|Home | About | Journals | Submit | Contact Us | Français|
We previously reported a randomized phase II clinical trial combining a poxvirus-based vaccine encoding PSA with radiotherapy in patients with localized prostate cancer. Here we investigate whether vaccination against PSA induced immune responses to additional tumor-associated antigens and how this influenced clinical outcome.
Pre- and post-treatment serum samples from patients treated with vaccine + external beam radiation therapy (EBRT) versus EBRT alone were evaluated by Western blot and serological screening of a prostate cancer cDNA expression library (SEREX) to assess the development of treatment-associated autoantibody responses.
Western blotting revealed treatment-associated autoantibody responses in 15/33 (45.5%) patients treated with vaccine + EBRT versus 1/8 (12.5%) treated with EBRT alone. SEREX screening identified 18 antigens, which were assembled on an antigen array with 16 previously identified antigens. Antigen array screening revealed that seven of 33 patients (21.2%) treated with vaccine + EBRT demonstrated a vaccine-associated autoantibody response to four ubiquitously expressed self antigens: DIRC2, NDUFS1, MRFAP1 and MATN2. These responses were not seen in patients treated with EBRT alone, or other control groups. Patients with autoantibody responses to this panel of antigens had a trend towards decreased biochemical-free survival.
Vaccine + EBRT induced antigen spreading in a large proportion of patients. A subset of patients developed autoantibodies to a panel of four self antigens and showed a trend toward inferior outcomes. Thus, cancer vaccines directed against tumor-specific antigens can trigger autoantibody responses to self proteins, which may influence the efficacy of vaccination.
Prostate cancer is the most frequently diagnosed cancer in North American men and remains the second leading cause of cancer-related death in men despite the improvement of early detection using PSA screening. Choice of curative treatment for localized disease is challenging and is based on several factors, including stage and grade of disease, PSA value, co-morbidities and patient preference. In general, localized disease is treated with either surgery or radiation therapy (external beam or brachytherapy), with the addition of androgen-deprivation therapy (ADT) for higher risk patients. Unfortunately, about 35% of patients fail definitive local therapy within 10 years as evidenced by an increase in PSA, often due to occult metastatic disease (1). Most recurrences are treated with ADT, which while initially efficacious inevitably develops into castration-resistant disease. Although chemotherapy can provide modest benefit at this stage, recurrent cancers are incurable. Thus there is a clear need for more effective therapies for both primary and recurrent disease.
There is considerable evidence that the immune system recognizes prostate tumors in a significant proportion of patients. Serum antibody responses to a number of different antigens have been detected in patients with prostate cancer (2-9). The presence of tumor-infiltrating lymphocytes (TIL) provides additional evidence that prostate tumors are recognized by the immune system. Vesalainen et al. (10) found that tumors with lower densities of TIL were associated with higher risk of tumor progression and poor prognosis, however in another study, the presence of CD4+, CD8+ and CD20+ TIL in primary prostate tumors was associated with inferior outcomes (11). Thus the immune system responds to prostate tumors, but it is unclear whether these responses are beneficial or detrimental to patients.
Although much is known about immune responses at the time of diagnosis, far less is known about how tumor immunity changes when patients undergo treatment. We have recently shown that ADT and radiation therapy (RT) induce autoantibody responses to a variety of tumor-associated antigens in 17-30% of prostate cancer patients (12). Although we do not yet know the influence of these treatment-associated immune responses on clinical outcomes, we examined this issue in a mouse model. Using the androgen-dependent Shionogi carcinoma model, we found that castration (the experimental equivalent of ADT) induced autoantibody and T cell responses to the self antigen PABPN1 in about 50% of animals. Remarkably, mice that developed autoantibody and T cell responses to PABPN1 showed a higher rate and shorter latency of tumor recurrence (13). Thus, castration-induced autoantibody responses are associated with inferior outcomes in the Shionogi tumor model, consistent with a growing number of studies indicating that anti-tumor immunity can sometimes promote tumor growth (14-17).
