Tumor Immunology and the Battle of Waterloo

The desire to eliminate cancer using the immune response of the patient has been a goal of tumor immunologists for the last century, since the initial use of Coley’s toxins. With the advent of radiotherapy and chemotherapy, immunotherapy was all but forgotten for more than half a century. Starting again in the 1970s, cancer immunotherapy has captured the interest of numerous immunologists and clinicians. But to apply immunotherapy effectively, many elements, similar to a military campaign, must be considered and applied effectively. On the following pages, we attempt to describe the various choices facing immunotherapists and, as for designing a battle strategy, some of the decisions that must be made.

In recent years numerous target antigens have been defined in human tumors. Most vaccine immunotherapy protocols have centered around melanoma antigens used to vaccinate patients with metastatic melanoma. The reasons for the choice of melanoma are 2-fold. Melanoma does not respond to chemotherapy, and it is one of the two indications where treatment with immunotherapy (IL-2)

² has shown some success. In addition, a variety of melanoma-associated antigens and CTL epitopes ³ from these antigens have been melanoma-associated antigens, and CTL epitopes from them have been identified. These antigens can be used for the immunization of patients, and epitopes from them can be used in one of the recently developed and very sensitive assays for the detection of antigen-specific T-cell activity, such as ELISpot, intracellular cytokine induction, and tetramer analysis.

Tumor antigens fall into three main categories.

First are those coded for by viral genomes. In principle these are attractive targets for immunotherapeutic attack because there should be no reason why the cells capable of responding to these antigens should have been effectively removed from the repertoire by central tolerance-inducing mechanisms. The immune response to these exogenously coded antigens should be vigorous; therefore, interference by other factors (such as peripheral tolerance or escape mechanisms) may be minimal. The success of therapy directed at EBV antigens in transplant patients suggests that under ideal circumstances, this type of response can indeed be effective (1).

The second category of antigens are self antigens altered by genetic changes. Most if not all tumors accumulate multiple mutations during the process of malignant transformation and, in principle, these can provide targets for antitumor response. This type of target has two inherent complications and difficulties for immunotherapists. The first is that the mutations may be immunologically silent because they do not generate new peptides with high affinity for self-MHC molecules. Even if they do, the second difficulty operates, in that each mutation is generally unique so that a generic tumor antigen for immunotherapy is difficult to generate. A separate group of interesting mutations are those that result in a frameshift. Most often, the frameshifted sequence results in a stop codon and, therefore, a truncated protein. However, some frameshift mutations result in novel sequences, which can be very antigenic (2).

Another type of altered self-antigen is exemplified by MUC1. Here, the altered glycosylation is presumably caused by genetic changes affecting glycosylation. Just how distinct these neoepitopes of MUC1 are is perhaps called into question by the evidence that most serologically detected epitopes on tumor mucins can equally be seen in the lactating breast. In practice there is little firm evidence for the development of high frequencies of MUC1-reactive T cells in tumor-bearing patients or even in those immunized with MUC1. Nevertheless, the overexpression of MUC1 by tumor cells and evidence for the generation of MUC1-specific T cells (3) in response to vaccination with MUC1 suggests that this may be a good target antigen for vaccine immunotherapy of MUC1-expressing cancers.

The remaining category of tumor antigens, which have been most often detected by Boon et al. (4), are unaltered self-antigens. These may be either differentiation antigens such as tyrosinase, or gene products found in the testis and fetus such as melanoma-associated antigens. Here the question of self-tolerance remains open. The fact that most of the antigens have been defined using T-cell clones derived from tumor-bearing patients shows that tolerance to these self-antigens is not the result of clonal deletion. The question of whether high-affinity responding cells have been removed from the repertoire leaving responsive cells that are suboptimally activated by the tumor antigens and easily anergized remains to be determined. Nevertheless, in principle, high-affinity ligands exist for every T-cell receptor, and it is also known that T cells activated by an optimal (high-affinity) ligand can recognize targets displaying suboptimal ligands. This has led to the idea of using altered peptide ligands for therapy. Results from an initial trial with an altered carcinoembryonic antigen peptide have already been reported and are encouraging. Strong CTL responses were seen in several patients, and in vitro and clinical response correlated, an unusual finding in most immunotherapeutic trials (5).

