Phase II Study of Personalized Peptide Vaccination for Previously Treated Advanced Colorectal Cancer

Colorectal cancer is one of the major causes of cancer-related death in the world. Although recent advances in chemotherapy and/or targeted therapy have helped to improve the clinical outcomes of patients with advanced colorectal cancer (aCRC), the prognosis still remains poor (1). Therefore, development of new therapeutic approaches, including immunotherapy, would be highly desirable. However, limited numbers of clinical trials of immunotherapies have been reported for patients with aCRC (2, 3).

We have developed a novel approach of cancer immunotherapy, named personalized peptide vaccination (PPV), in which vaccine peptides were selected from 31 cytotoxic T lymphocyte (CTL) epitope peptides derived from 15 tumor-associated antigens (TAA), based on both HLA class I types and preexisting host immunity (4, 5). Recently conducted clinical trials of PPV for patients with various types of cancers demonstrated the feasibility of this new approach (4–7). For patients with aCRC, phase I studies showed the safety and immunogenicity of PPV combined with chemotherapeutic agents, along with possible prolongation of survival time in immunologic responders (8, 9). We conducted a phase II study to examine the feasibility of PPV and to identify biomarkers that would be useful for prediction of overall survival (OS) in previously treated patients with aCRC.

Previously treated patients with aCRC, who had failed at least one regimen of standard chemotherapy and/or targeted therapy, were eligible for inclusion in this study, if they had positive humoral responses as determined by the peptide-specific IgG titers to at least two of the 31 different candidate vaccine peptides (Supplementary Table S1; refs. 4–9). Other inclusion and exclusion criteria are shown in Supplementary Methods. The protocol was approved by the Kurume University Ethics Committee and was registered in the University Hospital Medical Information Network (UMIN) Clinical Trials Registry (UMIN000006493). After a full explanation of the protocol, written informed consent was obtained from all patients before enrollment.

This was an open-label phase II study in which the endpoints were to analyze the clinical feasibility and safety of PPV and to identify biomarkers useful for prediction of OS after PPV in patients with aCRC. Thirty-one vaccine peptide candidates, whose safety and immunologic effects had been confirmed in clinical studies conducted previously (4–9), were prepared under the conditions of Good Manufacturing Practice (GMP) by the PolyPeptide Laboratories and American Peptide Company. Expressions of vaccine antigens in colorectal cancer tissues were examined by immunohistochemistry (Supplementary Fig. S1). Of the 15 vaccine antigens used for PPV, 13 were detectable in colorectal cancer tissues tested, but not the two prostate-related antigens (PSA and PSMA; Supplementary Table S1).

The protocol consisted of two cycles of six vaccinations. Two to four HLA-matched peptides were selected from the 31 peptides in individual patients, based on preexisting host immunity before vaccination by assessing the titers of IgG specific to each peptide, as described previously (4–9). The peptides derived from PSA and PSMA were selected only when preexisting IgG responses to other remaining peptides were absent. The selected peptides (3 mg/each peptide) were administered subcutaneously with incomplete Freund adjuvant (Montanide ISA51; Seppic) once a week for 6 consecutive weeks. After the completion of the first cycle of six vaccinations, IgG titers specific to each of 31 peptide candidates in plasma from vaccinated patients were measured again, and two to four HLA-matched peptides with higher specific IgG titers were selected and administered six times every 2 weeks for the second vaccination cycle. After the second cycle, vaccinations were maintained, if the patients wished; two to four antigen peptides, which were reselected on the basis of the titers of peptide-specific IgG at every cycle of six vaccinations, were administered every 4 weeks until uncontrollable disease progression. Combined chemotherapies and/or targeted therapies were allowed during the vaccination period. Adverse events (AE) were monitored according to the National Cancer Institute Common Terminology Criteria for Adverse Events (NCI-CTCAE) version 4.0. Complete blood counts and serum biochemistry tests were performed before and after every six vaccinations.

Peripheral blood (30 mL) was obtained from the vaccinated patients before and after each cycle of six vaccinations. After centrifugation, plasma was separated and stored frozen until analysis. Peripheral blood mononuclear cells (PBMC) were separated by density gradient centrifugation with Ficoll-Paque Plus (GE Healthcare) and stored frozen until analysis. Postvaccination blood samples were available from 51 and 35 patients at the end of the first (6 vaccinations) and second (12 vaccinations) cycles, respectively.

