| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Ludwig Institute for Cancer Research, Austin & Repatriation Medical Centre, Heidelberg, Victoria, Australia
Correspondence: Ian D. Davis, Ludwig Institute for Cancer Research, Austin & Repatriation Medical Centre, Studley Rd., Heidelberg, Victoria 3084, Australia. E-mail: Ian.Davis{at}ludwig.edu.au
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONTHE ERA OF EMPIRICISMTHE ERA OF RATIONAL...THE ERA OF EFFECTIVE...REFERENCES |
|---|
Key Words: review • vaccines • dendritic cells • T cells • antigens • peptides
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONTHE ERA OF EMPIRICISMTHE ERA OF RATIONAL...THE ERA OF EFFECTIVE...REFERENCES |
|---|
At that time, the only treatment for inoperable cancer was the newly discovered power of radiation. Within a short time of its discovery, radiation was being used for the treatment of many different types of cancer. Despite the impressive successes of this new treatment, many patients did not respond or relapsed later, ultimately doomed to die from their disease. Systemic chemotherapy was not to be described until the 1940s [2 , 3 ].
William Coley, a surgeon at Memorial Hospital in New York, noted that some patients with cancer who developed and survived erysipelas (an often fatal disease at that time) went on to exhibit regression of their cancer. Coley believed that some property of the microorganisms involved in erysipelas mediated this antitumor effect, perhaps by vaccinating the patients against a putative infectious cause for their cancer. He developed an extract of Streptococcus and Serratia, "Coley’s mixed toxins," with which he treated an initial series of 10 patients [4 ] and eventually over 800 [5 ]. Although this treatment was crude and modern clinical trial methodology was still many decades away, a significant proportion of Coley’s patients showed regressions of their tumors and some of these regressions lasted for several decades. As late as 1934, "Coley’s toxins" was the only known systemic treatment for cancer [6 ]. However, the mechanism of action of "Coley’s toxins" was and remains unknown. It is probably a result of several factors that combine to induce a cytokine cascade, leading to specific and nonspecific immune responses.
In retrospect, "Coley’s toxins" was the first effective immunotherapeutic intervention for cancer and was based on an astute observation. Immunotherapy for cancer developed in this way for much of the remainder of the 20th century, as observations made in the fields of cancer or infectious disease or in transplantation medicine provided insights into potential ways of treating cancer patients using immunological approaches. However, without the appropriate tools and reagents, it was not possible to develop an understanding of the basic biology of these responses nor to understand why some patients responded and some did not. This was the era of empiricism in cancer immunotherapy.
In the late 1980s, the first T cell antigens for human cancer were described [7 ]. At the same time, Steinman and colleagues described the dendritic cell (DC) and began to elucidate its pivotal role in the control and maintenance of the immune response [8 ]. This led to a rapid expansion in the field, as new antigens were described. New technologies in molecular biology and the availability of reagents (particularly antibodies) to characterize cell-surface molecules allowed the mechanisms underlying antitumor immune responses to be revealed. The ensuing decade saw the evolution of a wide range of rationally designed, immunotherapeutic approaches for cancer, taking into account the expanding understanding of the field and the availability of new tools and reagents. This relatively short period has been the era of scientifically based translational clinical cancer immunotherapy. The intensity of this research has been striking, and vast resources have been invested in it. Although much progress has been made, this era has not been without its pitfalls and unproductive sidetracks and so far very few patients have been cured with these approaches. Nevertheless, 10 years later, we are able to look back and see that more progress has been made in that time than in the previous century.
We now stand at the threshold of a new era: effective cancer immunotherapy, taking its place alongside conventional treatment modalities. For this to be successful, immunotherapy will need to be based strongly in biology and the processes underlying cancer growth and remission need to be fully understood. At the beginning of the 21st century, it is timely to review the status of the field with respect to its strengths, weaknesses, and likely areas of development.
|
THE ERA OF EMPIRICISM |
|---|
|
TOPABSTRACTINTRODUCTIONTHE ERA OF EMPIRICISMTHE ERA OF RATIONAL...THE ERA OF EFFECTIVE...REFERENCES |
|---|
The apparent success of Coley’s treatment encouraged other workers to try to expand on his work with microorganisms. Over subsequent decades, further work showed that other microorganisms could also mediate useful clinical responses. When injected directly into tumors, agents such as Bacille Calmette-Guerin (BCG) or Corynebacterium parvum could mediate tumor regression. In the case of melanoma, injection of BCG into melanoma metastases caused them to decrease in size, and significantly, a further 15% of distant lesions also regressed [9 ]. This suggested that the effect of this treatment was systemic as well as local. If a response was seen, 27% of patients survived beyond 5 years [10 ], well beyond the median survival expected for patients with metastatic melanoma. This was encouraging but unfortunately was not confirmed in randomized studies. Similarly, although favorable responses with microorganisms have also been reported for other cancer types, this also has not been confirmed in larger studies.
BCG now has an established role in the treatment of superficial bladder cancer. The rate of recurrence of such tumors is reduced from 42% to 100% to 15% to 40% when BCG is given intravesically [11 ]. This is superior to the effects of agents such as doxorubicin [11 ]. Although less strong, there are data to suggest that BCG may also have activity even in the setting of residual tumor [12 ].
Certain viruses have also been shown to be useful. Vaccinia virus has been administered directly into tumors and shown to mediate regressions [13 ]. More recently, viruses such as vaccinia, herpesviruses, or adenoviruses have been used as vectors to deliver genes for cytokines or tumor antigens into tumors. Early clinical studies using adenovirus, vaccinia virus, and avipox have demonstrated the safety of this approach [14 15 16 17 18 19 ]. The dominant immune response in this setting is often against proteins derived from the viral genome. This may be detrimental in that it leads to rapid clearance of the vector; however, it is possible that it may also be useful in providing xenoantigenic, helper epitopes to enhance the immune response against tumor antigens [20 ].
These results are interesting and have provided an additional treatment modality in certain circumstances such as superficial bladder cancer. However, without an understanding of the mechanisms of the effects and in particular why such treatment did not always work, these approaches could not advance the overall science of cancer immunotherapy.
Hematopoietic stem-cell transplantation
The evolution of organ transplantation and the associated need to induce tolerance against allogeneic organs have provided important insights into the mechanisms underlying the tolerance exhibited by the body against cancer and how to break that tolerance. Graft-versus-host disease (GVHD), where the engrafted T cells delivered as part of the transplant begin to attack the host tissues, is a common cause of morbidity and death in organ allograft recipients [21
]. Treatment involves reduction of immunosuppression with a consequent increased risk of graft rejection. However, patients with mild GVHD in the setting of bone marrow or peripheral blood progenitor cell autografts for leukemia have a reduced risk of recurrence of their malignancy. This has been termed the "graft-versus-leukemia" effect [22
, 23
]. A similar effect is seen less frequently in lymphoma.
This phenomenon is thought to be due to a population of engrafted donor cells recognizing and rejecting residual host leukemia cells. As a result, some investigators have studied the use of donor leukocytes in patients with relapsed malignancy and have found that although the recipients develop GVHD, there are also useful antitumor effects [24 ]. To reduce the risk of morbidity and mortality from GVHD, some groups have isolated donor T cells with specificity for the recipient’s malignant cells [25 , 26 ]. A useful model for this is in the setting of Epstein-Barr virus (EBV)-related post-transplant lymphoma, where the EBV antigens provide a target for immunotherapy.
