Phase I Trial of TGF-β2 Antisense GM-CSF Gene-Modified Autologous Tumor Cell (TAG) Vaccine

Despite decades of immune-based therapies in solid tumor patients (1–7), only a minority have effectively translated into the clinic (8). Among the hypotheses accounting for this are the ineffective priming of tumor-specific T cells, a lack of high avidity of primed, tumor-specific T cells, and physical or functional suppression of primed, tumor-specific T-cell activity by host- and/or tumor-related mechanisms. Tumor-infiltrating lymphocytes (TIL) include immunosuppressive regulatory T cells (Treg; CD4⁺ CD25⁺ Fox3p⁺; ref. 9) which, rather than cross-priming CD8⁺ cytotoxic T cells, inhibit T-effector antitumor activities (9–11) in part through TGF-β–dependent suppression of antigen-presenting dendritic cell (DC) processes (12). Further, both tumor-infiltrating, tolerogenic DCs and suppressor T lymphocytes express the TGF-β receptor 1 (TGF-βR1) and are therefore susceptible to immunosuppressive modulations by TGF-β1 and TGF-β2 produced by tumor cells (13–18). Circulating Th1-suppressive cytokines, including TGF-β and IL-10, that are frequently elevated in patients with advanced cancer mediate immune suppression in tumor-bearing animal models by downregulating antigen recognition, Th1 activation, and antitumor immune effector functions (19–21).

We have previously demonstrated safety as well as response and survival benefits in 2 studies of gene-based vaccines in non–small cell lung cancer (NSCLC) patients (22–26). In the GVAX trial using GM-CSF–secreting allogeneic tumor cells, we documented complete responses (CR) in 3 of 33 patients. In a second randomized study, we observed a dose-related survival advantage in patients receiving an antisense (AS) TGF-β2 knockdown allogeneic vaccine (Belagenpumatucel-L). Enhanced tumor antigen recognition correlated with clinical benefit.

By inhibiting TGF-β2 expression, Belagenpumatucel-L reduces the cytokine-associated immune suppression that is well documented in cancer patients (13, 14, 27–30). TGF-β2 also antagonizes natural killer (NK) cells, lymphokine-activated killer (LAK) cells, and DC function (20, 21, 31–34). GVAX was shown to enhance tumor antigen expression and DC migration to the vaccination site (35–37). However, GM-CSF–induced maturation can be blocked by TGF-β (38). On the basis of the hypothesis that the combined expression of GM-CSF and TGF-β2 AS transgenes can optimally stimulate immune responses to cancer-associated antigens, we have constructed an expression plasmid [the TAG (tumor-associated glycoprotein) plasmid] that coexpresses GM-CSF and TGF-β2 AS nucleotide sequences.

The construction and cGMP manufacturing of TAG has been described (39). TAG vector utilizes the pUMVC3 vector backbone that contains an origin of replication, kanamycin resistance gene, a CMV (cytomegalovirus) promoter, and intron A driving the hGM-CSF cDNA and a 930-base pair AS fragment of the hTGF-β2 cDNA.

Freshly harvested autologous tumor cells were mechanically and enzymatically disaggregated, then electroporated with the TAG vector. The resulting vaccine was irradiated, then aliquoted and cryopreserved until the time of injection (40).

The primary objective of this trial was to evaluate the safety of TAG vaccine in advanced solid tumor patients without alternative standard therapy options. Following progression on preceding therapy, cancer patients were entered into 1 of 2 cohorts depending on tumor harvest and vaccine manufacturing cell yield. Cohort 1 patients received 1 × 10⁷ cells per injection and cohort 2 patients received 2.5 × 10⁷ cells per injection. A maximum of 12 intradermal injections were administered monthly alternating between the right and left upper arms. A safety assessment was made after 3 patients were entered into each cohort.

Study participation requirements included a minimum of 5 manufactured vaccine doses. Patients not meeting this requirement were deemed ineligible. Treatment of eligible patients was continued until progressive disease (PD) or up to a maximum of 12 injections.

