The intersection of cancer, cancer stem cells, and the immune system: therapeutic opportunities

Abstract

During brain neoplasia, malignant cells subjugate the immune system to provide an environment that favors tumor growth. These mechanisms capitalize on tumor-promoting functions of various immune cell types and typically result in suppression of tumor immune rejection. Immunotherapy efforts are underway to disrupt these mechanisms and turn the immune system against developing tumors. While many of these therapies are already in early-stage clinical trials, understanding how these therapies impact various tumor cell populations, including self-renewing cancer stem cells, may help to predict their efficacy and clarify their mechanisms of action. Moreover, interrogating the biology of glioma cell, cancer stem cell, and immune cell interactions may provide additional therapeutic targets to leverage against disease progression. In this review, we begin by highlighting a series of investigations into immune cell-mediated tumor promotion that do not parse the tumor into stem and non-stem components. We then take a closer look at the immune-suppressive mechanisms derived specifically from cancer stem cell interactions with the immune system and end with an update on immunotherapy and cancer stem cell-directed clinical trials in glioblastoma.

Despite aggressive clinical intervention including surgery, radiation, and chemotherapy, malignant brain tumors such as glioblastoma (GBM) universally progress, leading to patient mortality in an average of 12–15 months.¹ Many factors subvert promising therapies. Cancer stem cells (CSCs) and the immune-suppressed glioma microenvironment represent 2 such factors that have recently emerged at the center of a series of investigations. On the CSC side, efforts are primarily focused on targeting critical factors required for the maintenance and function of these intractable populations. On the immunotherapy side, preclinical and clinical evaluations intended to undermine tumor-immune suppression and augment tumor-immune rejection are ongoing. These are timely efforts that offer novel alternatives to current molecular targeting and antiangiogenic attempts at therapy; however, these strategies remain largely disconnected from one another. As the field continues to reveal facets of a reciprocal relationship between CSCs and the tumor-immune microenvironment, we foresee opportunities to integrate CSC-targeted approaches with immunotherapy in combinatorial treatment strategies. In this perspective, we will begin with a general discussion of tumor cell/immune cell interactions. We will then examine the distinct relationship between CSCs and microenvironmental immune cells in detail. We will present the results of recent CSC and immunotherapy clinical trials and speculate how the combination CSC and immunotherapies might be leveraged to develop more effective treatments for GBM.

Immune Activity and Tumor Progression

The interactions between macrophages/microglia and tumor cells are the most extensively characterized immune cell/tumor cell interactions within the glioma microenvironment. They are typically considered facilitators of tumor cell invasion as well as promoters of tumor cell proliferation. Although there are examples of tumor-associated macrophages/microglia (TAMs) involved with noninvasive tumor growth.² Bettinger et al first described microglial enhancement of glioma cell migration using an in vitro transwell migration assay. They demonstrated that microglia outperformed other neural cell types in their ability to promote glioma cell migration. This effect could be further potentiated by activating the microglia with lipopolysaccharide or granulocyte-macrophage colony-stimulating factor (GM-CSF).³ Several molecules, linked to a number of different pathways, mediate cross talk between TAMs and invading glioma cells. The Kettenmann laboratory demonstrated an interactive cascade beginning with glioma-secreted versican acting as a ligand for TLR2 (Toll-like receptor 2)⁴ expressed exclusively by TAMs.⁵ This TLR2 activation triggered phosphorylation of p38 MAPK and subsequently increased presentation of matrix metalloproteinase 14 (MMP14, MT1-MMP) on the TAM cell surface. MMP14 serves to facilitate extracellular matrix (ECM) degradation and promotes glioma tissue infiltration. Additionally, microglial MMP14 activates MMP2⁶ and MMP9⁷ (secreted by invasive glioma cells), compounding ECM degradation and further easing tumor invasion.

Other groups have identified additional TAM-secreted factors that modulate glioma cell MMP expression, subsequent ECM degradation, and resultant tumor cell infiltration. For instance, Ye et al revealed that TGFβ (transforming growth factor β) signaling triggered proliferation and intensified MMP9 expression when TAM-secreted TGFβ was bound by glioma cell-expressed TGFβR2 (transforming growth factor β receptor II).⁸ Similarly, Fonseca et al demonstrated that exposure to TAM-secreted STI1 (stress inducible protein 1) resulted in increased MMP9 expression in adjacent glioma cells through a currently unknown signaling cascade.⁹ Taken together, these studies suggest that TAMs enable glioma invasion by stimulating protease activation and subsequent ECM degradation proximal to the developing glioma.