This work illustrates the principle that immunocompetent donor T cells can mediate regression of cancers in recipients. Purely allogeneic responses do not suffice to explain this phenomenon, as it can be distinguished from a generalised graft-versus-host response and is specific for the malignant cells.
Cytokines
Despite advances made in the pharmacological treatment of infectious diseases, the mechanisms by which the immune system fights microorganisms remained poorly understood for much of the 20th century. In the case of cancer, what was the link between the microorganism and the final clinical response? It was unlikely that the organisms mediated a direct antitumor effect. It is now known that intravesical BCG elicits cystitis, which leads to the elaboration of proinflammatory cytokines, causing in turn an influx of inflammatory cells, including lymphocytes as well as granulocytes, and the subsequent antitumor effect may well be a bystander effect of the production of inflammatory mediators [27
].
Since the 1930s, it had been known that cells infected with a virus were better able to resist a subsequent viral infection [28 ]. The responsible factor, interferon (IFN), was eventually isolated from cell-culture experiments [29 , 30 ] and later from peripheral blood lymphocytes [31 ]. Gresser et al. [32 ] later went on to show that IFN could mediate antitumor effects in mice and suddenly the world’s attention was focused on this molecule (actually a family of molecules). At around the same time, a factor in pus responsible for producing fever was extracted from neutrophils and named "endogenous pyrogen" [33 ]. This factor was later renamed interleukin-1 (IL-1), the first of a large family of cytokine molecules. The field of modern cytokine biology was launched. These two molecules, IFN and IL-1, provided the detonator for the subsequent explosion in cytokine research resulting in today’s much greater understanding of the cytokine network. The interactions among all of these factors, invading pathogens, host cells, and/or tumor cells, are now much better understood. However, in the absence of this knowledge, all clinical research involving cancer immunotherapy had of necessity to be empirical.
In 1984, Quesada et al. [34 ] published a paper describing a small series of patients with hairy cell leukemia who had been treated with and responded to partially purified IFN. IFN now has an established role in several malignancies, including chronic myeloid leukemia, multiple myeloma, melanoma, and renal cell carcinoma, and has also found a role in the treatment of some viral diseases such as hepatitis B, completing the circle that began in 1937.
In 1975, Old and colleagues [35 ] used a Meth A sarcoma model and demonstrated that hemorrhagic necrosis could be induced when mice were injected with BCG and later with lipopolysaccharide. They isolated a factor they termed tumor necrosis factor (TNF). In this model, the treatment was as lethal as the disease; however, the principle was established. TNF was subsequently cloned, and a new field of investigation was born.
Another prototypic cytokine is IL-2, which was first described in 1976 [36 ] and entered clinical trials in 1984 [37 ]. IL-2 has potent mitogenic and activating effects on T cells and natural killer (NK) cells as well as other tissues such as vasculature. It has found a clinical niche in the treatment of metastatic melanoma [38 , 39 ] and advanced renal cell carcinoma [38 , 40 ]. In both of these disorders, although responses to IL-2 are uncommon, if they occur then the patient will often experience a prolonged period of remission. In the case of melanoma, of the 4–7% of patients who attain a complete response to high-dose IL-2, most will not relapse [41 , 42 ]. When patients with melanoma receive chemotherapy, the median duration of response is only 9 months and the 5-year survival is 6% [43 ]. These long-term survivors after chemotherapy often have a slowly progressive, metastatic disease, in contrast to the IL-2 responders who usually remain disease-free. This major qualitative difference in outcome suggests that the agent has modified the body’s immune recognition of the tumor such that any recurring tumor cells are rapidly recognized and deleted. Unfortunately, to attain such responses, very high doses of IL-2 must be used, causing toxicity sufficient to warrant admission to the Intensive Care Unit in a large proportion of patients, depending on the treatment regimen used [38 , 44 , 45 ]. This toxicity is probably mediated by the induction of other cytokines such as TNF [46 ] or factors such as nitric oxide [47 ].
Despite the toxicity of high-dose IL-2, it has been approved for use in the treatment of metastatic melanoma or renal cell carcinoma in several countries. The clinical development of IL-2 illustrates the nature of cytokine clinical research in the 1980s and 1990s, where the biology was not well understood and clinical investigation was influenced by previous experience with cytotoxic agents. In these studies, the doses used and shown to be effective are clearly pharmacological rather than physiological. Indeed, when low doses of IL-2 are used or any regimen that abrogates the toxicity of IL-2, the efficacy of treatment is lost. The mechanism of action of the drug is therefore difficult to explain on an immunological basis. The low response rate raises the question as to why the agent does not have greater efficacy. A large number of subsequent studies have been performed using variations on the theme of IL-2, including various dosing regimens combining IL-2 with other agents such as IFN or lymphokine-activated killer cells, the addition of chemotherapy, and other interventions. None of these approaches has been shown to be superior. However, for over a decade, the field in general was sidetracked by this large number of clinical trials, from which copious amounts of data but little information were derived. The field is now once again starting to move ahead.
The way in which IL-2 is used is now being re-evaluated based on a greater understanding of its mechanisms of action. For example, low doses that can be tolerated for longer periods of time have been shown to induce NK expansion and activation [48 , 49 ]. The serum concentrations of IL-2 achieved by such regimens are likely to be able to mediate antibody-dependent cell-mediated cytotoxicity (ADCC). As a result, there is now increasing interest in the use of low-dose IL-2 to enhance the efficacy of monoclonal antibody (mAb) therapy. Some such trials have already been performed [50 , 51 ], although the IL-2-dosing regimens used in those studies were chosen on the basis of previous IL-2 trials and not with the intent of using the agent in a more physiological way. Our group is performing a clinical trial using the mAb cG250, which recognizes an antigen expressed in most clear cell-renal cell carcinomas [52 ] in combination with low-dose IL-2 given continuously for 6 weeks (trial LUD 00–010). One of the primary endpoints of this study is to examine the immunological effects of the combination. The results of such studies will shed light on the physiology of IL-2 and may suggest more logical ways of using this agent.
Since the discovery and clinical use of IL-2, numerous other cytokines have been discovered. At the time of writing, the tally of human ILs had risen to include IL-27 (http://www.gene.ucl.ac.uk/nomenclature/genefamily/interleukins.html). Many cytokines have entered the clinic, some with high hopes attached to them, as was the case with IL-12. IL-12 has not lived up to its initial promise. This cytokine has effects on T cells and NK cells and also has antiangiogenic activity. The results of preclinical studies suggested that it should be useful as an anticancer agent. Phase I studies in humans indicated that it was safe [53
] and that the expected immunological effects were seen [54
]. Subsequent phase II studies eliminated the initial test dose used in the phase I trials. This resulted in unexpected, severe toxicity including several deaths [55
]. This toxicity has been attributed to the much higher levels of IFN-
seen in patients who did not receive the test dose [55
]. As a result of these bad clinical experiences and also a lack of efficacy, the clinical development of IL-12 has been largely abandoned. This is unfortunate, as there may still be an important role for this agent.
