- Email: [email protected]
Cancer vaccines: challenges and potential solutions
Cancer vaccines: challenges and Almost a century has passed since immunotherapy of cancer was first attempted using cancer immunogens (vaccines); however, its clinical impact remains modest. Although initial concerns about a lack of human tumor antigens have decreased, prevailing issues include inefficient procedures for immunization and downregulated expression of major histocompatibility complex (MHC) class I molecules in tumor cells. While immunization can be improved, deficient MHC class I expression remains a problem, because it hampers the ability of tumor cells to present antigens for killing by CD8’ T cells. These are the major mediators of tumor destruction, and they have little or no activity against antigen-negative bystander cells. However, there are reasons to be optimistic that therapeutic vaccination against cancer antigens might become a reality at last. Therapeutically relevant tumor antigens were first demonstrated in the 1950s (Ref. 1). Mice were transplanted with cells from a chemically induced, syngeneic sarcoma. This was followed, a few days later, by surgical removal of the outgrowing tumor nodules and challenge with a small number of cells from the same sarcoma. These were rejected while cells from a different sarcoma were not. The data thus implied that there were tumor-specific transplantation (or rejection) antigens and that these (in the models studied) were unique for each tumor. Subsequent studies demonstrated rejection antigens among a large variety of sarcomas, carcinomas and lymphomas that had been chemically induced in inbred mice and rats. Tumors induced by certain oncogenic viruses, such as polyoma or SV40, were also found to have rejection antigens, which were shown to cross-react among tumors induced by the same virus.
Do naturally occurring tumors lack tumor antigens? Those tumors that were rejected by the immune system in the initial studies had been induced by 286
either a very large dose of chemical carcinogen or by a laboratory virus, such as polyoma, that does not cause tumors in nature. When the same immunization procedures were applied to mouse tumors that had been induced by small doses of chemical carcinogens, or against spontaneously occurring cancers, no rejection was observed. This led to the belief that naturally occurring neoplasms do not possess antigens that can serve as inducers and/or targets for a tumor-destructive immune response, although immunological reactions mediated by either lymphocytes or antibodies to cultivated human tumors had already been reported in the late 1960s (Ref. 1).
Improved immunization reveals that many tumors are immunogenic Methods to detect tumor rejection antigens have been vastly improved over the past 10-l 5 year@ by immunizing animals with tumor cells that have been experimentally manipulated to become more antigenically ‘foreign’, followed by testing against the original (wild-type) tumors. Such manipulations have included (1) treatment with certain mutagens, (2) infection with a virus Copyright
01997
Elsewer
Science
(such as Newcastle disease virus), or (3) transfection of tumor cells with a gene encoding a lymphokine [such as interleukin 2 (IL-2)] or a costimulatory molecule (such as 87-l; see below). As a result of using improved immunogens, many tumors that had been considered nonimmunogenie were found to be rejectable by the immune system.
Tumor peptides as T-cell epitopes A large variety of peptide epitopes for CD8+ T cells have been demonstrated in both animal and human cancers as potential targets for an in vitro cytolytic T lymphocyte (CTL) activity. Several of these can be present on the same tumor cell and some have high, or even absolute, tumor specificity2+. Examples include peptide epitopes encoded by: (1) human papillomavirus 16 (HPV16); (2) a mutated form of the cellular oncogene RAS; (3) mutated forms of the TP53 tumor suppressor gene: (4) a mutated form of the cyclin CDK4; and (5) an amplified form of the HER-2 oncogene. Other epitopes are associated with differentiation antigens, such as MAGE-1; without being entirely tumor specific, several differentiation antigens
Ltd. All nghth reserved.
1357
4310/97/$17
00
PH: Sl357-4310(97)01048-4
MOLECULAR
MEDICINE
TODAY.
JUL.Y
Opinion
1‘197
are sufficiently tumor selective to provide therapeutic targets. Epitopes encoded by genes involved in neoplastic transformation are of the greatest interest, however, because they are least likely to be lost during tumor progression6.
How antigens induce a tumordestructive immune response Presentationby MHC class1 Both CD4+ and CD8’T cells (and, to a lesser extent, antibodies) are involved in a tumordestructive immune response. The CD4+ cells (particularly those belonging to the helper 1 family, Thl) produce lymphokines (such as IL-2) that expand CD8’ T-cell populations. They also produce interferon y (IFN-?I), which activates natural killer (NK) cells and macrophages and can upregulate MHC class I expression. The CD& T cells have attracted particular attention because of the demonstration, in various animal models, that such lymphocytes with in vitro CTL activity play the major role in tumor rejection. These cells recognize peptide epitopes presented by MHC class I molecules, to kill epitope-positive cells specifically. As a first step, intracellular antigen is broken down by proteasomes into short peptides, which are usually nine residues long. These peptides are inserted into the endoplasmic reticulum by the transporters associated with antigen presentation (TAP-l and TAPB), where they form ternary complexes with p2 microglobulin (&m) and MHC class I heavy chains. These are then carried to the cell surface. The effective amount of a properly folded, peptide-loaded MHC class I molecule is a function of the concentration of each of the three components of the complex. A deficiency in any of several genes encoding proteins in this pathway can hamper the ability of a cell to present antigen to CD8+ T cells. Cells that are deficient in TAP-l, for example, have reduced MHC class I expression at their surface. The addition of large concentrations of ‘presentable’ peptide stabilizes the MHC class Ipeptide complex in aTAP-mutant B-cell line’. The products of different MHC alleles differ in their ability to bind TAP-1 and TAP-P. For example, the human leukocyte antigens HLA-A and HLA-C bind tightly to TAP while HLA- B products, in general, do note. A few antigens, most notably the peptide repeats that constitute part of mucin antigens, can serve as CTL targets without MHC class I restrictiong.
