The Influence of Granulocyte Macrophage Colony-Stimulating Factor and Prior Chemotherapy on the Immunological Response to a Vaccine (ALVAC-CEA B7.1) in Patients with Metastatic Carcinoma¹

Preclinical models of vaccines against tumor-associated antigens have demonstrated the ability of some vaccines to induce an effective immune response against the tumor-associated antigen and lead to the eradication of tumors bearing the antigen (1, 2). In such models, the development of a cytotoxic T-cell response has been paramount. To enhance the response of the immune system to a vaccine, therapy has often involved the coadministration of an adjuvant with the vaccine. Adjuvants may function in several ways: (a) to change the character and number of APCs³ in the area of the vaccination; (b) to act as a depot for the vaccine, thus prolonging the time it is presented to APCs; or (c) by altering the pathway by which the protein being presented is processed. Cytokines are of particular interest because they may affect the arm of the immune system that is stimulated.

In preclinical studies of vaccines, GM-CSF has been shown to be an effective adjuvant. GM-CSF has a variety of effects on the immune response. It can up-regulate major histocompatibility class II expression on macrophages, enhance the maturation of DCs, stimulate their migration, and produce a localized inflammatory response at the site of injection as well as a systemic response in the bone marrow (3). GM-CSF has been shown to enhance primary in vitro immune responses attributable to enhanced APC efficiency (4). It stimulates the growth of APCs such as DCs and macrophages. Irradiated tumor cells transfected with GM-CSF have been used as vaccines in murine models and have elicited enhanced immune responses against tumors, even upon rechallenge with nontransfected tumor cells (5). In a rat model, GM-CSF injected i.d. with tetanus toxoid demonstrated increased numbers of class II APCs in the draining lymph nodes of the injection site (6). There were greater numbers of APCs and higher Ab titers against tetanus toxoid with i.d. injection as compared with s.c. injections. Animals had similar DTH responses to vaccinations with GM-CSF, as compared with a “classical” adjuvant. Lastly, using this rat model, vaccinations with peptides derived from the rat neu protein with adjuvant GM-CSF led to the development of peptide-specific DTH responses, demonstrating the ability of GM-CSF to promote vaccination against a self-antigen.

Another method to enhance immune responses to vaccines has been the incorporation of a costimulatory signal. Without such costimulatory signal, presentation of the antigen to the T cell results in anergy (7). One such signal is B7.1. Binding of B7.1 to CD28 on T cells results in the production of multiple cytokines, including IL-2 and IFN-γ by CD4⁺ and CD8⁺ T cells. The importance of B7.1 as a costimulatory molecule for the development of an effective antitumor immune response has been demonstrated by transfecting B7.1 into nonimmunogenic tumor cells. B7.1-transfected tumor cells are rejected and also stimulate lasting immunity against wild-type tumor cells in murine models (8, 9, 10, 11, 12). Analysis of this response at the cellular level revealed production of various cytokines but not IL-4, consistent with a TH₁ or cytotoxic T-cell response (13).

CEA is a M_r 180,000 glycoprotein self-antigen present on endodermally derived neoplasms and in the digestive organs of the human fetus (1, 2, 14). Patients with breast, lung, gastric, colon, and ovarian cancers have elevated serum levels of CEA (15), and the majority of adenocarcinomas have CEA detected by immunohistochemistry (16, 17, 18, 19). Although CEA is expressed on some normal colonic mucosa, the level of expression is far less than that of colon carcinomas (20). There are several CEA-like antigens in normal tissues (21). Adenocarcinomas expressing this antigen do not stimulate a significant immune response against CEA, probably because of its expression during fetal development. Despite this, CEA has been evaluated as an immunogen for antitumor vaccines because there is minimal expression of CEA in adult tissues, limiting the possibility of vaccination leading to a potentially harmful autoimmune response. Various agents that vaccinate against CEA have not caused toxicity to normal organs (22, 23, 24, 25, 26, 27, 28, 29).

We reported previously (30) the results of 39 patients treated with ALVAC-CEA B7.1 alone. Patients with metastatic CEA-expressing adenocarcinomas received vaccine i.d. every 2 weeks for four injections. Nine patients were treated in an initial dose escalation phase. The final 30 patients received 4.5 × 10⁸ pfu ALVAC-CEA B7.1. Of the patients, 27% had disease stabilization after four vaccinations and continued on treatment. Six of 31 patients with elevated serum CEA levels had temporary declines in CEA. Also, HLA-A2 positive patients demonstrated increased CEA-specific T-cell frequencies vaccine after three vaccinations.

