Covert Operations: Cancer's Many Subversive Tactics in Overcoming Host Defenses

Abstract

In an effort to use the patient's T cells to fight his or her own cancer, we inadvertently discovered a distinctive form of tumor-induced immune suppression. T cells from tumor-bearing patients are often defective in signaling. They lack the zeta chain of the T-cell receptor and the src kinases crucial for its downstream effects including lck. They truncate the carboxy terminal of the p50 NF-kappaB transcription factor. At the population level, CD4 T cells are polarized toward the Th2 subtype and inhibitory Tregs expand. These T cells can recover after several days of culture outside of the tumor-bearing host environment. The effect is mediated by one or more factors made by the tumor given that the same T-cell defects occur in mice with tumors implanted in hollow fibers that never directly contact cells of the host. Several promising strategies may overcome these immunosuppressive effects.

Efforts to exploit the awesome destructive power of the immune system to fight cancer have had only a modest yield to date. The study of host/tumor interactions has shed considerable light on the many sources of difficulty that must be overcome (1–3). Tumors induce the expression of scavenger receptors on antigen-presenting cells that cause them to ingest large quantities of lipid, which interferes with antigen presentation. Tumors downregulate their major histocompatibility complex (MHC) antigens, which makes them invisible to T cells. Inhibitory subsets of T cells (so called T-regulatory cells) expand in patients with cancer and suppress T-cell responses. Myeloid-derived suppressor cells develop over time to inhibit tumor recognition and rejection. Tumors also directly interact with the T cells to disable them in various ways. They produce ligands to activate inhibitory receptors like CTLA-4 and PD-1. They make FAS ligand to induce T-cell apoptosis. They make poorly defined substances that alter T-cell signal transduction including the induction of enzymes that degrade lck and the CD3-zeta chain, and alter the structure of transcription factors such as NF-kappaB. They metabolize tryptophan into immunosuppressive kynurenine. Despite all of these (and more) actions of tumors to prevent their elimination by host defenses, substantial recent progress promises to allow T cells to join the therapeutic armamentarium. For example, the introduction of specialized signaling molecules into T cells and their adoptive transfer into patients with cancer has had dramatic antitumor effects in a few early studies. Antibodies that block the effects of the inhibitory molecules CTLA-4 and PD-1 appear to boost T cell antitumor effects. New vaccine strategies have also shown promising clinical effects. It appears we are nearing a time when we can add immune activation to the list of tools that can be used to fight cancer.

Many interpret the development of a cancer as a failure of the immune system. The concept of immune surveillance suggests that cancers are likely developing in each of us; a healthy immune system defeats most of these cancers before they have a chance to take hold. When immune surveillance fails, cancer develops.

The immune system has many effector mechanisms that can be deployed to kill or alter cancer cells, but much attention has been focused on T cells. One school of thought has been that one needs to use activated T cells that recognize tumor-specific antigens. This concept has been applied in several ways, including efforts to actively induce antigen-specific immunity by vaccinating the patient with tumor antigens (4) and by harvesting and growing T cells that infiltrate tumors in vivo, known as tumor-infiltrating lymphocytes (5). However, in any T-cell infiltrate, whether in tumors or in delayed-type hypersensitivity responses to cutaneous antigens, only a minority of the cells in the infiltrate are specific for the inciting antigen. T cells have effective measures of recruiting other T cells that are not necessarily antigen-specific. Studies in animal tumor models noted that polyclonal T cells could be activated in vitro with anti-CD3 epsilon antibodies and administered together with interleukin-2 (IL-2) to achieve effective antitumor responses in the setting of metastatic disease (6).

When this approach was tested in a clinical trial, the results were surprising (7). The in vivo expansion of T cells with an activated phenotype (Ia-positive, CD25- and CD69-positive) far exceeded expectations. The median peak circulating lymphocyte count during treatment was 38,000/mm3 and one patient had nearly 200,000 T cells/mm3 prompting concerns about blood viscosity (the count decreased swiftly when the IL-2 was stopped). However, only approximately 15% of the patients experienced tumor responses. Some responses were dramatic (Fig. 1), but the overall antitumor effect was disappointing.

Fig. 1.

Chest CT scan of a patient with metastatic melanoma before and after treatment with anti-CD3–activated autologous T cells showing a partial response of lung metastases.

Then we asked a question. What is the difference between the animal experiments that showed dramatic antitumor effects with these anti-CD3–activated T cells and the human clinical trial that did not? One answer that we considered was that in the animal model, the T-cell donor was a healthy genetically identical mouse; whereas in the clinical trial, the T-cell donor was a cancer-bearing patient. Therefore, we redid the mouse experiments using mouse donors that contained growing tumor cells. The first surprise was that animals bearing tumor for only a week (short term) seemed to have enhanced antitumor activity. We have not yet had a chance to follow-up on this interesting lead. We have been focused on defects in T cells from tumor-bearing animals. As shown in Fig. 2, splenocytes from mice that had growing tumor for 26 days (long term) were ineffective in mediating antitumor effects in adoptive transfer (8). Cytotoxic T cells from long-term tumor-bearing mice functioned poorly and expressed reduced levels of granzyme B.

