSLHIPNCD), p266–280 (KHGLYNLKQCKMSLN), p291–305 (PNTGKLIQGAPTIRG), and p307–321 (PECHLFYNEQQEARG), or IGF-IR class II binding peptides used in a pool: p1196–1210 (WSFGVVLWEIATLAE), p1242–1256 (FELMRMCWQYNPKMR), and p1332–1355 (GVLVLRASFDERQPYAHMNGGRKN; ref. 17).

Cryopreserved peripheral blood mononuclear cells (PBMC) collected at enrollment and 4 months after enrollment were thawed and washed according to our published methods (19), then incubated with 100 mcg of 10% normal mouse serum in PBS at room temperature for 30 minutes to block nonspecific binding. After washing, the cells were stained with fluorochrome-conjugated mAbs for phenotyping analyses. We performed intercellular staining for detecting Th1/Th2 transcription factors and regulatory T cells. The cells were surface-stained with CD3 PE-Cy5 and CD4 APC Cy7. After washing, the cells were stained with Tbet PE-Cy7, Gata 3 Alexa Fluor 647, and FOXP3 Alex 488 according to eBioscience FOXP3 staining protocol. The percent of Tbet-, Gata3-, and FOXP3-positive cells were analyzed among CD3⁺CD4⁺ and CD3⁺CD8⁺ T cells. Additional markers analyzed included: CD3 FITC, CD8 PE-Cy7, CD69 PE-Cy5, CD279 (PD-1) PE, CD28 PE-CF594, HLA-DR PE-Cy7, CD3 FITC, and CD4 APC. 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 was analyzed using the FlowJo software (RRID:SCR_008520).

T-cell receptor beta chain CDR3 regions were sequenced from PBMC collected at enrollment and 4 months after enrollment by ImmunoSeq (Adaptive Biotechnologies), with primers annealing to V and J segments, resulting in amplification of rearranged VDJ segments from each cell. Clonality values were obtained through the ImmunoSeq Analyzer software. Clonality was measured as 1 − (entropy)/log2 (number of productive unique sequences), with entropy considering the clone frequency. Morisita Overlap, a population overlap metric relating to the dispersion of clones in the samples, was calculated using the ImmunoSeq Analyzer software (Adaptive Technologies).

FACS sorting was performed on PBMC collected at enrollment and 4 months after enrollment using the Aria Cell Sorter (BD Biosciences) to purify activated CD4 cells (Live CD3⁺CD4⁺CD25⁺CD127^hi). Frozen cells were sent to NanoString Technologies for RNA extraction and profiling using the PanCancer IO 360 Panel Gene Expression Panel and analyzed on the nCounter MAX Analysis System (NanoString Technologies). Samples were analyzed using the Advanced Analysis Module of the nSolver software (NanoString Technologies). Samples were normalized against positive controls and selected housekeeping genes using the geometric mean. Ideal normalization genes were determined automatically by selecting those that minimize the pairwise variation statistic. 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 PBMC at enrollment and 4 months after enrollment.

For ELISPOT and flow cytometry, statistical analysis was performed using GraphPad Prism version 8 (GraphPad Software; RRID:SCR_002798). Data were compared using a paired t test (two-tailed) or the Wilcoxon matched-pairs signed rank test.

Patient characteristics are shown in Table 1. The median age of the patient population was 63 years, the median time from diagnosis was 25 months, and the median time from treatment was 7 months. Fifty-six percent of patients had only one prior chemotherapy regimen whereas 44% had undergone two or more prior regimens prior to enrollment.

Two hundred and three adverse events (AE) were collected during the course of the study. Ninety-nine percent of AEs were grades 1 and 2, with one grade 3 transient lymphopenia reported (Table 2; Supplementary Table S1). The most commonly reported possibly, probably, or definitely related AEs were injection site reaction (17%), fatigue (14%), flu-like symptoms temporally related to vaccination (7%), and arthralgia and decreased lymphocyte count (6%; Table 2). Three patients developed asymptomatic transient elevations in autoimmune serologies, all grade 1 or 2, which resolved without intervention during the course of the study.

