State of the Art ReviewCancer

On being less tolerant: Enhanced cancer immunosurveillance enabled by targeting checkpoints and agonists of T cell activation

See allHide authors and affiliations

Science Translational Medicine  25 Mar 2015:
Vol. 7, Issue 280, pp. 280sr1
DOI: 10.1126/scitranslmed.3010274


The recent approvals of two drugs that block the function of the immune checkpoint programmed cell death 1 (PD-1) have firmly planted tumor immunotherapy in the mainstream of clinical oncology. We provide a historical and immunologic context for these recent advances and discuss translational studies that provide insight into the efficacy of cancer immunotherapy at the individual patient level.


Paul Ehrlich initially postulated in 1909 that cancer would arise with greater frequency in the absence of immune system control, but it was not until experimental systems unequivocally demonstrated tumor-specific immunity nearly 45 years later that the theory of immunosurveillance against cancer was formally proposed by Foley (1) and Burnet (2). More recent and detailed studies using genetic models of immunodeficiency and carcinogen-induced tumors have elegantly demonstrated that the immune system does not prevent but acts to sculpt or edit chemically induced sarcomas that develop in wild-type mice (3, 4). What emerges is a model of Darwinian selection facilitated by oncogenic mutations, genomic instability, and evasion of immune detection as hallmarks of cancer when it presents as clinically apparent disease (5).

Initial successful clinical attempts at cancer immunotherapy occurred in the late 19th century on the basis of empiric observations made by William B. Coley, a surgeon at Memorial Hospital in New York. He injected mixtures of killed Streptococcus pyogenes and Serratia marcescens (termed Coley’s toxins) directly into tumor sites, inducing fevers and, in some cases, therapeutic benefit (6). We now understand that bacterial products (endotoxin) stimulated innate arms of the immune system, leading to direct tumor killing via perforin and secretion of tumor necrosis factor–α (TNF-α), and that in the context of Coley’s toxins, other cytokines such as interleukin-12 (IL-12) promoted subsequent adaptive immune responses to tumor antigens (7). Since Coley’s time, efforts at cancer immunotherapy have been buttressed by epidemiologic associations between the immune infiltrate and cancer outcomes, identification of T cells and antibodies reactive to endogenous tumor antigens, modest yet reproducible responses to cytokines [IL-2 and IFN-α (interferon-α)], and durable graft versus leukemia effects seen in the context of allogeneic stem cell transplantation (813). However, until recently, efforts at using the immune system to treat cancer have been largely discouraging.

One major challenge to successful tumor immunotherapy is immune tolerance. This includes central tolerance via the deletion of highly self-reactive T cells in the thymus during T cell development and peripheral tolerance via ignorance, anergy, exhaustion, and suppression of T cells with low avidity to self-antigens, which escape thymic deletion (14, 15). Ignorance occurs when T cells fail to encounter their cognate peptide and remain in a naïve state because of antigenic sequestration in a privileged site or inadequate presentation in draining lymph nodes (16, 17). Anergy occurs when T cells encounter antigen in the absence of costimulatory molecules necessary for full activation and become refractory to subsequent activation even when the proper signals are present. Such encounters can occur when antigen-presenting cells (APCs) such as dendritic cells (DCs) engage T cells while in an immature state (18, 19). Exhaustion occurs when T cells are chronically exposed to antigen stimulation (as can occur in malignancy) and lose the ability to proliferate and produce cytokines associated with effector function (20). Suppression is an active process in which one population of immune cells inhibits the proliferation and effector function of another. Several suppressor cell populations that can limit tumor immunity have been identified, including myeloid suppressor cells, regulatory T cells, and subpopulations of both natural killer T cells and CD8+ T cells (21).

Therapeutics aimed at overcoming immune tolerance resulted in the successes that now fuel enthusiasm about cancer immunotherapy. Attention has been focused on two major areas: (i) engineered T cell therapies aimed at overcoming both thymic deletion and peripheral tolerance, which were reviewed extensively elsewhere (22, 23), and (ii) T cell costimulation. Herein, we focus on T cell costimulation, which takes advantage of the multiple signals required to fine-tune T cell function after antigen recognition by the T cell receptor (TCR). T cell costimulation aims to overcome the mechanisms of peripheral tolerance such as anergy and exhaustion.