The second stage of this study, which is reported here, used GM-CSF as an adjuvant with the vaccine therapy. We hypothesized that the addition of GM-CSF would enhance the immunological responses observed in the HLA-A2 positive patients who received ALVAC-CEA B7.1. Preclinical studies vaccinating CEA transgenic mice with avipox-CEA and GM-CSF resulted in a 2.5-fold increase in T-cell proliferative response to soluble CEA in comparison with mice treated only with avipox-CEA. In addition, CEA-transgenic mice that received s.c. inoculations with CEA-expressing tumors had evidence of tumor reductions when vaccinated with avipox-CEA in combination with GM-CSF but not when vaccinated with avipox-CEA alone. Four of 14 animals had complete eradication of their tumors, with three of those animals protected upon further rechallenge (2). Thirty patients received GM-CSF plus 4.5 × 10⁸ pfu ALVAC-CEA B7.1 vaccine, were evaluated for clinical response, toxicity, and immunological responses in the HLA-A2 positive patients, and were compared with patients who received ALVAC-CEA B7.1 alone. In addition, vaccine site biopsies obtained 48 h after the initial vaccination were analyzed, and the two treatment groups were compared.

The details of patient inclusion for this clinical trial have been reported previously (30). Patients with advanced or metastatic CEA-expressing adenocarcinoma who had failed standard therapy were eligible for participation after signing informed consent. Patients were required to have an elevated serum CEA or immunohistochemical evidence of CEA expression in archival tumor samples. Patients were required to be 18 years of age or older, have an Eastern Cooperative Oncology Group performance status of 0–1, and have adequate hematological, renal, and hepatic function. In patients with known liver metastases, transaminase elevations up to 3 times the upper limit of normal and total bilirubin up to 1.5 times the upper limit of normal were permitted. All of the patients had anergy panels performed to assess DTH responses but were not excluded from participation in the study if they were anergic. All of the patients were HLA-typed to identify HLA-A2 positive patients, but HLA-A2 negative patients were not excluded. A lapse of at least 4–6 weeks from all of prior anticancer therapy was required. Measurable and evaluable disease was allowed. Patients were excluded if they had evidence of any immune compromise such as a known history of HIV infection, active eczema, atopic dermatitis, or other autoimmune disease, prior radiation to greater than 50% of all of the nodal groups, a prior splenectomy or concurrent use of systemic steroids. Patients were also excluded if there was a history of another malignant process active within the last 2 years.

All of the patients received 4.5 × 10⁸ pfu ALVAC-CEA B7.1. The vaccine was given as an i.d. injection every 2 weeks for four injections. Patients who received GM-CSF with vaccine started the GM-CSF 2 days before vaccination and continued for a total of 5 days. Daily, 250 μg was injected s.c. in the region to be vaccinated. Then, on day 3 of protocol therapy, patients received 4.5 × 10⁸ pfu ALVAC-CEA B7.1. The vaccine was manufactured by Pasteur-Merieux Sera et Vaccins, Marcy, France/Troy, New York and was supplied by the Cancer Therapy Evaluation Program, National Cancer Institute. Patients were evaluated for toxicity using National Cancer Institute Common Toxicity Criteria and for clinical response using standard response criteria (31). Patients with evidence of objective response or stable disease at 8 weeks were allowed to continue on study receiving boost injections every 4 weeks, with reevaluation every 8 weeks. Patients were removed from study for disease progression. While on the study, patients were followed with routine laboratory testing as well as serum CEA evaluations.

Patients had 3-mm punch biopsies of the vaccine site performed 48 h after the first vaccination. H&E-stained slides were evaluated for evidence of necrosis, dermal leukocytic infiltration, and perivascular inflammation. The inflammation intensity was scored using a scale of 0–3, no inflammation to severe inflammation. The percentage of the biopsy area involved with inflammation was determined. Multiplying the inflammation intensity by the percentage of infiltration resulted in the infiltrate score. The assessment of inflammation was performed by one investigator (H. S. C.) who was blind to the therapy received by the patient.