Tumors also harness the intrinsic phagocytic activity of TAMs to clear apoptotic cells from the tumor microenvironment. This canonical macrophage behavior is essential for the maintenance and homeostasis of healthy tissues and is preserved during tumorigenesis to the benefit of a developing neoplasia.^10,11 The concept that clearance of apoptotic cells might be required to promote tumorigenesis is somewhat counterintuitive when one considers that tumor cell persistence in spite of apoptotic signals (ie, through the loss of the p53 tumor suppressor) is a primary hallmark of cancer.¹² Convincing evidence presented recently by Ford et al suggests that rampant cell death is highly tumor promoting when partnered with efficient clearing by TAMs.¹³ When Ford et al overexpressed the antiapoptotic factors Bcl-2 (B-cell lymphoma-2) or Bcl-x_L (B-cell lymphoma-extra large) in the overall tumor cell population, they noted striking reductions in tumor proliferation and angiogenesis as well as significantly diminished TAM infiltration. In contrast, tumors that undergo abundant apoptosis were heavily infiltrated by TAMs. These tumors presented brisk mitotic indices, robust angiogenesis paired with decreased levels of hypoxia, and significantly shortened survival. It is worth noting that TAM clearing of apoptotic cells has not yet been demonstrated in GBM. Ford et al demonstrated this effect in models of non-Hodgkin's lymphoma and melanoma. However, like non-Hodgkin's lymphoma and melanoma, GBM cells actively recruit TAMs,^14,15 and targeting TAMs in GBM has recently been identified as an attractive therapeutic strategy. Pyonteck et al demonstrated that targeting TAMs via colony-stimulating factor 1 receptor (CSF-1R) inhibition in a mouse model of proneural GBM increased survival and caused regression of established tumors. Interestingly, TAMs were not depleted upon CSF-1R inhibition, but the tumor-promoting functions associated with activated M2 macrophages were markedly diminished.¹⁶

Regulatory T cells (T_Regs) have provided another example of tumors subjugating the immune system and co-opting immune function in the service of tumorigenesis. Under normal physiological conditions, the peripheral T cell repertoire is biased toward low-affinity self-reactivity but is held in check against autoimmunity by a population of immune-suppressive CD4⁺, CD25⁺, and Foxp3⁺ T_Reg cells.¹⁷ Interestingly, it has become increasingly clear that a variety of tumor types are actually highly immunogenic and can even be heavily infiltrated by effector T cells. These infiltrating T cells may be poised to mount an immune response; however, T_Reg cells present within the tumor microenvironment suppress this potential tumor immune rejection. Evidence suggests that T_Reg infiltration is an early event during tumor formation and that this cell population continually accumulates as the tumor develops. As the T_Reg infiltrate builds, it inhibits the accumulation of CD8⁺ T cells and the concomitant secretion of inflammatory cytokines such as interferon γ (IFNγ), tumor necrosis factor, interleukin 6, and chemokine (C-C motif) ligand 2 (CCL2) required to mount an adaptive immune response. T_Reg cells provide a continuously expanding cloak that shields the developing tumor from immune attack. Almost 10 years ago, Yu et al demonstrated the power of T_Reg-mediated immune-suppression and the potential in targeting this unique vulnerability. By administering an anti-CD25 antibody, these authors effectively cleared T_Reg cells from advanced fibrosarcomatous lesions. This treatment reversed T_Reg-mediated immune suppression and provocatively resulted in complete immune rejection of established tumors.¹⁸ T_Reg cells have been identified within the glioma microenvironment, and their selective depletion confers a similar, albeit less spectacular, survival advantage in mouse models of glioma.¹⁹ These less remarkable results are partially explained by the existence of additional immune-suppressive cell types within the GBM microenvironment. GBM patients have pathologically high levels of immune-suppressive myeloid-derived suppressor cells (MDSCs) in their circulation.²⁰ MDSCs infiltrate the GBM microenvironment, exclude effector lymphocytes,²¹ and are not affected by anti-CD25 administration.