Activation of naïve T cells requires interaction between a TCR and a unique peptide presented in the context of an MHC (major histocompatibility complex) molecule on APCs in addition to a second costimulatory signal (24). These activating signals are balanced and regulated by inhibitory receptors, sometimes referred to as checkpoint molecules. Two major families of co-receptors, the immunoglobulin-like superfamily (IgSF) and the TNF receptor superfamily (TNFRSF), further define the extent of T cell activation, function, and survival (Fig. 1) (25). The IgSF members include the costimulatory receptors CD28 and ICOS (inducible costimulator), and the inhibitory receptors CTLA-4 (cytotoxic T lymphocyte antigen 4), PD-1 (programmed cell death 1), LAG3 (lymphocyte activation gene 3 protein), and BTLA (B and T lymphocyte attenuator), among others. The TNFRSF includes the costimulatory receptors 4-1BB, OX40, CD40, CD27, and GITR (glucocorticoid-induced TNFR-related protein), among others. Additional pathways can be co-inhibitory or costimulatory. For example, herpes virus entry mediator (HVEM) and its ligands CD160, LIGHT, LTα (lymphotoxin-α), and BTLA regulate T cell activation in costimulatory or co-inhibitory fashion, depending on the context and specific ligand interaction (25). Therapeutically manipulating these inhibitory and costimulatory receptors by either blocking the inhibitory receptors or using agonist antibodies directed against the stimulatory receptors has proven to be a promising area of clinical development for immunotherapies against cancer (Table 1).

Fig. 1. Multiple cosignaling molecules regulate all phases of the T cell life cycle.

In addition to TCR interaction with its specific peptide in the context of a major histocompatibility molecule, the outcome of T cell activation is finely regulated by multiple cosignaling molecules. The sum total of inhibitory and stimulatory interactions determines the intensity, duration, and outcome of the immune response. The endogenous immune response is thus determined by duration of receptor expression and available ligand interaction in the microenvironment. CEACAM, carcinoembryonic antigen–related cell adhesion molecules; GAL-9, galectin-9; TIM-3, type I transmembrane immunoglobulin and mucin 3.

Table 1. Costimulatory and co-inhibitory receptors and ligands in clinical development.

N/A, none available.

View this table:


CTLA-4 is expressed on activated T cells and acts as a T cell–intrinsic negative regulator of activation within the immune synapse (Fig. 2) (2630). It was the first immunologic checkpoint molecule to demonstrate clinical potential as an anticancer therapy, and one CTLA-4–blocking antibody, ipilimumab, has been approved by the U.S. Food and Drug Administration (FDA) for the treatment of advanced melanoma. Building upon the success of CTLA-4–blocking agents, agents blocking an expanding number of checkpoint molecules are now being tested in the clinic.

Fig. 2. CTLA-4 and PD-1 pathways function at distinct points in the immune response.

Naïve T cells engage APCs via a specific interaction between the TCR and peptide presented by MHC. A second costimulatory signal is delivered via an interaction between CD80 or CD86 on the APC and CD28 on the T cell. This leads to translocation of intracellular CTLA-4 to the cell surface and/or up-regulation of PD-1 on the T cell. CTLA-4 has a much higher affinity for CD80 and CD86 and displaces CD28 from the immune synapse. PD-1/PD-L1 or PD-L2 interactions further shape initial immune activation. Active T cells see antigen in the tumor microenvironment. Effector cytokine production, such as IFN-γ, leads to up-regulation of PD-L1 on tumor cells and/or PD-L1/PD-L2 on tumor-infiltrating cells in the microenvironment, promoting PD-1/PD-L pathway–mediated adaptive resistance.

The story of CTLA-4 as a target for cancer therapy began in the laboratory of James Allison. Using antibodies that block the engagement of CTLA-4 with its ligands B7-1 and B7-2, Allison and colleagues were able to successfully treat mouse models of sarcoma and colorectal cancer (31). Subsequent studies expanded these findings to diverse mouse tumor models (32). The potential of these preclinical findings spurred the clinical development of CTLA-4–blocking antibodies for use in patients. Two such antibodies have been developed and tested in the clinic: ipilimumab (Bristol-Myers Squibb), a fully human monoclonal IgG1 κ antibody, and tremelimumab (previously Pfizer and now MedImmune Inc.), a fully human monoclonal IgG2 antibody. Ipilimumab was approved by the FDA in 2011 and is currently a first-line treatment option for patients with advanced melanoma. Tremelimumab is still in clinical development.