All eligible patients were treated in the outpatient facilities of Mary Crowley Cancer Research Centers (MCCRC). Inclusion criteria were as follows: a histologically confirmed, advanced or metastatic noncurable solid tumor following completion of 1 or more diseases appropriate standard of care therapies and recovery from all treatment-related toxicities to grade 1 or lower (except alopecia); availability of tumor in sufficient quantity (i.e., ≥2 g) for vaccine processing; history of brain metastases allowed if treatment completed month 2 or later prior to enrollment with MRI confirmation of no active disease; presence of 1 or more measurable or evaluable lesion; patients of age 18 years and above; ECOG performance status of 0 to 1; a signed, IRB-approved, protocol-specific written informed consent document; a negative pregnancy test for women of child-bearing potential; and normal organ and marrow function defined as follows: absolute granulocyte count (≥1,500/mm³), platelets (≥100,000/mm³), total bilirubin (≤2 mg/dL), AST (SGOT)/ALT (SGPT; ≤2× institutional upper limit of normal), and creatinine (<1.5 mg/dL).

Exclusion criteria included the following: surgery involving general anesthesia, chemotherapy, radiotherapy, steroid therapy greater than 2-mg prednisone equivalents per day, or immunotherapy within 4 weeks of study entry; use of other investigational agents within 30 days prior to study entry; mucinous adenocarcinoma; prior splenectomy; prior malignancy (excluding nonmelanoma skin cancer) unless in remission for 2 years or more; Kaposi's sarcoma; uncontrolled intercurrent illness, for example, ongoing or active infection, symptomatic congestive heart failure, unstable angina pectoris, cardiac arrhythmia, or psychiatric illness/social situations that would limit compliance with study requirements; or confirmation that patient was pregnant or nursing, HIV positive, known to have chronic hepatitis B or C infection, or a history of autoimmune diseases.

Eligible patients received monthly intradermal injections of TAG vaccine (either 1 × 10⁷ or 2.5 × 10⁷ cells per injection for ≥5 doses). Sites of injection were rotated between the right and left upper arms. If the ipsilateral axillary lymph nodes were radiated during prior therapy, alternative sites (e.g., anterior thigh) were used. Patients were observed for at least 30 minutes following vaccination with vital signs monitored every 10 minutes. If patients were deemed clinically stable, vaccine administration was continued for up to 12 monthly doses as long as sufficient material was available.

The following evaluations were performed within 2 weeks prior to therapy: a complete medical history, physical examination, ECOG assessment, and chest X-ray, chest/abdominal/pelvic computed tomography (CT) or magnetic resonance imaging (MRI), brain MRI or CT, and radionuclide bone scan, if indicated. A complete blood count (CBC) with differential and platelet count was also performed, as well as serum chemistries [creatinine, glucose, total protein, blood urea nitrogen (BUN), total carbon dioxide (CO₂), albumin, total and direct bilirubin, alkaline phosphatase, and AST and/or ALT] and electrolytes (total calcium, chloride, potassium, sodium). Urinalysis, pregnancy test for females of child-bearing potential, EKG (electrocardiogram), and immune function analysis, including enzyme-linked immunospot (ELISPOT) analysis of cytotoxic T-cell function to autologous tumor antigens, TGF-β2 levels and GM-CSF levels were also obtained.

Evaluations performed every 28 ± 3 days during therapy included the following: physical examination; ECOG performance status assessment; CBC with differential and platelet count; serum chemistry and electrolytes; toxicity assessment; and clinical assessment of tumor response. Radiological assessments of tumors were done quarterly. Immune function, specifically serum cytokine analysis (IFNγ, IL-4, IL-6), DC–mediated response (ELISPOT assay of CD8⁺ cells to autologous tumor antigens) and TGF-β2 and GM-CSF levels, was monitored at baseline and at months 3 and 6.

Tumor response in patients with measurable disease was reported using RECIST (response evaluation criteria in solid tumors) 1.0 criteria. Any objective response to treatment required confirmatory scans at least 4 weeks apart.