Sections were stained for CEA with a 1:4000 dilution of a rabbit polyclonal Ab (DAKO Corp., Carpinteria, CA) using the Techmate 1000 automated stainer manufacturer procedure. Slides were pretreated with heat-induced epitope retrieval in steamer with citrate buffer. Areas of CEA-positive staining were determined by light microscopy. Negative control sections were stained in the same manner with normal rabbit IgG, diluted to the same protein concentration as the rabbit anti-CEA.

Assays were only done on samples from patients positive for the HLA-A2 allele. Samples were analyzed using a modification of the method described by Scheibenbogen et al. (32) and Arlen et al. (33). Ficoll-PBMCs were washed three times with PBS, viably frozen at approximately 1 × 10⁷ cells/ml in 10% DMSO in heat-inactivated AB serum, and thawed just before use (34). PBMCs obtained before vaccination and 2 weeks after the fourth vaccination were evaluated. Ninety-six-well MultiScreen-HA plates (Millipore Corporation, Bedford, MA) were coated with 100 μl of capture Ab against IFN-γ at a concentration of 10 μg/ml. After a 24-h incubation at room temperature, plates were blocked for 30 min with RPMI 1640 containing 10% human pool AB serum. Added to each well were 1 × 10⁵ cells to be assayed. For each patient, between 5 × 10⁵ and 5 × 10⁶ total cells were analyzed, and the results were expressed as number of spots/5 × 10⁵. CIR-A2 cells pulsed with 25 μg/ml of the 9-mer CEA agonist peptide CAP1–6D (YLSGADLNL; Ref. 35) were added to each well as APCs at an effector:APC ratio of 1:3. Unpulsed CIR-A2 cells were used as a negative control. HLA-A2-binding Flu matrix peptide 58-66 (GILGFVFTL) was added to identical wells at 5 μg/ml and was used as a peptide control. The responding cells were determined by the use of a Domino Image Analyzer (Optomax, Hollis, NH).

In a preliminary analysis, the assumption that a continuous variable followed a normal distribution was examined by a normal plot and Shapiro-Wilks W test. For continuous variables, univariate comparisons between two groups were performed by two-sample t test or by nonparametric Mann-Whitney test. In case of repeated measurements (e.g., baseline and final evaluation), a one-sample paired t test was used. Pairwise associations between categorical variables were evaluated by contingency table analysis using χ² analysis or an exact test. For continuous variables, pairwise associations were examined by Pearson’s correlation coefficient and a t test for linear correlation. Forward stepwise multiple linear regression analysis was used to select variables with the greatest predictive value. ANOVA with repeated measurements was used to model the effect of GM-CSF cycles in peripheral WBCs. This was also done for the absolute neutrophil, absolute lymphocyte, and absolute monocyte counts and the cell counts obtained after Ficoll-separation of HLA-A2 positive patients who had ELISPOT assays performed. Two factors and their interaction were considered: treatment [coded as treated with GM-CSF (G) or not (N)] and time [coded as baseline (B), vaccine (V), and in between vaccines (I)]. The significance of a posteriori multiple comparisons was adjusted according to Scheffe’s method. The statistical analysis was performed using standard computer software statistical packages (SAS; Minitab). The critical significance level of 5% was chosen.

Two groups of 30 patients each were treated with 4.5 × 10⁸ pfu ALVAC-CEA B7.1 alone or with 250 μg of recombinant GM-CSF. Patient characteristics are outlined in Table 1. The majority of the patients had metastatic colorectal cancer, with other patients having lung, pancreas, breast, gall bladder, esophageal, jejunal, and thyroid cancer. The majority of patients in both groups had received prior therapy. All but four of the patients had received prior chemotherapy, one in the vaccine-alone group and three in the vaccine with GM-CSF cohort. The median number of prior regimens was three and two, with a range of one to five and one to seven in the GM-CSF plus vaccine and vaccine alone cohorts, respectively. Other prior therapies included radiation therapy and immunotherapy, as well as hormonal therapy for breast cancer patients in the vaccine-alone group. All of the patients had metastatic disease, and all of the patients except for one in each cohort had measurable disease. Thirteen of 30 patients were HLA-A2 positive in the GM-CSF arm versus 20 of 30 in the vaccine-alone arm. Patient characteristics were comparable between the patients treated with vaccine alone and those treated with vaccine and GM-CSF. Seven of the 30 patients in the GM-CSF cohort developed a >2-cm reaction to one or more of the antigens of the CMI multi test compared with 14 of 30 patients in the vaccine alone cohort. Seventy-seven percent of patients receiving vaccine with GM-CSF and 53% of the patients receiving vaccine alone were anergic, which did not reach statistical significance (P = 0.1163).