Once again, the experience with IL-12 illustrates a fundamental principle in the biological treatment of cancer: Maximal tolerated doses of a biological agent may not be the optimal doses. It is interesting that in the case of IL-12, there is evidence that higher doses may in fact be less effective than lower doses [56 ]. It is important to determine the "optimal biological dose", which may be quite different from the maximum tolerated dose. To determine this, however, appropriate biological, surrogate endpoints need to be identified and it is likely that these will differ depending on the agent used and the clinical scenario. Clearly, there remains substantial room for the further development of agents such as IL-12. Perhaps more novel approaches based on the understanding of IL-12 biology, such as the use of gene transfer to obtain high-tissue concentrations of the agent at the site of antigen [57 ] or combinations of IL-12 with mAb [58 ] or heat-shock proteins [59 ], may be a more logical manner in which to proceed.
Enhancing vaccine efficacy: adjuvants
There are obvious parallels between the field of cancer immunotherapy and the science of vaccine technology for infectious diseases. Vaccines can act by stimulating an immune response against the organism itself or against the toxins that produce disease. The latter case presupposes that the infection per se is not important and will be cleared in time as long as the host is able to survive the effects of the toxins. This is illustrated by the protective effects of vaccination against tetanus or diphtheria: In both cases, the immune response is not directed against the organism but against its toxin.
In the case of viral infections, disease is usually caused by the cytotoxic effects of the viral infection itself. Vaccines against viruses predominantly elicit humoral responses, which are most useful to protect against the establishment of infection by eliminating the circulating microorganism at an early stage. However, once an infection is established, antibody responses are unlikely to lead to its clearance. T cell responses are required to eliminate cells harboring viruses. Vaccine adjuvants in common use do not usually stimulate potent T cell responses.
Cancers behave in a similar way to viral infections, in that the malignant process is located inside a host cell and is not generally accessible to circulating antibodies. It is reasonable to assume that T cell responses are critical in the eradication of tumor cells. However, to elicit such responses, two things are required: a source of antigen and an adjuvant that can stimulate T cell responses.
Adjuvants such as alum have been in widespread use for many years and more recently other chemical adjuvants such as the microemulsion MF59 have been approved for use [60 ]. Unfortunately, when used with a variety of sources of tumor antigen including whole irradiated tumor cells, lysed tumor cells, or tumor extracts, these agents have provided disappointing results. Other, newer agents such as the saponin complex QS21 [60 ] or immunostimulatory particles such as ISCOM® [61 , 62 ] have rekindled interest in this area, which has also seen the resurgence even of older agents such as incomplete Freund’s adjuvant [63 64 65 ]. Sequences of immunostimulatory DNA containing CpG motifs have also been studied and have been shown to elicit T cell responses against specific antigens [66 67 68 ]. It is interesting to speculate whether such molecules may have been one of the active components of Coley’s toxins. However, even the best adjuvant will be of no use if an appropriate target cannot be found.
What didn’t work and why not?
Cancer immunotherapists are almost by definition extreme optimists. Cancer is an antigenically diverse disease, tumors are often bulky, cancer patients are often inherently or iatrogenically immunosuppressed, and cancers can continue to grow despite detectable immune responses. Still, if a transplanted liver weighing up to 2 kg can be rejected, why not a cancer?
The difficulty is that cancers usually arise from the patient’s own cells. Although cancer cells are abnormal, a vigorous rejection response is not triggered. This may be because of passive factors such as low levels of expression of tumor antigens or human leukocyte antigen (HLA) molecules, poor accessibility of effector molecules or cells, or rapid growth rate outstripping the immunological killing of cells; or active factors, such as the secretion by tumor cells or associated phagocytes of reactive oxygen and nitrogen radicals [69 ] or cytokines that down-regulate the effectiveness of incoming effector cells [70 ] or surface expression of molecules such as Fas ligand [71 ], which induce apoptosis of incoming T cells.
This diversity of escape mechanisms has triggered numerous approaches to the immunotherapy of cancer. The results of decades of work, thousands of patients, and millions of dollars can be summed up as follows: Almost nothing worked, but very occasionally a good response was seen for reasons that were not clear, providing enough incentive for the whole field to keep moving. What conclusions can be drawn from this extensive body of work?
First, technology is no substitute for a carefully thought out and planned experiment. For example, when efficient gene-transfer methods became available, extensive resources were invested in this field. It was an exciting and interesting time but there were unexpected problems. The early vectors often gave low levels of expression as a result of suboptimal promoters. Immune responses against the viral vectors or against the gene product occurred frequently [16 ], suggesting that sustained expression in vivo might be low and that repeated treatments were not practicable. There was perhaps insufficient thought about the biology of the cancer: Just because a particular gene product was known to be important in some part of the immune response in a defined anatomical portion of a lymph node did not mean it was sensible to overexpress that factor at high levels in the periphery at the tumor site; indeed, in some cases this was harmful [72 ]. Fortunately, this initial zeal has now worn off to a large extent and the majority of work currently being performed has a sound, scientific basis.
Second, the field at times became side-tracked. This was usually because of an apparent success in one direction, leading other investigators to follow that path. An example has been the recent rise and fall in enthusiasm for high-dose IFN for the adjuvant treatment of high-risk malignant melanoma. An initial study from the Eastern Cooperative Oncology Group (ECOG), Trial EST 1684, suggested improved disease-free and overall survival in the subset of melanoma patients with disease metastatic to regional lymph nodes who received very high doses of IFN-
[73
]. As a result of this work, IFN- was approved for this indication in several countries including the United States and it became the de facto standard of treatment for these patients. Indeed, until recently many thought it was unethical to have a "no-treatment" control arm in clinical trials involving these patients. The subsequent ECOG study E1690 did not confirm the overall survival benefit of IFN in these patients, although the effects on disease-free survival remained. There are several possible reasons for this discrepancy, including the fact that the advent of sentinel lymph node mapping led to "stage migration" and a correspondingly better outcome for the control group and also that many patients in the control group subsequently received IFN at relapse. Nevertheless, the original study, performed with the best intentions and planning, significantly altered the management of this disease and has influenced clinical research for almost a decade. A recent overview has concluded that there is no clear benefit from IFN- in melanoma [74
]. This is not to say that it is ineffective; rather, more work in better-designed trials needs to be done, building on the lessons learned from the previous studies and from advances in understanding basic biology. Similar problems have occurred in other areas, such as the use of melanoma oncolysates, which initially appeared to give promising results [75
76
77
78
79
] that were not confirmed when later studies were performed [80
].
Our patients are the reason for our work and in many ways are our greatest resource but one that needs to be used efficiently. Sometimes large trials may not be the most efficient way to use this resource, contrary to conventional wisdom. The U.S. National Cancer Institute (NCI, Bethesda, MD) has led the field of gene transfer of tumor antigen genes using viral vectors [65 , 81 ] for many years. There is no question that this was a good idea. As a result, several hundred patients were treated in this way at the NCI. Unfortunately, clinical responses were scarce, and immunological responses were essentially not seen [82 ]. The NCI group has learned from this experience and has since had greater success by altering the vectors used and by the addition of IL-2 [82 ]. Perhaps a series of smaller studies initially may have allowed these patient resources to be used in a more efficient way, at the NCI or at other centers.
