Biologic Activity of Autologous, Granulocyte–Macrophage Colony-Stimulating Factor Secreting Alveolar Soft-Part Sarcoma and Clear Cell Sarcoma Vaccines

Alveolar soft-part sarcoma (ASPS) and clear cell sarcoma (CCS) are rare cancers of mesenchymal origin that primarily affect young adults (1–6). These neoplasms frequently arise in the head and neck region or extremities and frequently metastasize hematogenously to the lungs, central nervous system, and other sites. The cell of origin remains under active study, with some evidence suggesting neural crest–derived lineages as a source (7, 8). Surgical excision may be curative for localized lesions, but disseminated disease, which may be relatively slow growing, is largely refractory to cytotoxic therapies and eventually proves fatal in most cases. The rarity of the neoplasms together with their indolent and variable natural histories present significant challenges for understanding disease pathogenesis and identifying treatment regimens that might induce durable clinical benefits.

Genetic investigations have provided important insights into the biology of ASPS and CCS. Both tumors are driven by chromosomal translocations that activate members of the microphthalmia transcription factor (MITF) family, a group of basic helix–loop–helix leucine zipper transcription factors that bind to a canonical CANNTG motif (9). ASPS is characterized by the nonreciprocal translocation t(X;17)(p11;q25) that fuses the gene encoding transcription factor E3 (TFE3), a member of the MITF family, on Xp11.2 to the gene encoding native alveolar soft-part sarcoma chromosome region candidate 1 (ASPL) on 17q25, which appears to modulate the glucose transporter GLUT4 (10). The chimeric ASPL-TFE3 protein retains the nuclear import signal and DNA-binding domain of TFE3, but replaces the N-terminal sequences with ASPL, thereby perturbing the normal regulation of TFE3 transcriptional activity. CCS is characterized by the balanced translocation t(12;22)(q13;12) that fuses the Ewing sarcoma gene EWS on 22q12 with the cyclic AMP (cAMP)–regulated transcription factor ATF1 on 12q13 (11). The chimeric EWS-ATF1 protein retains the DNA binding and heterodimerization domains of ATF, but replaces the N-terminal regulatory sequences with EWS, thereby constitutively upregulating the transcription of cAMP-responsive genes. Among these is the melanocyte master transcription factor MITF, which also plays a critical role in the pathogenesis of malignant melanoma (12).

Because TFE3 and MITF bind to the same DNA motif, their dysregulation triggers alterations in oncogenic pathways that are shared between ASPS and CCS. For example, both transcription factors upregulate the receptor tyrosine kinase product of the c-Met proto-oncogene, which upon engagement by the cognate ligand hepatocyte growth factor signals to promote tumor cell proliferation, survival, and invasion (13, 14). Accordingly, tivantinib, a small-molecule inhibitor of c-Met, has demonstrated some antitumor activity in ASPS and CCS patients, although the frequency and duration of clinical responses are modest (15, 16). The hypoxia-inducible factor (HIF)-1α is a second target for both TFE3 and MITF, and activation of this transcription factor triggers the expression of an array of proteins critical to angiogenesis, including vascular endothelial growth factor-A (VEGF-A), angiopoietin-1/2, platelet-derived growth factor, and their cognate receptors (7, 17–21). Consistent with these findings, antiangiogenic treatment with sunitinib shows clear antitumor activity in both ASPS and CCS patients (22–24). In addition, cediranib, a small-molecule inhibitor of all three VEGF receptors induced an impressive 35% response and 84% disease control rate at 24 weeks of therapy in a cohort of 46 ASPS patients (25, 26). Cederanib and sunitinib are now being tested in a randomized trial targeting advanced ASPS (ClinicalTrials.gov identifier NCT01391962).