The immune attack on any of these targets can have some impact, but the targeting of specific antigens alone may not be sufficient to eliminate tumors completely. In reality, the tumor has a high capacity to deploy new and often decoy molecules. In addition, the heterogeneity of the MHC means that there are no universal peptide “epitopes” (targets) that can be administered for the stimulation of an efficient and specific immune response by the immune system of each patient. Therefore, larger peptides and proteins, which in turn can be specifically adapted by the antigen-presenting cells of the individual patients, are being assessed alone and in combination with traditional therapies, and may prove more effective in clinical cancer therapy.

The speed with which a battle is engaged is an important strategic consideration and applies to cancer immunotherapy. “The Congress of Vienna, alarmed by Napoleon’s return to power, had reacted quickly to the crisis. In Paris, Napoleon, learning of the allies invasion plan, quickly determined to attack the allies on their own ground before their army could take shape.” In animal models of tumor rejection, the tumor has been established shortly before, and often after, the treatment is initiated, whereas advanced disease is usually the norm in clinical trials. By this stage a major pitfall is tumor cell evasion of an effective immune response. Ideally, clinical vaccine trials would be carried out in the adjuvant setting where the tumor burden is small.

Salvadori et al. (6) report that the transfer of spleen-derived T cells from a tumor-bearing animal is ineffective in conferring tumor protection on recipient mice. However, the previous resection of the solid tumor in the donor mouse reversed the donor T-cell defects, restored protective immunity, and protected T-cell recipient mice against tumor challenge. With advancing tumor stage, antigen presentation is often camouflaged, the loss of MHC molecules is demonstrated in many human tumor types (7), and down-regulation of costimulatory molecules occurs. A failure in cellular communication can lead to an immune setting toward tolerance by deletion or anergy of reactive T cells. This is paralleled by a progressive increase of the expression of inflammatory cytokines and transcription factors, sabotaging effective immunization (8). Emphasis is on the importance of the right cytokine environment and on the deleterious effects of IL-10, IL-6, VEGF, and colony-stimulating factor-1 among others (9–11).

For all of these reasons, variables such as type of antigen, minimal effective dose, rhythm of injections, or optimal site(s) of vaccination (12) are ideally explored in preclinical animal studies with localized and recently established tumors. Thus, intralymphatic DNA vaccination was shown to enhance immunogenicity by up to 1000-fold over the i.m. or i.d. routes in mice. Valuable information may be gained rapidly, cost effectively, and with no risk to patients.

The progress achieved in tumor immunology over a century spans from Coley’s toxins to the sophistication of modern cancer vaccines. Early attempts at immunotherapy involved stimulating the immune system intensely but not specifically with IFN and IL-2, resulting in significant toxicity but still showing effectiveness in a minority of melanoma and renal cell cancer patients. Emphasis has since shifted toward immunization with specific peptide antigenic epitopes. More recently, because of the growing awareness of the multitude of the tumor escape mechanisms, more complex strategies are being deployed, aiming to attack cancer on several fronts.

Powerful weapons are being developed using modified viral vectors for the transfer of specific genes designed to enhance the tumor antigen-specific immune response. A number of vectors have been tested, among which vaccinia virus-derived strategies have been brought to clinical development by a number of companies such as Transgène (Strasbourg, France), Therion (Cambridge, MA), and Xenova (Berkshire, United Kingdom). The use of vaccinia vector-based tumor vaccines was emphasized recently at the Miami, FL, AACR-EORTC-NCI meeting in October 2001 by Lattime

⁴ from the Cancer Institute of New Jersey. The large potential size (25 kb) of the insert, the absence of viral integration into the cellular genome, and the excellent immune stimulation by this virus all make it an excellent candidate for immune stimulation in cancer. Vaccinia virus infects all of the cells, and the immune response to the vector does not abrogate the immune response to the tumor after repeated injections. Additional attenuation of the virus (TK-, modified vaccinia Ankara) make vaccinia an attractive vector for vaccination studies in immunodelicate cancer patients. Adenoviral vectors also have a large transgene capacity, a high level of expression, and can infect a large variety of cell types, but limitations are the deficiency of adenoviral receptor expression in certain cell types and the pre-existing immunity limiting transgene expression. Herpesvirus vectors with a tropism for nerve cells have been advocated in the treatment of brain tumors. A number of other viral (adenoassociated virus) and nonviral vectors, such as bacterial (Salmonella or shigella), or lysosomal vectors are currently under scrutiny.