Complete response was defined as disappearance of all target lesions; partial response (PR), a 30% decrease in the sum of the longest diameters (SLD) of target lesions, taking as reference the baseline SLD; and PD, a 20% increase in the SLD of target lesions, taking as reference the nadir SLD recorded since the treatment started; or the appearance of ≥1 new lesion. SD met neither PD nor PR criteria.

ELISPOT assay was performed using ELISPOT Assay for Interferon Gamma (BD Biosciences). Ninety-six–well plates from the kit were coated with primary anti-IFNγ monoclonal antibody and incubated overnight at 4°C. Cyropreserved pretreatment and posttreatment peripheral blood mononuclear cells (PBMC) collected at weeks 12 and 24 after vaccination were thawed and revitalized in culture medium overnight at 37°C stimulation, then dispensed at 1 ×10⁴ cells per well into microwells that were previously treated with human AB serum (2 hours). Coincubation was carried out with the mitogen phorbol myristate acetate and the Ca²⁺ ionophore ionomycin (PMA-I; 5 and 500 ng/mL, respectively) or target cells (TAG autologous vaccine; 3 × 10⁴ cells per well) for 48 hours at 37°C to attain target to effector cell ratios of 3:1. The wells were washed and incubated with a biotinylated detection antibody (BD Biosciences ELISPOT Set), enzyme reagent, and the chromogenic substrate according to manufacturer's protocol. Positive reactions were analyzed using ELISPOT reader system (Carl Zeiss) with KS ELISPOT Software 4.9, service provided by ZellNet Consulting, Inc.

TAG vaccine manufacturing was successful in 32 of 38 patients. Five vaccines were rejected due to contaminants (n = 3; all from resections involving colon wall) or inadequate harvest of tumor cells (n = 2); 1 vaccine was intentionally manufactured for research purposes. Twenty-three patients received at least a single vaccine since June 2, 2008, in the outpatient facilities of MCCRC. Table 1 provides demographics of treated patients as well as tissue site, dose level, number of vials manufactured, cell viability, GM-CSF expression, and percent TGF-β2 and TGF-β1 knockdown of transfected product.

There was no obvious difference in the rate of serious adverse events across the 2 dose cohorts. No grade 3/4 treatment-related events were observed. Grade 1/2–related events and serious adverse events are shown in Tables 2 and 3, respectively.

To date, 23 patients have received vaccine (Table 4). Seventeen of 21 patients had SD at month 2 or later as their best response. Two withdrew early for personal reasons with SD after 1 cycle and were not considered evaluable; 2 had PD at month 2 of assessment; and 1 with PD at month 3 withdrew consent.

One 78-year-old male patient with metastatic melanoma (013) achieved a CR of all target (abdominal, lymph node) and nontarget (bone metastases) lesions (Fig. 1). He had previously failed standard therapy. The CR was ascertained after completion of 11 low-dose vaccine injections and confirmed 3 months later by repeat PET-CT (Fig. 1). Subsequent PET-CT scans continue to show disease-free status.

Two vaccine recipients (041 and 043) are currently receiving treatment on trial. Seven of 20 evaluable patients have survived for more than 1 year following initiation of treatment. Four of these patients (008, 013, 023, and 037) have successfully received all manufactured vaccines.

IFNγ ELISPOT assessments were retrospectively performed to compare quantified CD8⁺ T-cell activity in weeks 12 and 24 postvaccination blood to baseline samples from 16 patients (008, 009, 010, 012, 013, 014, 017, 023, 024, 026, 029, 031, 032, 033, 034, and 035) in a blinded fashion. In vitro IFNγ production was determined following phorbol myristate acetate (PMA)/ionomycin–induced polyclonal T-cell differentiation (41, 42) or coincubation with the patient's irradiated, autologous TAG vaccine (Fig. 2A and B, respectively).