The involvement of MITF family members in the pathogenesis of ASPS and CCS highlights a potential relationship with malignant melanoma. Indeed, genomic profiling revealed shared mRNA expression patterns between CCS and melanoma that included upregulation of melanocyte differentiation antigens, the transcription factor SOX10, and the growth factor receptors ERBB3 and FGFR1 (27). Correspondingly, transcriptional analysis of ASPS demonstrated increased expression of melanoma inhibitor of apoptosis protein (ML-IAP), a MITF target gene that promotes melanoma cell survival (17, 28). Although these findings underscore common aspects of ASPS, CCS, and melanoma biology, they also raise the possibility that elements of the host response to these neoplasms might also be similar. However, while malignant melanoma is perhaps the most intensively studied cancer from an immunologic perspective, little is known regarding the antitumor response to ASPS or CCS. A recent genomic analysis of ASPS uncovered high-level expression of the innate activating receptors TLR2 and TLR9, suggesting that host factors might be involved in disease pathogenesis (7).

We previously reported that vaccination with irradiated autologous tumor cells engineered to secrete granulocyte–macrophage colony-stimulating factor (GM-CSF) enhances cellular and humoral antimelanoma responses in some patients with advanced disease (29, 30). Metastatic lesions resected after, but not before therapy manifested dense intratumoral cellular infiltrates composed of CD4⁺ and CD8⁺ T lymphocytes and CD20⁺ B cells that effectuated tumor destruction. On the basis of the ability of this vaccination strategy to augment immunity in advanced melanoma patients and the shared biology of MITF-related tumors, we undertook a phase I trial of autologous, GM-CSF–secreting tumor cell vaccines in patients harboring advanced ASPS or CCS.

The clinical protocol received approval from the Dana-Farber/Harvard Cancer Center Institutional Review Board, the Food and Drug Administration, and the Recombinant DNA Advisory Committee. The study was registered at ClinicalTrials.gov (NCT00258687) and conducted at the Dana-Farber Cancer Institute (Boston, MA), Brigham and Women's Hospital (Boston, MA), and Boston Children's Hospital (Boston, MA) according to the institutional and federal guidelines. Written informed consent was obtained from all patients before study participation. For patients under the age of 18 years for whom it was developmentally appropriate, assent was also obtained. Patients of any age with histologically confirmed CCS or ASPS were eligible if they were considered to have unresectable and thereby incurable disease. The trial was also open to rare patients with translocation-associated renal cell carcinoma that involves activating mutations in TFE3 (9) and patients with melanoma less than 18 years of age, but no patients with these diseases were enrolled. Subjects were allowed to have any number of prior therapies, provided they were more than 4 weeks from the last treatment. Additional key inclusion criteria were: Eastern Cooperative Oncology Group (ECOG) performance status 0 to 1; estimated life expectancy of at least 6 months; and adequate hematological, hepatic, and renal function. Major exclusion criteria included pregnant or nursing mothers and infections with HIV, hepatitis B, or hepatitis C. Patients with brain metastases were excluded unless these were stable at least 3 months off of treatment, and the patient had no neurologic symptoms.

Solid tumors were placed into sterile media and transported on ice to the Cell Manipulation Core Facility at the Dana-Farber Cancer Institute, where they were dissected into small fragments and processed to single-cell suspension with collagenase and mechanical digestion. When sufficient cells were obtained (see below), 2 × 10⁶ tumor cells were irradiated (10,000 rads) and cryopreserved (90% fetal calf serum, 10% DMSO) in 1 × 10⁶ cell aliquots for use in delayed-type hypersensitivity testing. The remaining tumor cells were placed in media (α-MEM, 10% fetal calf serum, gentamicin) and infected overnight at 37°C with a replication-defective adenoviral vector encoding human GM-CSF (Ad-GM) at a multiplicity of infection of 10. Ad-GM contains a GM-CSF expression cassette in the E1 region of adenovirus type 5 and a second deletion in the E3 region. The GM-CSF expression cassette contains the cytomegalovirus (CMV) immediate early promoter/enhancer, a shortened human β-globin second intron, the human GM-CSF gene, and the β-globin polyadenylation signal and 3′-untranslated region (29). After overnight infection, the tumor cells were extensively washed and irradiated (10,000 rads). Cells (1 × 10⁶) were placed into culture for 48 hours, the supernatants were collected, and GM-CSF levels were determined with an ELISA (Endogen EH-GMCSF) according to the manufacturer's instructions. Individual vaccine aliquots were cryopreserved on the basis of overall tumor cell yield as follows: ≥6 × 10⁵ to <6 × 10⁶ total, 1 × 10⁵ aliquots (dose level 1); ≥6 × 10⁶ to ≤3 × 10⁷ total, 1 × 10⁶ aliquots (dose level 2); 3 × 10⁷ to 1 × 10⁸ total, 4 × 10⁶ aliquots (dose level 3); ≥1 × 10⁸ total, 1 × 10⁷ aliquots (dose level 4). Samples of nontransfected and infected tumor cells were tested for sterility, endotoxin, and Mycoplasma. Prior to clinical administration, cryopreserved cells were thawed, washed extensively, and resuspended in 1 mL of sterile saline for the vaccines and 0.5 mL for the nontransduced cells used for delayed-type hypersensitivity analysis.

