Autologous heat-shock protein vaccines

Keywords: :

• gp96
• hsp70
• hsp90
• heat shock proteins
• cancer vaccines

Dear Sir,

In a recent article, Reitsma and Combest discuss the challenges in the development of an autologous heat shock protein (hsp) based anti-tumor vaccine.¹ They detail the clinical experience with Vitespen and highlight a number of issues with the various clinical trials that could explain some of the disappointing results obtained.¹ I would like to point out a more fundamental issue with the Vitespen vaccine that could explain the inconsistent results obtained in moving from experimental systems to clinical application. Vitespen is an hsp-peptide complex purified ex vivo from an individual patient’s tumor cells.² The approach is based on a number of animal model studies showing that hsp-peptide complexes can be utilized by antigen presenting cells, including the key dendritic cells (DCs), for the capture and presentation of antigens to efficiently elicit antigen-specific immunity.^3–5 However, it should be noted that while the earlier studies use glycoprotein 96 (gp96) as the hsp carrier,⁴ the majority of immunological studies use the abundant stress protein, hsp70, instead and it is the latter studies that have shown the unequivocal activation of both innate and acquired immune responses leading to the use of the term “chaperokine” to describe the immunity of hsp-peptide complexes.⁶ The importance of distinguishing between studies of gp96 and hsp70 is that they have distinct biological and biochemical properties that preclude the extrapolation between studies using these two different hsps.^4–6 Thus, while hsp70 is a non-glycosylated cytoplasmic protein,⁵ gp96 is a glycosylated organellar protein found in the lumen of the endoplasmic reticulum (ER).^4,6 Moreover, while gp96 is classified as a member of the hsp90 family on the basis of sequence homology,⁶ it should be noted that hsp90, like hsp70, is a non-glycosylated cytoplasmic protein^6,7 and it may also be erroneous to extrapolate conclusions derived from studies on hsp90 to gp96.

The origins of Vitespen are in early studies on tumor rejection antigens (TRAs) from chemically induced sarcomas.^8,9 These were shown to be tumor specific and shared a number of common properties as non-glycosylated cytoplasmic proteins that were also found on the cell surface of the tumor cells. Biochemical dissection of tumor cytosolic fractions resulted in the characterization of a number of TRAs and in 1986 two papers identified gp96 as a tumor specific antigen.^8,9. The first (from the Appella lab) identified a tumor-specific transplantation antigen as an abundant cytoplasmic protein homologous to hsp90 on the basis of N-terminal sequence identity,⁸ while the second (from the Old lab) identified a 96kD cytoplasmic TRA that showed antibody cross- reactivity between tumors, suggesting the presence of structurally related molecules expressed in different tumors.⁹ Molecular cloning of gp96, however, revealed only a single gene for the protein leading to the intriguing suggestion that the tumor specificity resided in peptides associated with gp96, rather than the hsp itself.¹⁰ However, it should be noted that although the Srivastava group showed that peptides were found associated with gp96¹¹, their studies on the immunogenicity of the hsp-peptide complexes utilized hsp70 from the tumors instead.¹² Moreover, while the initial studies on gp96 identified it as a cytoplasmic protein,^8,9 presumably non-glycosylated, subsequent studies culminating in the development of Vitespen used lectin affinity chromatography to purify gp96, clearly defining it as a glycoprotein.^2,3

This apparent paradox can be resolved if one assumes that the initial studies reported were on bona fide hsp90, a non-glycosylated cytoplasmic protein, while the subsequent studies were on the glycosylated ER-luminal protein gp96, a chaperone also studied as endoplasmin.⁶ This distinction between the two proteins is more consistent with the different results from the various studies reported using hsp90 or gp96.^4,6 Most importantly, their distinct functions as cytoplasmic and ER-luminal chaperones would mean that they bind quite different client proteins.^4,6 Moreover, as hsp90 binds a variety of cytoplasmic, non-glycosylated client proteins,^6,7 it is quite likely that isolation of this hsp from tumor cells will result in the purification of intact hsp-client protein complexes. In contrast, however, gp96 binds glycosylated membrane proteins such as the integrins and Toll-like receptors (TLRs)¹³ and, as the isolation methods do not use any detergents to solubilize membrane-bound proteins, it is unlikely that the purification of gp96 will yield hsp-client protein complexes.^2,3 It is thus more likely that any peptides found bound to gp96 in Vitespen are a result of promiscuous peptide binding rather than the carriage of true client proteins from the tumor.^6,14 In this context, it should be noted that as the pre-clinical experimental models used genetically homogeneous tumor tissue, the promiscuous peptides bound would be likely to show apparent consistency.^6,14 In contrast, the clinical trials use heterogeneous patient tumors and, as such, only a pool of variable peptides would be available for binding to gp96, thus eliciting only variable immunogenicity as reported.¹ This alternative explanation raises the possibility that the equivocal clinical results reported with Vitespen may reflect a more fundamental scientific issue, resulting from the product itself rather than clinical trial design. It is thus likely that the results from parallel clinical trials using hsp70 isolated from patient tumors,¹⁵ which should carry true client peptides, will shed light on the validity of the use of tumor-derive hsps as cancer vaccines.

Camilo Colaco