While the foregoing discussion concerns the effects of standard treatments on tumor immunity, many groups are attempting to use vaccines to deliberately induce an immune response to prostate cancer. We conducted a clinical trial in which patients with localized prostate cancer were treated with a combination of external beam radiation therapy (EBRT) (with or without neoadjuvant ADT) and a poxvirus-based vaccine encoding PSA (18, 19). Initially, 30 patients were randomized to receive EBRT with PSA vaccine (vaccine + EBRT; n=19) or EBRT alone (n=11) (18). Patients in the vaccine arm also received granulocyte-macrophage colony stimulating factor (GM-CSF) and low-dose IL-2 as immunological adjuvants. The primary endpoint was immunological response specific to PSA. We found that 13 of 17 patients treated with vaccine + EBRT had significantly increased PSA-specific T cell responses compared to 0 of 8 patients treated with EBRT alone. However, most patients on the vaccine arm were not able to receive the full dose of IL-2 due to significant adverse events. Therefore, we treated an additional 18 patients with vaccine + EBRT and lowered the dose of IL-2 from 4 MIU/M2/day for 5 days to 0.6 MIU/M2/day for 14 days (19). Use of very low dose IL-2 significantly reduced the number of adverse events while still promoting vaccine-induced T cell responses.
In addition to the significant increases in PSA-specific CD8+ T cells seen in the majority of vaccinated patients, there was also evidence of antigen spreading in 9 of 13 vaccinated patients, as manifested by the development of T cell responses to antigens not contained in the vaccine, including PSMA, PAP, PSCA, MUC-1, XAGE-1 and PAGE-4 (18, 19). Such responses were also seen in 1 of 6 evaluated patients receiving EBRT alone. These positive results with a limited set of known prostate cancer antigens lead us to hypothesize that antigen spreading may have extended to a broader repertoire of prostate cancer antigens. To test this hypothesis, we used Western blotting and SEREX immunoscreening to broadly assess the development of autoantibody responses to prostate antigens in patients from this trial.
This study included 41 prostate cancer patients from a clinical trial (00-C-0154) conducted at the National Cancer Institute (18, 19). The clinical trial protocol was approved by the Institutional Review Board of the National Cancer Institute, and the current immunological study was approved by the University of British Columbia/British Columbia Cancer Agency (UBC/BCCA) Research Ethics Board. On this trial, men with previously untreated, biopsy-confirmed prostate cancer were treated with EBRT or vaccine + EBRT. ADT was also given to most patients (n=34) at the discretion of the attending radiation oncologist. The first vaccination cycle consisted of a priming dose of vaccinia PSA admixed with vaccinia B7.1, given on Day 2. Patients then received seven boosts with fowlpox PSA on a 28-day cycle. GM-CSF was given at 100 μg/d at the vaccination site on days 1 to 4 of each cycle. IL-2 was given s.c. in the abdomen either at 4 MIU/M2 on days 8 to 12 (n=17) or at 0.6 MIU/M2 on days 8 to 21 (n=16). EBRT was administered after the third cycle of vaccine and consisted of a total dose of 70 Gy, with 1.8 – 2.0 Gy fractions given over 2 months (over cycles 4 and 5 of vaccine). In the vaccine arm, blood samples were collected before vaccine treatment was initiated, after every vaccine cycle and after all treatment was completed. In the EBRT arm, blood samples were collected before EBRT, immediately after, and 3 months after completion of EBRT.
An additional 24 prostate cancer patients receiving standard treatment at the BC Cancer Agency in Victoria, BC, Canada served as controls. Of these, 15 patients were treated with ADT + EBRT, while nine patients chose active surveillance. Fifteen age-matched men with no personal history of prostate cancer served as an additional control group. Blood samples were collected with informed consent and approval from the UBC/BCCA Research Ethics Board. Patient characteristics and treatment categories are summarized in Table 1.