Irradiated, autologous, engineered cellular vaccines were administered intradermally (0.5 mL) and subcutaneously (0.5 mL) into normal skin on the limbs and abdomen on a rotating basis. Injections were given weekly three times and then every other week until the vaccine supply was exhausted or the patient was removed from study. A minimum of six immunizations were required to consider a patient evaluable for biologic activity. Disease evaluation was conducted at baseline, week 10, and then at 4-month intervals or whenever clinically indicated. Blood was drawn for immune monitoring before, during, and after administration of the vaccine. Irradiated, dissociated, nontransduced tumor cells were injected intradermally (0.5 mL) into normal skin at the time of beginning vaccination and with the fifth vaccination to evaluate delayed-type hypersensitivity. Punch biopsies were obtained 2 to 3 days after injections. When possible, distant metastases were biopsied after vaccination to assess immune infiltrates.

Tissues were fixed in 10% neutral buffered formalin, processed routinely, and embedded in paraffin. Immunohistochemistry was performed using standard techniques with monoclonal antibodies to CD1a, CD11c, CD4, CD8, CD20 (all from Ventana), FoxP3 (Abcam), PD-1, and PD ligand-1 (PD-L1; ref. 31). Vaccination and delayed-type hypersensitivity responses were graded on a semiquantitative scale (0–4+) based on the presence of: mononuclear cells admixed with eosinophils and basophils accumulated around blood vessels; endothelial cells that were swollen or necrotic, or showing vessel luminal occlusion; and dermal edema and fibrin exudation. The scoring was graded according to the density of the mononuclear cells. Null referred to no cells per high-power field (HPF); trace to the presence of a rare cell per HPF; 1+ to 3–5 cells around a vessel or scattered between vessels; 2+ to approximately 10 cells per HPF; 3+ to >10–19 cells per HPF; 4+ to greater that 20 cells per HPF.

The construction of the K008 melanoma-derived cDNA expression library was described previously (32). Postvaccination sera from a long-term surviving patient was precleared against Escherichia coli and lambda phage lysates and used at a 1:1,000 dilution in TBST (50 mmol/L Tris/138 mmol/L NaCl/2.7 mmol/L KCl/0.05% Tween 20, pH 8.0). Positive plaques were detected with an alkaline phosphatase–conjugated polyclonal goat anti-human pan-IgG antibody (Jackson ImmunoResearch) and 5-bromo-4-chloro-3-indolylphosphate/nitroblue tetrazolium (BCIP/NBT; Promega). Reactive clones were plaque-purified and the inserts matched to the NCBI Entrez Nucleotide database.