Humoral immune responses specific to the vaccine peptides were determined by peptide-specific IgG titers using a bead-based multiplex assay with the Luminex 200 system (Luminex), as reported previously (10, 11). CTL responses specific to the vaccine peptides were evaluated by the IFNγ ELISPOT assay. The detailed procedures are shown in the Supplementary Methods. When spot numbers in response to specific peptides were significantly higher (P < 0.05 by Student t test) than those in response to the control peptides, antigen-specific CTL responses were shown as the differences between them (means of the triplicate samples).

Levels of C-reactive protein (CRP), serum-amyloid A (SAA), and IL6 in prevaccination plasma were examined by ELISA using kits from R&D Systems, Life Technologies, and eBioscience, respectively. Bead-based multiplex assays were used to measure cytokines, including IL4, IL13, IL21, IP-10 (IFNγ-induced protein 10), BAFF (B-cell activating factor), and TGFβ, with the Luminex 200 system. Prevaccination plasma from 1 patient was unavailable for this analysis (n = 59). Frozen plasma samples were thawed, diluted, and assayed in duplicate in accordance with the manufacturer's instructions. Means of the duplicate samples were used for statistical analysis.

DNA was extracted from thawed PBMCs using a QIAamp Blood kit (Qiagen) and stored at −80°C until analysis. To investigate the IL6 −634G>C (rs1800796), CRP 1846C>T (rs1205), and IL6R 48892A>C (rs8192284, Asp358Ala) genetic polymorphisms with the extracted DNA, genotyping was performed using the polymerase chain reaction-restriction fragment length polymorphism method, as reported previously (12, 13).

OS time was defined as duration from the first date of peptide vaccination or that of the first-line chemotherapy until the date of death and was censored by the last date of contact for patients alive at the last follow-up. The survival function, including survival rates, for OS was estimated by the Kaplan–Meier method with the Greenwood variance estimates. In addition, exploratory analyses, which were not predefined in the protocol, were performed to examine association among biomarkers, immune responses, and OS. Association between prevaccination biomarkers and OS were evaluated by univariate and multivariate analyses with the Cox proportional hazards regression model. In applying Cox regression, the transformation of log(biomarker + 1) was used because the distribution of each biomarker was highly skewed. Statistically significant biomarkers (P < 0.1) in the univariate analysis were included in the multivariate analysis. The Spearman rank correlation among these biomarkers was estimated to avoid collinearity.

Humoral and cellular immune responses were determined by IgG and CTL responses specific to the administered peptides, respectively. IgG responses were defined as positive if IgG titers specific to at least one of the administered peptides in the postvaccination plasma were more than two times higher than those in the prevaccination plasma, and as negative otherwise. CTL responses were defined as positive if CTL responses to at least one of the administered peptides in the postvaccination PBMCs were greater than those in the prevaccination PBMCs and as negative otherwise. Association between IgG or CTL responses and other prognostic factors was examined by logistic regression analysis. Association between IgG or CTL responses and OS was examined by the Kaplan–Meier method with the log-rank test and the Cox regression analysis. The relationship between IgG and CTL responses was evaluated by the χ² test. The prognostic significance of genetic polymorphisms was analyzed by the Kaplan–Meier survival curves with the log-rank test. All statistical tests were conducted at the two-sided 5% significance level, unless indicated. Because of the exploratory nature of biomarker analyses, any multiplicity adjustment was not applied. All statistical analyses were conducted using the JMP version 10 or SAS version 9.3 software package (SAS Institute Inc.).