Nonmutated tumor antigen vaccines have been shown to elicit both Th1 and Th2 responses (20). The IGFBP-2 N-terminus vaccine was specifically designed to be Th1 selective so we questioned to what level type I (antigen-specific IFNγ secretion) and type II (antigen-specific IL10 secretion) T cells might be boosted (11). The median antigen-specific IFNγ response before the first vaccine was 168 (P range 0–1,095) cSPW/10⁶ PBMC and the median response generated after vaccine was 301 (range 0–1,065) cSPW/10⁶ PBMC (Fig. 1A). Ten (50%) of 20 patients significantly augmented immunity and were considered to be responders. Three (15%) patients did not augment IGFBP-2 Th1 immunity, including the 1 patient who had greater than 3 previous lines of treatment, and 7 patients (35%) who had a decrease in antigen-specific immunity with immunization, all considered nonresponders. Of note, the presence of preexistent IGFBP-2 Th1 immunity was associated with lack of response to vaccination as compared with those patients without preexistent IGFBP-2 responses (P = 0.07). Thirteen (65%) of 20 patients demonstrated a preexisting IGFBP-2–specific IFNγ response. Only 3 (23%) of those patents with preexisting immunity generated significant antigen-specific immune responses after vaccination. The median antigen-specific IL10 response prior to immunization was 0 (range 0–370) cSPW/10⁶ PBMC and the median response generated after vaccine was 0 (range 0–438) cSPW/10⁶ PBMC. Twenty (100%) of 20 patients did not induce a statistically significant IGFBP-2 IL10-specific T-cell response (Fig. 1B).

We further analyzed the modulation of Th1 (Tbet) and Th2 (Gata3) cells by evaluating circulating CD4 and CD8 T cells before and after vaccination. There were significantly increased CD4⁺Tbet⁺ cells in the peripheral blood of most patients after vaccination (mean 9.6%, range 0.7%–36.4%) as compared with pre-vaccination (mean 5.6%, range 1%–19.2%; P = 0.04; Fig. 1C). There was no significant difference in CD4⁺Gata3⁺ levels in PBMC post-immunization (mean 2.5%, range 0.6%–7.3%) as compared with pre-vaccination (mean 2.6%, range 0.4%–7.6%; P = 0.34). There were significantly increased CD8⁺Tbet⁺ cells in the peripheral blood after vaccination as well (mean 43.0%, range 6.3%–92.1%) as compared with pre-vaccination (mean 29.6%, range 9%–42.6%; P = 0.01; Fig. 1D). There was no significant difference in CD8⁺Gata3⁺ levels in PBMC post-immunization (mean 1.9%, range 0.3%–6.8%) as compared with pre-vaccination (mean 1.7%, range 0.4%–6.4%; P = 0.45).

An increase in T-cell clonality after vaccination has been associated with improved survival of patients with cancer (21). After vaccination, there was a significant increase in T-cell receptor beta (TCRβ) clonality in our population (P = 0.03; Fig. 2A). The mean clonality prior to immunization was 0.063 (range 0.023–0.207) and after vaccination, the mean clonality was 0.079 (range 0.012–0.234). The development of multiple clonal populations could be an indication of epitope spreading, so we evaluated T-cell responses to the C-terminus of IGFBP-2 as a measure of intramolecular epitope spreading and to IGF-IR as an assessment of intermolecular epitope spreading.

Patients generated a significantly increased IFNγ response to the C-terminal epitopes of IGFBP-2 after N-terminus vaccination (P = 0.03; Fig. 2B). The median antigen-specific response before immunization was 71 (range 0–536) cSPW/10⁶ PBMC and the median response generated after vaccine was 83 (range 0–1,747) cSPW/10⁶ PBMC. Ten (50%) of 20 patients significantly augmented C-terminus Th1 immunity; 5 of these patients also developed Th1 responses to the N-terminus. Patients also generated a significantly increased IFNγ response to IGF-IR epitopes after vaccination (P = 0.02; Fig. 2C). The median IGF-IT–specific response pre-immunization was 2.8 (range 0–869) cSPW/10⁶ PBMC and the median response generated after vaccine was 52 (range 0–1,847) cSPW/10⁶ PBMC. Fourteen (70%) of 20 patients significantly augmented immunity. Six (30%) of 20 patients developed statistically significant immune responses to both of these antigens, the C-terminus and IGF-IR, after IGFBP-2 N-terminus immunization. Of note, there is only 8% sequence homology between IGFBP-2 (NCBI ascension number: AAA36048.1) and IGF-IR (NCBI ascension number: NP_000866.1).