Most patients (11/16) responded to PMA plus ionomycin stimulation (≥50% increase at week 12), including the 2 patients with prolonged SD (008 and 023) and 1 patient (013) with CR (Fig. 2A). Therefore, a majority of the advanced cancer patients retained immunoresponsiveness by these in vitro criteria.

Six of 13 patients displayed a positive ELISPOT response to autologous TAG vaccine (>10-fold increase over baseline) at week 12 postvaccination (Fig. 2B). These included the 3 patients with prolonged SD or CR, who also demonstrated further elevated ELISPOT activities at week 24 (4-, 5-, and 14-fold increase over the week 12 values). Of the 3 other ELISPOT-responsive patients, 2 (032 and 033) survived through week 24, as compared with none of the 7 ELISPOT-negative patients.

Safety and tolerability of the TAG vaccine have been demonstrated in the 23 reported patients. Despite the poor prognosis of advanced cancer patients enrolled into trial, an unexpectedly high proportion of patients survived with SD of 3 months or longer and survival of more than 1 year was observed in 35% of the evaluable patients. Moreover, the CR observed in a patient with extensive metastatic melanoma confirms the clinical activity of the TAG vaccination.

Rejection of antigen-expressing tumor cells is effected primarily by specific host CTL (cytotoxic T lymphocyte; refs. 43, 44). TILs have been shown to mediate durable regression of established tumors in mice with advanced tumor burdens (45, 46). In patients with metastatic tumors, various investigators have documented the existence of antitumor CTL effectors in PBMC and TIL that are able to lyse autologous tumor cells, but not NK targets, allogeneic tumors cells, or autologous fibroblasts (47–51). These findings support the premise that tumor-associated antigens expressed by metastatic human tumors can stimulate a specific T-cell response by immune effectors that can be expanded ex vivo to achieve clinical objective responses. This is particularly evident within the subset of patients with prolonged SD or CR postvaccination, who displayed progressively elevated recall responses at weeks 12 and 24 as compared with prevaccination levels (Fig. 2B).

Overexpression of 2 or more of the TGF-β isoforms has been demonstrated in melanoma, gliomas, prostate, gastric, colorectal, ovarian, and gastric cancers (15–17, 52). TGF-β1 and TGF-β2 bind to TGF-β receptor 2 (TGF-βR2) which phosphorylates TGF-βR1. This heterotetrameric receptor complex is then able to suppress DC and helper T-cell function through regulation of the Smad complex and non-Smad MAP (mitogen-activated protein) kinases (12, 18). Polak et al. recently showed that tumor-infiltrating, tolerogenic DCs and suppressor T-cell lymphocytes in malignant melanoma correlate with immunosuppressive TGF-β1, TGF-β2, and IL-10 expression (15). This mechanism of tumor-associated immunosuppression is likely to contribute to tumor escape.

Genetic modification with a TGF-β2 AS-encoding plasmid represents 1 of many approaches currently being evaluated to inhibit local TGF-β activity. Others include the use of neutralizing antibodies, soluble receptors, receptor kinase antagonist drugs, AS reagents, and a number of less-specific drugs such as angiotensin II antagonist and tranilast (53). Systemic TGF-β blockade could potentially interfere with healing, exacerbate the inflammatory disease response, or neutralize the TGF-β tumor suppressive effect in normal cells (54). Our results suggest the effectiveness of incorporating a de novo gene-modifying moiety into the vaccinating cancer cells ex vivo to produce localized TGF-β blockade so as to enhance activation of the afferent arm of the immune response while bypassing the pharmacokinetic and toxicity concerns potentially associated with systemic anti-TGF-β neutralization. We observed an objective response in only 1 of 23 patients entered into trial, despite encouraging SD achievement. Although these results would suggest an induced state of immunoequilibrium, another factor affecting this balance is the immunosuppressive activity of other TGF-β isoforms (TGF-β1 and TGF-β3) that may override TGF-β2 knockdown. Thus, an alternate gene-modifying moiety that can globally attenuate TGF-β1, TGF-β2, and TGF-β3 may further potentiate the knockdown effects of all TGF-β–related immunosuppressive activities of the malignant cell. We are currently investigating this strategy for future vaccine development.