Between January 2009 and November 2012, 60 patients with aCRC were enrolled in this study. Table 1 summarizes the clinicopathologic characteristics of the enrolled patients. There were 33 male and 27 female subjects with a median age of 60 years, ranging from 35 to 83 years. All patients (stage IV, n = 26; recurrent, n = 34) were refractory to at least one regimen of chemotherapies and/or targeted therapies. The location of original tumor was right-sided colon (n = 14) or left-sided colon/rectum (n = 46). All patients had metastatic tumors; liver (n = 33), lung (n = 31), peritoneal dissemination (n = 23), or lymph nodes (n = 14). The number of metastatic organs per patient was one (n = 29), two (n = 21), or three (n = 10). Before enrollment, the patients had failed to respond to one (n = 17), two (n = 15), three (n = 9), four (n = 13), or five (n = 6) regimens of chemotherapies, targeted therapies, and/or combinations of them. The median duration of these preceding regimens before PPV was 552.5 days, ranging from 9 to 1,819 days. The median time from patient enrolment to first vaccination was 13.5 days, ranging from 7 to 27 days. The numbers of peptides used for vaccination during the first cycle were four peptides in 36 patients, three in 16 patients, and two in 8 patients. Among the 60 patients, 51 (85%) completed the first cycle of six vaccinations, and the remaining 9 patients failed to do so due to rapid disease progression. The median number of vaccinations was 12, with a range of 2 to 33. During the PPV, 49 patients (82%) received combined chemotherapies and/or targeted therapies, including FOLFOX/XELOX with bevacizumab (n = 10), FOLFIRI with bevacizumab (n = 5), FOLFIRI (n = 5), S-1 (n = 5), irinotecan with cetuximab (n = 5), cetuximab (n = 5), FOLFOX/XELOX (n = 2), FOLFIRI with cetuximab (n = 2), or other regimens (n = 10). The remaining 11 patients (18%) had no options for combined chemotherapies or were unable to tolerate them.

Toxicities are shown in Supplementary Table S2. The most frequent AEs were dermatologic reactions at the injection sites (n = 55; 92%), anemia (n = 27; 45%), lymphopenia (n = 23; 38%), and hypoalbuminemia (n = 20; 33%). Grade 4 anemia was noted in 2 patients. Grade 3 serious AEs (SAE) comprised leukocytopenia (n = 3), lymphopenia (n = 2), increased γ-glutamyl transpeptidase (GGT; n = 2), hyponatremia (n = 2), ileus (n = 2), increased aspartate aminotransferase (AST; n = 1), hyperglycemia (n = 1), hypercholesteremia (n = 1), and rash (n = 1). However, according to the evaluation by the independent safety evaluation committee for this trial, all the grade 3 or 4 SAEs were concluded to be not directly associated with the vaccinations, but with other causes, such as combined chemotherapies and/or targeted therapies and cancer progression.

Median OS time (MST) for the 60 patients from the first vaccination was 498 days [95% confidence interval (CI), 233–654 days] with 1- and 2-year survival rates of 53% and 22%, respectively (Fig. 1A). When calculated from the first date of the first-line chemotherapy, MST was 1,179 days (95% CI, 885–1,272 days) with 1-, 2-, 3-, 4-, and 5-year survival rates of 97%, 77%, 53%, 24%, and 15%, respectively (data not shown). Of note, among the enrolled 60 patients, 32 patients, who had a treatment history of two or more regimens of standard chemotherapy and were refractory or intolerant to all of irinotecan, oxaliplatin, and fluoropyrimidines before enrollment, showed MST of 375 days (95% CI, 191–561 days) from the first vaccination, with 1-year survival rate of 51% (Fig. 1B).

The Cox proportional hazards model was used to identify factors that were significantly associated with OS, from prevaccination clinical findings or laboratory data. As shown in Table 2, univariate analysis using prevaccination clinical findings showed that the number of previous chemotherapy regimens were potentially prognostic factors (P = 0.067). In addition, albumin, carcinoembryonic antigen (CEA), CRP, SAA, IL6, IP-10, and BAFF in prevaccination blood were significantly prognostic of OS by univariate analysis (P = 0.012, P = 0.002, P <0.001, P < 0.001, P < 0.001, P = 0.018, and P = 0.005, respectively). However, none of the other factors examined were significantly correlated with OS.

