Despite the improvement in the outcome of anticancer therapy achieved during the last few years, several metastatic tumors remain resistant to therapy. This is particularly true for metastatic melanoma, renal and lung carcinoma, and, although to a lesser extent, for colorectal carcinoma. For these tumors, several biological therapies have been proposed, including vaccination by a variety of approaches. Over the past decade, it has become increasingly apparent that the host immune system can recognize and destroy cancer cells, although this will only rarely translate into a significant clinical benefit (1) . Consequently, antitumor therapies have focused on exploiting the immune response in an attempt to increase its strength and specificity to better control the disease.
Many tumor-associated antigens (TAAs) are now available in the form of single peptide epitopes restricted by known HLA class I or II alleles (see ref. 2 ). However, all of the TAAs used in clinical protocols belonged to the group of normal (i.e., self) peptides/proteins (e.g., gp100, tyrosinase, CEA) shown to be weakly immunogenic in vivo. This may be one important reason for the limited clinical response rate observed in these studies, in addition to mechanisms of immune evasion, which have been reported previously (3) .
Thus, we believe that a new group of TAAs needs to be explored as for purposes of anticancer vaccine development, namely the individual (unique) TAAs that often result from mutations of a variety of proteins. In animal models, these TAAs enable rejection of tumors more potent than that obtained by shared TAAs (4) .
Such individual antigens do exist also in human tumors (see ref. 2 ) and appear to be associated with long-term disease-free condition in the few patients that could be studied (5, 6, 7) . However, with today’s technology, it still requires several months before novel TAAs are identified and characterized for the use in cancer vaccines. In addition, the heterogeneity of human cancers with regard to their molecular background and the resulting antigenic repertoire warrants a polyvalent approach to efficiently address larger numbers of patients within a given disease population. An alternative approach, which carries the promise to circumvent the above challenges, is offered by the use of tumor-derived heat shock protein-peptide complexes (HSPPCs) as cancer vaccines. In fact, Srivastava et al. (8) have shown that tumor-derived autologous heat shock proteins (HSPs) function as chaperones of TAA peptides, and the use of HSPPCs allows for immunization of the host against a large repertoire of individual TAAs. Moreover, mice immunized with heat shock proteins (HSPs) in a therapeutic setting often showed retarded progression of primary cancers and reduced metastatic load (9) . In the following sections, we shall discuss how this principle has been applied for designing new clinical trials of cancer vaccines.
Preclinical Characterization of HSPs as Cancer Vaccines
The molecular chaperones HSP70, Grp94/gp96, calreticulin, and gp110 are well characterized within the family of HSPs. HSP70 has diversified functions, which include chaperoning protein into degradation pathways, binding of new synthesized amino acid chains on ribosomes, and maintaining translocation-competent folding of estrogen receptor and mitochondrial precursor proteins in the cytosol. As for the estrogen receptor resident Grp94/gp96, several authors showed that also this protein displays intrinsic polypeptide-binding activity (see ref. 10 ). Gp110 is related to HSP70 in structure and functions, but it appears more effective in binding and chaperoning full-length proteins (11) .
However, not much is known on the structural requirements for peptide binding to HSPs because the primary sequences of very few peptides directly eluted from HSP70 and gp96 are available (12 , 13) . An important new feature of HSPs is that they bind specifically to antigen-presenting cells through toll-like receptors (TLRs), thus triggering the innate immunity signaling pathways (14 , 15) .
To investigate whether gp96 and HSP70 purified either from human tumors or in vitro cultured melanoma lines could indeed contain TAA peptides, we assessed the ability of tumor-derived gp96 and HSP70 to reconstitute the epitopes recognized by CD8+ T cells specific for melanoma or colorectal carcinoma antigens. For gp96, experiments showed that HLA-A*0201–restricted CD8+ T cells recognizing melanoma (e.g., Melan-A/MART-1) or colorectal carcinoma (CEA and EpCAM) antigens were triggered to release IFN-γ by HLA-A*0201–matched antigen-presenting cells pulsed with HSP96 purified from tumor cells expressing the relevant TAA (16) .