Previously described procedures were used for the ELISAs with some modifications (32, 33). EIA/RIA plates (Corning Incorporated) were coated with 100 μL of purified recombinant protein at a concentration of 5 μg/mL in a coating buffer (0.05% sodium azide containing PBS) overnight at 4°C. Angiopoietin-1 and angiopoietin-2 were from R&D Systems, tissue-type plasminogen activator (tTPA) was from Abnova, and recombinant ML-IAP and NY-ESO-1 were prepared in house. The plates were washed with PBST (0.05% Tween-20 containing PBS) and blocked for 2 hours at room temperature with 200 μL/well blocking solution (PBST, 2% nonfat milk, 0.05% sodium azide). After the plates were again washed, longitudinal sera samples were added at a final dilution of 1:500 in blocking solution (100 μL/well) and incubated at 4°C overnight. After several further washes, the plates were incubated with 100 μL/well of a 1:2,000 diluted alkaline phosphatase–conjugated goat anti-human IgG antibody (Jackson ImmunoResearch Laboratories, Inc.) for 1 hour at room temperature. Finally, the plates were washed again, incubated with 100 μL/well of the PNPP substrate (Sigma) for 25 minutes at room temperature, and then the OD (405 nm) was determined (Spectramax 190 Microplate Reader; Molecular Devices).

The main considerations for this single stage design study were the feasibility of vaccine manufacture and the safety of administration of the agent. Given limited information about the patient population, which was composed of patients with rare cancers, we assumed that the feasibility and safety of administering GVAX would not differ between CCS and ASPS patients. The trial was originally designed to include 20 patients, but was stopped early at 12 patients due to slow accrual of patients with these rare tumors. Feasibility goals were set at 90% for vaccine manufacture, 85% for vaccine initiation, and 75% for delivering at least six vaccines.

Twelve patients were enrolled onto this phase I vaccine study (Table 1). Three subjects had CCS and 9 had ASPS. There were 4 females and 8 males with a mean age of 25.1 years (range, 10–50). All patients had metastatic disease and failed at least one prior therapy. Involved sites included soft tissues, lung, central nervous system, bone, viscera, retroperitoneum, adrenal, breast, and spleen.

Metastatic lesions were harvested from all subjects for vaccine manufacture. Tumors were most commonly obtained from the lungs or soft tissues. Resected metastases were processed to single-cell suspensions with mechanical and enzymatic digestion, transduced overnight with a replication-defective adenoviral vector encoding GM-CSF at a multiplicity of infection of 10, irradiated at 10 cGy, and cryopreserved in individual aliquots for subsequent administration. At least six vaccines were successfully produced for 11 patients (92%). The tumor preparation for ASPS-12 was contaminated with Proprionibacterium acnes, likely from a skin source, and was not administered; the subject was withdrawn from study. Vaccines were manufactured at dose level I (1 × 10⁵) in one case, dose level III (4 × 10⁶) in two cases, and dose level IV (1 × 10⁷) in nine cases (including ASPS-12). The average GM-CSF secretion rate was 347 ng/10⁶ cells/24 hours. Other than the samples from ASPS-12, all vaccine preparations were negative for endotoxin, Mycoplasma, and sterility testing. The average prefreeze viability was 84%.

Vaccines were administered intradermally and subcutaneously into the limbs or trunk on a rotating basis every week three times and then every other week until the supply was exhausted or the patient removed from study. Six vaccinations were required to determine biologic activity. Eleven subjects received at least one vaccine and were evaluable for toxicity (92%). Rapid disease progression resulted in the withdrawal of one subject (CCS-2) after the third vaccination. Ten patients received at least six vaccines and were evaluable for biologic activity (83%; 90% exact binomial confidence interval, 56%–97%). The largest number of vaccines received by any one subject was 13.

Vaccination consistently induced grade 1–2 erythema and induration at injection sites. Mild local pruritus was easily controlled with emollients. Occasional grade 1–2 fatigue and flu-like symptoms were reported. One subject was withdrawn after six vaccines for grade 2 urticaria deemed possibly related to vaccination, but this resolved with local therapy over several weeks. There were no significant hepatic, renal, pulmonary, cardiac, hematologic, gastrointestinal, or neurologic toxicities attributable to immunization. One patient (CCS-8) was diagnosed with type I insulin-dependent diabetes 1 year after completing a course of 12 vaccinations. A cryopreserved serum sample obtained prior to treatment revealed the presence of anti-β cell antibodies, demonstrating that the autoimmunity predated vaccination. Whether immunization affected the kinetics of disease development remains unclear, but the diabetes was effectively managed with an insulin pump. No other autoimmune toxicities were observed.

