Caveats important to consider when evaluating prior work

Several caveats should be kept in mind when reviewing the results of prior experimentation in glioma immunology. Recent insight into the endogenous suppressors of cellular immunity, increasing awareness of the complexity of regulatory circuits in immunologic systems, and more common acceptance of the limitations of glioma cell lines and animal models require more critical retrospection of prior experimentation implicating a wide spectrum of glioma-associated factors in the suppression of antitumor immunity.

Cellular heterogeneity is a sine qua non of GBM, with histologic analysis demonstrating an amalgam of presumed tumor cells, reactive astrocytes, tumor-associated microglia, trapped neurons, and various cells infiltrating from the peripheral circulation (including lymphocytes and macrophages). Study of bulk (ie, “unsorted”) tumor specimens does not take into consideration potentially significant differences in expression patterns amongst these varied cell types, rendering RNA and protein analysis of limited usefulness. These differences are of critical importance when attempting to categorize phenotypic and functional characteristics of cells involved with the cellular immune response. Although immunohistochemistry (IHC) can provide some specificity of cellular expression and subtumoral anatomic localization, this technique is not particularly conducive to the concurrent detection and quantitation of multiple markers necessary for identification and subclassification of relevant regulatory cells.

The use of glioma cell lines has been a mainstay of research evaluating expression patterns of immunologically relevant markers from malignant brain tumors. However, recent comparative analyses have suggested that there are in fact dissimilarities between commonly used glioma cell lines and direct ex vivo tumor specimens. A seminal series of experiments, recently reported by Howard Fine’s group, used comparative chromosomal analysis and microarray expression profiling between a large number of primary GBM specimens and a range of commonly used glioma cell lines (including U87, U251, and T98G). They found that the overall spectrum of genomic alterations differed substantially between primary tumors and the tested cell lines. Their analysis demonstrated that observed expression patterns and genomic abnormalities within tested glioma cell lines were more congruent to abnormalities observed in other nonglioma cell lines, suggesting that selection pressure of in vitro culture resulted in activation or alteration of common “transformation” pathways. Similar analysis of immunologically relevant markers in short passage cultures of fresh human GBM specimens has demonstrated a rapid shift in the expression of immunologically relevant markers, including downregulation of major histocompatibility complex (MHC)-I and increased expression of TGF-β. Recent data supporting the role of “cancer stem cells” found at high frequencies in human GBM and proven to harbor a tumorigenic phenotypic in xenograft models mandate that future in vitro study includes these cell populations. Cells with stemlike characteristics are notoriously sensitive to culture conditions for maintenance of a limited-differentiation phenotype, a consideration which sets in vitro experimentation with these populations apart from standardized culture cocktails used in most prior in vitro glioma experimentation. Although such observations do not obviate the results of prior studies investigating the expression of immunologically relevant factors through in vitro culture, these studies mandate that the results of in vitro testing are further confirmed in more biologically relevant systems.

Investigators have used many animal models for comparative analysis in the study of immune responses to malignant brain tumors. These models have been primarily developed in rodents, using stereotactic injection of glioma cell lines into the brains of test animals, allowing for consistent and predictable tumor growth. However, these models often bear little pathologic and immunologic similarity to human GBM. Many of the commonly used rat glioma lines grow as “pushing masses” rather than infiltrating neoplasms, and formed tumors often do not demonstrate the pathologic hallmarks of human malignant glioma. From an immunologic standpoint, these models are often more representative of a transplant system, in some cases even demonstrating genetic mismatch between transferred cell lines and the host animal (rendering the immunologic relevance of these models limited at best). In addition, it is possible that the procedure involved with stereotactic injection of tumor cells may alter the immunologic microenvironment of the brain, providing an increased “danger” signal and promotion of a proinflammatory environment. More recently, xenograft systems using the injection of dissociated human GBM specimens or human glioma cell lines into immunoincompetent rodents has become a commonly accepted method for studying glioma biology. Deficiencies in immunity that allow for engrafting of tumor in these models concurrently prevent any meaningful study into tumor-specific immune responses. In addition, the evaluation of expression patterns of tumors initially derived from human glioma cell lines subsequently xenografted into immunodeficient animals confirms a closer correlation of derived tumors with other long-term cultured cell lines, rather than ex vivo human glioma specimens. More recently, tumor models driven by retroviral targeting of white-matter progenitor cells in rodents have been developed that closely recapitulate the pathologic hallmarks of human GBM (including pseudopallisading necrosis, microvascular proliferation, and infiltrative growth). Preliminary studies have suggested that the frequency and phenotype of tumor-infiltrating lymphocytes (TIL) within these lesions more closely resemble profiles of TIL within human GBM than do other rodent models (B. Killory and D. Fusco, unpublished data, 2009). Although preliminary, such models may be instrumental in the further study of relevant immune responses to malignant brain tumors.