When considering the level of immunity generated in responders, the fold change in IGFBP-2–specific IFNγ secreting T cells was striking (mean, 178; range 33–425) as compared with nonresponders (mean, −0.01; range 0.05–1.3; P < 0.0001; Fig. 3A). We questioned whether we could determine any characteristics in the T cells between populations which would predict a lack of response to the immunizing antigen after vaccination. We evaluated the Morisita index to assess TCRβ sequences pre- and post-vaccine to determine similarities in the clonal repertoire between the populations (Fig. 3B). There was a significant decrease in the index in responders (mean, 0.83; range 0.654–0.969) as compared with the nonresponders (mean, 0.926; range 0.866–0.992; P = 0.04; Fig. 3B), indicating responders employed different TCRβs than nonresponders.

Evaluating different T-cell subsets, we did not observe a significant difference in levels of FOXP3⁺ regulatory T cells, or activated CD28⁺ or CD69⁺ T-cell subsets in the PBMC before as compared with after vaccination in responders or nonresponders (all P > 0.05). IGFBP-2 N-terminus responders, however, had significantly fewer PD1⁺CD4⁺ cells (mean, decrease 8%; range decrease 41%–increase 1%) post-vaccination as compared with those who had no or a decreased response to IGFBP-2 (mean, increase 20%; range 1% decrease–46% increase; P = 0.02; Fig. 3C). There was no difference in PD1⁺CD8⁺ levels after vaccination in either group (Fig. 3C).

PBMC were sorted for the CD4⁺CD127⁺CD25⁻ subset, which included IL2-producing naive and central memory T-helper cells, and a panel of genes associated with T-cell function and immunity was analyzed. Differential expression analyses from pre- and post-vaccination gene signatures from nonresponders revealed a decrease in expression of genes associated with T-cell activation (F2RL1, SOCS1, STAT3, ICOS, IL2RB) and an increase in the gene IFI35, which is a negative regulator of interferon (Fig. 3D). There was no difference in expression of these same genes in responding patients pre- and post-vaccination.

There is a concern that immunity to nonmutated tumor antigens or the use of DNA-based vaccines may be associated with longer-term toxicities including autoimmunity (22). The median follow-up for the study population was 43 months (range 12–84). Three adverse events were reported in long-term follow-up. One patient developed an elevated TSH (5.52) which resolved. A second patient developed breast cancer and a third patient experienced a lichen planus flare. No other toxicities were reported. An additional concern was immunization with the entire N-terminus domain of an oncogene. For this reason, we evaluated disease outcome. The median progression-free survival was 12 months with 8 of 23 patients (34%) showing no progression at the time of this report (Fig. 4A). The median overall survival was 48 months (Fig. 4B). Twelve of 23 (52%) patients are still alive at a median of 52 months after entry to the study (range 49–84 months).

Vaccines targeting CD4 T cells can be designed to be universal to most HLA-DR types by identifying T-cell epitopes that bind promiscuously across multiple HLA-DR alleles (23). We use a multi-algorithm approach for the identification of pan-HLA-DR binding epitopes based on a scoring system, which prioritizes high binding affinity across the 15 most common MHC class II alleles (10). We further introduced functional screening to identify epitopes which selectively induce type I cytokine secretion in response to an antigen such as IGFBP-2 (24). IGFBP-2 is upregulated in many common solid tumors and promotes oncogenic pathways such as epithelial to mesenchymal transformation, cell migration, and metastasis (25). Vaccination with IGFBP-2 epitopes designed to elicit antigen-specific Th1 cells inhibits tumor growth in mouse models via the generation of both tumor-specific CD4⁺ and CD8⁺ T cells (10). Data presented here demonstrate Th1 selective immunity is safely generated against IGFBP-2, that epitope spreading is induced in a majority of patients and is the predominant type of immune response elicited, and that preexistent immunity and specific T-cell functional defects may be associated with a lack of T-cell response to the immunizing antigen.