Western blots were performed as previously described (12). Briefly, 400 μg of protein isolated from LNCaP and PC3 cells was separated by standard PAGE and transferred to nitrocellulose. Sera were diluted 1/500 in Blotto (5% dry milk powder; 0.1% Tween 20; 50 mmol/L Tris; 150 mmol/L NaCl) and incubated for 1 h at room temperature using the Mini Protean II MultiScreen multichannel immunoblotting device (BioRad, Mississauga, Canada). The membrane was then incubated for 1 h at room temperature with HRP-conjugated goat anti-human IgG secondary antibody (H+L; Jackson ImmunoResearch, West Grove, PA), diluted to 1/10000 and visualized by enhanced chemiluminescence. Pre- and post-treatment sera were run against both LNCaP and PC3 protein lysate a minimum of two times each.
SEREX screening of a previously described prostate cancer phage cDNA expression library (12) was performed using post-treatment serum samples from six patients who showed treatment-associated serological changes by Western blot. To make screening more efficient, sera were pooled in pairs, with each serum represented at a 1/200 dilution. Approximately 3.0-4.8 × 104 clones were screened with each patient serum. Screening of the prostate cancer library and antigen arrays was carried out as previously described (12, 20). Antigen arrays were screened with a 1/200 dilution of pre- and post-treatment serum from the 33 vaccine + EBRT, 8 EBRT without vaccine, 15 ADT+EBRT and 9 active surveillance prostate cancer patients as well as 2 serial serum samples from 15 cancer-free controls.
To broadly assess vaccine-induced immune responses, patient sera were first analyzed by Western blot against lysates from the prostate cancer cell lines LNCaP and PC3, as per Nesslinger et al. (12, 21). Two groups of patients were assessed: those who received vaccine + EBRT (n=33), and those who received EBRT without vaccine (n=8). Of the patients receiving vaccine + EBRT, 15 of 33 (45.5%) demonstrated a serological change when comparing the pre-vaccine and post-vaccine serum samples (Figure 1). In contrast, only 1 of 8 (12.5%) patients who received EBRT without vaccine showed a serological change during EBRT, in accord with our previous study of an independent cohort of patients receiving EBRT (12). The difference between the vaccine + EBRT versus EBRT without vaccine groups was not statistically significant (p=0.12, Fisher’s exact test), likely due to the low number of patients in the EBRT without vaccine group. Indeed, if we increase the sample size of the EBRT without vaccine group by including patients from our prior study (12), the trend reaches significance (p=0.003, Fisher’s exact test; 15 of 33 vaccine + EBRT versus 5 of 40 EBRT without vaccine). Thus, vaccination appears to induce serological changes beyond those induced by EBRT.
A wide variety of antigens appeared to underlie the vaccine-associated serological changes, as evidenced by the diversity of protein sizes seen on Western blot (Figure 1). While the majority of serological changes were seen with both LNCaP and PC3 protein lysates, four serological changes were seen with one protein lysate but not the other, indicating these two cell lines have somewhat distinct antigen repertoires.
To identify the antigens underlying the above serological changes, we performed SEREX immunoscreening of a cDNA expression library derived from the prostate cancer cell lines LNCaP, PC3 and DU-145 (12). The library was screened with post-treatment serum samples from six patients who had demonstrated vaccine-associated serological changes by Western blot (patients NIH-04, NIH-32, NIH-33, NIH-37, NIH-43, NIH-10). Eighteen antigens were identified, representing a wide range of structural and functional protein classes (Supplemental Table 1). To assess serological responses to these antigens across the entire patient cohort, SEREX antigen arrays were constructed containing these 18 antigens plus 16 antigens derived from previous SEREX screens of the same prostate cancer library, a cancer-testis library and an ovarian cancer library (12) (Supplemental Table 1). Antigen arrays were screened with pre-treatment serum (serum collected before initiation of vaccine or EBRT) and post-treatment serum (i.e., serum collected after completion of all 8 cycles of vaccine and/or EBRT) from all patients from the clinical trial (33 vaccine + EBRT patients, and 8 EBRT without vaccine patients). Arrays were also screened with pre- and post-treatment sera from an additional group of patients treated with standard ADT + EBRT. Finally, as additional controls, antigen arrays were screened with serial serum samples from prostate cancer patients undergoing active surveillance (n = 9), as well as age-matched men with no personal history of prostate cancer (n = 15).