Another area of difficulty has been the albatross of the traditional, oncological, clinical trial design that still hangs around the necks of those involved in the development of new biological approaches. Biological therapies including immunotherapy cannot be compared with cytotoxic therapies. The results of cytotoxic therapies can usually be understood in the context of the pharmacokinetics of the agent, and for most cytotoxic drugs, it is clear that responses occur early rather than late, a concept directly derived from work in other areas of therapeutics such as infectious diseases or hypertension. As a result, response rates have become a surrogate marker for clinical outcome and this is now so pervasive that the response rate has become for some people more important than overall survival or quality of life. This is clearly wrong. In contrast, the effects of biological agents are not usually related to their pharmacokinetics. The biological effect does not follow an exponential decay with a half-life that can be calculated. In the case of immunotherapy, often tumors will progress before a response is seen [83 ], as an effective immune response (if it occurs) may take several months to become apparent. If effectiveness were measured by response rate alone at conventional time points, then these agents would be thought to fail even if they could ultimately lead to a cure of the disease.
This raises a new concept for the biological treatment of cancer: Cancer is a chronic disease. Patients might be able to live for long periods with cancer, provided the cancer is under control. A realistic goal may be to convert a progressive and destructive cancer into an indolent and innocuous one, although it may not be eradicated. Response rate may no longer be important: A tumor may not reduce in size by 50%, but if it never progresses any further, then the patient has achieved a benefit. Progression-free survival, overall survival, and quality of life now become the significant endpoints of treatment. Treatment may need to be given for long periods of time and perhaps even life-long, meaning that toxicity must be low and the treatment must be relatively simple. Most cancers do not have a single pathogenesis and it may be naïve to assume that treatment directed against a single aspect of their biology will be effective. Rather, approaches that take advantage of knowledge of the biology of the tumor cell will allow rational design of treatments with new levels of efficacy. This is already occurring in some areas [84 ].
Some of the issues to be considered in cancer immunotherapy studies are summarized in Table 1 .
|
|
THE ERA OF RATIONAL TRANSLATIONAL RESEARCH |
|---|
|
TOPABSTRACTINTRODUCTIONTHE ERA OF EMPIRICISMTHE ERA OF RATIONAL...THE ERA OF EFFECTIVE...REFERENCES |
|---|
-fetoprotein in hepatocellular carcinoma [87
]. Animal studies had allowed the definition of antigens in artificial situations of chemical or viral carcinogenesis, but until that point, it was thought that human tumors were relatively nonimmunogenic. In 1989, Knuth et al. [7
] described the first of what would turn out to be a large family of MAGE genes and showed that this antigen could elicit a T cell response. The discovery of human tumor antigens was a defining moment in tumor immunology. This was the first step to rational design of cancer immunotherapy: With a target in sight, we were no longer shooting blind. Having a defined target also led the way to optimization of assays for the measurement of immunological endpoints, as discussed below.
Several methods have been used to define human tumor antigens. Initially, T cells were isolated from a melanoma patient who had been repeatedly immunized with mutagenized cancer cells [88 ]. An important and successful method has been to transfect genomic DNA or cDNA libraries into cells expressing the appropriate major histocompatibility complex (MHC) molecule and then to identify transfectants using cytokine release or lysis by human T cells with specific antitumor activity [87 88 89 90 91 92 ]. Alternatively, peptides can be eluted from tumor cells or from the MHC molecules of tumor cells and then be pulsed onto antigen-presenting cells (APC). Fractions capable of stimulating antitumor T cells are then examined further and the responsible peptide is identified [93 , 94 ].
Another approach that has been very successful has been to use serum from cancer patients to detect antigens derived from tumor libraries [95
, 96
]. In the SEREX (serological identification of antigens by recombinant expression cloning) approach, a cDNA library is constructed from fresh tumor specimens, packaged into 
-phage vectors, and expressed in Escherichia coli. Recombinant proteins are transferred onto nitrocellulose membranes and identified as antigens by their reactivity with high-titer immunoglobulin G (IgG) antibodies present in the patient’s serum. Positive colonies are cloned and the nucleotide sequence of the inserted cDNA is determined [95
]. The antigen NY-ESO-1 (discussed below) was first detected by this approach [96
], although NY-ESO-1 is known to have epitopes recognized by CD8+ [97
] and CD4+ T cells [98
]. More than 1500 antigens have been described in this way (http://www.licr.org/SEREX.html).
Some have suggested that the common techniques for identification of tumor antigens, involving prolonged in vitro culture of T cells in the presence of tumor cells, may skew toward the identification of inferior tumor antigens [99 ]. High-avidity cytotoxic T lymphocytes (CTL) corresponding to immunodominant epitopes are less likely to survive this process [100 ]. Similarly, the fact that CTL used to identify tumor antigens are generated from patients with advanced cancer, who by definition have failed to reject their cancer, may also skew toward the isolation of weaker tumor-rejection antigens. Anichini and colleagues [101 ] have shown that most of the HLA-A2.1-restricted immune repertoire to melanoma is directed against epitopes expressed in the neoplastic but not in the normal cells of the melanocyte lineage and that many CTL respond to as-yet uncharacterized antigens.
More recently, tumor antigens have been detected from the study of candidate antigens [102 103 104 ]. T cells are generated against candidate antigens and then tested for their ability to recognize intact tumor cells. Candidate peptide epitopes for T cells can be identified based on predicted binding affinities of peptide to MHC and assessed for immunogenicity based on the functional capacity of experimentally generated, peptide-specific T lymphocytes. This approach was used to identify immunogenic epitopes associated with the telomerase catalytic subunit, which appears to be a frequently expressed and potent tumor antigen, with little expression in normal tissues [105 ]. Scanlan and colleagues [106 ] recently identified several novel tumor antigens by database mining and mRNA expression analysis. The authors specifically sought the identification of novel "cancer-testis" (CT) antigens by mining of the Unigene database for gene clusters containing expressed sequence tags derived from normal testis and tumor-derived cDNA libraries.
Many cancer antigen peptides have been defined based on predicted binding sequences for known MHC molecules. The danger with this approach is that although the predicted epitopes may bind to MHC and be able to elicit T cell responses in vivo and in vitro, the peptide may not be generated naturally by the proteasomal machinery within the tumor cell, and so the resulting CTL will be unable to recognize the tumor cell [107 ]. However, this has led to a more rational method of predicting naturally processed epitopes by means of proteasomal processing of proteins or long peptides [107 108 109 ].
Published clinical trials using HLA class I peptides in cancer treatment are summarized in Table 2 .
|
In general terms, antigens can be classified in several ways. Many antigens are "shared," i.e., expressed commonly in cancers of that particular type [99 ]. These include antigens that are restricted to cells of a particular lineage, for example, the melanocyte antigens tyrosinase, Melan-A/MART-1, or gp100, which are expressed in normal and malignant melanocytes. A second group is the so-called CT antigens, expressed in a wide variety of cancer types but in very few normal tissues. A third group includes antigens derived from viruses that may be involved in the oncogenic process, such as EBV in some lymphomas, human papillomavirus in the case of cervical or anal carcinoma, or hepatitis B in some cases of hepatocellular carcinoma. A fourth group is that of overexpressed or mutated, normal cellular components, such as p53 or ras. Finally, a very large group is that of the "unique" antigens, specific for a given individual’s tumor but not necessarily generalizable to all patients with that tumor type; an example of such an antigen is a clonal idiotype protein from a B cell malignancy.
Several authors have recently reviewed the range of known human tumor antigens [91 , 149 , 150 ]. Gilboa [99 ] has discussed the distinction between a tumor antigen and a tumor-rejection antigen. An important consideration regarding the potency of a tumor antigen is the avidity of cognate T cells that can be activated against it and against tumors bearing the antigen. Gilboa suggests that tumor antigens that are patient-specific, tumor-specific, or highly restricted shared antigens (such as the CT antigens) are more likely to have been previously ignored by the immune system and can serve as effective tumor-rejection antigens.