Although in vitro and animal data are important components of ongoing immunologic research in glioma, increased emphasis must be placed on the further development of more immunologically relevant models and renewed study of fresh human tissues. This article focuses on immunoregulatory factors explored or confirmed to be present within human tissue.

Observations of cellular immunity in patients with malignant glioma

What is known about cellular immunity and the antitumor immune response in patients with malignant glioma? T cells can be found within human GBM specimens, although these cells are present at low frequencies and are primarily restricted to the perivascular spaces. These intratumoral populations also demonstrate a skew from the expected CD8+ (“effector”) phenotype towards a more dominant CD4+ representation. Studies have also demonstrated the presence of a limited repertoire of T-cell receptor expression on TIL within GBM specimens, and the presence of a predominantly CD45RA-negative (or “memory”) phenotype, suggesting that there may be an antigen-specific expansion of T-cell subsets within these lesions. In addition, tumor antigen-specific T cells can be generated from bulk peripheral blood mononuclear cell (PBMC) samples from affected patients, providing evidence for potential tumor-specific cellular immunity. However, these potentially tumor-specific T cells are ultimately ineffective at providing any benefit in regards to tumor control. Certain T-cell abnormalities can be temporarily reversed following tumor resection, and subsequently return with tumor recurrence, potentially confirming an association between tumor and the suppressive phenotype rather than a global defect in cellular immunity.

Circulating T cells in patients with glioma-associated immunosuppression seem to be globally abnormal through in vitro analysis. Although not systemically immunocompromised, these patients generally exhibit a loss of CD4+ cells, resulting in a CD4/CD8 ratio closer to 1 (rather than the usual 2:1) and a total mild lymphopenia. CD4+ and CD8+ T cells are handicapped in their response to in vitro mitogenic stimulation, a phenomenon that has been partially attributed to decreased production of and hyporesponsiveness to interleukin 2 (IL-2). Aside from widely reported anomalies of T-cell number and function in patients with GBM, further detailed analyses of various signaling pathways involved with T-cell activation have demonstrated significant abnormalities in expression levels and poststimulation phosphorylation patterns of proteins downstream from the T-cell receptor, including PLCγ1, pp100 and p56lck.

The aforementioned findings offer critical insights into the nature of GBM-associated immunosuppression. First, T cells can enter the tumor microenvironment and potentially respond specifically to antigen, but these cells are somehow blocked in their ability to expand into effector populations and provide tumor clearance. Second, considering that most peripheral/circulating T cells are never physically exposed to tumor cells, and assuming that affected patients are not globally immunocompromised, the presumed tumor-specific immunosuppressive factor must extend far beyond the tumor microenvironment. Third, considering that relative immunosuppression can be transiently reversed following tumor extirpation, and returns with tumor recurrence, the tumor itself provides a driving force behind the generation of the immunosuppressive effect.

Observations of cellular immunity in patients with malignant glioma

What is known about cellular immunity and the antitumor immune response in patients with malignant glioma? T cells can be found within human GBM specimens, although these cells are present at low frequencies and are primarily restricted to the perivascular spaces. These intratumoral populations also demonstrate a skew from the expected CD8+ (“effector”) phenotype towards a more dominant CD4+ representation. Studies have also demonstrated the presence of a limited repertoire of T-cell receptor expression on TIL within GBM specimens, and the presence of a predominantly CD45RA-negative (or “memory”) phenotype, suggesting that there may be an antigen-specific expansion of T-cell subsets within these lesions. In addition, tumor antigen-specific T cells can be generated from bulk peripheral blood mononuclear cell (PBMC) samples from affected patients, providing evidence for potential tumor-specific cellular immunity. However, these potentially tumor-specific T cells are ultimately ineffective at providing any benefit in regards to tumor control. Certain T-cell abnormalities can be temporarily reversed following tumor resection, and subsequently return with tumor recurrence, potentially confirming an association between tumor and the suppressive phenotype rather than a global defect in cellular immunity.

Circulating T cells in patients with glioma-associated immunosuppression seem to be globally abnormal through in vitro analysis. Although not systemically immunocompromised, these patients generally exhibit a loss of CD4+ cells, resulting in a CD4/CD8 ratio closer to 1 (rather than the usual 2:1) and a total mild lymphopenia. CD4+ and CD8+ T cells are handicapped in their response to in vitro mitogenic stimulation, a phenomenon that has been partially attributed to decreased production of and hyporesponsiveness to interleukin 2 (IL-2). Aside from widely reported anomalies of T-cell number and function in patients with GBM, further detailed analyses of various signaling pathways involved with T-cell activation have demonstrated significant abnormalities in expression levels and poststimulation phosphorylation patterns of proteins downstream from the T-cell receptor, including PLCγ1, pp100 and p56lck.