Using a similar approach, we and others showed that also HSP70, purified from a human melanoma line, included peptides from Melan-A/MART-1, gp100, TRP-2, and tyrosinase (17 , 18) . Of note, not only HLA-A*0201 but also HLA-A*0301 and HLA-C*0801–restricted CTLs were specifically triggered by HSP70 derived from the autologous melanoma line 15392, thus indicating that different epitopes were chaperoned by HSP70 deriving from a single tumor (17) .
Unfortunately, because T-cell clones specific for unique class I HLA-restricted melanoma TAAs were not available, the presence of peptides derived from these antigens inside the HSP70 chaperoned peptides could not be assessed, and therefore, the relation of shared versus unique class I HLA-restricted epitopes in human HSP70-chaperoned peptide repertoire is unknown. Altogether, more studies are necessary to prove that unique TAA peptides can bind different members of the HSP families.
Tumor-derived HSPPC-96 appeared to activate TAA-specific T cells and expand them in vivo. In fact, immunization with autologous tumor-derived HSPPC-96 induced a significant increase in the frequency of T lymphocytes recognizing Melan-A/MART-127–35, CEA571–579, and EpCAM263–271 in a subset of HLA-A*0201 melanoma and colorectal carcinoma patients, respectively, as detected by enzyme-linked immunospot and HLA/tetramer staining (16) . These increases in TAA-specific T-cell responses were associated with a favorable disease course after HSPPC-96 vaccination (19 , 20) .
Thus, our data provide evidence that HSPPC-96 derived from human tumors can present TAA peptides to CD8+ T cells and activate them both in vitro and in vivo.
HSPPCs have been shown to display another crucial immunologic function, i.e., the activation of innate immune responses by promoting the maturation of antigen-presenting cells in vitro and in vivo (21 , 22) by binding specific TLRs (22 , 23) . Moreover, a 14-amino acid sequence of HSP70 has been identified that is specifically recognized by human natural killer cells (24) .
This conclusion, however, has been criticized because of the possibility that (a) minute amounts of endotoxins may be responsible for this phenomenon despite the use of different methods aimed at eliminating or reducing such contaminants, or (b) HSPs themselves directly bind bioactive lipids such as lipopolysaccharide and transport it to the antigen-presenting cell receptor (see ref. 25 ). Although it is almost impossible to exclude that undetectable amounts of endotoxins may contaminate or bind HSPs, several findings do not corroborate such an explanation for the antigen-presenting cell activation function of HSPs. In fact, in our ex vivo studies, HSP-96 obtained by the same procedure from histologically different or antigenically unrelated human tumors were used as negative controls, and antigen-presenting cells pulsed with such preparations were not recognized by TAA-specific T-cell clones (16 , 17) . Moreover, recent studies in animal models indicate that HSPs can stimulate in vivo a CD8+ T-cell response through activation of dendritic cells, which is independent of lipopolysaccharide and occurs via the TLR4 (26) . A separation between antigen delivery and dendritic cell stimulation in HSP70 mycobacterial binding domain has also been reported previously (27) .
Thus, although endotoxin contaminants cannot be completely ruled out in the biological activity of HSPs, the role of these contaminants, if any, cannot explain the bulk of the important immunologic functions exerted by these proteins particularly in chaperoning TAA peptides.
Taken together, the above findings, along with the demonstration of anticancer activity of HSPPCs in animal models, provided the rationale for HSPPC-based vaccination in cancer patients.
Clinical Trials of Vaccination with HSPPC-96 in Metastatic Melanoma and Colorectal Carcinoma Patients
The first human clinical study with autologous HSPs was conducted by Janetzki et al. (28) in a small group of miscellaneous tumors. These authors noted signs of clinical responses and found that a T-cell reaction could be generated in a fraction of patients.