The first studies using defined antigens were done with HLA-restricted peptides. Clinical cancer vaccine trials with an emphasis on peptide vaccines have recently been reviewed [151 152 153 ]. Tumor regressions have been infrequent but have been described. In a study of 39 patients with metastatic melanoma vaccinated with HLA A1-restricted MAGE peptides, seven patients displayed significant regressions with three complete responses [111 ]. No CTL responses were detected, although it is not clear whether this negative result simply reflects the insensitive assays used at that time. Other studies have found a correlation between tumor regression and immunological endpoints [154 ]. The addition of adjuvants such as granulocyte macrophage-colony stimulating factor (GM-CSF) [155 ] or IL-2 [126 ] to peptide vaccination may augment anti-tumor effects.
However, peptide presentation on the surface of APC comprises more than simply adding peptide to cells. There is a complex equilibrium between peptide and the class I molecule as well as between the peptide-MHC complex and the T cell receptor. Some cells may also take up exogenous peptide and present it on the class I pathway, although the efficiency of this process is probably affected by the peptide/MHC avidity [156 ]. When peptides are modified to allow greater binding to MHC molecules, more efficient and increased induction of antigen-specific CTL can be achieved [157 158 159 160 ]. Such issues are important, as the nature of the presentation of peptide may lead to immunogenicity or to tolerance [161 ].
This plethora of antigens poses new dilemmas. Which should be used in clinical immunotherapy trials? As more and more antigens are defined, it is reasonable to assume that the most important ones may have been defined earliest and most often, as it is these antigens by definition that most commonly elicit immune responses. It is therefore logical to proceed to develop these antigens as vaccine targets. Should single antigens or multiple antigens be used? A polyvalent approach may ultimately be the most effective, as tumor escape by selection in vivo for clones lacking expression of the antigen will be less likely. This may also allow a wider variety of tumor types to be targeted. The exception to this rule would be if the antigenic target is essential to the neoplastic process so that the tumor cell cannot simply switch off expression of the antigen. Few of the currently known antigens have this property, with the possible exception of the telomerase catalytic subunit, which is ubiquitously expressed and is essential for the malignant phenotype.
However, there may be situations in which the tumor antigen is not known. In this situation, the source of antigen may of necessity be less refined. Several possible approaches have been suggested and used. These include the use of tumor lysates, eluted peptides, whole tumor cell vaccines, genetically modified tumor cells, and tumor cell/DC fusions. The potency of the allogeneic response may be able to be used to help eradicate autologous tumor [162 ]. However, when these approaches are used, care must be taken not to revert to the old empirical ways of thought. The field can only progress if there is understanding of the process.
Antigen presentation and DC
The other major turning point underlying the modern era of tumor immunotherapy was the recognition of DC as the chief orchestrators of the adaptive-immune response and the elucidation of the complex interaction between DC and T and B cells. DC also interact with NK cells [163
] and react to proinflammatory factors [164
], thus providing a crucial link between the adaptive and innate immune systems. DC are uniquely able to prime a naïve CD4 or CD8 T cell response (reviewed in ref. [165
]). Because of their pivotal position in immunity, DC have been extensively studied as antitumor adjuvants (reviewed in refs. [166
] and [167
]). Reported clinical trials have established the feasibility and safety of DC vaccination approaches and have frequently been associated with immune responses. Some studies have documented encouraging clinical responses. These studies are summarized in Table 3
.
|
The methods of preparation and the quality of DC preparations have varied considerably. Immature and mature DC have been used, an important distinction depending on the tasks required of the cells (for review, see ref. [166 ]). DC may be loaded with antigen by a variety of techniques. Peptides may be pulsed directly onto DC with the assumption that they will displace native peptides from the MHC complex or occupy empty MHC. Immature DC may process and present whole protein molecules. DC may also be transfected with tumor RNA [205 206 207 208 ] or transduced with genes by a variety of methods [209 ]. DC may also acquire antigen from apoptotic cells [210 , 211 ] or secreted "exosomes" [212 , 213 ]. Various routes of delivery of DC have also been used, such as SC, ID, IV, or intranodal injection. The dose and timing of injections have varied widely. Such a range of variables makes it very difficult to compare different strategies and to draw generalizable conclusions. A recent paper has reviewed many of these issues and called for greater standardization in trial design, including measurement of immune responses [214 ].
Although DC trials have been encouraging, it is important to note that DC may not always generate a strong antitumor response. For example, in one study, injection of immature DC led to the induction of tolerance [215 ]. Hawiger et al. have also reported that at steady-state, DC can induce unresponsiveness of T cells to presented antigen [216 ].
These results could not be explained by the traditional "self/nonself" theory of antigen recognition, which was also lacking in some other respects. By what means could a T cell know that a given peptide came from "nonself", and from where would the precursors for such a cell come? Why were adjuvants required for an optimal immune response (the phenomenon termed "immunology’s dirty little secret" by Janeway, ref. [217 ]). Simply responding to a previously unseen peptide would be dangerous: The risk of life-threatening autoimmunity as a result of inadvertent presentation of critical self-determinants in certain circumstances would be too high. The "self/nonself" model has now been superseded by the "danger" model [218 , 219 ]. In this model, the source of the antigen is less important than the context in which it is presented to the immune system. With the appropriate stimuli, any antigen can potentially be seen as "dangerous" and stimulate an immune response. This provided the long-awaited link between acquired and innate immunity: Proinflammatory cytokines and chemokines elaborated by cells of the innate immune system or by stromal cells could provide the appropriate context to elicit a "danger" signal. These signals commonly occur in early phases of microbial infection or transplant rejection and infiltration by specific T cells occurs only later.
Recent work has explored the reasons why some tumor cells escape immune recognition. Part of this mechanism may be that tumor cells do not provide appropriate "danger" signals that can be picked up initially by cells of the innate-immune system [220 ]. The importance of stress-inducible ligands on tumor cells that can be recognized by nonspecific immune effectors is now recognized. An old concept, that of regulatory T cells that can downmodulate specific immune responses, is also now being resurrected. The complex interaction among acute inflammatory cells, NK cells, DC, stroma, and T and B cells is now much more clearly understood. As a result, it has been possible to devise strategies to enhance these interactions.
Gene transfer
Gene transfer for cancer is discussed briefly here but has been reviewed recently [221
222
223
224
225
226
]. Many methods have been used and there is an increasing tendency now to use approaches based on known biology rather than simply inserting genes because it was possible to do so. For example, with a greater understanding of the mechanisms involved in antigen presentation, anergy, and how to sustain effector T cells in the periphery, tumor cells can be modified to enhance their immunogenicity, such as by enhancement of expression of MHC or costimulatory molecules on tumor cells or through the introduction of cytokine genes such as GM-CSF, IL-2, or IL-12. Several early-phase human studies of cytokine gene-modified tumor cells have been performed [227
228
229
230
231
232
]. These demonstrate that the approach is generally safe and that appropriate biological endpoints can be measured, although clinical responses have been few.