We observed consistent GM-CSF expression by manufactured autologous TAG vaccines per 1 × 10⁶ cells at 24 hours, albeit at lower levels than observed in our adenoviral transgene GVAX studies. Although in 1 study, there was a suggestive correlation between survival of patients who received GVAX with GM-CSF expression (22), this relationship was not subsequently confirmed (35). Variations in GM-CSF expression are likely attributable to differences in constructs (plasmid-based vs. viral-based expression). Nonetheless, local GM-CSF expression levels by the TAG vaccine are deemed clinically relevant as 1) use of a plasmid rather than a viral vector obviates the neutralizing effects of elicited antiviral antibodies, 2) use of a plasmid-based vaccine can minimize the development of humoral responses that interfere with long-term gene expression, and 3) concurrent suppression of TGF-β2 can abrogate tumor-induced inhibition of GM-CSF–dependent DC maturation (38). Conversely, we observed no objective tumor responses with the bystander GVAX vaccine that produced a 25-fold higher level of GM-CSF than the autologous vaccine, indicating that high levels of vaccine-based GM-CSF protein expression is not a necessary criterion for achieving clinical activity (55).

Mechanistically, tumor cells are able to promote the proliferation of immunosuppressive Treg cells directly via TGF-β production or through the conversion of DCs into regulatory cells that secrete TGF-β (56). As noted, GM-CSF is a key immunostimulatory agent in both antigen → DC interaction and DC ↔ T-cell activation (57, 58). It is also potentially the most active of the immunostimulatory cytokines tested (35, 57, 58). However, GM-CSF–induced DC maturation is effectively blocked by TGF-β as measured by decreased expression of ICAM-1, B7–2, and MHC class II expression (38). Furthermore, mixed lymphocyte reaction–stimulating activity of cells was blocked when TGF-β was added during the first 4 to 6 days to cells cultured in GM-CSF, where a dose-response effect was evident (38).

Others have also demonstrated that gene-based immunotherapy has the potential for local and systemic management of disease (59). Our initial phase I/II trial results with TGF-β2 AS gene vaccine suggested a dose-related survival response correlating with enhancement of tumor antigen recognition in advanced stage NSCLC patients. Although the dose of ≥2.5 × 10⁷ cells per injection, compared with 1 × 10⁷ cells per injection, given on a monthly basis in the Belagenpumatucel-L NSCLC study, was associated with a survival advantage, an earlier study in patients with recurrent glioma using the same vector, documented responses at doses at and below 1 × 10⁷ cells per injection (60). There was no correlation of survival and cell dose in GVAX trials at MCCRC at or above 1 × 10⁷ cells per injection (25, 26). Thus, 2 dose levels, 1 × 10⁷ cells per injection and 2.5 × 10⁷ cells per injection, were defined as acceptable for purposes of this protocol, especially given the presumptive GM-CSF function-modulating effect of TGF-β2 silencing. Our results, thus, support the hypothesis that a combined plasmid with GM-CSF gene and TGF-β2 AS is safe and, furthermore, provide preliminary evidence of durable clinical benefit justifying further clinical investigation. Insofar as TGF-β1 is the dominant immune inhibiting cytokine in most solid tumors, TGF-β blockade limited to the TGF-β2 isoform with TAG vaccine may limit immunomodulatory potency (15, 61). Further elucidation of post–vaccination-activated immune subsets (DCs, T effectors, Tregs) will serve to address the mechanistic role of GM-CSF plus TGF-β–AS modifications in promoting a clinically relevant antitumor immune response. Methods to block both TGF-β1 and TGF-β2 are under current clinical investigation by our team.

No potential conflicts of interest were disclosed.

We gratefully acknowledge the generous support of the Jasper L. and Jack Denton Wilson Foundation. The authors thank Susan W. Mill for her competent and knowledgeable assistance in the preparation of the manuscript.

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.