The largest body of evidence supporting the anticancer activity of CTLA-4–blocking antibodies is reflected in the clinical development of ipilimumab for the treatment of advanced melanoma. Some of the first reports of responses to ipilimumab emerged from phase 1 studies in patients with melanoma (33). Early studies helped establish the remarkable durability of responses among the minority of patients who benefitted from partial or complete responses or, in some cases, durable stable disease. The phase 1 studies did not identify a maximum tolerated dose (MTD), but they characterized some new toxicities associated with this class of agents, as described in more detail below. Subsequent phase 2 studies explored the dose and schedule for ipilimumab. A randomized phase 2 study found an apparent dose-response relationship, with higher response rates observed for patients treated with higher doses of drug (34). As might be anticipated, higher doses were also associated with a higher rate of toxicity. A randomized phase 3 study comparing the doses of 3 and 10 mg/kg has completed accrual, and results are awaited (NCT01515189). The FDA has approved ipilimumab given at a dose of 3 mg/kg once every 3 weeks for four doses on the basis of two randomized, placebo-controlled phase 3 studies (35, 36). One study compared ipilimumab alone to ipilimumab combined with a peptide vaccine or the peptide vaccine alone and demonstrated a survival benefit for patients who received ipilimumab (10.1 months versus 6.4 months). The response rate for patients treated with ipilimumab was relatively modest (~11%). However, a clear and durable survival benefit was identified for patients who received ipilimumab versus the peptide vaccine: 45.6% versus 25.3% at 1 year and 23.8% versus 16.3% at 2 years. A second phase 3 study comparing patients receiving ipilimumab in combination with dacarbazine versus dacarbazine alone confirmed a survival benefit for patients receiving ipilimumab and dacarbazine. Whether dacarbazine added to the efficacy of ipilimumab alone remains unclear, because there was no comparator arm with ipilimumab alone in this phase 3 study. Nevertheless, there were large survival advantages for patients who received ipilimumab at 1 year (47.3% versus 36.3%), 2 years (28.5% versus 17.9%), and 3 years (20.8% versus 12.2%).

Responses to tremelimumab have also been observed in phase 1 and 2 studies in patients with advanced melanoma, suggesting similar clinical activity to ipilimumab. However, an open-label phase 3 study of tremelimumab compared to chemotherapy failed to pass a futility boundary at an interim analysis and was halted (37). Survival analysis of the study demonstrated a nonsignificant trend in favor of an overall survival benefit for patients who received tremelimumab, as well as a longer duration of response. Questions about the dosing, schedule, postprogression therapy with ipilimumab, and study design have been raised in this study, and clinical testing of tremelimumab is ongoing.

Outside of melanoma, the largest clinical experience with CTLA-4 blockade has focused on metastatic prostate cancer. Clinical activity, primarily reflected in biochemical responses, has been observed in phase 1 and 2 studies of ipilimumab in patients with advanced prostate cancer (38, 39). However, a placebo-controlled phase 3 study comparing ipilimumab versus placebo after bone-directed radiotherapy in patients with advanced prostate cancer failed to meet its primary endpoint of overall survival (40). In a post hoc subgroup analysis, the authors found that patients with more favorable clinical features (such as lack of visceral metastases) had a survival benefit, suggesting new hypotheses amenable to clinical exploration in future studies. Clinical data in additional solid tumors, including non–small cell lung cancer, renal cell cancer, and pancreatic cancer, and hematologic malignancies have been generally modest (4145).