Co-stimulatory molecules For an antigen to induce an immune response that can lead to tumor destruction, it must be presented in the presence of appropriate costimulatory, or second signal(s). The interaction between CD28 on T cells and its ligands 87-l
and/or 87-2 on antigen-presenting cells (APCs) provides a particularly important signal. Antigen presented in its absence will be ‘ignored’ by the immune system, and it can even induce unresponsiveness or anergylO,‘l. A tumor can present its antigens in two ways to T cells: (1) directly, via MHC molecules at the tumor-cell surface; and (2) indirectly, after the antigens have been taken up by professional APCs, such as dendritic cells, which possess various co-stimulatoty molecules, including 87-l and 87-2. Most tumors fail to express molecules of the 87 family, i.e. they present antigen in an ineffective and potentially anergyinducing way. This might explain why even highly antigenic tumors can avoid the immune system of an immunocompetent host, and has contributed to the difficulties detecting tumor antigens, which can induce a rejection response3. Methods have been developed to circumvent this problem3. They include immunization with tumor cells transfected with genes encoding 87-l or 87-2, or with a recombinant virus that can deliver a tumor epitope together with 87. By using such immunogens, the outgrowth of wildtype (i.e. B7-) tumor cells can be prevented, and mice with multiple, but small, actively growing, syngeneic tumors (in the lung, for example) can be cured. Further improvement of cancer vaccines can be accomplished by using immunogen in which tumor antigen is combined with two different types of co-stimulatory molecules. When 87-i was transfected together with CD48 (the mouse counterpart of human CD58 and CD59) into Ag104 mouse sarcoma cells, which have very low immunogenicity, the doubly transfected cells induced rejection of wild-type Ag104 tumors. By contrast, transfection of either B7-1 or CD48 alone was not efficient3. Another approach to obtain appropriate costimulatory signals is to deliver tumor antigen to professional APCs effectively, such as dendritic cells, which express 87-l and other costimulatory molecules and can present antigen exogenously via MHC class I and class II molecules.
Loss of MHC class I expression in tumors lmmunohistologyand fluorescence-activated cell sorting (FACS) analysis, which are two assays based on antibody binding, have been widely used to analyze the presence of MHC class I molecules in human cancer cells12-14.According to these assays, most human carcinomas growing in sifuexpress MHC class I molecules. However, this expression is commonly downregulated, or even lost, as the tumors grow invasively and
metastasize, although this occurs to a different extent for different tumors. There is often a correlation between downregulated MHC class I expression and poor prognosis. Loss of an entire MHC class I haplotype has been observed in some animal and human tumor cell lines, although a partial loss is more frequent. In some cases, MHC class I losses are detected in most or even all cells of a tumor while in other cases the MHC class l-deficient cells constitute only a few per cent. Loss of enzymes involved in antigen processing and presentation, such as TAP-I, has also been demonstrated, and we suggest that such loss is frequently responsible for the downregulated expression of MHC class I molecules at the cell surface. If the TAP expression in a tumor is low, there will be a selective deficiency of HLA-B locus proteins. This is predicted because, while most HLA-A and -B locus products require TAP for charging with peptide and for transport to the cell membrane, HLA-B locus products show a lower affinity for TAP. Thus, as tumor cells progress and produce less TAP, the HLA-B locus products lose the competition for the limited supply of these TAP moJecules12. Conversely, HLAA2 products are less sensitive to shortage of TAP because they can use their own signal peptides in place of an immunogenic peptide; their expression is therefore TAP independenF and is rarely lost by tumor cells. Loss of MHC class I expression can be the outcome of a series of mutations followed by the selection of tumor cells that are resistant to lysis by T cells. Restifo et al. infected cultured cells from various tumors with the influenza virus and found that some tumor lines could present the influenza-virus-derived epitopes needed for lysis by influenza-specific and MHC class l-matched CTLs, while other lines had lost this ability16. Lehman eta/. reported that tumor metastases occurring after otherwise successful immunotherapy for melanoma were composed of MHC class I- cells, indicating that there was immunoselection for MHC class I- (and hence non-antigenic) tumor ceW?
Growth rate of cells and de ltovo production of MHC class I molecules Although loss of MHC class I molecules in humans can be a consequence of immunoselection, it is contrary to the fact that even highly antigenic rodent tumors propagated by transplantation in syngeneic, immunocompetent hosts can present their rejection antigens to serve as immunological targets (see below). Therefore, we suggest an additional explanation based on certain inherent differences between tumor cells and professional APCs. Professional APCs, such 287
Opinion
0 What are the mechanisms responsible for the low major histocompatibility complex (MHC) class I expression in human cancers,and can they be overcome? l How often is a systemic, dowaregulated abiity to present tumor peptides an earlv event in tumar formation an& if so, why? l To what extent will losses of MHC class I expression invalidate active immunotherapy of early and late cancers? l How~~~~~~~planted mouse tumors to understand immune responsesto hun3ancancer? 0 IfanapparentlackofMHCclassI on tumor cells is actually due to reduced function of the transporter associated with antigen beg (TAP), can peptides be designed FM, as is the case for HLA-A2, are present4 in a TAPindependent manner?
as dendritic cells, are not dividing while they are presenting antigen, i.e. the ratio of recycled versus newly synthesized MHC class I molecules is high, which is favorable for antigen presentation. Conversely, proliferating tumor cells need to produce large amounts of new MHC class I molecules to produce amounts equivalent to those on professional APCs. This alone might account for the reported deficiency of MHC class I expression in tumor cells. Although tumor cells can upregulate the synthesis of MHC class I heavy chain in response to a shorter cell-cycle time, they might not be able to alter the inherent timing of post-translational events, such as peptide loading onto MHC molecules. Unloaded MHC class I heavy chain could therefore be degraded before reaching the cell surface’*. Moreover, the production of proteasome proteins and TAPS might not be upregulated to the same extent as the synthesis of MHC class I molecules. This suggests that newly synthesized MHC class I molecules will not be properly loaded with peptide and will thus be degraded in the endoplasmic reticuIum18. As a tumor progresses towards faster growth, which is commonly the case in metastases, the reduction in MHC class I expression could worsen, particularly for peptides loaded via HLA-B. The observed loss of MHC class I molecules raises concern that therapeutic procedures aimed to activate CD8+ T cells might be ineffective, particularly in metastatic cancer. However, 288
we have a guarded optimism that the problem might be less severe than many anticipate it to be. In the remainder of this article, we summarize why we hold this view.