Biopsies of vaccination sites 2 to 3 days after the first and fifth inoculations were obtained in 8 patients. The reactions to the fifth immunization were more intense. Histopathology revealed brisk infiltrates of dendritic cells, macrophages, eosinophils, neutrophils, and lymphocytes that extended throughout the dermis and sometimes into the subcutaneous fat (Fig. 1). Endothelial cell activation and damage were observed in the superficial venules of the upper dermis, and there was evidence of dermal edema and fibrin deposition. A semiquantitative scoring system that integrated these morphologic features (range, 0–4+) indicated that reactions in 6 of 8 subjects evaluated were of 3+ or 4+ intensities (Table 1). Dendritic cells were identified in H&E sections based on an ovoid or dendritic shape with prominent pale-gray cytoplasm; an oval, sometimes indented nucleus with clear nucleoplasm; and a single, small blue nucleolus often apposed to a delicate nuclear membrane. Immunohistochemistry for CD1a and CD11c confirmed the strong dendritic cell response. Significant numbers of CD4⁺ and CD8⁺ T cells and, to a lesser extent, FoxP3⁺ regulatory T cells (Tregs) were present, but CD20⁺ B cells were rare (not shown).

Irradiated, autologous nontransfected sarcoma cells were available for delayed-type hypersensitivity testing in 7 patients (insufficient cells precluded these studies in 4 patients). Injections of nontransfected sarcoma cells failed to elicit significant cellular infiltrates (or clinical reactions) in all patients tested at the time of beginning treatment. However, biopsies of the injection sites at the time of the fifth vaccination demonstrated responses that were graded at least 1+ intensity (range, 0–4+) in all 7 patients tested (Table 1). The more modest reactivity compared with vaccine sites might reflect the injection of smaller numbers of tumor cells and/or the absence of enforced GM-CSF expression. Histopathologically, the responses were characterized by brisk infiltrates of T lymphocytes, eosinophils, and macrophages throughout the dermis (Fig. 2). Immunohistochemistry revealed the presence of CD11c⁺ dendritic cells, CD4⁺ and CD8⁺ T cells, and scattered FoxP3⁺ T cells.

Prior studies of autologous GM-CSF–secreting tumor cell vaccines in advanced melanoma patients revealed the development of B-cell infiltrates in distant metastases (29, 30). Consistent with these results, analysis of a pulmonary metastasis resected after completion of 12 vaccines in patient CCS-8 disclosed a prominent plasma cell component, with the characteristic eccentric nuclei and cartwheel-like chromatin pattern (Fig. 3A). We thus investigated whether the vaccinated ASPS and CCS patients generated antibodies to ML-IAP or NY-ES0-1, two targets of humoral immunity in melanoma and sarcoma (34–36). However, analysis of sera samples obtained longitudinally revealed only low levels of antibodies to ML-IAP in one patient and no responses to NY-ES0-1 (not shown).

To characterize the antibody responses in more detail, we sought to identify targets using an unbiased approach. Toward this end, we screened a tumor-derived cDNA expression library with postvaccination sera obtained from a long-term surviving patient (ASPS-6). Because no ASPS or CCS cDNA expression library was available, we used a previously constructed melanoma-derived cDNA expression library (K008) that has proved informative for antigen discovery efforts in several other tumor types (32, 33, 37, 38). Although the use of the K008 library limits the ability to detect sarcoma-specific antigens, it favors the identification of shared tumor antigens that may include proteins commonly involved in transformation by the MITF-related transcription factor family.

The library screening yielded 13 distinct gene products, 12 of which encode known proteins (Table 2). As in prior studies of advanced melanoma patients, the antibody targets were primarily intracellular proteins that functioned in fundamental aspects of cancer cell biology, such as transcription/translation, signaling, cell division, metabolism, and intracellular trafficking. Of particular interest was the identification of tTPA as an antibody target. tTPA is a secreted protein that plays a critical role in fibrinolysis, which is involved in angiogenesis and tumor cell invasion (39, 40). The development of an ELISA with recombinant tTPA protein confirmed the presence of specific antibodies in ASPS-6, but the titers were not altered with vaccination (Fig. 3B). Evaluation of the entire cohort demonstrated that several patients harbored higher anti-tTPA antibody titers than ASPS-6, but no impact of immunization could be discerned.