Multivariate Cox regression analysis was performed to evaluate the influence of each of the factors that had been shown to be significantly associated with OS in the univariate analysis (P < 0.1). SAA and CRP were not included in this analysis, because the level of SAA and CRP was highly correlated with that of IL6 (SAA vs. IL6: Spearman rank correlation coefficient = 0.482; CRP vs. IL6: Spearman rank correlation coefficient = 0.653). As shown in Table 2, higher IL6 and IP-10 levels and a lower BAFF level in prevaccination plasma were significantly predictive of unfavorable OS [hazard ratio (HR) for the unit of 1 SD, 1.508, 95% CI, 1.014–2.245, P = 0.043; HR, 1.579, 95% CI, 1.062–2.347, P = 0.024; HR, 0.509, 95% CI, 0.329–0.787, P = 0.002, respectively]. The other factors showed no statistically significant association.

Because inflammation markers, IL6 and CRP, were potentially prognostic in patients treated with PPV, we examined genetic polymorphisms of related genes, IL6 −634G>C, CRP 1846C>T, and IL6R 48892A>C (Supplementary Table S3). There was no statistically significant relationship between IL6 634G>C polymorphism and OS (P = 0.319). However, CRP 1846C>T and IL6R 48892A>C polymorphisms tended to show a statistically significant effect on OS (P = 0.069 and 0.085, respectively). Patients carrying the CRP 1846C/C genotype had a potentially better prognosis than those carrying the CRP 1846C/T or those carrying the CRP 1846T/T genotype (P = 0.029 or 0.054, respectively; Fig. 2A). In addition, patients carrying the IL6R 48892C/C or 48892A/C genotypes tended to show a better prognosis than those carrying the IL6R 48892A/A genotype (P = 0.059; Fig. 2B). This genetic polymorphism was further evaluated in patients positive or negative for IL6 in prevaccination plasma (Fig. 2C). Of note, the difference between patients carrying the IL6R 48892C/C or A/C genotypes and the IL6R 48892A/A genotype was statistically significant in patients negative for plasma IL6 (P = 0.025), but not in those positive for plasma IL6 (P = 0.118).

IgG responses specific to at least one of the administered peptides were increased in 25 of 51 patients (49%) and in 33 of 35 patients (94%) at the end of the first and second cycles of vaccinations, respectively (Supplementary Table S4). CTL responses specific to at least one of the administered peptides that were evaluated by IFNγ ELISPOT assay were increased in 32 of 51 patients (63%) at the end of the first cycle of vaccinations (Supplementary Table S4). A representative result of IFNγ ELISPOT assay with PBMCs before and after vaccination is shown in Fig. 3A. According to the χ² test, increased CTL responses against administered peptides after the first cycle of vaccinations were significantly associated with increased IgG responses (P = 0.002).

Relationship between the increase in peptide-specific CTL or IgG responses after vaccination and other potential prognostic factors, including prevaccination IL6, IP-10, and BAFF levels (Table 2), were examined by logistic regression analysis. As shown in Table 3, the level of IP-10 was predictive of the increase in CTL and IgG responses (OR, 0.427; 95% CI, 0.191–0.957; P = 0.039; OR, 0.354; 95% CI, 0.127–0.982; P = 0.046; respectively), whereas other factors, including IL6 and BAFF levels, were not predictive.

The prognostic significance of successful boosting of peptide-specific CTL or IgG responses was analyzed by the Kaplan–Meier survival curves with the log-rank test. This analysis showed a statistically significant association between increased CTL or IgG responses and OS (P = 0.025 and 0.022, respectively; Fig. 3B and C). Patients with both CTL and IgG responses (P = 0.010), but not those with CTL responses alone (P = 0.138) or IgG responses alone (P = 0.351), showed significantly better prognosis than those without CTL or IgG responses (Supplementary Fig. S2).

In addition, multivariate Cox regression analysis with peptide-specific CTL or IgG responses (positive or negative) and other potential prognostic factors (Table 2) was performed. IP-10 was not included in this analysis because the CTL and IgG responses were significantly associated with plasma IP-10 level (Table 3). As shown in Table 4, increased CTL responses after vaccination were significantly associated with favorable OS (HR, 0.231; 95% CI, 0.067–0.803; P = 0.021) independently of other factors, whereas IgG responses after vaccination were not significantly predictive of favorable OS (HR, 0.790; 95% CI, 0.285–2.188; P = 0.650). Furthermore, to analyze association of the magnitude of CTL responses with OS, the number of peptides, to which CTL responses were increased after vaccination, was evaluated by multivariate analysis. As shown in Supplementary Table S5, the number of peptides with increased CTL responses after vaccination was also significantly predictive of favorable OS (HR, 0.216; 95% CI, 0.077–0.604; P = 0.004).