Total loss of MHC class I is rare A key question is the extent to which tumors, growing in viva, completely lose MHC alleles and/or other molecules responsible for antigen processing and intracellular transport to make them resistant to any response mediated by CD& T cells with in vitro CTL activity. Studies in syngeneic rodents can provide opportunities to detect such complete losses, to an extent not feasible by applying an approach based on antibody-binding assays. They can also be useful to assess the sensitivity of tumor cells to destruction by T cells sensitized to tumor antigens. There is surprisingly little recent information on this; however, two different sets of experiments with mouse tumors are pertinent, although they were published more than 30 years ago. Klein et al. performed transplantation experiments with tumors from several congeneic mouse strains that differed only at the H-2 (i.e. MHC class I) locus, using immunoselection by the highly sensitive allograft reaction to detect variants that had lost their expression of H-2 alleleslg. They found that tumor cells originating in Fl hybrids between two strains that differed at H-2 could sometimes lose the H-2 alleles from one of the two parental strains and, furthermore, that a loss of some individual H-2 components occurred with a frequency suggesting that it was based on mutation and/or chromosome deletion. However, no variants were ever obtained that had completely lost H-2 and were therefore capable of growing in preimmunized mice (Fig. 1). This is noteworthy because some 100 different neoplasms, including sarcomas, carcinomas and lymphomas, were tested repeatedly for outgrowth in various congeneic strains. Experiments performed by Sjijgren on mouse polyoma tumors also have a bearing on the present discussionzO. Polyoma tumors are normally rejected when transplanted into syngeneic mice preimmunized against the polyoma-virusencoded rejection antigen, while control tumors grow just as well in preimmunized mice as they do in non-immune mice. The rejection, like that of other DNA-virus-induced neoplasms, is mediated primarily by CD8’T cells that are MHC class I restrictedzl. Sjogren attempted to select for cells from polyoma tumors that had lost their rejection antigen by transplanting sufficiently large doses of polyoma-induced tumor cells into polyomaimmune mice; this resulted in a few successful transplants that could override the immune response. From these successful engraftments,
cells were further transplanted to polyomaimmune mice, a procedure that was repeated for >20 passages. No variants could be selected that had lost the polyoma tumor rejection antigen (Fig. l), i.e. the tumors retained MHC class I molecules to the extent needed to present this antigen.
Low expression of tumor peptides is sufficient for lysis by activated T cells Naive T cells need to be exposed, via MHC molecules, to a relatively high level of antigenic peptide for induction of an immune response. By contrast, a low level of peptide presented by MHC class I molecules is sufficient for lysis of target cells by activated CTLs. This probably explains why mouse tumor cells that appear to be MHC class I- when assayed by immunohistology and FACS, and that cannot induce a T-cell response, are often killed as targets of sensitized CD8 T cells. Porgador et al. published a set of experiments that strikingly demonstrate this situation**. They worked with an MHC class If, highly metastatic mouse tumor that was nonimmunogenic when tested by a standard protocol. Transfection of the missing MHC class I allele into this tumor made it both nonmetastatic and capable of inducing an immune response that could eradicate wild-type (MHC class I-) cells from the same tumor. The same could be accomplished by transfection of a gene encoding IFN-y. Although the level of MHC class I expression by wild-type tumor cells was too low for detection by binding assays, it was sufficient for the presentation of tumor peptide necessary for destruction by CD8’ T cells, both in vitro and in viva. Similar results have been obtained in other systems23~24.
Differences between mouse models and human cancer Several factors might contribute to the discrepancy between the retention of MHC class I among mouse tumors and its apparent loss in naturally occurring tumors in humans. Differences in the techniques employed to detect MHC class I provide one explanation. This is because small amounts of any H-2 (i.e. MHC class I) component might lead to the rejection of tumor cells transplanted to a mouse preimmunized against the whole H-2, whereas cells that have only undergone partial losses remain undetected. Differences might also be accentuated by the fact that many experiments with human material have been performed using cultured cells in vitro; by contrast, the mouse tumors were propagated in t&o, where MHC class I- cells are probably inhibited by exposure to NK cellsz5. We have already argued that unoccupied MHC class
.MOLECULAR
MEDICINE
TODAY.
,,I,.,’
Opinion
,9‘17
a
b
H-2a,
.
H-2b
,/ /
\:
in non-immune
Figure 1. Schematic representation of the experiments of (a) Klein et a/.l9 and (b) Sj6greP. (a) Klein et a/. performed transplantation experiments with tumors from several congeneic mouse strains that differed only at the H-2 locus (the mouse equivalent of MHC class I), and detected variants that had lost their expression of H-2 alleles. Tumor cells originating in Fl hybrids between two strains that differed at H-2 sometimes lost the H-2 alleles from one of the two parental strains, but no variants were ever obtained that grew in preimmunized mice because they had completely lost H-2. (b) Sjiigren attempted to select for cells
I is, in part, the result of a failure of TAP or proteasome elements to ‘keep up’ with rapidly growing tumor cells. Tumors commonly do not grow as fast in vivo as do cell lines propagated in vitro, and therefore have a greater opportunity to be loaded with antigen. There might also be differences in the extent to which MHC class I expression depends on the activity of TAP and proteasomes. The affinity of mouse MHC heavy chain for p,m is lower than that for the homologous pair in humans and this might lead to differences in the sensitivity of tumors from mice versus humans to lysis by CTLs. The extent to which MHC class I losses will invalidate those forms of therapy that are primarily mediated by CTLs might not be known until
from polyoma-virus-induced tumors that had lost their rejection antigen by transplanting large doses of polyoma-induced tumor cells into polyoma-immune mice; this resulted in a few successful transplants that could override the immune response. From these successful engraftments, cells were further transplanted to polyoma-immune mice, a procedure that was repeated for >20 passages. No variants could be selected that had lost the polyoma tumor rejection antigen, i.e. the tumors retained MHC class I molecules to the extent needed to present this antigen.
immunogens that effectively activate CD8+ T cells have been evaluated in humans. Our prediction is that losses of individual MHC class I sequences will be observed relatively frequently while losses of the entire MHC class I repertoire will be rare. If that is the case, ‘cocktails’ of peptides or the use of whole antigen proteins might be necessary (and sufficient) to induce therapeutic benefit.
blood lymphocytes, to help select immunogens containing peptides that can be presented by the MHC class I molecules present in an individual patient. Because of the downregulation of MHC class I molecules during tumor progression, tumor cells from patients who will be subjected to T cell-based immunotherapy should also be typed.