Fig. 2.

Histogram showing number of hepatic metastases from MCA-38 in animals treated with anti-CD3-activated T cells plus IL-2. The antitumor effects were influenced by the status of the donor. T cells from animals without tumor or bearing tumor for 7 days showed antitumor activity when transferred to an adoptive host with cancer. Indeed, T cells from animals bearing tumor for a week appear to have enhanced antitumor effects, a finding that must be further pursued and clarified. By contrast, T cells from animals that had been inoculated with tumor cells 26 days earlier and had actively growing macroscopic tumor showed no antitumor effects.

On a cell basis, the T cells had normal levels of CD3 on the surface, but they had impaired calcium release upon activation, one of the earliest events in the process of activation, and analysis of the structure of the T-cell receptor revealed dramatic changes. The normal structure of the T-cell antigen receptor/CD3 complex is shown in Fig. 3. The zeta chain is generally the rate-limiting step in assembly of the complex. It is made in the smallest amount and assembles the complex in a stoichiometrically correct fashion at the cell surface. Zeta also connects the complex to its downstream tyrosine kinase enzymes such as lck and fyn. When T cells from tumor-bearing mice were examined for the structural integrity of the receptor, a number of abnormalities were noted (Fig. 4). Zeta chain was absent and gamma chain was nearly absent. In some cases, a zeta chain family member, Fc-epsilon gamma, had substituted for zeta with an alteration in the complex (9). Downstream signaling molecules (lck and fyn) were also absent. Structural alterations were also noted in transcription factors of the NFkappaB/rel family (10). All of these molecular abnormalities reversed when it was possible to effectively treat the murine tumor.

Fig. 3.

Schematic diagram of the multichain T-cell receptor signaling complex. The dark bands indicate where downstream signaling molecules attach to the receptor. CD8 tends to associate with the CD3-gamma chain and CD4 with the delta chain. Src family kinases, lck and fyn, associate with the zeta chain, which is also involved in the assembly of the complex at the cell surface.

Fig. 4.

Two-dimensional gel electrophoresis of the CD3 complex from T cells from normal mice and from tumor-bearing mice. The CD3 complexes on T cells from tumor-bearing mice have a number of alterations including loss of zeta chain and a decrease in gamma chain. In some cases, a zeta family member, Fc-epsilon-gamma, may be substituted for the zeta chain (arrow in the right panel indicating the spot migrating below the diagonal).

On a population basis, a gradual loss of Th1 helper T cells was noted in the spleens of mice during progressive tumor growth, a change that compromised the generation of an effective antitumor cytotoxic T-cell response (11). The immunologic defects began in the tumor itself affecting tumor infiltrating lymphocytes and then spreading to draining lymph node T cells and ultimately systemically to affect the entire immune system (12). These changes were also documented in patients with renal cell cancer (13), and subsequently in melanoma (14), colorectal cancer (15), lymphoma, and other tumor types. Although definitive large-scale human studies have not been completed, it appears in pilot studies that patients with cancer whose T cells lack zeta chain have a poorer overall survival than those with normal T cells. These findings have been widely reproduced (16–20).

What is the mechanism of this tumor-induced immune suppression? At least one mechanism contributing to the loss of zeta chain and the signaling molecules is protein degradation. Zeta-less T cells from tumor-bearing animals have increased levels of zeta chain messenger RNA; thus, the absence of the protein does not appear to be transcriptional. Definitive answers have been elusive because cell culture models take more than 2 weeks to develop the characteristic abnormalities when exposed to cell-free tumor cell supernatants. However, several inferences have been made and many known factors that were candidates to mediate the effect have been ruled out. First, we believe the factor is a tumor product smaller than 100 kD. The evidence for this statement is shown in Fig. 5. When hollow fibers containing the murine renal cell cancer, RENCA, are placed into the peritoneal cavity of naïve mice and allowed to grow, the mice develop the same immune defects as animals in which the tumor is growing in contact with host tissues. The hollow fibers allow substances smaller than 100 kD, but not larger ones, to permeate. Therefore, we assume the tumor factor or factors are less than 100kD. The factor is destroyed by heat and trypsin; therefore, we believe it is protein in nature. The factor is not IL-10 or transforming growth factor-beta. Additional studies to characterize the factor responsible for the immune suppression are ongoing.

Fig. 5.

Tumor cells can be grown in vivo in hollow fibers that permit tumor products to seep out of the fibers and exert influences on the host. In this experiment, a murine renal cell cancer, RENCA, was grown in hollow fibers that allow the passage of molecules smaller than 100 kD. It shows that the tumor does not have to be in direct contact with any cell of the host to exert its effects on lck and zeta chain expression.

Others are convinced that the mechanism of zeta chain destruction is based on oxidative stress. Some have suggested that macrophages make hydrogen peroxide that causes the defect (21). Others have suggested that the uptake of arginine by macrophages induces the production of factors that destroy zeta chain (22). A tumor-derived factor has not been ruled out as an explanation (23).