This analysis revealed several patterns of autoantibody response. First, some antigens were recognized at baseline (i.e., pre-treatment) by multiple cancer patients and cancer-free controls, suggesting they represent normal autoantigens (Supplemental Table 2). Second, some antigens demonstrated baseline responses in prostate cancer patients only (Supplemental Table 2). Third, and of greatest interest, some antigens were associated with treatment-induced autoantibody responses. Specifically, 7 of 33 (21.2%) patients in the vaccine + EBRT group (compared to 0 of 8 patients in the EBRT without vaccine group) showed treatment-induced autoantibody responses by antigen array (Table 2). Of these seven patients, five had shown a response by Western blot in the preceding section, whereas two had not. Thus, when the results of the Western blot and the antigen array were combined, the total number of treatment-induced autoantibody responses was 17/33 (51.5%) in the vaccine + EBRT group and 1 of 8 (12.5%) in the EBRT without vaccine group (p=0.059, Fisher’s exact test; Table 2). Again, this reached statistical significance when we included additional EBRT without vaccine patients from our previous study (12); (17/33 vaccine + EBRT versus 6/40 EBRT without vaccine; p=0.0009, Fisher’s exact test).
Remarkably, all seven patients from the vaccine + EBRT group who demonstrated a treatment-induced autoantibody response recognized the same set of four antigens; DIRC2, NDUFS1, MRFAP1 and MATN2 (Figures (Figures22 and and3).3). These serological responses were unique to vaccinated patients, as they were not seen in any other treatment or control group. This is the first time in our experience that multiple individuals demonstrated treatment-induced responses to a common set of antigens; therefore, we hypothesized these responses may have been induced by vaccination. Indeed, analysis of serial serum samples revealed that, in all seven patients, autoantibody responses to DIRC2, NDUFS1, MRFAP1 and MATN2 arose after 3 cycles of vaccination but prior to EBRT (Figure 2B). Thus, this vaccine regimen induces autoantibody responses to a common set of autoantigens in a significant proportion of patients.
In an attempt to explain why patients might demonstrate autoantibody responses to a common set of antigens, we considered the possibility that the observed autoantibodies might be cross-reactive. To assess this, we performed sequence homology comparisons between the four antigens, as well as the viral vectors contained in the priming and boosting vaccines. mRNA and protein sequences for DIRC2, NDUFS1, MRFAP1, MATN2, PSA, B7.1, GM-CSF and IL-2 were obtained from the GenBank database, as were the complete nucleotide sequences of the vaccinia and fowlpox virus genomes. The sequence alignment program ClustalW2 was used to cross-compare all of these sequences (22). There were stretches of homology between the viral vectors and each of the antigens; however, these were short (less than 8 nucleotides) and were not shared between the antigens. Similarly there was little homology between the four antigens and the vaccine components at either the nucleotide or amino acid level (less than 8 nucleotides or 3 amino acids). Alignment between the four antigens themselves revealed no homology at the nucleotide or amino acid level, making it unlikely that the observed autoantibody responses resulted from cross-reactivity.
In our prior study, we reported that 9/13 vaccinated patients from this clinical trial had shown evidence of antigen spreading, as manifested by the development of T cell responses to prostate-associated antigens not contained in the vaccine, including PSMA, PAP, PSCA, MUC-1, XAGE-1 and PAGE-4 (18, 19). Of these 9 patients, 3 also demonstrated treatment-associated autoantibody responses by either Western blot or antigen array in the present study. Conversely, of the 17 patients who demonstrated autoantibody responses, 3 also demonstrated antigen spreading by T cell assays. Thus, although there was some overlap between autoantibody and T cell responses, these appeared to be largely independent phenomena (Supplemental Table 3).