Conclusions and future directions Matching peptide immunogens patient’s haplotype
to a
A different problem arises when a cancer patient lacks the MHC haplotype that is needed for presentation of a particular tumor-derived peptide. Obviously, it is necessary to HLA-type peripheral
Experiments in animals indicate that cancer cells can present antigen, to the extent needed to serve as targets for CD8+ CTLs, more often than is indicated by immunohistological analysis of MHC class I expression on the tumor cells. Nevertheless, deficient presentation of tumor 289
Opinion
antigen is probably an obstacle to successful CTL-based therapy of many cases of metastatic cancer. Because losses of individual MHC class I alleles are more likely than complete losses of MHC class I, a combination of many different peptides will probably be needed as a tumor immunogen to allow for presentation by many different MHC class I sequences. The problem of MHC class I loss is less important for carcinomas in situ, and preventive vaccination might become a reality for some neoplasms; for example, epitopes encoded by the E6 or E7 genes of HPVIG, or those encoded by some mutated cellular genes closely associated with neoplastic transformation, might be included in prophylactic vaccines. It should be possible to explore the type and frequency of events that interfere with the ability of tumor cells to present peptides at the level needed for their destruction. For example, the relative amounts of TAP proteins in various tumors can be compared with that in professional APCs, and antibodies are available to study whether there is a direct relationship between tumor immunogenicity and the levels of TAP-1 and TAP-P. Proteasome activity can now be measured using specific synthetic protease substrates. It will also be worthwhile to investigate how to upregulate the presentation of T-ceil epitopes by cancer cells, to induce and target immune responses. Systemic administration of IFN-y can upregulate a low level of antigen presentation, but it is not likely to rectify a complete lack of MHC class I expression. It should be borne in mind that MHC class I(and therefore CTL-resistant) tumor cells usually remain sensitive to destruction by immunological effecters other than CTLs. These include NK ceils, macrophages, antibodies and various lymphokines such as tumor necrosis factor. CD4’ Thi cells also have therapeutic potential because they recognize tumor antigens presented by tumor-infiltratingAPC9. ‘Targeting’of an antitumor response, using bispecific antibodies or fusion proteins comprising an antibody sequence recognizing a tumor antigen combined with an immunopotentiator (such as granulocytemacrophage colony-stimulating factor or IL-12) can provide an alternative to activate tumoricidal mechanisms at the tumor site. Because one method by which tumors escape from immune destruction is the production of T-cell inhibitory molecules such as transforming growth factor p and Fas ligand, drugs that can abrogate such inhibition should further improve the effects of therapeutic tumor vaccination. References
290
4 5
6
7
8
9
10
11 12
13
14
15
16
17
18
1 HellstrBm, K. and Hellstriim, munity against tumor Cancer Res. 12,167-223
3
I. (1969) Cellular
specific antigens,
imAdv.
19
Boon, T. (1992) Toward a genetic analysis of tumor rejection antigens, Ann. Inst. Pasteur Immunol. 58,177-210 Hellstrijm, K.E., Hellstriim, I. and Chen, L. (1995) Can co-stimulated tumor immunity be therapeutically efficacious? Immunol. Rev. 145, 123-145 Pardoll, D.M. (1994) Tumor antigens: a new look for the 198Os, Nature 369, 357-358 Boon, T. and van der Bruggen, P. (1996) Human tumorantigens recognized by 1 lymphocytes, J. Exp. Med. 183, 1173-l 183 Hellstriim, K.E. and Hellstrdm, I. (1989) Oncogeneassociated tumor antigens as targets for immunotherapy, FASEB J. 3, 1715-l 720 Mellins, E. eta/. (1991) Agene required for class II-restricted antigen presentation maps to the major histocompatibility complex, J. Exp. Med. 174,1607-1615 Neisig, J. eta/. (1996) Allele-specific differences in the interaction of MHC class I molecules with transporters associated with antigen processing, J. Immunol. 156,3196-3206 Finn, O.J. et a/. (1995) MUG1 epithelial tumor mucin-based immunity and cancer vaccines, Immunol. Rev. 145,61-89 Linsley, P.S. and Ledbetter, J.A. (1993) The role of the CD28 receptor during T cell responses to antigen, Annu. Rev. Immunol. 11, 191-212 June, C.H. eta/. (1994) The B7 and CD28 recep tor families, Immunol. Today 15, 321-331 Marincola, F.M. et a/. (1994) Loss of HLA haplotype and B locus down-regulation in melanoma cell lines, J. Immunol. 153, 1225-1237 Garrido, F. et a/. (1993) Natural history of HLA expression during tumor development, Immunol. Today 14,491-499 Stern, P.L. and Duggan-Keen, M. (1994) MHC expression in the natural history of cervical cancer, in Human Papi//omaviruses and Cervical Cancer: Bio/ogy and Immunology (Stern, P. and Stanley, M., eds), pp. 162-176, Oxford University Press Wei, M.L. and Cresswell, P. (1992) HLA-A2 molecules in an antigen-processing mutant cell contain signal sequence-derived peptides, Nature 356, 443-446 Restifo, N. P. el al. (1993) Identification of human cancers deficient in antigen processing, J. Exp. Med. 177,265-272 Lehmann, F. eta/. (1995) Differences in the antigens recognized by cytolytic T cells on successive metastases of a melanoma patient are consistent with immune selection, Eur. J. Immunol. 25,340-347 Elliott, T. (1996) Tapping into tumours, Naf. Genet. 13, 139-l 40 Klein, E., Klein, G. and Hellstrijm, K.E. (1960) Further studies on isoantigenic variation in
20
21
22
23
24
25
26
mouse carcinomas and sarcomas, J. Nat/. Cancer Insf. 25,271-274 SjGgren, H.O. (1965) Transplantation methods as a tool for detection of tumor specific antigens, Prog. Exp. Tumor Res. 6289-296 Berke, Z. et a/. (1996) A short peptide eluted from the H-2Kb molecule of a polyomaviruspositive tumor corresponds to polyomavirus large T antigen peptide at amino acids 578 to 585 and induces polyomavirus-specific immunity, J. Viral. 70, 3093-3097 Porgador, A. ef a/. (1995) Combined vaccination with major histocompatibility class I and interleukin 2 gene-transduced melanoma cells synergizes the cure of postsurgical established lung metastases, Cancer Res. 55, 4941-4949 Restifo, N.P. et a/. (1992) A nonimmunogenic sarcoma transduced with the cDNA for interferon y elicits CD8+ Tcells against the wild-type tumor: correlation with antigen presentation capability, J. Exp. Med. 175, 1423-1431 Van den Driessche, T. et a/. (1994) Metastasis of mouse T lymphoma cells is controlled by the level of major histocompatibility complex class I H-2Dk antigens, Int. J. Cancer 58,217-225 Karre, K. (1995) Express yourself or die: peptides, MHC molecules, and NK cells, Science 267,978-979 Greenberg, P.D. (1991) Adoptive T cell therapy of tumors: mechanisms operative in the recognition and elimination of tumor cells, Adv. Immunol. 49,281-355
Do naturally occurring tumors lack tumor antigens? Those tumors that were rejected by the immune system in the initial studies had been induced by 286
either a very large dose of chemical carcinogen or by a laboratory virus, such as polyoma, that does not cause tumors in nature. When the same immunization procedures were applied to mouse tumors that had been induced by small doses of chemical carcinogens, or against spontaneously occurring cancers, no rejection was observed. This led to the belief that naturally occurring neoplasms do not possess antigens that can serve as inducers and/or targets for a tumor-destructive immune response, although immunological reactions mediated by either lymphocytes or antibodies to cultivated human tumors had already been reported in the late 1960s (Ref. 1).