Several new observations related to the kinetics and clinical features of response to checkpoint blockade emerged from the clinical development of ipilimumab and tremelimumab. Perhaps the most clinically exciting observation has been the remarkable durability of responses to CTLA-4 blockade. Although responses are restricted to a minority of treated patients, those who respond are likely to enjoy a durable response, often measured in years. Several studies underscore the long duration of response, including a retrospective analysis of patients treated on some of the first phase 1 studies and a pooled analysis of 5-year survival follow-up from patients treated on phase 2 studies of ipilimumab (46, 47). Additionally, in some of the first studies of ipilimumab, it was noted that patients who appeared to benefit from this agent might have apparent disease progression before showing evidence of a response or a relatively delayed response to treatment after a period of disease stability. These kinetics of response are different from those described for typical tumor-directed cytotoxic chemotherapies. These observations prompted investigators to ask whether the radiographic response criteria such as Response Evaluation Criteria in Solid Tumors (RECIST) and modified World Health Organization (mWHO), developed to measure responses to chemotherapy, might be inadequate for capturing the clinical activity of checkpoint-blocking antibodies. In a large retrospective analysis of the experience with ipilimumab, investigators categorized patients according to traditional response criteria and newly defined immune-related response criteria (irRC) (48). In this analysis, four distinct radiographic patterns associated with favorable clinical outcomes were observed and captured: (i) decrease in baseline lesions without new lesions (classical response), (ii) durable stable disease, (iii) initial increase in total tumor burden with later response (flare), and (iv) reduction in the size of baseline lesions concurrent with the growth of new lesions (mixed pattern). As expected, patients who responded according to traditional criteria showed a survival benefit compared to nonresponding patients. Patients who responded according to irRC (including those who did not respond according to traditional criteria) also had a favorable survival compared to nonresponding patients, suggesting that irRC may be able to identify additional patients deriving benefit from immunotherapeutic agents. Prospective evaluation of irRC is under way in numerous ongoing trials using these agents (NCT01295827) (49).


CTLA-4 was the first immunologic checkpoint to be clinically targeted. Experience targeting CTLA-4 increased the interest in pursuing clinical approaches targeting additional immunologic checkpoints, such as the PD-1/PD-L1 (programmed cell death ligand 1) pathway. Potential mechanisms of immunosuppression mediated by this pathway are extensively reviewed elsewhere (50). Briefly, PD-1 is an inhibitory receptor expressed on activated T cells, B cells, and natural killer cells (51). The ligands for PD-1 are PD-L1 (B7-H1), which is expressed on hematopoietic cells and in peripheral tissues including tumor, and PD-L2 (B7-DC), whose expression is more restricted to hematopoietic cells (50, 51). Tumor expression of PD-L1 both in the microenvironment and on tumor cells acts to protect against T cell–mediated killing by promoting T cell exhaustion and through direct antiapoptotic signaling mediated by the cytoplasmic tail of PD-L1 (Fig. 2) (50). There are now hundreds of studies investigating PD-1 targeting and PD-L1 targeting agents registered on the Web site. Agents that target PD-1/PD-L1 have been investigated in a number of tumor types and in a number of disease settings. Here, we focus on the most thoroughly documented experiences of clinical efficacy, side effects, and biomarkers associated with benefit of anti–PD-1/PD-L1 antibodies.

The anti–PD-1 antibody with the longest follow-up data is nivolumab (formerly MDX1106/BMS-956558, Bristol-Myers Squibb). Nivolumab was first evaluated in large phase 1 studies, where responses were seen in a variety of solid tumors, including renal cell carcinoma, melanoma, non–small cell lung cancer, prostate cancer, and colorectal cancer (52, 53). Long-term follow-up data from these phase 1 studies show that responses can be durable, with a median duration of response in patients with melanoma being 2 years (54). More recently, data from randomized trials involving nivolumab in patients with melanoma have begun to emerge. In the front-line setting, nivolumab was shown to improve overall survival compared to dacarbazine chemotherapy (55). For patients with melanoma who have progressed after ipilimumab, nivolumab had a higher response rate and less high-grade toxicity than chemotherapy (56). On the basis of these favorable data, nivolumab is now approved for patients with advanced melanoma in both the United States and Japan. Large phase 3 trials of nivolumab in patients with other solid tumors, such as non–small cell lung cancer and renal cell carcinoma, are ongoing, although one trial of nivolumab versus chemotherapy was recently stopped because of early evidence of survival benefit in the nivolumab arm (NCT01642004, NCT01673867, NCT02066636, NCT02041533, NCT01668784, and NCT02231749).