We considered whether the seven patients who demonstrated treatment-induced responses against the four antigens had any clinical characteristics in common. Their ages (range 51-73 years) and ADT use (5/7 patients) were typical of the entire cohort. Five of the patients were treated with IL-2 at the 4 MIU/M2 dose while the remaining two received the very low 0.6 MIU/M2 dose. Finally, the seven patients showed no unusual features with respect to the severity of their prostate cancer: Gleason scores ranged from 6 to 8 (median 6.5); stages ranged from T1c to T3b; risk stratification ranged from low to high; and PSA values at diagnosis ranged from 5.7 to 206 (Supplemental Table 4). Thus, these patients had no obvious clinical characteristics that distinguished them from the rest of the cohort.
Kaplan-Meier analysis was performed to compare the biochemical-free survival of patients in the vaccine + EBRT group who did or did not demonstrate treatment-induced autoantibody responses by Western blot or antigen array. There was no trend or significant difference between the two groups when autoantibody responses to all antigens were considered (p=0.9537, log rank test; Figure 4A). Similarly, there was no significant difference between those patients that had a positive or negative T cell response to PSA and biochemical-free survival (p= 0.5339, log rank test). However, the seven patients who demonstrated autoantibody responses to the panel of four common antigens (DIRC2, NDUFS1, MRFAP1 and MATN2) showed a trend toward decreased survival (p=0.1136, log rank test; Figure 4B). Indeed, 4/7 of these patients demonstrated biochemical failure, compared to only 1/8 non-vaccinated patients. While not statistically significant, these results suggest that vaccination might not only be ineffective but actually detrimental to patients in those cases in which autoantibody responses to ubiquitous antigens are induced.
The use of vaccines to augment the immune response to cancer has been the focus of intense investigation for several decades. Although there have been occasional clinical successes, overall results have been disappointing. The reason for vaccine failure is not known in most cases; however, known barriers to success include tolerance to self antigens, generation of antigen-loss variants, inadequate immunological danger signals and expression of immunosuppressive factors in the tumor microenvironment (23, 24). This study focused on patients with localized prostate cancer who were treated with a recombinant poxvirus-based vaccine containing PSA in combination with radiation therapy (18, 19). By serological analyses, we found that 17 of 33 (51.5%) of patients developed autoantibody responses after vaccination. Remarkably, seven of these patients developed autoantibody responses to a common set of four antigens. This phenomenon appeared to be unique to vaccination, as it was not observed in patients receiving EBRT without vaccine in this study or our previous study (12). Patients who developed an autoantibody response to the panel of four antigens showed a trend toward reduced biochemical-free survival, suggesting the vaccine could potentially have induced a detrimental immune response in these patients.
A priori, autoantibodies could play either a beneficial or detrimental role in cancer depending on a number of factors, including the particular antigen. For example, one could imagine that autoantibody responses to tumor-specific antigens might be beneficial, as they could specifically target tumor cells for destruction by complement or antibody-dependent cellular cytotoxicity. By contrast, autoantibody responses to broadly expressed self antigens (as reported here) could potentially be detrimental, as they might provoke or reflect tolerizing immune responses that serve to minimize autoimmune damage. In reality, many tumor antigens lie somewhere between “tumor-specific” and “self” antigens, and there is no clear consensus on their association with clinical outcomes. The significance of autoantibodies against p53 has been most extensively studied, with some studies finding a positive correlation with clinical outcome (25, 26) and others a negative correlation with outcome (27-31). Autoantibodies to other antigens have been associated with improved prognosis in glioblastoma (32), melanoma (33), gastric cancer (34) and breast cancer (35). Although most of these studies examined the levels of autoantibodies at the time of diagnosis, some examined the development of autoantibodies during treatment. For example, in a clinical trial of melanoma patients treated with IFN-α2B, the development of autoantibodies and clinical signs of autoimmunity were associated with prolonged survival (36).