Improved immunization reveals that many tumors are immunogenic Methods to detect tumor rejection antigens have been vastly improved over the past 10-l 5 year@ by immunizing animals with tumor cells that have been experimentally manipulated to become more antigenically ‘foreign’, followed by testing against the original (wild-type) tumors. Such manipulations have included (1) treatment with certain mutagens, (2) infection with a virus Copyright
01997
Elsewer
Science
(such as Newcastle disease virus), or (3) transfection of tumor cells with a gene encoding a lymphokine [such as interleukin 2 (IL-2)] or a costimulatory molecule (such as 87-l; see below). As a result of using improved immunogens, many tumors that had been considered nonimmunogenie were found to be rejectable by the immune system.
Tumor peptides as T-cell epitopes A large variety of peptide epitopes for CD8+ T cells have been demonstrated in both animal and human cancers as potential targets for an in vitro cytolytic T lymphocyte (CTL) activity. Several of these can be present on the same tumor cell and some have high, or even absolute, tumor specificity2+. Examples include peptide epitopes encoded by: (1) human papillomavirus 16 (HPV16); (2) a mutated form of the cellular oncogene RAS; (3) mutated forms of the TP53 tumor suppressor gene: (4) a mutated form of the cyclin CDK4; and (5) an amplified form of the HER-2 oncogene. Other epitopes are associated with differentiation antigens, such as MAGE-1; without being entirely tumor specific, several differentiation antigens
Ltd. All nghth reserved.
1357
4310/97/$17
00
PH: Sl357-4310(97)01048-4
MOLECULAR
MEDICINE
TODAY.
JUL.Y
Opinion
1‘197
are sufficiently tumor selective to provide therapeutic targets. Epitopes encoded by genes involved in neoplastic transformation are of the greatest interest, however, because they are least likely to be lost during tumor progression6.
How antigens induce a tumordestructive immune response Presentationby MHC class1 Both CD4+ and CD8’T cells (and, to a lesser extent, antibodies) are involved in a tumordestructive immune response. The CD4+ cells (particularly those belonging to the helper 1 family, Thl) produce lymphokines (such as IL-2) that expand CD8’ T-cell populations. They also produce interferon y (IFN-?I), which activates natural killer (NK) cells and macrophages and can upregulate MHC class I expression. The CD& T cells have attracted particular attention because of the demonstration, in various animal models, that such lymphocytes with in vitro CTL activity play the major role in tumor rejection. These cells recognize peptide epitopes presented by MHC class I molecules, to kill epitope-positive cells specifically. As a first step, intracellular antigen is broken down by proteasomes into short peptides, which are usually nine residues long. These peptides are inserted into the endoplasmic reticulum by the transporters associated with antigen presentation (TAP-l and TAPB), where they form ternary complexes with p2 microglobulin (&m) and MHC class I heavy chains. These are then carried to the cell surface. The effective amount of a properly folded, peptide-loaded MHC class I molecule is a function of the concentration of each of the three components of the complex. A deficiency in any of several genes encoding proteins in this pathway can hamper the ability of a cell to present antigen to CD8+ T cells. Cells that are deficient in TAP-l, for example, have reduced MHC class I expression at their surface. The addition of large concentrations of ‘presentable’ peptide stabilizes the MHC class Ipeptide complex in aTAP-mutant B-cell line’. The products of different MHC alleles differ in their ability to bind TAP-1 and TAP-P. For example, the human leukocyte antigens HLA-A and HLA-C bind tightly to TAP while HLA- B products, in general, do note. A few antigens, most notably the peptide repeats that constitute part of mucin antigens, can serve as CTL targets without MHC class I restrictiong.
Co-stimulatory molecules For an antigen to induce an immune response that can lead to tumor destruction, it must be presented in the presence of appropriate costimulatory, or second signal(s). The interaction between CD28 on T cells and its ligands 87-l
and/or 87-2 on antigen-presenting cells (APCs) provides a particularly important signal. Antigen presented in its absence will be ‘ignored’ by the immune system, and it can even induce unresponsiveness or anergylO,‘l. A tumor can present its antigens in two ways to T cells: (1) directly, via MHC molecules at the tumor-cell surface; and (2) indirectly, after the antigens have been taken up by professional APCs, such as dendritic cells, which possess various co-stimulatoty molecules, including 87-l and 87-2. Most tumors fail to express molecules of the 87 family, i.e. they present antigen in an ineffective and potentially anergyinducing way. This might explain why even highly antigenic tumors can avoid the immune system of an immunocompetent host, and has contributed to the difficulties detecting tumor antigens, which can induce a rejection response3. Methods have been developed to circumvent this problem3. They include immunization with tumor cells transfected with genes encoding 87-l or 87-2, or with a recombinant virus that can deliver a tumor epitope together with 87. By using such immunogens, the outgrowth of wildtype (i.e. B7-) tumor cells can be prevented, and mice with multiple, but small, actively growing, syngeneic tumors (in the lung, for example) can be cured. Further improvement of cancer vaccines can be accomplished by using immunogen in which tumor antigen is combined with two different types of co-stimulatory molecules. When 87-i was transfected together with CD48 (the mouse counterpart of human CD58 and CD59) into Ag104 mouse sarcoma cells, which have very low immunogenicity, the doubly transfected cells induced rejection of wild-type Ag104 tumors. By contrast, transfection of either B7-1 or CD48 alone was not efficient3. Another approach to obtain appropriate costimulatory signals is to deliver tumor antigen to professional APCs effectively, such as dendritic cells, which express 87-l and other costimulatory molecules and can present antigen exogenously via MHC class I and class II molecules.