Synthetic genes for tumor antigens can be transfected directly into APC or delivered by in vivo gene transfer. Different vectors allow the use of different priming and boosting mechanisms, and synthetic genes allow specific epitopes or costimulatory molecules to be delivered [233 ]. Alternatively, genetic material can be derived from primary tumor tissue using a more direct approach. For example, RNA can be derived from tumor cells and transferred into APC such as DC, the intention being that the DC will present peptides from the translated gene product [205 , 207 , 234 , 235 ]. Similarly, genomic DNA can be derived from a tumor and transfected into a fibroblast cell line to generate a vaccine [236 ].
The latter approaches have several potential advantages. If the relevant antigen or MHC determinant is unknown, RNA or DNA allows all potential antigens to be expressed and processed appropriately by the APC. It is often possible to do this even if only a small amount of tumor is available, for example, by using microdissection of tumor tissue and subsequent amplification of tumor RNA [235 ]. However, there are also several potential concerns. Genes that specify normal cellular constituents may also be expressed in the recipient cell and presented in a context that may lead to autoimmunity. It is also possible that the cells used for the vaccine may be transformed or transfer an oncogene or defective tumor-suppressor gene to a normal host cell provoking a neoplasm [223 ].
Monoclonal antibodies
Monoclonal antibodies have been one of the most successful modalities of immunological treatment for cancer developed over the last 10–15 years. Although many tumor-specific antigens are intracellular and therefore, only able to be recognized by T cells (as peptide/MHC complexes), some targets are expressed on the cell surface and are appropriate targets for antibody therapy. Several recent reviews of antibody therapy for hematologic and solid tumors have been published [237
238
239
].
mAb may be used alone ("naked" antibody) or used to convey conjugated toxins, radioisotopes, or other agents to the tumor site [240 ]. Naked antibodies rely for their cytotoxic effects on the induction of ADCC or the induction of an inhibitory or apoptotic signal as a result of binding to a cell-surface target. A growing range of antigen targets suitable for mAb therapy has been identified [240 ]. Molecular engineering techniques have led to the development of modified antibodies with improved affinity and/or reduced immunogenicity, allowing repeated dosing with less probability of development of antibodies to the injected antibody. Other recent approaches include bispecific antibodies to simultaneously target tumor as well as T or NK cells, single-chain antibodies, or fusion molecules [241 ].
Her2/neu (c-erbB-2 oncoprotein) is a membrane tyrosine kinase receptor, present in normal tissues but overexpressed in a significant number of breast and ovarian cancers [242 ]. Because of its cell membrane localization, it may induce an antibody response [243 , 244 ] as well as a cellular response, either spontaneously [245 246 247 ] or following vaccination [141 , 142 , 248 ]. Murine models have shown that anti-Her2/neu antibodies protect against a later challenge with tumor cells expressing Her2/neu [249 ]. For women with breast cancer, the addition to chemotherapy of trastuzumab [250 ], a recombinant mAb against Her2, gave a higher rate of objective response, longer duration of response, improved 1-year survival, and longer overall survival [251 ]. An unexpected, adverse effect of trastuzumab has been cardiac toxicity, which in turn, has stimulated new areas of research.
A second mAb targeting CD20 (IDEC-C2B8, Rituximab/Rituxan) has also entered standard clinical practice. CD20 is an antigen present on mature B cells [252 ] as well as over 90% of B cell lymphomas [253 ]. CD20 is not shed or modified following encounter with the antibody [254 , 255 ] and hence is considered a good target antigen. The mode of action of rituximab is unclear: ADCC is probably important, although receptor-mediated effects may also be involved [256 ]. Clinical trials have confirmed the therapeutic effect of rituximab [257 ] in low-grade and intermediate-grade CD20+ non-Hodgkin’s lymphoma and also in combination with radioisotopes [258 ].
Other mAb have also recently been approved by the U.S. Food and Drug Administration (Rockville, MD). MylotargTM (gemtuzumab ozogamicin) is a humanized anti-CD33 antibody conjugated to calicheamicin and is approved for patients with CD33-positive acute myeloid leukemia [259 ]. Campath® (alemtuzumab) is an anti-CD52 humanized mAb approved for patients with B cell chronic lymphocytic leukemia who have been treated with alkylating agents and have failed fludarabine therapy [260 ].
Other mAb are also in various stages of development, including several being developed by our own institute for use in colorectal cancer, renal cell carcinoma, breast cancer, and others. This field will continue to move forward rapidly as new targets and newer ways of using these agents are explored.
T cell help
There has been a recent reappraisal of the role and importance of CD4+ helper T (Th) cells in antitumor responses. Previously, attention has concentrated on CD8+ CTL, as most tumors express MHC class I molecules but do not express MHC class II. CD8+ CTL are able to lyse tumors directly and until recently few antigenic epitopes have been identified that are able to stimulate a CD4+ Th cell response [92
, 261
].
CD4+ Th cells have an important role in priming to CD8+ CTL response, through expression of CD40 ligand and binding of CD40 on APC [262 , 263 ]. In vitro, the need for Th cells can be bypassed by activation of DC through CD40 [264 ]. Under noninflammatory circumstances, CTL responses are more dependent on Th cell involvement, as DC need to be activated first by specific CD4+ T cells before they can trigger CTL responses [261 ]. CD4+ T cells are also implicated in the activation of tumoricidal macrophages that are involved in tumor clearance [265 ]. Cytokines produced by CD4+T cells can recruit and activate macrophages and eosinophils [266 ], again linking the T cell response with the innate immune response.
To obtain a high titer IgG response, T cell help is required, so the identification of SEREX-defined antigens implies recognition by CD4+ Th cells. Consequently, the SEREX repertoire can be considered to be a reflection of the CD4+ T cell repertoire [267 ]. SEREX-defined antigens have recently been shown to intensify the response of CD8+ T cells to unrelated tumor-specific peptides through stimulation of a CD4+ Th cell response in a murine system [267 ]. Coimmunization with plasmids encoding SEREX-defined wild-type molecules and unrelated tumor-specific antigens resulted in a profound increase in CD8+ T cells specific for the tumor antigens. This response depended on CD4+ T cells and copresentation of SEREX-defined molecules and CD8+ T cell epitopes. In a tumor protection assay, the combination of SEREX-defined molecules and CD8+ T cell epitopes was protective, whereas either alone was not. These experiments further reinforce the importance of T cell help in vaccine strategies and suggest new clinical approaches using helper epitopes to enhance CTL responses.
Heat-shock proteins
Heat-shock proteins are a family of molecules involved in chaperoning partially unfolded or denatured proteins and peptides [268
] that also have an important role in the antigen presentation pathway [269
]. Heat-shock proteins can carry tumor antigens and MHC molecules from tumor cells to APC [270
]. Heat-shock proteins can be derived relatively simply from small amounts of viable tumor. A number of clinical trials are currently underway in several indications, including non-Hodgkin’s lymphoma, sarcoma, melanoma, gastric, renal, and pancreatic cancers [271
272
273
].
Oncolytic viruses
Interest has rekindled in the use of viruses directly for cancer treatment. This field has recently been reviewed [224
]. Oncolytic viruses have been engineered to replicate selectively within tumor cells. As these viruses replicate, they should then be able to destroy tumors specifically, both by a direct cytolytic effect of the virus and by its ability to spread within the tumor from one cell to another, sparing normal cells. Clinical experience in humans is most advanced with Onyx-015. This is an E1B 55-kDa gene-deleted adenovirus, believed to selectively replicate in and lyse p53-deficient cells. Several clinical studies have demonstrated the safety and clinical efficacy of Onyx-015 in patients with advanced, recurrent, or refractory head and neck cancers, alone or in combination with systemic chemotherapy [274
275
276
277
278
]. These responses are encouraging and further phase III trials of this agent are warranted. Other oncolytic viruses are also under development [279
, 280
].