Vaccination with ALVAC-CEA B7.1 alone and with GM-CSF was well tolerated by most patients. The observed toxicities are outlined in Table 2. The most frequent toxicity was local erythema and swelling at the vaccine site. The majority of patients had grade 1 toxicity. Six patients in the combined vaccine and GM-CSF group had grade 2 toxicity compared with only one patient in the vaccine-alone group. The local toxicity attributable to vaccine was worse with the first vaccination. Seven patients had reactions at the GM-CSF s.c. injection sites, presenting with erythema and swelling before vaccination. The increased grade 2 local toxicity may have been attributable to the combined reaction to vaccine and GM-CSF, which were injected in the same areas. More constitutional complaints were reported in the cohort of patients who received vaccine with GM-CSF compared with those who received vaccine alone. Fatigue was seen in 15 of 30 patients with GM-CSF and 8 of 30 patients without GM-CSF (P = 0.0776), and fever was seen in 12 of 30 patients with GM-CSF and 7 of 30 patients without GM-CSF (P = 0.1722). Flu-like symptoms were noted in 14 of 30 patients with GM-CSF and 7 of 30 patients without GM-CSF (P = 0.0742). There was no associated increase in the grade of the toxicities seen with the addition of GM-CSF. Symptoms tended to recur with subsequent vaccinations in the GM-CSF cohort, unlike the pattern seen in the patients who received vaccine alone, where the symptoms were primarily limited to the first vaccination only.

Hematological laboratory abnormalities were noted with greater frequency in the patients receiving vaccine and GM-CSF. One patient with a normal WBC count at baseline had grade 1 leukopenia only on the day of the first vaccination. Overall, there was an increase in peripheral WBC counts in patients treated with GM-CSF compared with those treated with vaccine alone (P = 0.0167). The time factor (P < 0.0001) and the interaction between treatment [GM-CSF (G) versus no GM-CSF (N)] and time (P < 0.0001) were also significantly different. Overall, baseline (B), vaccine (V), and in between vaccine (I) counts were significantly different pairwise: B versus V (P < 0.0001), B versus I (P = 0.0025), and V versus I (P < 0.0001). Considering the interaction between treatment and time, the following comparisons were significantly different pairwise: B-N versus V-G, B-G versus V-G, B-G versus I-G, V-N versus V-G, V-G versus I-N, and V-G versus I-G. In the weeks between vaccines, there was no statistically significant difference in WBC counts between the two groups (P = 0.9964). Nine of 30 patients treated with vaccine and GM-CSF had anemia, in comparison with 1 of 39 patients who received vaccine alone (P = 0.0017).

Gastrointestinal side effects such as nausea and vomiting were reported in three patients in both treatment cohorts. There were two patients with colon cancer and one patient with lung cancer found to have brain metastases and removed from the study in the GM-CSF group, and three patients with colorectal cancer in the vaccine-alone group. The patient with grade 3 nausea/vomiting had bowel obstruction because of progressive disease. Forty-three percent of patients had elevations in transaminases as well as in alkaline phosphatase, which was not different from 40% of the patients treated with vaccine alone (P = 0.3629). These patients had colon, rectal, and jejunal carcinomas. Ten of the GM-CSF-treated patients exhibited increased liver function tests in association with disease progression within the liver, whereas one patient had baseline elevations that remained unchanged throughout treatment. Two patients with no evidence of liver metastases had treatment-related increases in their liver function tests. Another patient with a history of a T₃N₀M₀ adenocarcinoma of the colon with lung metastases but no liver metastases developed increased γ-glutamyl transferase, alkaline phosphatase, and lactate dehydrogenase before the initial vaccination. γ-Glutamyl transferase remained elevated throughout treatment, whereas lactate dehydrogenase and alkaline phosphatase returned to normal limits. The patient was not taking any other medications except for GM-CSF, which had been started 2 days before documenting elevations in the liver function tests. The patients treated with vaccine alone who experienced elevations in their liver function tests primarily had progressive liver metastases, with one additional patient each with lung metastases and bone metastases.