Loss of MHC class I expression in tumors lmmunohistologyand fluorescence-activated cell sorting (FACS) analysis, which are two assays based on antibody binding, have been widely used to analyze the presence of MHC class I molecules in human cancer cells12-14.According to these assays, most human carcinomas growing in sifuexpress MHC class I molecules. However, this expression is commonly downregulated, or even lost, as the tumors grow invasively and
metastasize, although this occurs to a different extent for different tumors. There is often a correlation between downregulated MHC class I expression and poor prognosis. Loss of an entire MHC class I haplotype has been observed in some animal and human tumor cell lines, although a partial loss is more frequent. In some cases, MHC class I losses are detected in most or even all cells of a tumor while in other cases the MHC class l-deficient cells constitute only a few per cent. Loss of enzymes involved in antigen processing and presentation, such as TAP-I, has also been demonstrated, and we suggest that such loss is frequently responsible for the downregulated expression of MHC class I molecules at the cell surface. If the TAP expression in a tumor is low, there will be a selective deficiency of HLA-B locus proteins. This is predicted because, while most HLA-A and -B locus products require TAP for charging with peptide and for transport to the cell membrane, HLA-B locus products show a lower affinity for TAP. Thus, as tumor cells progress and produce less TAP, the HLA-B locus products lose the competition for the limited supply of these TAP moJecules12. Conversely, HLAA2 products are less sensitive to shortage of TAP because they can use their own signal peptides in place of an immunogenic peptide; their expression is therefore TAP independenF and is rarely lost by tumor cells. Loss of MHC class I expression can be the outcome of a series of mutations followed by the selection of tumor cells that are resistant to lysis by T cells. Restifo et al. infected cultured cells from various tumors with the influenza virus and found that some tumor lines could present the influenza-virus-derived epitopes needed for lysis by influenza-specific and MHC class l-matched CTLs, while other lines had lost this ability16. Lehman eta/. reported that tumor metastases occurring after otherwise successful immunotherapy for melanoma were composed of MHC class I- cells, indicating that there was immunoselection for MHC class I- (and hence non-antigenic) tumor ceW?
Growth rate of cells and de ltovo production of MHC class I molecules Although loss of MHC class I molecules in humans can be a consequence of immunoselection, it is contrary to the fact that even highly antigenic rodent tumors propagated by transplantation in syngeneic, immunocompetent hosts can present their rejection antigens to serve as immunological targets (see below). Therefore, we suggest an additional explanation based on certain inherent differences between tumor cells and professional APCs. Professional APCs, such 287
Opinion
0 What are the mechanisms responsible for the low major histocompatibility complex (MHC) class I expression in human cancers,and can they be overcome? l How often is a systemic, dowaregulated abiity to present tumor peptides an earlv event in tumar formation an& if so, why? l To what extent will losses of MHC class I expression invalidate active immunotherapy of early and late cancers? l How~~~~~~~planted mouse tumors to understand immune responsesto hun3ancancer? 0 IfanapparentlackofMHCclassI on tumor cells is actually due to reduced function of the transporter associated with antigen beg (TAP), can peptides be designed FM, as is the case for HLA-A2, are present4 in a TAPindependent manner?
as dendritic cells, are not dividing while they are presenting antigen, i.e. the ratio of recycled versus newly synthesized MHC class I molecules is high, which is favorable for antigen presentation. Conversely, proliferating tumor cells need to produce large amounts of new MHC class I molecules to produce amounts equivalent to those on professional APCs. This alone might account for the reported deficiency of MHC class I expression in tumor cells. Although tumor cells can upregulate the synthesis of MHC class I heavy chain in response to a shorter cell-cycle time, they might not be able to alter the inherent timing of post-translational events, such as peptide loading onto MHC molecules. Unloaded MHC class I heavy chain could therefore be degraded before reaching the cell surface’*. Moreover, the production of proteasome proteins and TAPS might not be upregulated to the same extent as the synthesis of MHC class I molecules. This suggests that newly synthesized MHC class I molecules will not be properly loaded with peptide and will thus be degraded in the endoplasmic reticuIum18. As a tumor progresses towards faster growth, which is commonly the case in metastases, the reduction in MHC class I expression could worsen, particularly for peptides loaded via HLA-B. The observed loss of MHC class I molecules raises concern that therapeutic procedures aimed to activate CD8+ T cells might be ineffective, particularly in metastatic cancer. However, 288
we have a guarded optimism that the problem might be less severe than many anticipate it to be. In the remainder of this article, we summarize why we hold this view.