In the setting of prostate cancer, several groups have conducted Phase II clinical trials using GVAX, an allogeneic vaccine comprised of the LNCaP and PC3 prostate cancer cell lines engineered to secrete GM-CSF (37-39). By immunoblot analysis, vaccination induced the development of autoantibody responses to LNCaP or PC3 antigens in 79% of patients in one study (38) and 43%-89% of patients (depending on dose) in a second study (39). In general, these responses were directed against a diverse set of antigens. However, a subset of patients exhibited new or enhanced autoantibody responses to the self protein filamin-B, which was associated with favorable changes in PSA kinetics (37-39). By contrast, in the present study, the development of autoantibodies during vaccination appeared to be associated with inferior outcomes. Although this trend did not reach statistical significance, possibly due to insufficient sample size, it supports the idea that vaccine-induced autoantibody responses might be detrimental under some conditions.
There are several differences between the present study and the GVAX studies that could account for the different associations between autoantibodies and outcome. With the GVAX approach, autoantibodies are likely induced by a classic, direct immunization mechanism, since the cell-based vaccine presented a full complement of human proteins in the context of a strong adjuvant. Thus, the development of autoantibody responses may reflect a successful vaccination overall, hence the correlation to favorable outcomes. By contrast, with the viral vectors used in our trials, autoantibody responses likely developed by an indirect mechanism, since the target antigens were not contained within the vaccine. In this scenario, the autoantibody response might arise as a consequence of inflammation at the tumor site, secondary to T cell recognition of tumor cells expressing the target antigen PSA. One could further speculate that this indirect form of autoantibody induction might be associated with ineffective T cell responses, such as Th2 responses, rather than cytolytic Th1 responses, which are generally more effective against tumors.
An alternative explanation for the observed trend between autoantibody responses and inferior outcome is that tumors of higher malignant potential could potentially be more immunogenic. However, we saw no obvious differences between the relevant characteristics of patients who developed autoantibody responses to broadly expressed self antigens and the remainder of the patients (Supplemental Table 4).
The present results are reminiscent of our recent findings in the Shionogi mouse model (12, 13). The Shionogi tumor line is highly androgen-dependent and forms adenocarcinomas when implanted in male mice. Castration of host mice results in rapid regression of established tumors, similar to hormone therapy in human prostate cancer. In 50-75% of mice, castration-induced tumor regression is accompanied by the development of autoantibody and T cell responses to a self antigen we identified as Poly(A) Binding Protein N1 (PABPN1). Remarkably, mice that developed autoantibody and T cell responses to PABPN1 demonstrated inferior outcomes, as manifested by an increased rate of tumor recurrence (13). We are currently investigating whether B and/or T cell responses to PABPN1 actively promote tumor recurrence in this model, or simply serve as markers of other biological processes associated with tumor recurrence. In support of the former possibility, it has recently been shown in a mouse model of prostate cancer that androgen ablation results in infiltration of regressing tumors with leukocytes, including B cells (40). Importantly, tumor-infiltrating B cells were shown to produce cytokines such as lymphotoxin that activate IKK-alpha and STAT3 in prostate tumor cells, thereby enhancing hormone-free survival and the transition to androgen-independence (40). Other studies have also demonstrated a pro-tumorigenic role for B cell responses. For example, B cells (or serum antibodies) were shown to promote primary tumor formation in a transgenic mouse model of inflammation-associated carcinogenesis (17). Furthermore, B cells were shown to inhibit T cell-mediated rejection of thymoma, melanoma and colon carcinoma cells in other murine studies (41, 42). Thus, there is abundant evidence that B cell responses can promote the development or progression of cancer.