To be evaluable for response, patients had to receive four injections of vaccine. Of the 30 patients treated with vaccine and GM-CSF, 5 patients were not evaluable. The 5 patients were found to have progressive disease after one (n = 2) or two (n = 3) vaccinations. Rapid progression of disease within 8 weeks was observed in two HLA-A2 negative patients and three HLA-A2 positive patients, which was not statistically different (P = 0.2729). The 25 evaluable patients received four vaccinations, with 11 patients demonstrating stable disease at reevaluation. These patients, 7 with colorectal cancer, 2 with non-small cell lung cancer, and 1 with pancreatic cancer, received 1–11 monthly boost injections with continued stabilization of disease. One patient continued on study for 13 months until he was found to have progressive disease. This patient has a history of T₃N₁M₀ colon cancer treated with 6 months of neoadjuvant and postoperative chemotherapy, radiation therapy, and hemicolectomy (from July 1996 to January 1997) and intra-abdominal recurrence in January 1998 treated by photodynamic therapy. A subsequent laparotomy revealed residual disease, and the patient started vaccine plus GM-CSF in May 1999 and continued without evidence of progressive disease for 13 months. There were no partial or complete responses. However, one patient had a mixed response consisting of more than a 50% reduction in known metastatic lymphadenopathy but progression at other disease sites. The clinical stabilization of disease seen in the patients receiving GM-CSF contrasts with the experience in the 30 patients who received vaccine alone. Six patients had progressive disease before completing the initial four vaccines, one patient refused further therapy after three vaccines, and an additional patient was removed for noncompliance. At the initial tumor evaluation, only 6 of 22 evaluable patients had stable disease and continued on study for an additional 1–6 months of therapy (P = 0.1213). However, there were 4 of 23 patients with elevated CEA at baseline who received vaccine alone and had declines in their serum CEA, in contrast with no decreases in serum CEA in the patients receiving vaccine with GM-CSF.

Most patients treated with 4.5 × 10⁸ pfu of ALVAC-CEA B7.1 alone (n = 28) and GM-CSF (n = 28) had a punch biopsy of the vaccine site 48 h after the first vaccination. All of the biopsies have shown leukocytic infiltrates in the epidermis and perivascular regions. To assess the infiltrate, the biopsies were evaluated in a blinded fashion to the treatment received. The intensity of leukocytic infiltration was rated from 0–3 (0 = no infiltrate; 3 = severe), and the percentage of the biopsy specimen containing infiltrate was determined. Then, an infiltrate score was determined by multiplying the two variables. Infiltrates that had a distinct component compromised predominantly of lymphocytes were scored separately. The character of the infiltrate was altered by the use of GM-CSF when compared with the biopsies obtained in the patients who received vaccine alone. As can be seen in Fig. 1, the median infiltrate score for patients treated with vaccine alone was 5. In contrast, patients receiving GM-CSF and vaccine had a median score of 30 (P = 0.0127). Eleven patients had infiltrate scores of 50 or greater in the GM-CSF group, compared with only 3 patients in the cohort receiving only the vaccine (P = 0.0141). Also, a greater number of biopsies with prominent lymphocytic infiltrates were noted in patients receiving vaccine alone, as seen in Fig. 2. All of the vaccine-site samples demonstrated CEA expression by immunohistochemistry in the inflammatory response in leukocytes, in spindle-shaped cells suggestive of dendritic cells, and in fibroblasts (data not shown). Negative control sections using an antirabbit IgG did not reveal any positive staining (data not shown).

Nine of the 13 patients who were HLA-A2 positive received four vaccinations and had baseline and post-vaccine 4 samples available for T-cell assays. An additional patient completed the four vaccinations but had insufficient material to perform ELISPOT assays. T-cell assays using the HLA-A2 class I allele 9-mer CEA peptide CAP1–6D and the HLA-A2 9-mer Flu matrix peptide were used to investigate T-cell responses in patients positive for the HLA-A2 allele. PBMCs were assayed after only 24 h in culture in the presence of peptide to rule out effects of in vitro selection of T-cell populations. The ELISPOT assays using CEA and Flu peptides and PBMCs from each patient before and after vaccination were always done simultaneously and coded to minimize interpretive bias. Results are expressed as a precursor frequency of IFN-γ-secreting cells in response to the given peptide. A smaller number in the denominator of the precursor frequency reflects a higher number of precursors.