Total loss of MHC class I is rare A key question is the extent to which tumors, growing in viva, completely lose MHC alleles and/or other molecules responsible for antigen processing and intracellular transport to make them resistant to any response mediated by CD& T cells with in vitro CTL activity. Studies in syngeneic rodents can provide opportunities to detect such complete losses, to an extent not feasible by applying an approach based on antibody-binding assays. They can also be useful to assess the sensitivity of tumor cells to destruction by T cells sensitized to tumor antigens. There is surprisingly little recent information on this; however, two different sets of experiments with mouse tumors are pertinent, although they were published more than 30 years ago. Klein et al. performed transplantation experiments with tumors from several congeneic mouse strains that differed only at the H-2 (i.e. MHC class I) locus, using immunoselection by the highly sensitive allograft reaction to detect variants that had lost their expression of H-2 alleleslg. They found that tumor cells originating in Fl hybrids between two strains that differed at H-2 could sometimes lose the H-2 alleles from one of the two parental strains and, furthermore, that a loss of some individual H-2 components occurred with a frequency suggesting that it was based on mutation and/or chromosome deletion. However, no variants were ever obtained that had completely lost H-2 and were therefore capable of growing in preimmunized mice (Fig. 1). This is noteworthy because some 100 different neoplasms, including sarcomas, carcinomas and lymphomas, were tested repeatedly for outgrowth in various congeneic strains. Experiments performed by Sjijgren on mouse polyoma tumors also have a bearing on the present discussionzO. Polyoma tumors are normally rejected when transplanted into syngeneic mice preimmunized against the polyoma-virusencoded rejection antigen, while control tumors grow just as well in preimmunized mice as they do in non-immune mice. The rejection, like that of other DNA-virus-induced neoplasms, is mediated primarily by CD8’T cells that are MHC class I restrictedzl. Sjogren attempted to select for cells from polyoma tumors that had lost their rejection antigen by transplanting sufficiently large doses of polyoma-induced tumor cells into polyomaimmune mice; this resulted in a few successful transplants that could override the immune response. From these successful engraftments,
cells were further transplanted to polyomaimmune mice, a procedure that was repeated for >20 passages. No variants could be selected that had lost the polyoma tumor rejection antigen (Fig. l), i.e. the tumors retained MHC class I molecules to the extent needed to present this antigen.
Low expression of tumor peptides is sufficient for lysis by activated T cells Naive T cells need to be exposed, via MHC molecules, to a relatively high level of antigenic peptide for induction of an immune response. By contrast, a low level of peptide presented by MHC class I molecules is sufficient for lysis of target cells by activated CTLs. This probably explains why mouse tumor cells that appear to be MHC class I- when assayed by immunohistology and FACS, and that cannot induce a T-cell response, are often killed as targets of sensitized CD8 T cells. Porgador et al. published a set of experiments that strikingly demonstrate this situation**. They worked with an MHC class If, highly metastatic mouse tumor that was nonimmunogenic when tested by a standard protocol. Transfection of the missing MHC class I allele into this tumor made it both nonmetastatic and capable of inducing an immune response that could eradicate wild-type (MHC class I-) cells from the same tumor. The same could be accomplished by transfection of a gene encoding IFN-y. Although the level of MHC class I expression by wild-type tumor cells was too low for detection by binding assays, it was sufficient for the presentation of tumor peptide necessary for destruction by CD8’ T cells, both in vitro and in viva. Similar results have been obtained in other systems23~24.
Differences between mouse models and human cancer Several factors might contribute to the discrepancy between the retention of MHC class I among mouse tumors and its apparent loss in naturally occurring tumors in humans. Differences in the techniques employed to detect MHC class I provide one explanation. This is because small amounts of any H-2 (i.e. MHC class I) component might lead to the rejection of tumor cells transplanted to a mouse preimmunized against the whole H-2, whereas cells that have only undergone partial losses remain undetected. Differences might also be accentuated by the fact that many experiments with human material have been performed using cultured cells in vitro; by contrast, the mouse tumors were propagated in t&o, where MHC class I- cells are probably inhibited by exposure to NK cellsz5. We have already argued that unoccupied MHC class
.MOLECULAR
MEDICINE
TODAY.
,,I,.,’
Opinion
,9‘17
a
b
H-2a,
.
H-2b
,/ /
\:
in non-immune
Figure 1. Schematic representation of the experiments of (a) Klein et a/.l9 and (b) Sj6greP. (a) Klein et a/. performed transplantation experiments with tumors from several congeneic mouse strains that differed only at the H-2 locus (the mouse equivalent of MHC class I), and detected variants that had lost their expression of H-2 alleles. Tumor cells originating in Fl hybrids between two strains that differed at H-2 sometimes lost the H-2 alleles from one of the two parental strains, but no variants were ever obtained that grew in preimmunized mice because they had completely lost H-2. (b) Sjiigren attempted to select for cells
I is, in part, the result of a failure of TAP or proteasome elements to ‘keep up’ with rapidly growing tumor cells. Tumors commonly do not grow as fast in vivo as do cell lines propagated in vitro, and therefore have a greater opportunity to be loaded with antigen. There might also be differences in the extent to which MHC class I expression depends on the activity of TAP and proteasomes. The affinity of mouse MHC heavy chain for p,m is lower than that for the homologous pair in humans and this might lead to differences in the sensitivity of tumors from mice versus humans to lysis by CTLs. The extent to which MHC class I losses will invalidate those forms of therapy that are primarily mediated by CTLs might not be known until
from polyoma-virus-induced tumors that had lost their rejection antigen by transplanting large doses of polyoma-induced tumor cells into polyoma-immune mice; this resulted in a few successful transplants that could override the immune response. From these successful engraftments, cells were further transplanted to polyoma-immune mice, a procedure that was repeated for >20 passages. No variants could be selected that had lost the polyoma tumor rejection antigen, i.e. the tumors retained MHC class I molecules to the extent needed to present this antigen.
immunogens that effectively activate CD8+ T cells have been evaluated in humans. Our prediction is that losses of individual MHC class I sequences will be observed relatively frequently while losses of the entire MHC class I repertoire will be rare. If that is the case, ‘cocktails’ of peptides or the use of whole antigen proteins might be necessary (and sufficient) to induce therapeutic benefit.
blood lymphocytes, to help select immunogens containing peptides that can be presented by the MHC class I molecules present in an individual patient. Because of the downregulation of MHC class I molecules during tumor progression, tumor cells from patients who will be subjected to T cell-based immunotherapy should also be typed.