Because tTPA is involved in angiogenesis, a prominent aspect of ASPS and CCS biology, and these tumors are sensitive to angiogenesis inhibition, we wondered whether the immune response might be directed toward other vascular targets. Indeed, prior work in advanced melanoma patients showed that GM-CSF–secreting tumor cell vaccines elicit antibodies to angiopoietin-1 and -2, and expression profiling analysis showed that these cytokines were elevated in the sarcomas (7, 17–21, 33). In accordance with these findings, 9 of the 10 ASPS and CCS patients evaluable for biologic activity generated antibodies to angiopoietin-1 and -2 as a function of treatment. Together with the histopathology, these results reveal the ability of GM-CSF–secreting tumor cell vaccines to elicit humoral immunity in sarcoma patients.

Notwithstanding the plasma cell infiltrates observed in the postvaccination resection sample from patient CCS-8, minimal tumor necrosis was detected. A breast metastasis obtained after vaccination on patient ASPS-6 similarly showed only modest tumor destruction. We thus considered whether immunosuppressive mechanisms operative in the tumor microenvironment might limit the activity of effector T cells (41). Immunohistochemical analysis of the metastasis from ASPS-6 showed scattered CD8⁺ T cells that were juxtaposed to sarcoma cells, but without clear evidence of cytotoxic effect (Fig. 4). However, the infiltrating CD8⁺ T cells were positive for the negative T-cell costimulatory receptor PD-1 (42, 43). Furthermore, the adjacent sarcoma cells expressed the cognate ligand PD-L1, with a pattern of staining that appeared to include both cytoplasmic and membrane compartments. Insufficient material was available for immunohistochemical evaluation of the postvaccination sample from patient CCS-8, but a prevaccination sample from patient ASPS-9 showed comparable PD-1–expressing CD8⁺ T cells and PD-L1–expressing sarcoma cells as the ASPS-6 specimen. Together, these results raise the possibility that the PD-1 pathway might contribute to immunosuppression in ASPS and CCS patients.

Restaging evaluation of the 10 patients who completed at least six vaccinations revealed 7 with stable and 3 with progressive disease at week 10. No tumor regressions were observed. The variable and sometimes indolent natural history of ASPS and CCS limits the ability to draw inferences regarding the impact of vaccination on disease activity or survival. Eight patients succumbed to progressive disease from 4 to 101 months after study entry. Patient ASPS-7 received ipilimumab and sunitinib after this vaccination trial. Two patients are alive at 103 and 104 months; ASPS-6 is currently being treated with cediranib. ASPS-11 was lost to follow-up 11 months after study entry, but was known to have progressive disease at that time.

Our studies were undertaken in an effort to learn more about the host response to ASPS and CCS. These soft tissue sarcomas often affect young adults and are usually fatal, but their rarity presents challenges to unraveling disease pathogenesis and testing the potential activity of novel treatments. Investigations of the characteristic chromosomal translocations that activate TFE3 and MITF have helped delineate key oncogenic mechanisms and guide the selection of targeted therapies such as small-molecule inhibitors of c-Met and angiogenesis (12, 22). Nonetheless, little is known about the immunogenicity or potential sensitivity to immunotherapy of ASPS and CCS. Because MITF is a major driver of malignant melanoma, we hypothesized that strategies that provided insights into antimelanoma immunity might similarly illuminate the host reaction to these genetically related sarcomas.