In this study, we demonstrated that successful boosting of peptide-specific CTL responses resulted in increased OS after PPV, suggesting a potential clinical benefit of PPV. The most unique aspect of PPV is the personalized selection of optimal antigen peptides for individual patients on the basis of preexisting host immunity before vaccination (4, 5). In view of the heterogeneity of tumors and the complexity and diversity of immune responses, we thought that this approach would be more rational than selecting nonpersonalized universal tumor antigens. Because tumor tissues were unavailable in most patients with aCRC, it was difficult to precisely characterize tumor cells in individual patients. Therefore, we selected and administered multiple (up to four) antigens to increase the possibility that the antigens used for vaccination were expressed in tumor cells.

We currently measure preexisting antigen-specific IgG responses, but not T-cell responses, for personalized selection of antigen peptides from a panel of candidate antigens, because antigen-specific T-cell assays often show limited sensitivity due to quite low frequencies of antigen-specific T cells before vaccinations, even after in vitro cell culture for expansion. Indeed, if the preexisting CTL responses in prevaccination PBMCs were used for selection of peptides in this study, much smaller numbers of peptides would be selected for vaccination (Supplementary Table S4). In contrast, the multiplex bead–based LUMINEX technology allows high-throughput screening of IgG responses specific to large numbers of peptide antigens with high accuracy (10, 11). Our previous studies suggested the clinical significance of antigen-specific IgG responses as a surrogate biomarker in monitoring vaccine-induced immune responses (14). In addition, this study demonstrated that increased IgG responses against administered peptides after vaccination were significantly associated with increased CTL responses. These results support our hypothesis that evaluation of IgG responses might be useful for predicting peptides that could induce specific CTL responses.

Because the vaccine peptides used for PPV are HLA-restricted CTL epitopes, they might act mainly through peptide-specific CTL responses. Indeed, peptide-specific CTL responses were significantly associated with OS (Table 4). Nevertheless, IgG responses to the vaccine peptides might also affect antitumor immunity. For example, in our preliminary study in mice, antibody complex with specific peptides facilitated the uptake of peptides and enhanced the cross-presentation of these peptides by antigen-presenting cells (S. Matsueda and colleagues; unpublished data). Further studies are currently in progress for clarification of the biologic functions of peptide-specific IgG.

Because not all patients show clinical benefits from cancer immunotherapies, it would be critical to identify prognostic or predictive biomarkers for patients receiving such therapies. Several postvaccination biomarkers have been reported to be associated with clinical responses (14–18), but there are currently no validated prevaccination predictive biomarkers. By multivariate analysis, higher IL6 and IP-10 and lower BAFF levels in prevaccination plasma were significantly associated with unfavorable OS, although these factors might be prognostic irrespective of treatment, and not necessarily predictive and unique to PPV. Of note, however, the IP-10 level was predictive of the increase in CTL responses, which was associated with improved OS, suggesting that IP-10 might be potentially useful for selecting patients with aCRC, who would benefit from PPV. To more clearly assess the causal relation of IP-10, CTL responses, and OS, and to elucidate prognostic versus predictive relevance of such biomarkers, future randomized, controlled clinical trials with or without PPV would be essential.

IL6 has been reported to induce suppressive immune cell subsets, such as myeloid-derived suppressor cells and Th17 cells (19–22). Therefore, high levels of IL6 might inhibit immune responses to cancer vaccines by inducing these suppressive cells. BAFF is a cytokine for the differentiation and survival of follicular B cells along with humoral response potentiation (23). As previously suggested (24–26), BAFF might induce beneficial humoral immune responses to vaccine antigens. IP-10 is a chemokine for attraction of human monocytes, activated T cells, and NK cells (27, 28). Although local production of IP-10 within tumor tissues has been reported to be associated with antitumor immunity, systemic inflammatory responses mediated by IP-10 might contribute to poorer immune responses to vaccines (27, 28). The precise mechanisms of IL6, BAFF, and IP-10 in immune responses after PPV remain to be determined.