Subsequently, a phase II trial was carried out in stage IV (American Joint Committee on Cancer) melanoma patients treated with autologous, tumor-derived HSPPC-96 (Oncophage, Antigenics, Inc., Lexington, MA; ref. 19 ). The trial was designed to investigate safety and immunogenicity and to demonstrate antitumor activity of HSPPC-96 in patients enrolled and treated in four Italian centers. Patients were randomly assigned to receive 5 or 50 μg of HSPPC-96 at weekly intervals starting 5 to 8 weeks after surgical resection of tumor metastases. The study included 45 patients, which were vaccinated, and 39 were considered evaluable for response after having received at least 4 weekly injections of vaccine (one cycle of treatment). Twenty-eight patients had residual measurable disease whereas 11 were disease free after surgery.
Among patients with measurable disease, two complete responses (24 and ≥48 months) and three long-term stable diseases (153, 191, and 272 days) were reported. Disease-free survival of the subgroup of vaccinated patients with no evidence of disease ranged from 29 to 642 days (median, 117 days). No clear correlation between clinical responses and dose of HSPPC-96 could be found; additionally, besides transient erythemas and/or induration at vaccination sites, patients did not experience treatment-related systemic toxicity (19) .
Enzyme-linked immunospot assays showed that vaccination with HSPPC-96 significantly increased the ability of peripheral blood mononuclear cells from 11 of 23 patients tested (47.8%) to release IFN-γ when exposed to autologous or HLA-A–matched melanoma cells. The T-cell involvement in this antitumor reaction was shown by blocking with the anti-HLA class I monoclonal antibody W6/32. These findings indicated that repeated administrations of HSPPC-96 vaccine increased the frequency of melanoma-specific T cells in a large proportion of immunized patients. Noteworthy, the two complete responses and two of the three long-term stable diseases were observed in the group of immunologic responders.
Taken together, the results of this initial study showed that vaccination with HSPPC-96 is feasible, safe, and apparently associated with clinical and immunologic responses in melanoma patients, providing the scientific and clinical rationale for additional trials. An increase of natural killer cell activity in most of the vaccinated patients was also observed, but the role of natural killer cells for the development of clinical responses could not be assessed.
Another phase II study was done by using HSPPC-96 to vaccinate patients after resection of liver metastases of colorectal carcinoma with a schedule similar to that used for melanoma (19) . Patients with colorectal carcinoma who had developed liver metastases after initial treatment of their primary tumors underwent complete surgical resection of their liver disease, and HSPPC-96 was prepared from liver lesions. Four to 6 weeks after surgery, patients were given HSPPC-96. The disease-free and overall survival now at a median follow-up of 48 months in 29 evaluable patients, who received at least four doses of HSPPC-96, were comparable with those reported by our and other groups for similar patients undergoing the same kind of surgery followed by different types of chemotherapeutic regimens. However, a different clinical outcome was reported between the group of subjects showing an increased, class I HLA-restricted T-cell response to colorectal carcinoma cells (n = 17) via the enzyme-linked immunospot assay and those who did not show such a response (n = 12). Immune responders had a statistically significant survival advantage for both disease-free and overall survival at 48 months (Table 1)⇓ . Although, because of the limited number of patients, a possible influence of other prognostic factors could not be excluded, these data suggest that HSPPC-96 vaccination not only generated an antitumor immune response in >50% of the subjects but that it may also have contributed to improve patient’s clinical outcome.
Other Current Clinical Trials with HSPPCs
To date, several other clinical studies have been completed or are ongoing investigating the clinical benefit of autologous cancer-derived HSPPC-96 for patients in either the adjuvant or metastatic disease settings (see ref. 29 ). In fact, there are currently >150 medical centers worldwide enrolling patients in randomized, controlled phase III clinical trials testing this vaccine for the treatment of renal cancer and melanoma. In addition, autologous HSPPCs are being tested in phase I and II trials of chronic myelogenous leukemia, non-Hodgkin’s lymphoma, non–small-cell lung, pancreatic, and gastric cancers (Table 2)⇓ .