Our phase I clinical trial established the feasibility and safety of irradiated, autologous GM-CSF–secreting ASPS and CCS vaccines. Metastatic lesions were processed to single-cell suspensions and efficiently transduced with a replication defective adenoviral vector encoding GM-CSF, resulting in sufficient numbers of cytokine-producing tumor cells to constitute at least six immunizations for all 12 subjects. One preparation was contaminated with Proprionibacterium acnes, likely reflecting colonization of the tumor, and was not administered. Patients received from three to 13 vaccinations, which were well tolerated. Toxicities were limited to mild or moderate local skin reactions and constitutional symptoms. One subject was diagnosed with type I diabetes a year after completing therapy, but autoantibodies to islet cell antigens were present before treatment, indicating that vaccination did not provoke the loss of tolerance. No other serious inflammatory pathologies were noted.

Tumor regressions were not observed on the trial, but several indices of antitumor immunity revealed the biologic activity of vaccination. Injections of irradiated, GM-CSF–secreting tumor cells strongly elicited local myeloid and lymphoid cell infiltrates. The prominent dendritic cell component together with CD4⁺ and CD8⁺ T cells suggest that immunization may have enhanced tumor antigen presentation and activated antitumor T cells (44). Consistent with this idea, vaccination triggered delayed-type hypersensitivity reactions to irradiated, autologous, nontransduced sarcoma cells. These responses were composed of CD4⁺ and CD8⁺ T cells, dendritic cells, macrophages, and eosinophils, a cellular profile that likely reflects the mixed Th1 and Th2 cytokine profile characteristic of GM-CSF–secreting tumor cell vaccines (30, 45).

Examination of vaccination and delayed-type hypersensitivity reactions also disclosed the presence of FoxP3⁺ Tregs, a distinct cell population that restrains T effectors (46). GM-CSF elicits Tregs through a mechanism that involves myeloid cell production of milk fat globule epidermal growth factor-8 (MFG-E8), a secreted protein that binds phosphatidylserine on the surface of apoptotic cells (47). MFG-E8 acts as a bridge that promotes the ingestion of apoptotic cells by mononuclear phagocytes, which, in turn, release TGF-β and CCL22 to support Treg homeostasis. In preclinical models, blockade of this GM-CSF–driven suppressive pathway using a dominant-negative MFG-E8 mutant intensifies tumor destruction through inhibition of Treg activity, and efforts to translate this combinatorial vaccine strategy to testing in patients are under way. Additional strategies that are being explored to enhance the potency of these cellular vaccines include the codelivery of other dendritic cell–activating agents such as Toll-like receptor ligands, type I interferon, or STING agonists (48).

Histopathologic analysis of a metastasis resected after therapy disclosed a prominent plasma cell infiltrate, raising the possibility that vaccination evoked a humoral response. Because minimal reactivity was detected against ML-IAP and NY-ESO-1, two immunogenic antigens in melanomas and sarcomas (34–36), we pursued an unbiased approach to target discovery and screened a melanoma-derived cDNA expression library with postvaccination sera from a long-term surviving patient. This work uncovered high titer antibodies against an array of gene products that participate in diverse aspects of tumor cell biology. Of particular interest was the identification of tTPA, a central regulator of fibrinolysis, given the prominent angiogenesis characteristic of ASPS and CCS (39, 40). Although antibody titers to tTPA were not affected with therapy, extension of the analysis to other tumor vasculature–associated factors revealed angiopoietin-1 and -2 as common targets of vaccine responses. Future studies will examine whether the antibodies block functional activities of the angiogenic cytokines, as we previously demonstrated for immunized melanoma patients (33).

The induction of humoral reactions to angiopoietin-1 and -2 might have therapeutic relevance in view of the sensitivity of ASPS and CCS to angiogenesis inhibition, particularly with the VEGFR antagonists sunitinib and cediranib (25). VEGF and angiopoietins cooperate during tumor angiogenesis and promote immunosuppression through skewing dendritic cells toward Treg stimulation (49, 50). Combination therapy with VEGFR blockade and vaccination might thus exert a potent effect on the tumor vasculature while intensifying antitumor immunity. In accordance with this idea, a recent phase I clinical trial that evaluated concurrent administration of blocking antibodies to VEGF-A and cytotoxic T lymphocyte–associated antigen-4 (CTLA-4) in advanced melanoma patients revealed marked tumor endothelial cell activation and high levels of antimelanoma cellular and humoral responses (51).