The aforementioned findings offer critical insights into the nature of GBM-associated immunosuppression. First, T cells can enter the tumor microenvironment and potentially respond specifically to antigen, but these cells are somehow blocked in their ability to expand into effector populations and provide tumor clearance. Second, considering that most peripheral/circulating T cells are never physically exposed to tumor cells, and assuming that affected patients are not globally immunocompromised, the presumed tumor-specific immunosuppressive factor must extend far beyond the tumor microenvironment. Third, considering that relative immunosuppression can be transiently reversed following tumor extirpation, and returns with tumor recurrence, the tumor itself provides a driving force behind the generation of the immunosuppressive effect.

Suppressive surface markers expressed on glioma cells

Several studies have explored the expression of surface markers on GBM cells with putative suppressive effects on cellular immunity. The most widely explored of these factors is likely Fas ligand (FasL or CD95L). Fas (CD95) is constitutively expressed on activated T cells in CD4 and CD8 compartments, and the Fas/FasL system is known to be an important regulator of activated T-cell populations. Binding of Fas and FasL is believed to be critical in mediating activation-induced cell death (AICD). AICD is essential for downregulating activated T-cell populations at the termination of immune responses, for maintaining T-cell homeostasis, and in the prevention of autoimmunity. Although FasL is primarily expressed on the surface of activated T cells, several groups have documented the expression of FasL on the surface of glioma cells. The resulting hypothesis focuses on the potential killing of activated tumor-specific T cells through cell-cell contact with FasL-expressing glioma cells, thereby resulting in effective escape from the cellular immune response. To provide potential functional relevance for this hypothesis, Ichinose and colleagues correlated the expression of FasL in 9 of 14 GBM specimens with decreased numbers of TILs in relevant samples. Yu and colleagues performed double-staining IHC for FasL and endothelial cell markers, finding that approximately 30% of tested GBM samples demonstrated positive FasL staining on tumor vasculature. In contrast to the findings of Ichinose and colleagues, these investigators did not find any correlation between FasL expression and extent of lymphocytic infiltration, although they did identify a shift in CD4/CD8 ratios in positive samples. Frankel and colleagues identified increased levels of soluble FasL in fluid aspirated from cystic malignant gliomas. Surface-bound FasL can be cleaved by metalloproteinases, resulting in a soluble yet functional form of the protein. Soluble FasL within cystic aspirates from astrocytoma was effective at inducing cell death in cultured Jurkat cells. These investigators suggested that elevated FasL in cyst fluid is representative of concentrations in the tumor extracellular microenvironment and therefore potentially effective at killing infiltrating tumor-specific T cells.

The expression of nonclassic human leukocyte antigen (HLA) molecules by glioma cells has also been explored as a potential source of immunosuppression. In contrast to classically described MHC-Ia molecules, which are expressed on cells of virtually all tissue types and primarily responsible for peptide presentation to CD8+ T cells, MHC-Ib molecules bind a restricted set of peptides and have been primarily associated with natural killer (NK) cell-mediated immunity through the binding of the CD94/NKG2 class of receptors. The expression of these markers, previously associated with suppression of cellular immunity, has been explored in GBM. The first of these is HLA-G, which has been identified on the surface of several glioma cell lines and a limited subset of fresh glioma samples. A potential functional role for HLA-G in glioma was demonstrated through gene transfer experiments, which suggested that forced expression of HLA-G by U87 glioma cells rendered the transfected cells resistant to killing by alloreactive PBMCs. Additional studies demonstrated that increased levels of soluble HLA-G can be detected in sera from patients with multiple malignancies, including glioma, although a functional role for this finding has not been defined. HLA-E, a second member of the MHC-Ib family, has been similarly associated with glioma. Advanced tumor grade is paralleled by increasing levels of tumor-specific HLA-E expression, although no functional relevance has been provided for this finding. A potentially contradictory association between expression of HLA-E in GBM specimens with increasing numbers of infiltrating CD8+ T cells has been reported, resulting in an unclear impact of these nonclassic HLA proteins on glioma-specific cellular immunity.