CpG oligodeoxynucleotides (ODN)
An intriguing new class of agent has emerged over recent years. CpG ODN contain motifs found preferentially in bacterial rather than vertebrate DNA and are able to stimulate potent immune responses. The type of response differs according to the nucleotide sequence. The biology of CpG has recently been reviewed [281
]. These agents have been shown to be potent vaccine adjuvants and interestingly, can elicit T cell as well as antibody responses [282
]. Clinical trials using CpG ODN are currently underway in the settings of non-Hodgkin’s lymphoma, melanoma, and breast cancer. Our group is performing a phase II study for patients with renal cell carcinoma. The results of these studies are awaited with interest.
Immunological assays and monitoring
Background.
The characterization of tumor antigens has made it possible to measure specific immune responses to these molecules. Following vaccination, an immune response against each tumor antigen is almost certainly necessary to achieve an objective clinical response; however, a measurable immune response may not be sufficient to achieve this. One of the major challenges of immune monitoring is to understand the relationship between immune and clinical responses. Immune responses are often considered to be surrogate endpoints; however, as surrogates they need to be validated against meaningful clinical outcomes. Unfortunately, at this stage in the development of the field there is little evidence to support a clear relationship between immune and clinical endpoints. Some reports have shown immune response, tumor regression, and in vivo "selection" of antigen-negative variants [283
], observations that taken together would suggest that there is a causal relationship between vaccination, immunity, and clinical response. However, some peptide vaccine studies have shown evidence of immune induction without detectable clinical responses [128
], or evidence of objective tumor regression without T cell induction [111
]. Perhaps these assays are therefore best seen as tools that can help us better understand the relationship between tumor and immune rejection and the requirements for tumor rejection.
It is likely that the clinical context in which cancer vaccination takes place will be important in this regard. Vaccination can be offered to patients who have evaluable cancer or to patients whose disease has been resected. In the presence of tumor (and therefore antigen), there is the potential for spontaneous immunity to develop. This is well demonstrated for the CT antigen NY-ESO-1, where the majority of patients bearing tumor have antibody against this antigen and where antibody titer appears to correspond to tumor burden [284 ]. Vaccination of such patients will therefore be in the setting of a preexisting immune response or a memory repertoire and may be qualitatively and quantitatively different to vaccination of a naïve subject. Furthermore, the persistence of antigen may be important in helping to maintain immunological memory [285 , 286 ]. The presence and response of evaluable cancer can also help establish the correlation between immune and clinical responses. Conversely, patients with advanced disease may not be good candidates for vaccination for other reasons: In malignant melanoma, the median survival of stage IV disease is 6 months [287 ]. As a result, some patients will not complete a course of vaccinations before running into problems because of progressive disease. In other patients, prior treatment with chemotherapy, radiotherapy, or their poor performance status may compromise immune responsiveness. Such patients may not be ideal candidates for studies where immune response is the primary endpoint. In the setting of resected disease, patients are likely to complete a full course of vaccination and prior or concurrent treatments are far less likely to compromise immunity. Nonetheless, as there may be no persistent source of antigen in those patients, spontaneous immune responses may be less frequent and a vaccination-induced immune response may not be sustained.
Methods of monitoring immune responses.
Many published methods have been used to evaluate immune responses (Table 4
). This field has been reviewed recently [299
]. These methods vary in complexity and reliability, and some assays require in vitro stimulation in the presence of antigen to amplify the response. Few laboratories have used them in a standardized manner, and all assays measure a different facet of the immune response. They cannot be validated as useful surrogates without having a clear understanding of how they relate to clinical endpoints, so the correlation of each method with clinical response needs to be tested prospectively. Unfortunately, the relationship between these endpoints and tumor regression is complex, as this will not only depend on immune responses but also on host factors. Tumors have demonstrated an extraordinary capacity to escape immune recognition [300
]. Important factors include antigen expression (which may be heterogeneous or downregulated), HLA expression by tumors and its down-regulation in response to an immune response, and access by effector cells to the tumor.
|
, which are in turn a lesser population than those that bind tetramer [302
]. Just as it can be difficult to demonstrate a clear-cut relationship between cellular assays and clinical outcomes, relatively few trials have demonstrated a tight correlation between one assay and the next. In a recent paper by Disis et al. [248
], the relationship between DTH testing and T cell stimulation index was described. In this study, a high degree of correlation was seen between these two methods but only when DTH measurements were greater than 10 mm.
Quantitation of CD8 responses.
CD8+ T lymphocytes are the key effectors for cellular killing, so the enumeration of these cells is widely recognized to be an important immunologic endpoint in cancer vaccine trials. There are a number of important issues that need to be addressed when one considers measuring antigen-specific T cells. These include the magnitude of the response, the tissue compartment in which the response is measured, the kinetics of the response, the mechanisms by which the response is maintained, and standardization of the assays used.
Magnitude of the response. Assuming that the effectors that can be detected are capable of killing, how big of a response is sufficient if one is aiming to get tumor eradication? In the human setting, antigen-specific CD8+ T cells have been adoptively transferred to treat or prevent EBV-associated lymphoproliferative disease. In one such study, viral load was monitored in response to T cell infusion after two to four infusions of 107/m2 EBV-specific CTL. Treatment starting from the time of maximal virus load resulted in a 2- to 3-log decrease of viral titers in three of nine patients [289 ].
Objective regression of post-transplant lymphoproliferative disease has been reported following such infusions, and the frequency of T cells (as determined by limiting dilution assay) was 
40 antigen-specific cells/107 peripheral blood mononuclear cells. This number was equivalent to the frequency seen in a nonimmunosuppressed EBV-seropositive individual [290
]. Using a similar approach, Yee et al. [292
] adoptively transferred T cells specific for Melan-A/MART-1 into patients with melanoma and followed this with systemic treatment with low-dose IL-2. High frequencies of tetramer-positive cells were detected in blood (up to 1.2% of all CD8+ cells). In parallel, there was impressive infiltration of skin and melanoma, where 28% and 37% of CD8+ cells were tetramer-positive. In association with these infiltrates, skin rash and tumor necrosis were seen, the rash consisting of an inflammatory reaction around pigmented nevi.
It is difficult to relate these CTL frequencies to those observed following anticancer vaccination. In general, few cancer vaccine trials have reported CD8+ cells at frequencies comparable with those seen following viral infections or adoptive transfer. Some studies using DC-based vaccines have reported detectable levels of antigen-specific T cells using quantitative assays and clinical tumor regressions. In one study with the peptide antigen MAGE-3.A1, 6 of 11 patients showed some clinical regression [113 ]. A semiquantitative method was used to assess MAGE 3.A1 CTL precursor frequency [302 ]. An immune response was detected in 8 of 11 patients; however, frequencies of CTL precursors were relatively low, comprising fewer than 0.1% of CD8+ cells in these patients. In another study using CEA, 2 of 12 patients showed a clinical or CEA response and one showed a mixed response [140 ]. Tetramer-positive cells increased to 1–2% of CD8+ T cells in responding patients.