Results from this study suggested that the CRP 1846C>T and IL6R 48892A>C polymorphisms might show a statistically significant effect on OS after PPV. Because the CRP 1846C>T polymorphism, which affects serum CRP levels (29), has been reported to be associated with advanced diseases in patients with colorectal cancer (30) and esophageal squamous cell carcinoma (13), it might be a prognostic factor irrespective of the therapeutic approach. In contrast, because the IL6R 48892A>C polymorphism has been reported to show no effects on prognosis in some types of cancers, such as esophageal squamous cell carcinoma and neuroblastoma, without cancer vaccines (12, 31), the prognostic significance of this polymorphism might be unique to PPV-vaccinated patients. The IL6R 48892C (358Ala) allele has been reported to affect proteolytic cleavage of the membrane-bound IL6R, leading to reduced numbers of the functioning IL6R (32). As a result, this genetic variant is suggested to contribute to anti-inflammatory effect through attenuation of IL6 signaling on cells expressing the membrane-bound IL6R (33–35). On the basis of our finding, the effect of reduced IL6R expression might be more prominent when the availability of IL6 is limited, whereas it might be overcome by overexpression of IL6.

Importantly, this study demonstrated that successful boosting of peptide-specific CTL responses was significantly predictive of favorable OS by multivariate analysis, suggesting a causal relationship between biologic and clinical efficacy of PPV. However, peptide-specific IgG responses were not significantly predictive of OS by multivariate analysis, although they were significantly associated with favorable OS by the Kaplan–Meier method with the log-rank test. This discrepancy might be explained by the speculation that IgG responses might be more strongly affected by other confounding factors, such as IL6 and BAFF, compared with CTL responses. Because IL6 and BAFF are known to play important roles in the differentiation and survival of B cells along with humoral response potentiation (19, 23), it is possible that they substantially affected IgG responses, but not CTL responses, after vaccination.

In summary, this study demonstrated that PPV-induced substantial immune responses to vaccine antigens without severe adverse events and showed potential clinical benefits in previously treated patients with aCRC, even in the refractory stage. Nevertheless, this study has several drawbacks. First, this is a small study with a limited number of patients, all of whom received PPV. Second, combined chemotherapies and/or targeted therapies during the vaccination period might affect the occurrence of immune responses and conclusion about the prognostic versus the predictive role of biomarkers. Therefore, clinical efficacy of PPV, as well as clinical utility of the identified biomarkers, in patients with aCRC remain to be confirmed in future larger scale, randomized trials of PPV without combined chemotherapies or targeted therapies.

T. Nomura is an employee of the Kyowa Hakko Kirin Co., Ltd. A. Yamada is a board member for Green Peptide Co., Ltd. K. Itoh reports receiving commercial research support from Taiho Pharmaceutical Co., Ltd., for which he has also received speakers bureau honoraria and is a consultant/advisory board member; he also has ownership interest (including patents) in Green Peptide Co., Ltd. No potential conflicts of interest were disclosed by the other authors.

Conception and design: S. Kibe, S. Yutani, K. Itoh, T. Sasada

Development of methodology: S. Yutani, K. Itoh

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Kibe, S. Yutani, S. Motoyama, N. Tanaka, A. Kawahara, N. Komatsu, M. Miura, K. Itoh, Y. Akagi, T. Sasada

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S. Kibe, T. Nomura, M. Miura, Y. Hinai, S. Hattori, M. Kage, T. Sasada

Writing, review, and/or revision of the manuscript: S. Kibe, K. Itoh, T. Sasada

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): N. Tanaka, T. Yamaguchi, S. Matsueda, N. Komatsu, A. Yamada, T. Sasada

Study supervision: M. Kage, Y. Akagi

The authors thank Ms. Emi Muta and Ms. Chieko Seki (Kurume University, Kurume, Japan) for their technical assistance.

This study was supported by a research program of the Project for Development of Innovative Research on Cancer Therapeutics (P-Direct), Ministry of Education, Culture, Sports, Science and Technology of Japan, and a research program of the Regional Innovation Cluster Program of the Ministry of Education, Culture, Sports, Science and Technology of Japan.

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.