Table 3 illustrates ELISPOT results from HLA-A2 positive patients who received ALVAC-CEA alone, and Table 4 illustrates results from patients receiving ALVAC-CEA B7.1 with GM-CSF. Statistical analyses of differences in precursor frequencies were done by analyzing all of the patients in each group. As seen in Tables 3 and 4, PBMCs from all of the HLA-A2-positive patients in both treatment cohorts showed a less than 2-fold difference in precursor frequency to the Flu 9-mer peptide at baseline and after four vaccinations with ALVAC-CEA-B7.1. No statistically significant difference was observed in the flu precursor frequencies by univariate paired t test (P = 0.0112 and 0.226, respectively). In the same assays, these PBMCs showed a statistically significant increase in the CEA-specific T-cell precursor frequency after vaccination compared with baseline for vaccine alone (P = 0.002) but no statistically significant increase in the patients treated with vaccine and GM-CSF (P = 0.073). Five of 9 patients demonstrated at least a 3-fold increase in T-cell precursors specific to CEA peptide compared with 7 of 12 patients receiving vaccine alone. Four of the patients treated with vaccine alone showed increases of at least 4-fold in after vaccination precursor frequency, whereas only one patient who received vaccine with GM-CSF (patient 44) showed this degree of an increase in CEA-specific precursors (P = 0.2189). Patient 44 had received adjuvant chemotherapy with 5-fluorouracil and leucovorin for only 2 months for a T₃N₂M₀ colon cancer. Therapy was terminated because of significant 5-fluorouracil-associated toxicity, and the patient had been off all therapy for 40 months. Patients 10, 11, 28, and 37 from the vaccine-alone cohort and patients 44, 52, 53, and 65 from the vaccine with GM-CSF cohort had clinically stable disease for 4, 4, 3, 8 and 4, 5, 4, and 8 months, respectively. The other patients progressed after the initial four vaccinations. Patients 17, 18, 30, and 34 in the vaccine-alone group and patients 43, 44, 52, 59, and 63 from the vaccine with GM-CSF group had negative anergy panels at baseline, whereas all of the other HLA-A2 patients had evidence of a DTH response.

The cohorts of patients treated with vaccine alone and vaccine with GM-CSF were compared for differences in anergy status, number of prior chemotherapy regimens, number of months since diagnosis, and the number of months since last therapy. There was no significant difference between the two groups. Therefore, all of the HLA-A2 patients treated in this study were combined in a multiple regression analysis to determine whether there was a correlation between the measured difference in CEA precursor frequencies, between baseline, and after the fourth vaccine and: (a) the number of prior chemotherapy regimens, (b) length of time in months after a cancer diagnosis, (c) number of months since the prior chemotherapy regimen, or (d) pre-vaccine anergy. The two significant predictors of a vaccine-induced increase in CEA-specific T-cell precursors were the number of prior chemotherapy regimens and the number of months since the prior chemotherapy regimen. The number of prior chemotherapy regimens was negatively correlated with the ability to increase the CEA-specific T-cell precursor frequency after vaccination (Fig. 3), whereas there was a positive correlation with the number of months since the previous chemotherapy regimen and a CEA-specific T-cell response (Fig. 4).

This is the first clinical study to characterize the immunological contribution of GM-CSF as an adjuvant to a viral vaccine designed to elicit a T-cell response against a known tumor-associated antigen. The initial stage of this trial accrued patients to ALVAC-CEA B7.1 vaccine alone. The second cohort of patients received vaccine with GM-CSF. The vaccine with GM-CSF was well tolerated. In contrast to patients receiving vaccine alone, there was an increase in the number of constitutional symptoms observed, but there was no increase in the severity of the toxicities. More patients in the vaccine with GM-CSF group experienced injection site reactions, some of which were clearly attributable to GM-CSF alone. There was no evidence of the induction of an autoimmune response (i.e., neutropenia, colitis, or cholangitis).