A series of immune modulatory molecules are being evaluated in preclinical settings. An effective immune response, as measured by IFN-γ production, could be restored through vaccination in the presence of anti-IL-10 antibodies (13). Other tactical approaches include the (adenoviral) transfection of DCs with the CD40 ligand coding sequence,

⁵ thereby reducing the need for T-cell help in immunization (14). Circulating CD4+ lymphocytes ⁶ are on average low in the presence of advanced stage disease. Suppression of tumor growth as well as improved survival could be shown after intratumor administration of these CD40 ligand-transfected DCs (special operations forces) in CD4 knockout mice, suggesting that there was no need for T-cell help in these engineered DCs. Additional immune enhancement strategies include the transfection of MIP-3α for the attraction of DCs to the injection site as well as transfection of a soluble flt receptor aiming to act as a dominant-negative effector for VEGF, thereby neutralizing VEGF activity. Genetic interventions using suicide genes together with radiosensitizing or immunomodulatory agents have been packaged in conditionally replicating or nonreplicating (herpes virus) vectors (15, 16). Genes encoding TK, connexin (bystander effect), tumor necrosis factor α, and green fluorescent protein have been inserted into nonreplicating herpes virus vectors, and early results suggest a therapeutic effect, with high survival rates and tumor protection in primate studies. The coexpression of tumor necrosis factor α and TK appeared to be important for tumor control, but efficacy was additionally enhanced through the association with radiation therapy (γ knife). Methods are needed to “protect” potential tumor-specific T cells from tumor-associated immune suppression, the way Wellington protected the feet of his troops from the “inflammatory” mud at Waterloo by the now famous “Wellington Rubber boot,” and an effective classical therapy may be useful to control the tumor while the immune system is being retrained to attack tumor cells.

Confirmation of the appearance of tumor-antigen-specific T cells as a consequence of vaccination has been obtained through a variety of sophisticated techniques. Clinical responses to vaccine immunotherapy in the form of measurable tumor shrinkage are also documented regularly, albeit at low incidence. The documentation of a clinical tumor response in correlation with convincing immunological data remains exceptional. Review of the literature concerning melanoma immunotherapy shows that a demonstrable immune response to the immunizing antigen, and a clinical response can be seemingly mutually exclusive (reviewed in Ref. 17).

If we want to assess which is the best vaccination strategy we may want to progressively increase the complexity of the antitumor design in small patient samples. To achieve this, a randomization between two or more arms is advocated to avoid selection bias. Each arm can be assessed on its own merit. Minimum as well as maximum criteria of toxicity or response can be established to determine whether or not the strategy warrants additional exploration. Because of small sample sizes no a priori statistical comparison can be planned. However, one of the two or several arms can be dropped at the end based on criteria such as:

• observed difference in toxicity (primary end point);

• observed difference in clinical responses time to progression (primary end point);

• observed difference in immunological responses (secondary end point); and

• correlation between observed immune response and clinical response (secondary end point).

At the recent AACR-NCI-EORTC meeting (October-November 2001) in Miami, FL, Pat Price from the University of Manchester, Manchester, United Kingdom (18), reviewed the most reliable methods to assess the efficacy of antiangiogenic compounds using either functional computed tomography, dynamic contrast-enhanced MRI, ultrasound, or ¹⁵O-positron emission tomography scanning techniques. Emphasis was on the complexity of the measurement and the need to validate these techniques as potential early surrogate end points for a variety of cancer treatments. The availability of an imaging tool combining either MRI, positron emission tomography, or ultrasound scanning with a biological readout system, which is able to correlate or even predict a successful tumor immune response, would greatly facilitate future clinical designs. Nevertheless, these are measurements of the “nonimmune” parameters. In immunotherapy it is important to be able to measure an “effective,” tumor-specific immune response.