Whatever the mechanism may be for this particular type of immune suppression, it is clear that it is associated with functional T-cell defects, structural changes in the T-cell receptor for antigen, alterations in T-cell signal transduction, absence of nuclear c-rel and NF-kappaB p65 after activation, loss of Th1 pattern nuclear binding proteins, and polarization of the helper cell population to Th2 rather than Th2 phenotype. The changes appear to correlate with the clinical course. Successful therapy may reverse the defects. The process appears to be mediated at least in part by soluble factors probably from the tumor.

Similar defects in T-cell zeta chain expression and signal transduction have been documented in patients who have chronic infections and inflammation (24–26), in autoimmunity (27, 28), and in pregnancy (29) prompting the speculation that zeta chain destruction may represent a mechanism of receptor desensitization to prevent over-reaction to a particular stimulus (30, 31). If the T-cell signaling defects induced by cancer are derived from physiologic downregulation of the immune response to chronic stimulation, they are inappropriate in the setting of growing cancer.

Figure 6 depicts a partial representation of host-tumor actions that serve to protect the tumor from host defenses. The induction of T-cell signaling defects is one of many such mechanisms. Their reversal by some intervention represents only one component of what may be necessary to harness the destructive power of the immune system against cancer.

Fig. 6.

Cartoon illustrating the many pathways in which tumors attempt to block host defenses. They produce suppressive cytokines, polarize the T cell response to Th2 help (antibody synthesis rather than cytotoxicity), interfere with antigen presentation, induce cells that suppress immunity and abet metastases, mask their presence by downregulating MHC molecules, and directly inhibit the function of T cells by blocking signaling and degrading signaling components.

DISCUSSION

Rich, Birmingham: These are fascinating observations. I am wondering if you've had any opportunity to look at the other phenotypic characteristics of the zeta-chain-absent T cells to see whether they can be characterized phenotypically as some of the other kinds of cells that we've recognized in the panoply of T cells that have regulatory effects.

Longo, Boston: The answer to that question is that they seem to produce Th2-type cytokines. The CD4 cells tend to be preferentially Th2. There is a relative deficiency of cytotoxic CD8-positive T cells. It's not clear that the CD8 isn't made, it's just not associated with the receptor and therefore the cells that might be cytotoxic don't seem to be effective in killing.

Billings, Baton Rouge: As a very much tangential thought, the nine-banded armadillo is the only mammal that also has been infected with leprosy and I am wondering if the T cells in that vicious beast have been looked at. You know, armadillos are born dead on the highway but I thought maybe you might see what's going on with the T cells that allows leprosy to be infectious in the armadillo and also man but there are no other mammals that have leprosy.

Longo, Boston: I am unaware of detailed study of the armadillo immune system but it's very interesting. People with leprosy have similar defects in their T cells; zeta chain is absent from the receptor.

Billings, Baton Rouge: I mean you might look at the T cells and see why it is that they do what they do or don't do.

Longo, Boston: It's an interesting idea. Much of the look at comparative immunology has been hampered by the fact that reagents we have to probe the immune system don't tend to cross-react across species as well but it's a very interesting point.

Falk, Chapel Hill: That was lovely. There is another interesting protein that's involved in that immunological synapse that you've so nicely described and one that coimmunoprecipitates with both the delta chain and the beta chain. I don't know whether it coimmunoprecipitates with the zeta chain and that's non-muscle MYH9. Some of your blots actually raise the possibility that MYH9 may be playing a role here. Have you explored that possibility that it's actually a defect in MYH9 that's driving this?

Longo, Boston: It's on the list of things to look at but we haven't gotten to that yet. It's a very good idea, very good.

Schreiner, Los Altos: Dan, rather than think of this as a system that cancer cells have developed out of a malignant intelligence, have you looked at whether or not this might represent an adaptation of other systems of T-cell desensitization? I'm thinking specifically of the growing awareness that cancer in many ways shares some biological similarities with embryonic development. There are recent data that women who have more children often are microchimerous showing evidence or incorporation actually of both son and daughter DNA carrying cells within their own body structures. Have you looked at whether or not the T-cell desensitization of a mother toward an embryo may share some qualities analogous to what you're observing in the tumor and that this really may represent an adaptive use of something that already occurs which is a selective desensitization of cell-mediated immunity in a natural setting?

Longo, Boston: We have looked at the circulating T cells of pregnant women and have not found zeta chain defects in particular. There are other interesting things that have happened to those cells biochemically. In fact, it looks as though the PI3 kinase pathway has been blocked by something so that instead of the signaling that they get from the baby, producing activation in memory, it induces tolerance. I don't know whether or not that can explain the symbiotic relationship and the tolerance of cells from the baby in the mother but it's a very important area for further investigation. I just don't think that the zeta chain downregulation is part of that process.

Howley, Boston: Your model includes the zeta chain and lck being degraded. I think you implicated a proteasome ubiquitin-mediated proteolysis in its degradation. As such, do proteasome inhibitors block it, given that proteasome inhibitors are now in the clinic?

Longo, Boston: Yes. Some of the protease activity is blockable with MG132. We have some evidence that zeta chain degradation in T cells from tumor-bearing animals may be inhibited somewhat by proteasome inhibitors, but so far in vivo data are not conclusive.