The hypothesis we set out to test was that the addition of adjuvant GM-CSF to vaccination with ALVAC-CEA B7.1 would enhance the induction of anti-CEA T-cell responses. Preclinical studies and clinical trials have demonstrated that GM-CSF can serve as an effective adjuvant with whole cell (5, 36), peptide (6), protein (37), anti-idiotypic (38), and DNA vaccines (39). However, some vaccine clinical trials have not demonstrated a benefit to GM-CSF as an adjuvant. In a study by Rosenberg et al. (40), the addition of IL-2 or GM-CSF to a peptide melanoma vaccine led to a decrease in the induction of specific T cells against melanoma antigens. GM-CSF was given at 100 μg or 500 μg as a s.c. injection starting 3 days before vaccination and continuing for a total of 6 days. Interestingly, in that study the reduction of peptide-specific precursors using IL-2 was also accompanied by an increased clinical response. Simmons et al. (41) reported a Phase II study using peptide-pulsed DCs alone or with 75 μg/m²/day GM-CSF s.c. for 7 days beginning the day of vaccine infusion in a group of patients with metastatic and locally recurrent prostate cancer. Patients who received the vaccine alone were more likely to experience a 50% reduction in prostate-specific antigen or radiographic improvement than did patients who received vaccine with GM-CSF. Also, a greater number of patients who received vaccine alone had an enhanced DTH response against peptides after vaccination. In contrast, this same group demonstrated enhanced T-cell and B-cell responses to an autologous tumor cell vaccine transfected with GM-CSF compared with a nontransfected vaccine in renal cell carcinoma patients (42). These studies have tested different types of vaccines, treated different tumor types, and used different GM-CSF schedules and injection routes, making comparisons between studies difficult.

The current study suggests that the addition of GM-CSF does not increase the number of CEA-specific T-cell precursors in HLA-A2 positive patients vaccinated with ALVAC-CEA B7.1. This disparity may be attributable to differences in the patient characteristics in the vaccine-alone versus vaccine plus GM-CSF groups. Regression analysis of HLA-A2 positive patients treated with vaccine alone suggested that the number of prior chemotherapy regimens correlated with a decreased induction of CEA-specific T cells after vaccination. Other factors such as anergy status, length of diagnosis, and time from prior therapy were not shown to be correlated. The correlation persisted when the HLA-A2 positive patients receiving GM-CSF were added. The patients receiving GM-CSF had a median of two prior chemotherapy regimens, whereas patients treated with vaccine alone had received a median of three prior chemotherapy regimens. Analysis of the entire HLA-A2 positive study population demonstrated a correlation with the number of months since the previous chemotherapy regimen and the ability to induce a CEA-specific T-cell response after four vaccinations. In the HLA-A2 positive patients treated with the vaccine plus GM-CSF, the number of months since prior chemotherapy ranged from 0 to 40 with a median of 5 months, which was not dissimilar to the HLA-A2 positive patients receiving vaccine alone (0–61 months with a median of 6 months). Another patient characteristic that may have impacted immune competence is anergy status. Four of 12 and 4 of 9 HLA-A2 positive patients evaluated for CEA response by ELISPOT were anergic in the vaccine-alone and the vaccine plus GM-CSF groups, respectively. Regression analysis of anergy status did not demonstrate a correlation between T-cell responsiveness for each individual cohort alone or for the entire study population.

The patients treated with GM-CSF had elevated peripheral WBC counts on the days they received vaccine therapy. This observation raises the possibility that the GM-CSF produced a systemic but not a local effect at the vaccination sites. The vaccine biopsy sites argue against this, because there was an increase in the leukocytic infiltrates observed in the patients who received GM-CSF. Patients receiving the vaccine alone had lower infiltrate scores, with some patients having a prominent lymphocytic infiltrate (Figs. 1 and 2). The different leukocytic infiltrates may reflect a different cytokine milieu at vaccine sites, resulting in altered APC or DC inhibition. Tumor necrosis factor-α has been shown to abrogate tumor-specific immunity by GM-CSF-stimulated DCs (43), and IL-1 (44) and IL-10 (45) can inhibit antigen presentation by DCs. Another explanation may be that the increased leukocytic infiltration resulted in enhanced viral clearance, limiting the ability of the vaccine viral construct to infect cells, which decreased the production and presentation of CEA and B7.1. Lastly, the greater peripheral WBCs observed in the patients receiving GM-CSF may have resulted in a dilutional effect when the ELISPOT assays were performed. Precursor frequencies are tabulated based on the number of peripheral lymphocytes that produce IFN-γ after stimulation with the CAP-6D peptide. The total number of cells is the same in every assay. Cells for the ELISPOT assay were Ficoll-separated from peripheral blood, yielding an enriched pool of lymphocytes and monocytes without granulocytes. A greater number of neutrophils might have resulted in contamination of the Ficoll separation, resulting in a decrease in the number of T lymphocytes present in the assay and lowering the precursor frequency observed. The absolute neutrophil count in the HLA-A2 positive patients who received vaccine plus GM-CSF was statistically higher than in the patients who received vaccine alone at baseline and after vaccine (P < 0.0001). There was no statistical difference in the absolute lymphocyte count (P = 0.4508) or the absolute monocyte count (P = 0.2969). However, there was a significant difference in the number of cells collected/ml after Ficoll separation (P = 0.0446) at baseline and after vaccine. This analysis suggests there may have been a dilutional effect.