If the immune response to a tumor can be compared with a battle between immune cells and tumor cells, it is evident that assessing immune responses by T cells taken from the peripheral blood is akin to sampling soldiers or messengers from the roads leading to a battlefield, such as Waterloo. Fig. 1A shows a map of the battleground at Waterloo. If a picture had been taken from the roads leading to the battlefield on the French side, the battle-ready troops of Commander D’Erlan would have been visualized. These troops, because of an unfortunate miscommunication, were going back and forth between Ligny and Quatre-bas, and as a result did not take part in the battle. As such, fresh troops, capable of recognizing and fighting the enemy, could be found on the road, but because of opposing orders from two commanders, these troops were of no help to Napoleon. If, on the other hand, we had surveyed the roads leading to the battlefield on the side of the allies, we would have seen messengers telling of the defeat of the Prussians at Charleroi and Gilly, foretelling a defeat of the allies under Wellington. Again, this information would not have accurately predicted the winner at Waterloo. However, if we had examined the battlefield itself, throughout the day of the battle, it might have become clear who was the victor.

In this context, the design of more sensitive methods to follow immune responses or the preferential analysis of lymphocytes taken from tumor draining lymph nodes have been suggested. Thus, combining the tetramer technology to immunohistological staining techniques allows the detection of the presence as well as the activation state of tumor antigen-specific T cells. Nevertheless, a biopsy is not always feasible in metastatic disease and remains a snapshot, and a single snapshot at noon or 4:00 p.m. may not have been discriminating enough to foresee the outcome at Waterloo.

What is clearly needed is a noninvasive real-time method allowing a series of observations of areas within the tumor and throughout the course of an immunotherapy protocol. Currently, this technology is not available, but interesting advances are being made. A recent paper by Louie et al. (19) shows that the MRI contrast reagent, gadolinium, can be encased in a molecular basket referred to as EgadMe, rendering it invisible to MRI. One of the bonds of the basket is susceptible to enzymatic activity, such that when the EgadMe complex is in the vicinity of this enzyme, the top of the basket is prized off and the gadolinium evokes a signal on the MRI scan. The enzyme used in this study was β-galactosidase. Enzyme activity could be seen in living embryos of transgenic Xenopus laevis. Evidently, the important advance represented by this system is the chemistry that went into constructing the EgadMe basket.

Several enzymes associated with CTL and natural killer cell activity have been identified, such as Granzyme B, substrates of which are known. Efforts could be made to incorporate a granzyme-susceptible substrate into an EgadMe-like basket around a contrast reagent such as gadolinium (or iron oxide). In this way, a contrast reagent could be devised, which would allow the detection of CTL and natural killer activity inside a tumor or the draining lymph nodes at a succession of time points after vaccination. It could also be imagined that specific antigens, as well as the susceptible bond, might be incorporated into the complex to allow the visualization of specific T-cell activity.

On balance, immunotherapy of cancer can be improved on several fronts, by: (a) improving the choice of patients eligible for individual immunotherapeutic strategies; (b) improvements to the immunotherapeutics themselves by rendering the target antigen more “antigenic” (application of altered peptide ligands), the weaponry more specific, and by designing complex tactical strategies; and (c) more efficient monitoring of the effectiveness of the immunization (and, therefore, of the immune response) by noninvasive, real-time imaging of specific immune responses inside the tumor, and its draining lymph nodes, which would enable the evaluation of any correlation between the observed immune response and clinical effectiveness of the vaccination strategy.

We feel that despite the emergence of promising data, there is yet a long way to go for immunotherapy, and there have been many false dawns. Even if altered ligands prove able reproducibly to induce responses capable of recognizing unaltered tumor antigens, therapy may still fail because of tumor escape. Effective immunotherapy will have to be instituted much earlier in the course of disease than is usually the case if it is to have a realistic chance of preventing escape. Still, whereas some of the initial battles between the immune response and tumors may not have resulted in clear victories for immunology, many improvements to the armory and battle strategies are possible, and their application will undoubtedly improve the outcome. Finally, if cancer vaccines in patients are to be rendered more effective, better means for monitoring the early stages of treatment need to be devised. It is essential that we can ascertain: (a) whether the patient has been “successfully” immunized; and: (b) whether the immune effects we see, do predict and correlate with a clinical response.