Immunization with the IGFBP-2 N-terminus vaccine elicited Th1 selective immunity with no evidence of the generation of regulatory T cells, which have been reported to be stimulated after vaccination with nonmutated tumor antigens. Vaccine-induced regulatory T cells will suppress immune responses (26). We have shown that within the natural sequence of nonmutated tumor antigens are CD4 T-cell epitopes that will elicit type I or type II antigen-specific cytokine secretion (11). Inclusion of the type II epitopes in a vaccine significantly reduces vaccine efficacy. The IGFBP-2 N-terminus vaccine was designed to be enriched for Th1 selective epitopes with type II cytokine inducing epitopes edited from the vaccine construct (11). As a result, the median level of immunity achieved after vaccination was 1:3,300 PBMC being an IFNγ secreting IGFBP-2–specific T cell. These levels are consistent with those achieved with mutated antigen vaccines. In a study of neoantigen vaccines deployed for the treatment of melanoma, immunization with numerous mutated antigenic peptides could produce T-cell levels as high as 1:2,000 PBMC responding to a neoantigen after immunization (27). Despite high levels of unopposed type I T cells generated with IGFBP-2 N-terminus vaccination, there was no evidence of autoimmunity either in short- or long-term follow-up. Abnormal overexpression of a nonmutated protein in oncogenesis enhances the immunogenicity of that protein (28). Subdominant epitopes, not present when the protein is expressed at basal levels, become unmasked with overexpression and are present only on the tumor (29). It may be for this reason, vaccines targeting overexpressed nonmutated antigens have been associated with minimal toxicity (30).

The majority of vaccinated patients demonstrated evidence of epitope spreading, a broadening of the immune response to new epitopes within an antigen or to other antigens not included in the vaccine. Even patients who did not show statistically significant immune response to the immunizing antigen, IGFBP-2, could develop epitope spreading. Tbet-expressing T cells were also elevated after Th1 selective immunization, suggesting proliferation of activated non-exhausted cytolytic T cells (31). Epitope spreading reflects enhanced cross-priming due to activation of antigen-presenting cells by tumor trafficking type I T cells (32). Type I cytokines released by the T cells increase the efficiency and efficacy of antigen processing by innate immune cells. Epitope spreading occurring in most patients suggests that at least some vaccine-induced Th1 cells could home to the tumor in the majority of vaccinated patients despite the inability to measure significant differences in the levels of IGFBP-2 Th1 at the prescribed time-points. The development of epitope spreading is associated with improved clinical outcomes after various forms of cancer immunotherapy. One report of the use of a dendritic cell–based vaccine targeting the MART-1 antigen suggested that favorable clinical outcomes after vaccination were not related to the levels of immunity achieved with the immunizing antigen, but rather to the development of epitope spreading to other melanoma-associated antigens (33). A study of adoptive transfer of Epstein Barr Virus–specific T cells for the treatment of lymphoma demonstrated that evidence of epitope spreading to nonviral tumor-associated antigens was only seen in those patients achieving a clinical response (34). A limitation of our study is that the small number of patients preclude our ability to significantly associate the development of epitope spreading with disease outcome. The assessment of epitope spreading, however, could be an important biomarker of response and will be evaluated in phase II trials of the vaccine.