Conclusions and future directions Matching peptide immunogens patient’s haplotype
to a
A different problem arises when a cancer patient lacks the MHC haplotype that is needed for presentation of a particular tumor-derived peptide. Obviously, it is necessary to HLA-type peripheral
Experiments in animals indicate that cancer cells can present antigen, to the extent needed to serve as targets for CD8+ CTLs, more often than is indicated by immunohistological analysis of MHC class I expression on the tumor cells. Nevertheless, deficient presentation of tumor 289
Opinion
antigen is probably an obstacle to successful CTL-based therapy of many cases of metastatic cancer. Because losses of individual MHC class I alleles are more likely than complete losses of MHC class I, a combination of many different peptides will probably be needed as a tumor immunogen to allow for presentation by many different MHC class I sequences. The problem of MHC class I loss is less important for carcinomas in situ, and preventive vaccination might become a reality for some neoplasms; for example, epitopes encoded by the E6 or E7 genes of HPVIG, or those encoded by some mutated cellular genes closely associated with neoplastic transformation, might be included in prophylactic vaccines. It should be possible to explore the type and frequency of events that interfere with the ability of tumor cells to present peptides at the level needed for their destruction. For example, the relative amounts of TAP proteins in various tumors can be compared with that in professional APCs, and antibodies are available to study whether there is a direct relationship between tumor immunogenicity and the levels of TAP-1 and TAP-P. Proteasome activity can now be measured using specific synthetic protease substrates. It will also be worthwhile to investigate how to upregulate the presentation of T-ceil epitopes by cancer cells, to induce and target immune responses. Systemic administration of IFN-y can upregulate a low level of antigen presentation, but it is not likely to rectify a complete lack of MHC class I expression. It should be borne in mind that MHC class I(and therefore CTL-resistant) tumor cells usually remain sensitive to destruction by immunological effecters other than CTLs. These include NK ceils, macrophages, antibodies and various lymphokines such as tumor necrosis factor. CD4’ Thi cells also have therapeutic potential because they recognize tumor antigens presented by tumor-infiltratingAPC9. ‘Targeting’of an antitumor response, using bispecific antibodies or fusion proteins comprising an antibody sequence recognizing a tumor antigen combined with an immunopotentiator (such as granulocytemacrophage colony-stimulating factor or IL-12) can provide an alternative to activate tumoricidal mechanisms at the tumor site. Because one method by which tumors escape from immune destruction is the production of T-cell inhibitory molecules such as transforming growth factor p and Fas ligand, drugs that can abrogate such inhibition should further improve the effects of therapeutic tumor vaccination. References
290
4 5
6
7
8
9
10
11 12
13
14
15
16
17
18
1 HellstrBm, K. and Hellstriim, munity against tumor Cancer Res. 12,167-223
3
I. (1969) Cellular
specific antigens,
imAdv.
19
Boon, T. (1992) Toward a genetic analysis of tumor rejection antigens, Ann. Inst. Pasteur Immunol. 58,177-210 Hellstrijm, K.E., Hellstriim, I. and Chen, L. (1995) Can co-stimulated tumor immunity be therapeutically efficacious? Immunol. Rev. 145, 123-145 Pardoll, D.M. (1994) Tumor antigens: a new look for the 198Os, Nature 369, 357-358 Boon, T. and van der Bruggen, P. (1996) Human tumorantigens recognized by 1 lymphocytes, J. Exp. Med. 183, 1173-l 183 Hellstriim, K.E. and Hellstrdm, I. (1989) Oncogeneassociated tumor antigens as targets for immunotherapy, FASEB J. 3, 1715-l 720 Mellins, E. eta/. (1991) Agene required for class II-restricted antigen presentation maps to the major histocompatibility complex, J. Exp. Med. 174,1607-1615 Neisig, J. eta/. (1996) Allele-specific differences in the interaction of MHC class I molecules with transporters associated with antigen processing, J. Immunol. 156,3196-3206 Finn, O.J. et a/. (1995) MUG1 epithelial tumor mucin-based immunity and cancer vaccines, Immunol. Rev. 145,61-89 Linsley, P.S. and Ledbetter, J.A. (1993) The role of the CD28 receptor during T cell responses to antigen, Annu. Rev. Immunol. 11, 191-212 June, C.H. eta/. (1994) The B7 and CD28 recep tor families, Immunol. Today 15, 321-331 Marincola, F.M. et a/. (1994) Loss of HLA haplotype and B locus down-regulation in melanoma cell lines, J. Immunol. 153, 1225-1237 Garrido, F. et a/. (1993) Natural history of HLA expression during tumor development, Immunol. Today 14,491-499 Stern, P.L. and Duggan-Keen, M. (1994) MHC expression in the natural history of cervical cancer, in Human Papi//omaviruses and Cervical Cancer: Bio/ogy and Immunology (Stern, P. and Stanley, M., eds), pp. 162-176, Oxford University Press Wei, M.L. and Cresswell, P. (1992) HLA-A2 molecules in an antigen-processing mutant cell contain signal sequence-derived peptides, Nature 356, 443-446 Restifo, N. P. el al. (1993) Identification of human cancers deficient in antigen processing, J. Exp. Med. 177,265-272 Lehmann, F. eta/. (1995) Differences in the antigens recognized by cytolytic T cells on successive metastases of a melanoma patient are consistent with immune selection, Eur. J. Immunol. 25,340-347 Elliott, T. (1996) Tapping into tumours, Naf. Genet. 13, 139-l 40 Klein, E., Klein, G. and Hellstrijm, K.E. (1960) Further studies on isoantigenic variation in
20
21
22
23
24
25
26
mouse carcinomas and sarcomas, J. Nat/. Cancer Insf. 25,271-274 SjGgren, H.O. (1965) Transplantation methods as a tool for detection of tumor specific antigens, Prog. Exp. Tumor Res. 6289-296 Berke, Z. et a/. (1996) A short peptide eluted from the H-2Kb molecule of a polyomaviruspositive tumor corresponds to polyomavirus large T antigen peptide at amino acids 578 to 585 and induces polyomavirus-specific immunity, J. Viral. 70, 3093-3097 Porgador, A. ef a/. (1995) Combined vaccination with major histocompatibility class I and interleukin 2 gene-transduced melanoma cells synergizes the cure of postsurgical established lung metastases, Cancer Res. 55, 4941-4949 Restifo, N.P. et a/. (1992) A nonimmunogenic sarcoma transduced with the cDNA for interferon y elicits CD8+ Tcells against the wild-type tumor: correlation with antigen presentation capability, J. Exp. Med. 175, 1423-1431 Van den Driessche, T. et a/. (1994) Metastasis of mouse T lymphoma cells is controlled by the level of major histocompatibility complex class I H-2Dk antigens, Int. J. Cancer 58,217-225 Karre, K. (1995) Express yourself or die: peptides, MHC molecules, and NK cells, Science 267,978-979 Greenberg, P.D. (1991) Adoptive T cell therapy of tumors: mechanisms operative in the recognition and elimination of tumor cells, Adv. Immunol. 49,281-355