In most trials, enumeration of T cells depends on assays that are not standardized. In the absence of commonly accepted criteria or standardized methods, it is therefore very difficult to compare the results of one trial directly with those of another. This will be discussed further below.
Tissue compartments. Circulating tumor antigen-specific T cells can be detected in blood samples; however, it is not clear that the profile of cells circulating in the blood is reflected in the tissues. Whereas the clearance of a systemic viral infection may require circulating T-lymphocytes, the rejection of a tumor localized in a tissue compartment will require cells that traffic into the tumor. There are no easy ways of characterizing such cells apart from tumor biopsy. Attempts at quantitating T cell activation within tumor based on cytokine (IFN-) production have been reported. Despite being able to demonstrate an increase in IFN- mRNA at tumor sites following vaccination, there was poor correlation between the response in tissue, response in blood, and objective tumor regression [303
]. In tumor-bearing lymph nodes, Romero et al. [302
] reported that high frequencies of antigen-specific CD8+ cells could be detected by tetramers. This approach has also been taken for evaluating responses following vaccination, where lymph nodes draining vaccine sites have been evaluated for the presence of antigen-specific cells [130
]. Not surprisingly, immune responses were detected more often in nodes than in peripheral blood (five of five vs. two of five patients). Unfortunately, an assay that relies on an invasive approach is not practical for widespread application. Perhaps a more accessible approach is to biopsy superficial metastases, vaccination sites, or DTH test sites [304
] to identify local infiltrating antigen-specific lymphocytes. Such studies also enable evaluation of responding lymphocyte populations by evaluating T cell receptor clonality. For example, following vaccination with a hapten-modified autologous tumor cell vaccine, clonal expansion of T cells was demonstrated in metastatic lesions following vaccination, although the vaccine did not consist of a defined antigen or antigens [293
].
Kinetics of the response. In contrast to viral infections, very little is known about the kinetics and trafficking of antigen-specific T-lymphocytes following cancer vaccines in humans. It is possible that the nature of the vaccine will influence this. For example, the kinetics of an immune response following a peptide vaccine may be different to that following vaccination with a full-length protein or a vaccine formulated using a viral vector. Furthermore, the kinetics of the response may differ depending on the antigen, whether a response is being primed from a naïve background or is a recall response in the context of pre-existing immune memory. If the kinetics and magnitude of the immune response vary depending on the nature of the vaccine and antigen, the timing of tests used to monitor each of these vaccines may need to be tailored. There have been no human cancer vaccine trials reported to date that have been designed specifically to identify the optimal time to sample blood or other tissues following vaccination. It is therefore possible that negative results in some trials simply reflect a failure of investigators to collect blood or tissue at the correct time. As clinical regressions have been seen in some patients in the absence of circulating effectors, it seems quite likely that effectors have been generated following a vaccine but were undetectable because of an inability to sample the right compartment, because frequencies in blood were simply too low, or because timing prevented the identification of antigen-specific T cells trafficking in peripheral blood.
Maintenance of the response. In a chronic disease such as cancer, sustaining an immune response over time may be particularly important. To maintain or expand an immune response, cytokines such as IL-2 have been used. In the study of Yee at al. [292 ] described above, the maintenance of circulating T cells depended on addition of systemic IL-2 treatment. Helper CD4+ cells have also been used to help maintain CD8+ T cell responses in the setting of adoptive transfer of cytomegalovirus-specific CD8+ T cells [294 ]. Other means of sustaining a response include repeated vaccination and the persistence of antigen in the context of tumor. Therefore, when monitoring immune responses, it may be important to take into account the duration of the response as well as its magnitude and the compartment in which it is found. How long does immunity need to be sustained? Perhaps this question can only be answered while simultaneously asking how quickly cancers evolve strategies for evading immune recognition.
Standardization of assays. One of the largest obstacles to obtaining good comparative data between different clinical trials is the lack of standardization of assay methods. As antigen-specific T cells are often present at very low frequencies, methods that amplify or expand these cell populations have been used. These methods depend on one or more rounds of stimulation in vitro. All methods that require amplification run the inherent risk of being poorly reproducible. This can be a problem even in a single laboratory where the same method is applied repeatedly. The problem is further exacerbated if assays are performed at different laboratories or if the method is altered.
Without performing a limiting dilution analysis on samples that have not undergone restimulation, it is often difficult to know to what extent the numbers of T cells generated in such an assay reflect the numbers in vivo. As a result, methods have been evaluated using samples derived directly ex vivo. Without amplification, it can be difficult to use such methods to assess cells that are present only at very low frequencies. Nonetheless, such assays may also have low background, so the signal-to-noise ratio may still be acceptable. To evaluate and standardize ELISpot assays, a European group has demonstrated that with an ex vivo assay it is possible to obtain reproducible data from more than one laboratory [295 ]. These studies were performed using antigens that can reliably give a robust immune response. It is not clear that such methods will be useful for antigens where a numerically less robust response is seen. For multi-center trials, standardized assays ideally should be as simple as possible. That probably means the assays should avoid amplification steps, as although amplification adds sensitivity, it will also add to variability, time, and cost. Consideration also needs to be given to centralizing the performance of assays so that valid comparisons can be made from study to study. If assays are not centralized for multi-site studies, a minimal requirement should be the distribution of standard cells or cell lines that can be used to calibrate assays between laboratories.
Many of these considerations relate to peptide antigens where epitopes are known. A number of vaccine studies are now also being undertaken using a whole protein antigen (such as MAGE-3 or NY-ESO-1) [296 , 297 ]. In these studies, the precise epitopes may not be known, and methods need to be used that can help identify as-yet undefined epitopes. Such assays may include the use of DC to process epitopes that fit the patient’s HLA type or the use of libraries such as overlapping panels of peptides.
Immunological monitoring—conclusions.
It is increasingly clear that laboratory immune surrogates need to be validated against clinical endpoints. The correlation of clinical and laboratory surrogates will potentially enable the use of surrogate endpoints to drive the process of identifying optimal vaccine strategies or optimizing vaccines. Once these strategies have been optimized, the true value of vaccination as a therapeutic modality can be assessed. The steps in optimizing this process could therefore be considered as follows: First, evaluate assays in terms of reproducibility and reliability. Second, correlate immune endpoints with clinical endpoints. Third, use immunological endpoints as a means to optimize clinical trial design. Fourth, test the clinical impact of a vaccine strategy using an optimized trial design, in the setting of advanced evaluable disease or in the setting of minimal residual disease.
What still needs to be changed?
Trial design
There is a clear need for improvement in clinical trial design for biological agents. Traditional endpoints of response (such as rates of tumor regression) need to be rethought, and more relevant endpoints such as clinical benefit, progression-free survival, overall survival, and quality of life will be important. In the case of immunotherapy, it is time to rethink the system by which various treatments are developed. The underlying assumptions are that a potent immunological treatment will result in measurable immunological outcomes and that the strongest immunological responses will correlate with good clinical outcomes. Conversely, patients with advanced cancer are often inherently immunosuppressed as a result of their disease or their treatment, and their cancers will often progress during treatment, so they are probably the worst patient group in which to work up and validate a given immunotherapeutic approach. These assumptions are rarely questioned and are probably reasonable but in fact have never been proven.
The solution to this is that a clear distinction should be made between trials performed to assess clinical outcomes and those