More recent data have explored the potential functional relevance of glioma-specific expression of B7-homolog 1 (B7-H1, also known as programmed death ligand 1, or PD-L1). The CD28-B7 family of costimulatory molecules has been expanded to include several receptor-ligand pairs that have potent regulatory effects on antigen-specific T cells. The classic role for the canonical members of this family, CD28 (expressed on T cells) and B7-1 (expressed on antigen-presenting cells), is to provide necessary costimulation in the process of antigen-specific activation through the T-cell receptor. Newer members of this family, including B7-H1, have been suggested to be critical regulators and suppressors of activated T-cell function. Involvement of B7-H1 in glioma-associated immunosuppression has been hypothesized and explored by several groups. Studies have demonstrated low-level expression of this marker on multiple glioma cell lines. In addition, single-color IHC has provided evidence for B7-H1 positivity in the small number of GBM samples tested. Further experimentation demonstrated that coculture of alloreactive T cells with B7-H1+ cell lines, in the presence of blocking antibodies, can drive increased expression of proinflammatory cytokines and T-cell activation markers. However, no evidence for increased T-cell death was identified in the absence of blocking antibodies, which argues against a baseline functional role in this system. A series of experiments recently reported by Parsa and colleagues identified a link between the commonly identified mutation in phosphatase and tensin homolog (PTEN) and increased expression of B7-H1. Expression levels of B7-H1 were directly correlated with loss of PTEN function, shown to be secondary to increased Akt activity and posttranscriptional upregulation of B7-H1 on the cell surface. PTEN-deleted cells were rendered less susceptible to cytotoxic T lymphocyte-mediated targeting, providing evidence for a potential functional connection between a commonly identified mutation in GBM and the immunosuppressive effect.

Although the expression of immunosuppressive factors on the surface of glioma cells may provide some explanation for local effects on activated T cells, this hypothesis does not adequately define a comprehensive solution to the systemic immunosuppression problem. Again, considering that most circulating T cells do not come into direct contact with tumor cells, additional far-reaching factors are necessary to explain the characteristically global defects in affected individuals.

Glioma-secreted factors with potential immunosuppressive effects

In 1984, Fontana and colleagues presented the first report providing evidence for a factor secreted from cultured glioma cells that inhibited IL-2-induced T-cell proliferation and blocked allogeneic T-cell responses in mixed lymphoid reactions. Over the next several years, continued study from this group, and others, confirmed the immunosuppressive effects of this factor. Purification and sequencing of the responsible protein, initially named GBM-derived T-cell suppressor factor (G-TsF), demonstrated significant homology with human transforming growth factor beta (TGF-β). Subsequent study found G-TsF to be identical to TGF-β2. Expression of all 3 isoforms of TGF-β by GBM was ultimately documented. This family of proteins has been shown to significantly impair activation of T cells and monocytes, and more recently TGF-β has been associated with the generation of CD4+ T regulatory cells (Treg, discussed later). Although most data have focused on the expression of TGF-β by glioma cells in culture, potential in vivo relevance has been provided through IHC analysis of GBM specimens, indicating that areas of high extracellular TGF-β concentration within tumor samples have fewer numbers of infiltrating lymphocytes than do regions with lower levels of TGF-β. Although the secretion of TGF-β provides a compelling explanation for glioma-associated immunosuppression, the potential for systemic effects is unclear. Several studies using comparative analysis of TGF-β levels in serum from GBM patients and controls identified no difference between the 2 groups, suggesting that systemic levels were not significantly altered by the presence of tumor.

Tumor-specific production of immunosuppressive prostanoids, most notably prostaglandin E2 (PGE-2), has also been associated with suppression of antitumor immunity. Studies have demonstrated high levels of expression of cyclooxygenase type 2 (COX-2) and arachadonic acid (the substrate for prostanoid production) in malignant glioma. Elevated intratumoral levels of these factors have been correlated with increasing pathologic grade and proliferative behavior. These findings have resulted in some interest in the use of selective COX-2 inhibitors for treatment of GBM. Production of PGE-2 has also been suggested to play a role in GBM-associated immunosuppression, although attempts to demonstrate PGE-2-specific immunosuppressive effects, using coculture experiments with mitogen-driven proliferation of PBMC in the presence of glioma supernatant, have provided mixed results. A more recent study suggested that COX-2 expression by GBM, with secondary induction of IL-10 production from intermediary dendritic cells, resulted in the generation of CD4+ Treg cells. This effect could be blocked by selective COX-2 inhibitors and subsequently rescued via addition of exogenous PGE-2, although in vivo correlates of these findings have yet to be provided.

Multiple studies have attempted to link expression of the immunosuppressive cytokine IL-10 with glioma-associated immune inhibition. Early reports used real-time polymerase chain reaction to demonstrate increased levels of IL-10 mRNA within bulk GBM specimens. However, subsequent studies using in-situ hybridization confirmed that IL-10 was not expressed by putative glioma cells. Rather, IL-10 expression could be specifically localized to circulating monocytes and tumor-associated microglia/infiltrating macrophages within these tumors. The role of immunosuppressive mononuclear cells is discussed in more detail later.

It remains difficult to conceptualize a mechanism by which factors secreted locally by glioma cells could reach serum concentrations sufficient to repress T-cell function globally. In addition, it seems unlikely that such a factor could be subsequently transferred into in vitro, tumor-free studies at a sufficient concentration to exert continued suppressive effects in experimental systems. However, GBM-secreted factors may play an important role in recruiting or driving the development of a secondary immunosuppressive agent.