Some patients were not effectively immunized against IGFBP-2 N-terminus. Nonresponders were more likely to have a preexistent immune response to IGFBP-2 at the start of the trial. Patients with ovarian cancer have been shown to have preexistent immune responses directed against many antigens expressed in the tumor (5). Presumably these T cells were antigen educated in the tumor microenvironment and exposed to chronic antigen stimulation. Although the presumption is that preexistent immune responses would be significantly boosted by immunization, there is some evidence in animal models that an existing response is negatively modulated by vaccination. One study vaccinated mice with established melanoma tumors with an antigen-specific vaccine (35). Preexisting intratumoral T cells could initially proliferate in response to vaccination, but over time became less functional with subsequent immunizations. An initial vaccine boost could explain why we observed a higher rate of epitope spreading than IGFBP-2–specific Th1 immunity. The time course of our immune monitoring could have missed an initial vaccine-induced immune response that was diminished by continuing immunization in some patients. Vaccination may have driven the existing antigen-specific T cells to an exhausted phenotype as evidenced by upregulation of PD-1. We have shown that upregulation of PD-1 on peripheral blood CD4 T cells after immune modulation of chest wall tumors with the topical toll-like receptor-7 agonist imiquimod is associated with lack of clinical response to the immune therapy (19). The T cells derived from nonresponders showed evidence of other functional defects. The CD4⁺ T cells had downregulated genes associated with T-cell activation as compared with vaccine responders. Upregulation of SOCS1 is associated with a high-magnitude type I T-cell response, which was lacking in these patients (36). Expression of ICOS and F2RL1 (PAR-2) stimulates T-cell activation and enhances effector function (37, 38). Signaling through IL2 receptor β increases proliferation of CD8 effector T cells (39) and STAT 3 is important for inflammatory T-cell differentiation (40). One upregulated differentially expressed gene between vaccine responders and nonresponders was IFI35 and encodes interferon inducible protein 35, which functions to attenuate the IFN response (41). This constellation of differentially expressed genes highlights T cells that are not capable of activation and tissue destruction.

Data shown in this report demonstrate that vaccines targeting nonmutated tumor antigens, such as IGFBP-2, can be constructed to generate high-level type I immune responses without significant evidence of self-regulation. Studies described here also highlight potential biomarkers that may be useful for correlation to clinical outcome, such as epitope spreading. We have also identified biomarkers which may predict patients who will respond to the immunizing antigen, such as evidence of significant preexistent immunity. Phase II studies will address the therapeutic efficacy of Th1 selective IGFBP-2 vaccination in advanced ovarian cancer.

D.L. Cecil reports grants from NIH during the conduct of the study, as well as grants from Department of Defense, Janssen, Komen, and Rivkin Foundation outside the submitted work; in addition, D.L. Cecil has a patent for 61/972,176 issued and a patent for 61/972,179 527 issued. J.B. Liao reports grants from Merck, Sanofi, Forty-Seven, Laekna Therapeutics, Sumitomo Dainippon, Harpoon Therapeutics, and Precigen outside the submitted work. A.L. Coveler reports grants from Amgen outside the submitted work, as well as institutional grants from Seagen, AbGenomics, Novocure, AstraZeneca/Medimmune, Amgen, Surface Oncology, and Nextrast. J.S. Childs reports grants from NCI during the conduct of the study. M.L. Disis reports other support from University of Washington and Epithany during the conduct of the study. M.L. Disis also reports other support from Epithany, as well as grants from Pfizer, Precigen, Bavarian Nordisk, EMD Serono, and Veanna outside the submitted work; in addition, M.L. Disis has a patent for University of Washington issued to Epithany. No disclosures were reported by the other authors.

D.L. Cecil: Formal analysis, investigation, methodology, writing–original draft, project administration, writing–review and editing. J.B. Liao: Resources, supervision, investigation. Y. Dang: Investigation. A.L. Coveler: Resources, investigation. A. Kask: Data curation, investigation. Y. Yang: Data curation, investigation. J.S. Childs: Resources, data curation, investigation. D.M. Higgins: Resources, investigation. M.L. Disis: Conceptualization, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

The work was supported by NIH-P50 CA083636 (funded the conduct of the study, as well as collection, management, analysis, and interpretation of the data), NIH-UL1 TR002319 (supported the facility where the patients were treated during the conduct of the study), the Helen B. Slonaker Endowed Professor for Cancer Research for M.L. Disis (supported data analysis and manuscript preparation), and The Ovarian Cancer Research Program, Early Investigator Award, OC130304 and Department of Defense, W81XWH-14–1-0161 for J.B. Liao (supported the clinical conduct of the study, data analysis, and manuscript preparation).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.