At the very least, these provocative immune/tumor interactions highlight facets of basic tumor biology that have been underappreciated and relatively untested clinically (see below for clinical efforts in GBM immunotherapy). Ideally, these interactions represent new, potentially targetable vulnerabilities of the growing tumor. However, a major caveat to the studies highlighted above is the basic assumption that the immune system acts on all cells of a developing tumor equally. The field understands that heterogeneity within the tumor microenvironment, and within the tumor itself, is the rule and not the exception. Dividing tumors into cancer stem cells (CSCs) and non-stem tumor cells (NSTCs) represents one effort to address tumor heterogeneity. GBM has been a prototypic tumor for CSC studies based on the observation that primary, patient-derived GBM cells can be cultured in stem cell conditions^22–24 and that a subpopulation of tumor cells possesses enhanced self-renewal and tumor-initiating capacities.^24,25 Functional studies have revealed that CSCs are active participants in key tumorigenic processes such as angiogenesis^26,27 and are uniquely refractory to conventional radiation²⁵ and chemotherapy^28,29 compared with NSTCs. In the sections that follow, we will move beyond treating the entire tumor as a single, uniform cell mass and highlight specific interactions between CSCs and the immune system.

Cancer Stem Cells and Immune Modulation

Recently, investigations have begun to reveal that—in addition to their direct contributions to tumor cell genesis, invasion, and therapeutic resistance— CSCs contribute to tumor development indirectly by attenuating immune surveillance within the tumor microenvironment. Furthermore, CSCs themselves are capable of evading the immune system, mainly by altering their immunogenicity to avoid immune-mediated rejection in vivo. A wide variety of preclinical models has linked CSCs with tumor progression and therapy resistance and has demonstrated that selective ablation of this cell population is effective for inhibiting tumorigenic growth.³⁰ CSCs impede tumor immunity through the secretion or expression of immunosuppressive factors (Fig. 1). Studies have shown that CSCs partially mimic antigen-presenting cells in their expression of major histocompatibility complex I (MHC I) and the inhibitory co-stimulating molecule B7 homolog 1 B7-H1 (also known as programmed death-ligand 1, PD-L1) but lack the activating co-stimulatory molecules CD40, CD80, and CD86. Without proper co-stimulating signals, effector T cells are induced into anergy following antigen presentation, rendering them incapable of becoming activated. Further, Parsa et al demonstrated that tumor cell-expressed PD-L1 facilitated cell-cell contact with host T cells, depressing effector T-cell activation and cytokine production.³¹ These observations correlate with similar reports demonstrating that CSCs secrete more TGFβ than their NSTC counterparts.³² In this context, TGFβ is attributed to the attenuation of MHC II expression and subsequent antigen processing³³ as well the gross expansion of the immune-suppressive T_Reg cell population.³⁴ Lastly, Wei et al presented an interesting feed-forward mechanism in which GBM CSC secretion of galectin-3 triggered apoptotic cell death in naïve and activated T cells, which in turn enabled the expansion of the CSC population, further depleting the intratumor effector immune cell population.³⁴

Fig. 1.

Cancer stem cell (CSC) interaction with immune cells. Schematic demonstrating mechanisms of CSC interactions with immune cells. CSCs are capable of driving tumor growth by attenuating immune surveillance through secretion or expression of immune-suppressive factors or by the recruitment of accessory cells that locally suppress the immune response. CSCs tailor their local microenvironment by secreting soluble factors such as arginase and periostin to recruit anti-inflammatory M2 tumor-associated macrophages/microglia (TAMs), which suppress both innate and adaptive immune responses. Likewise, CSCs induce T_Reg expansion and effector function through the production of TGF-β and pSTAT-3 while simultaneously triggering cytotoxic T cell apoptosis through the production of galectin-3. In addition, direct cell-to-cell contact between CSCs and cells of the immune system mediated through cell surface expression of PD-L1 suppresses immune cell function.

In addition to secreting immune-suppressive cytokines, GBM CSCs are capable of recruiting additional cells with tumor-supportive phenotypes. For instance, CSCs mediate recruitment of TAMs to the glioma microenvironment. Wu et al demonstrated significantly increased macrophage migration in vitro upon exposure to colony-stimulating factor-1 (CSF-1), C-C motif ligand 2 (CCL2), and macrophage inhibitory cytokine 1 (MIC-1), factors enriched in CSC-conditioned media. Further, these authors demonstrated that CSF-1- and CCL2-mediated recruitment resulted in the polarization of the monocyte pool toward the immune-suppressive M2 phenotype.³⁵ A similar study revealed that CSCs also mediate TAM recruitment to the tumor microenvironment via periostin secretion. Periostin expression levels were directly proportional to TAM density and inversely correlated with GBM patient survival. Disruption of periostin in CSCs diminished the frequency of M2-stage TAMs in the tumor microenvironment, thus abrogating their tumor supportive capability, inhibiting tumor growth, and increasing survival in mice with GBM CSC-derived xenografts.¹⁵