In an attempt to investigate options to increase the immunogenicity of HSPPC-96, we started a new trial in metastatic melanoma patients (30) by administering HSPPC-96 together with the granulocyte macrophage colony-stimulating factor given at the site of vaccination for 3 subsequent days (−1, 1, +1, 1 being the day of vaccine) and followed by two s.c. injections of 3 × 106 IU of IFN-α2b to up-regulate the expression of class I HLA often decreased in metastatic melanoma cells (3) . This regimen was given four times weekly, and after 4 weeks of rest, a second cycle of biweekly vaccinations was repeated. The immune response both in terms of frequency and intensity of induction of anti-TAA T cells was not increased as compared with the previous study carried out in similar melanoma patients. Table 3⇓ summarizes the anti-melanoma T-cell response, which was found either significantly increased or de novo activated after vaccination with HSPPC-96. With the possible exception of the recent study of HSPPC-96 given with the granulocyte macrophage colony-stimulating factor, the frequency of subjects showing an immune response was reproducible (48 to 52%), a finding of interest if one considers that most of the patients, particularly those with melanoma, had a significant tumor burden.
One of the unsolved issues of vaccination is the identification of surrogate markers predictive for clinical response, which may be found in one or more quantitative ex vivo immunologic assays. However, an association between T-cell response and clinical outcome begins to emerge from some of the most recent vaccination trials (19 , 20 , 31 , 32) . In our trials with HSPPC-96, an association was found between class I HLA-restricted antitumor T-cell responses and clinical outcome, although a higher number of clinically responding patients is necessary to corroborate these early findings (19 , 20) .
Safety has been carefully investigated in the studies of vaccination with HSPPCs because of the possibility that self-TAA contained in such preparations could potentially trigger autoimmune reactions. However, no significant laboratory or clinical signs of autoimmunity were reported in the clinical trials done, to date, in >800 patients treated. The safety profile of HSPPC-96 is mostly characterized by mild and transient side effects such as local reactions or low-grade fever. This allows a reasonably good quality of life during treatment.
Limitations of the HSP vaccination approach consist of the requirement of tumor tissue to be removed by surgery and the time necessary for vaccine manufacturing, which may delay or even prevent (in case of disease progression) the treatment. Strategies are being developed at the preclinical and clinical level to solve these problems even with the use of recombinant HSPs and gene manipulation of tumor cells (33 , 34) .
HSPs represent an interesting family of proteins that may function as in vivo chaperones of TAA epitopes carrying peptides, although such conclusion requires additional experimental evidence. However, it appears clear that at least some T-cell epitopes deriving from human TAAs are associated with gp96 and potentially endowed with the ability to stimulate antitumor T cells. The alternative explanation that this may occur through the activation of innate immunity that will then favor the triggering of TAA-specific T cells (35) is difficult to rule out in vivo but can be excluded in most in vitro representation assays with TAA-specific T-cell clones. Thus, at the moment, we conclude that gp96 and HSP70 preparations obtained from human tumors do contain several of the known TAA peptides, which can be recognized by the patients’ immune system.
However, the HSPPC vaccination approach is based on the principle that HSPPCs can potentially bind the full repertoire of the TAAs of a single tumor, including the individual strong antigens that make each tumor antigenically different from the other. By this mechanism, tumor-derived HSPPCs should provide vaccines tailored to the individual patient’s cancer.
We thank the expert technical work of Agata Cova, Gloria Sovena, and Paola Squarcina and the editing assistance of Grazia Barp.
Grant support: Antigenics, Inc. (New York, NY).
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.
Requests for reprints: Giorgio Parmiani, Unit of Immunotherapy of Human Tumors, Istituto Nazionale Tumori, Via Venezian 1, 20133 Milan, Italy. Phone: 39-0223902328; Fax: 39-0223902630; E-mail:
- Received June 29, 2004.
- Revision received August 27, 2004.
- Accepted September 2, 2004.