It remains unclear why the antigens DIRC2, NDUFS1, MRFAP1 and MATN2 were common targets of autoantibody responses. According to public databases, all four proteins are ubiquitously expressed in most normal and tumor tissues and only DIRC2 appears to be over-expressed in prostate cancer tissue compared to normal prostate tissue. As mentioned, the four antigens do not share significant sequence homology with each other or with other coding sequences in the viral vectors. Nonetheless, it is formally possible that these antigens share conformational similarities that are not evident from their primary sequence. In our prior study of ovarian cancer, we found that autoantibody responses were biased toward gene products on chromosome 17 (20); however, the four antigens identified here are located on chromosomes 3, 2, 4 and 8, respectively. In addition, the four antigens have unique cellular functions. DIRC2 (Disrupted in renal carcinoma protein 2) is a membrane-bound protein with homology to members of the major facilitator superfamily of transporters. Intriguingly, it was identified as a breakpoint spanning gene on chromosome 3 in a constitutional familial case of a t(2;3)(q35;q21) translocation resulting in hereditary renal cell carcinoma (43). MRFAP1 (Mof4 family associated protein 1) is an intracellular protein that associates with members of the MORF4/MRG (Mortality Factor on Chromosome 4/MORF4 related gene) family as well as the tumor suppressor protein Rb and thus may play a role in cell growth, immortalization and/or senescence (44). MATN2 (Matrilin-2) is a secreted adaptor protein of the extracellular matrix that plays a role in cell growth and tissue remodeling (45). Interestingly, we initially cloned MATN2 from an ovarian cancer library using serum from an ovarian cancer patient, suggesting that autoantibodies to this antigen are not unique to prostate cancer. Finally, NDUFS1 (NADH dehydrogenase (ubiquinone) Fe-S protein 1, 75kDa (NADH-coenzyme Q reductase)) is a subunit of respiratory complex I located at the mitochondrial inner membrane. Intriguingly, caspase-mediated cleavage of NDUFS1 is required for several mitochondrial changes during apoptosis (46), which fits with the general notion that caspase-mediated cleavage of proteins can generate neo-epitopes that trigger autoantibody responses (47-49). We are currently investigating whether these four proteins serve as autoantigens in the context of other cancer vaccines.
In summary, our results provide further evidence that cancer vaccines targeting a single antigen such as PSA can induce broader immune responses involving multiple host antigens. While this process of antigen spreading is generally considered a desirable outcome of vaccination, our findings suggest this assumption may not be universally true. We propose that further studies involving other vaccine strategies are warranted. For example, we vaccinated patients in conjunction with EBRT and, in some cases ADT, both of which can contribute to the development of autoantibody responses, albeit not to the four antigens described here (12). Although the sample size is small, it is interesting to note that of the patients with treatment-induced immune response to the panel of antigens, all four patients who recurred had received prior ADT (Supplemental Table 4). We also included GM-CSF in the vaccine formulation, which several recent studies suggest might be a suboptimal adjuvant for cancer vaccines (50, 51). Finally, the use of IL-2 in the vaccine formulation may have played a role; notably, all four autoantibody-positive patients who recurred received the higher dose of IL-2 (Supplementary Table 4). To better understand the contribution of GM-CSF and IL-2 to the development of autoantibody responses, we intend to analyze samples from patients who received poxvirus-based vaccines without these adjuvants (52, 53). Going forward, if autoantibody responses are indeed detrimental to the efficacy of cancer vaccines, it may be possible to dampen or re-direct these responses using other immunomodulatory agents that promote cytolytic over humoral immune responses.
While cancer vaccines offer promise as a new form of treatment, the results of most cancer vaccine trials to date have been disappointing. To improve on these results, we require a better understanding of the impact of vaccination on the patient’s immune system. This report is focused on patients with localized prostate cancer who were treated as part of a Phase II clinical trial with a recombinant poxvirus-based vaccine targeting PSA. In addition to inducing immunity to PSA, as intended, vaccination induced autoantibody responses to a panel of widely expressed self antigens in a significant proportion of patients. Moreover, these vaccine-induced autoantibody responses appeared to be associated with inferior outcomes. Therefore, tumor-specific vaccines can induce immune responses to self proteins in patients, which we hypothesize may compromise therapeutic efficacy by provoking regulatory responses by the host to minimize autoimmune damage.
Grant support: US Department of Defense; Grant Number: W81XWH-07-1-0259, British Columbia Cancer Foundation, The Capital Region Prostate Centre (Victoria, British Columbia), U.S. Army Medical Research and Materiel Command Prostate Cancer Research Program Grant DAMD17-02-IA-0004, Intramural Research Program of the Center for Cancer Research, NCI, NIH.