Clinical Trials and Therapeutic Opportunities

Cancer vaccines are increasingly used in the clinical care of cancer patients. Sipuleucel-T was the first therapeutic cancer vaccine approved for treatment of castrate-resistant prostate cancer in 2010.³⁶ A number of cancer vaccines are currently undergoing evaluation in clinical trials for diverse cancers including GBM. Tumor-associated antigens being targeted in GBM include epidermal growth factor receptor variant III (EGFRvIII), glycoprotein 100 (gp100), survivin, melanoma-associated antigen 1 (MAGE-1), human epidermal growth factor receptor 2 (HER-2), tyrosinase-related protein 2 (TRP-2), ephrin type-A receptor 2 (EphA2), Wilm's tumor protein (WT1), sex determining region Y-box 2 (SOX2), sex determining region Y-box 11 (SOX11),and interleukin 13 receptor alpha 2 (IL13α2) and absent in melanoma 2 (AIM-2). EGFRvIII is a truncated, constitutively active EGFR variant expressed in 30% of GBMs that drive cell proliferation, differentiation, invasion, and tumor cell survival and is linked to poor long-term survival. The EGFRvIII trial represents one attempt to combine immunotherapy with CSC targeting because EGFRvIII is highly co-expressed with the CD133⁺ CSC pool.³⁷ Rindopepimut is a 14-amino acid peptide (PEPvIII) conjugated to the foreign antigen keyhole limpet hemocyanin KLH that targets EGFRvIII. In the phase 2 study ReACT, rindopepimut (anti-EGFRvIII), and bevacizumab (anti-VEGF) were administered to patients with recurrent EGFRvIII⁺ GBM. ReACT treatment resulted in an objective response rate of 23% (6/26) versus 12% (3/25) in those patients treated with bevacizumab and placebo.³⁸ In the phase 2 trial Act III, 65 patients with newly diagnosed EGFRvIII-expressing GBM underwent gross total resection, radiation, combination rindopepimut, and adjuvant temozolomide chemotherapy resulting in a median overall survival (OS) of 21.8 months and a 3-year OS of 26%.³⁹ These encouraging results led to ACT IV, a phase 3 international, multicenter, double-blind trial currently underway of 700 patients with newly diagnosed and resected EGFRvIII⁺ GBM randomized to receive either rindopepimut/GM-CSF or control (KLH) in combination with standard adjuvant temozolomide (Tables 1 and 2).

Table 1.

Select completed vaccine therapy trials in GBM

Trial Name	Phase	N	Patient Population	Experimental Design	PFS (mo)	OS (mo)	Reference
Cancer Stem Cell Vaccine, ICT-1079	2	124	Newly diagnosed glioblastoma	Autologous DCs pulsed with immunogenic peptides from tumor antigens vs placebo	11.2 vs 9.0	18.3 vs 16.7	NCT01280552
ACT III	2	65	Newly diagnosed glioblastoma	Rindopepimut + GM-CSF + TMZ vs TMZ	12.3	24.6	NCT00458601
REACT	2	72	Recurrent glioblastoma	Rindopepimut + bevacizumab vs bevacizumab	–	12.0 vs 8.8	NCT01498328
ACT IV	3	440	Newly diagnosed glioblastoma	Vaccine + GM-CSF + TMZ vs TMZ and placebo	Completed patient recruitment, results pending	Completed patient recruitment, results pending	NCT01480479

Abbreviations: DC, dendritic cell; OS, overall survival; PFS, progression-free survival; TMZ, temozolomide.

Table 2.

Select ongoing vaccine therapy or immunotherapy trials in GBM

Trial Name/Agent	Phase	N	Patient Population	Experimental Design	Reference
Cancer stem cell vaccine (A2B5+)	2	100	Newly diagnosed glioblastoma	Autologous DCs loaded with stem cell-like antigens from irradiated GBM vs placebo	NCT01567202
DCVax-L	3	300	Newly diagnosed glioblastoma	Autologous DCs pulsed with tumor lysate antigen vs placebo	NCT00045968
Heat Shock Protein Peptide Complex-96 (HSPPC-96)	2	222	Recurrent glioblastoma	Protein peptide complex consisting of the 96 kDa heat shock protein, gp96 and an array of gp96-associated cellular peptides + bevacizumab vs bevacizumab alone	NCT01814813
Nivolumab	3	200	Recurrent glioblastoma	Anti-PD-1 (Nivolumab) vs bevacizumab	NCT02017717
Nivolumab or Ipilimumab	1	42	Newly diagnosed glioblastoma	Temozolomide in combination with ipilimumab or nivolumab	NCT02311920

Abbreviation: DC, dendritic cell.