Our analysis of ASPS biopsies also identified the PD-1 pathway as a potential contributor to immunosuppression (42, 43, 52). Scattered tumor-infiltrating CD8⁺ T cells expressed PD-1, whereas adjacent tumor cells showed cytoplasmic and surface PD-L1 staining. PD-1 is upregulated on T cells upon stimulation and is both a marker for a nascent antitumor response as well as a mediator of T-cell exhaustion. Whether PD-L1 expression on sarcomas reflects cell autonomous oncogenic signaling or induction through interferon-producing infiltrating CD8⁺ T cells will require further study. In either case, PD-1 engagement restricts effector T-cell proliferation, cytokine production, and cytotoxicity. Blocking antibodies to PD-1 have accomplished durable regressions in multiple cancer types, and our results suggest that this treatment strategy should also be investigated in ASPS and CCS patients.

The variable natural history of the sarcomas complicates interpretation of survival data in this phase I trial. Nonetheless, the survival of five subjects with advanced disease for at least 2 years after study enrollment together with the safety and immunologic activity of this vaccination scheme should motivate more detailed evaluation of immunotherapy for these rare tumors. Our clinical and laboratory investigations suggest that a combination of autologous cancer vaccination, VEGFR inhibition, and PD-1 blockade might effectively antagonize major host factors that impede immune-mediated tumor destruction.

J.M. Goldberg is a consultant/advisory board member for Advaxis. G.J. Freeman holds ownership interest (including patents) in Amplimmune, Boehringer-Ingelheim, Bristol-Myers Squibb, EMD Serono, Merck, Novartis, and Roche and is a consultant/advisory board member for Novartis, Roche, and Surface Oncology. M. Mihm is a consultant/advisory board member for Caliber ID, and Melasciences. F.S. Hodi reports receiving clinical trial support, through his institution, from Sanofi. No potential conflicts of interest were disclosed by the other authors.

Conception and design: J.M. Goldberg, D.E. Fisher, G.D. Demetri, D. Neuberg, F.S. Hodi, G. Dranoff

Development of methodology: J.M. Goldberg, G.D. Demetri, C. Fonseca, D. Nemer, G.J. Freeman, F.S. Hodi, G. Dranoff

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.M. Goldberg, G.D. Demetri, S.A. Allsop, Y. Nakazaki, C.P. Raut, S. George, J.A. Morgan, A.J. Wagner, J. Ritz, C. Lezcano, M. Mihm, C. Canning, G. Dranoff

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.M. Goldberg, G.D. Demetri, D. Neuberg, S.A. Allsop, M. Mihm, F.S. Hodi

Writing, review, and/or revision of the manuscript: J.M. Goldberg, G.D. Demetri, D. Neuberg, C.P. Raut, J.A. Morgan, A.J. Wagner, G.J. Freeman, M. Mihm, F.S. Hodi, G. Dranoff

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): J.M. Goldberg, G.D. Demetri, J. Ritz, C. Canning, F.S. Hodi

Study supervision: J.M. Goldberg, D.E. Fisher, G.D. Demetri, J. Ritz, C. Canning, F.S. Hodi, G. Dranoff

Other (special immunoperoxidase study interpretation): M. Mihm

The authors thank the staff of the Dana-Farber Cancer Institute Cell Manipulation Core Facility for vaccine manufacturing and sample processing. The authors also thank Holcombe E. Grier, Christopher Weldon, Monica M. Bertagnolli, Katherine Janeway, Karen Albritton, Mark Gebhardt, and Robert C. Shamberger for patient care and George Murphy and Christine Lian for help with the immunohistochemistry.

This work was supported by NCI grants P01-CA163222, U54CA163125, R01 AR043369, R01 CA111506, and philanthropic support from the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation (to D.E. Fisher and G.D. Demetri), iCureASPS (to G.D. Demetri and G. Dranoff), and the Knockin' Down ASPS Committee (to G.D. Demetri and G. Dranoff).

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.