In conclusion, this pilot study, in which patients were vaccinated with advanced CEA-expressing adenocarcinomas, has demonstrated that ALVAC-CEA B7.1 is safe when used alone or in combination with GM-CSF. There was a statistically significant increase in CEA-specific T-cell precursors in HLA-A2 positive patients after vaccination with ALVAC-CEA B7.1 alone but not with ALVAC-CEA B7.1 and GM-CSF. Serum CEA values did not decrease after vaccination in the patients receiving GM-CSF, whereas 4 of 23 patients who were treated with vaccine alone had declines lasting 4–12 weeks. However, 11 of 25 patients treated with vaccine and GM-CSF exhibited stable disease and one mixed response compared with only 6 of 22 patients with stable disease who received vaccine alone (P = 0.1213). These clinical effects did not correlate with improvements in anti-CEA ELISPOT, raising the question of the appropriateness of this immunological assay as a predictor of clinical benefit. Also, regression analysis demonstrated the negative impact of the number of prior chemotherapies on immune response as well as the positive correlation between the number of months between initiation of vaccine and previous chemotherapy. Patient 44, who had only received one prior chemotherapy regimen, 2 months of adjuvant 5-fluorouracil, which was terminated 40 months before initiating vaccine therapy, had the greatest increase in anti-CEA T-cell precursors highlighting the correlation. These observations underscore the need to evaluate vaccine strategies in patient cohorts that are not heavily pretreated. Melanoma vaccine studies have demonstrated clinical efficacy. Melanoma has few effective chemotherapeutic options, raising the possibility that patients treated on melanoma studies are a less heavily pretreated cohort than those with recurrent metastatic adenocarcinoma.

The dose and schedule of GM-CSF used in this study did not enhance the induction of anti-CEA-specific T cells. In another study (46) using ALVAC-CEA in patients with recurrent CEA-expressing colorectal cancer, the addition of GM-CSF results in enhanced T-cell responses. In that clinical trial, the dose of GM-CSF is 100 μg given s.c. for 4 days, beginning the day of vaccination. Moreover, the vaccine was administered monthly rather than biweekly, as in the current study. The differences in dose and schedule do not allow for direct comparison. However, the dose of GM-CSF may be important for its adjuvant effects; e.g., in a study using an anti-idiotype vaccine in a murine model, 50,000 units of GM-CSF were less effective than 10,000 units in antitumor effects (47). It is possible that prolonged stimulation of DCs by GM-CSF may result in maturation of the DC, resulting in the loss of the ability of the cell to process antigen. Alternatively, an altered cytokine milieu may have resulted in loss or inhibition of DC function. Also, the effect of GM-CSF may be dependent on the dose used and the temporal relationship to the administration of vaccines. Finally, our results may indicate that in the setting of costimulation, GM-CSF may not have an additional effect on vaccination. An upcoming clinical trial will further explore the role of costimulation with and without GM-CSF, using a vaccine construct that encodes CEA in addition to the three costimulatory molecules, B7.1, LFA-3, and ICAM (48).

We thank Jonathan Cheng, Paul F. Engstrom, Lori J. Goldstein, Scott Kindsfather, Corey J. Langer, Elizabeth Rosvold, Russell Scher, and Christine Szarka for enrolling and caring for the patients who participated in this study, Josephine Schultz, Anne Amoroso, and Jonathan Cheng for their assistance with Ficoll separation of patient samples, and Christine Densmore for coordination of patient sample transfers from Fox Chase Cancer Center to the National Cancer Institute.