In animal studies, dendritic cell (DC) vaccines using lysate from CSCs rather than whole tumor lysate produced greater T-cell responses. ICT-107 is an autologous peripheral blood mononuclear cell (PBMC)-derived DC vaccine pulsed with 6 synthetic peptide CTL epitopes that target the GBM tumor and tumor stem cell-associated antigens MAGE-1, HER-2, AIM-2, TRP-2, gp100, and IL-13Rα2.⁴⁰ In a randomized phase 2 trial, 124 patients with newly diagnosed GBM were randomized 2:1 to receive ICT-107 or its matching control (unpulsed DC). Median progression-free survival (PFS) was significantly improved in the ICT-107 treatment group, and a future phase 3 trial for this patient population is anticipated because those in the HLA-A2 MGMT subgroup also demonstrated a derived clinical benefit. A recent study in mice and patients showed that a strategy utilizing tetanus booster vaccine could set off an inflammatory response to prepare the immune system for increasing DC migration to lymph nodes when used in combination with a vaccine. This enhanced the effectiveness of a DC vaccine that targeted an antigen from cytomegalovirus and significantly improved survival.⁴¹

There are no FDA-approved monoclonal antibodies that directly target CSCs; however, disruption of the tumor microenvironment may indirectly result in death of CSCs. Bevacizumab, a monoclonal antibody against vascular endothelial growth factor (VEGF) causes disruption of the highly regulated perivascular niche where CSCs reside,²⁷ an indirect effect that may assist in the clinical management of GBM. CD47 is a cell surface protein that is expressed at an increased rate within a CD133, CD15-enriched CSC population.⁴² There is an ongoing clinical trial of a humanized mAb against CD47. There has been considerable interest and excitement about immune checkpoint blockade and its potential as a therapeutic option in cancer. In 2010, the FDA approved ipilimumab, a monoclonal antibody against CTLA-4 for treatment of metastatic melanoma.⁴³ More recently, pembrolizumab and nivolumab (2 monoclonal antibodies against PD1) were FDA approved for metastatic melanoma, ushering in a new era for cancer immunotherapy.^44–46 Since, studies have shown that CSCs express MHC I and PD-L1, it will be interesting to see the outcomes of the clinical trials of anti-PD-1 including a randomized study of nivolumab versus bevacizumab in recurrent GBM and a phase 1 trial of temozolomide with ipilimumab and/or nivolumab in newly diagnosed GBM (Table 2 and Fig. 2).

Fig. 2.

Cell-type specific expression of immune checkpoint receptors and ligands in the glioblastoma (GBM) microenvironment. PD-L1 expression has been described in GBM cells and neurons. PD-L1 binds its receptor, PD1, found on multiple cells of the immune system including cytotoxic T cells, dendritic cells, and NK cells. PD-L1/PD1 signaling is thought to play a large part in inhibiting lymphocyte proliferation, thus enabling tumor cells to evade immune surveillance. In addition, CTLA-4 is expressed on T cells, while its ligands CD80/CD86 are expressed on antigen-presenting cells (APCs). CTLA-4 signaling is thought to play a role in regulating immune responses by transmitting an inhibitory signal to T cells, thus decreasing their effector functions.

Future Directions

It is becoming increasingly apparent that a single therapeutic target is unlikely to lead to effective treatment for GBM. Exemplified by antiangiogenic and antiproliferative treatment failures, despite their initial promise, recurrence is all but inevitable. We speculate that this phenomenon is mediated by a small, treatment-resistant and immune-silent CSC population capable of giving rise to heterogeneous tumor cells with aggressive phenotypes. In the interest of treating both the CSC and NSTC populations, it is critical that we come to understand the similarities and differences in the basic biology of CSCs and overall immune evasion. The immune system has the capacity to be an extraordinarily powerful tool, demonstrating complete immune rejection of established tumors in the most provocative of preclinical cases. As our understanding of tumor biology deepens and our technology advances, we foresee immunotherapy approaches poised at the forefront of modern clinical oncology.

Funding

This work was funded by the Sontag Foundation (J.D.L.), Blast GBM (J.D.L., M.V.), Cleveland Clinic VeloSano Bike Race (J.D.L., M.V.), and Cleveland Clinic (M.A.).