Pembrolizumab (formerly MK-3475, Merck), another anti–PD-1 antibody, has similarly demonstrated a high response rate in patients with melanoma (57, 58). Pembrolizumab, similar to nivolumab, is FDA-approved for the treatment of patients with melanoma who have received ipilimumab and, if relevant, a BRAF inhibitor. As with nivolumab, randomized data in patients with melanoma are demonstrating the superiority of pembrolizumab to chemotherapy (59). Other tumor types such as gastric cancer and non–small cell lung cancer have also been shown to respond to pembrolizumab (NCT01848834 and NCT01295827). Although no data directly comparing these two agents are available, pembrolizumab and nivolumab are believed to be similarly effective but differ in the dose and frequency of administration. Both are administered continuously, but it remains unknown whether treatment can be safely discontinued in some patients with excellent responses.

Although most data for anti–PD-1/PD-L1 therapies have been generated in patients with solid tumors, anti–PD-1 approaches have also shown promise in patients with hematologic malignancies. A recently published study of a small group of patients with refractory Hodgkin’s lymphoma showed that treatment with nivolumab resulted in a very high response rate (60). A high response rate in Hodgkin’s lymphoma has also been reported with pembrolizumab (61). Nivolumab treatment resulted in encouraging clinical responses in patients with other types of B and T cell non-Hodgkin’s lymphomas as well (62). Another anti–PD-1 antibody, pidilizumab (formerly CT-011, CureTech), has similarly shown efficacy in patients with hematologic malignancies (6365).

Although the data are generally less mature compared to those for anti–PD-1 antibodies, anti–PD-L1 antibodies are similarly showing promising efficacy in a variety of tumor types. BMS-956559 (Bristol-Myers Squibb, human IgG4, anti–PD-L1) first demonstrated the efficacy of targeting PD-L1 (66). Subsequently, many tumor types, such as bladder cancer (67), head and neck cancer, melanoma, non–small cell lung cancer, and renal cell cancer, have been shown to respond to the anti–PD-L1 antibody, MPDL3280A (Genentech, human Fc-engineered IgG1) (68). MEDI4736 (MedImmune, human Fc-engineered IgG1, anti–PD-L1) is similarly showing promise (69), and MSB0010718C (EMD Serono, human IgG1, anti–PD-L1) is also undergoing evaluation (70).


The clinical development of CTLA-4–blocking antibodies also defined a new set of mechanism-based toxicities associated with checkpoint-blocking agents and not typical for traditional chemotherapies. The label immune-related adverse events (irAEs) has been applied to describe these toxicities, which are presumed to reflect the activation of the immune system in response to checkpoint blockade. This immune activation can result in a pattern of tissue-specific inflammation that can target practically any tissue but appears more likely to affect certain tissues such as the gastrointestinal tract (diarrhea and colitis), the liver (hepatitis and elevated liver enzymes), the skin (rash, pruritus, and vitiligo), and endocrine glands (hypophysitis, hypothyroidism, thyroiditis, and adrenal insufficiency) (35). Other rare irAEs include uveitis, conjunctivitis, pancreatitis, cytopenias, neuropathy, myopathy, pneumonitis, and nephritis, among others (7175). Guidelines for the management of irAEs associated with ipilimumab are available (76). Broadly speaking, these toxicities are typically managed and are reversible with interruption in dosing or discontinuation of CTLA-4 blockade in combination with immunosuppressive drugs. Drugs that are most commonly used for the treatment of irAEs are corticosteroids; more rarely, TNF-blocking antibodies (for colitis), mycophenolate mofetil (for hepatitis), or other immunosuppressing approaches have been used (77, 78). Clinical responses can persist despite a period of immunosuppression. For some patients who develop endocrinopathies, such as hypophysitis or hypothyroidism, irreversible damage during the period of tissue inflammation may result in the need for long-term hormone supplementation (7981).

The anti–PD-1/PD-L1 approaches have also been associated with some side effects of an inflammatory nature (irAEs as previously described). Generally, high-grade irAEs are believed to occur much less frequently with anti–PD-1/PD-L1 therapy than with anti–CTLA-4 antibodies. Rashes and arthralgias are the most commonly reported side effects of anti–PD-1/PD-L1 therapy, presumably due to an immunemediated etiology. Although diarrhea/colitis and hepatitis can occur, the incidence of severe cases is believed to be much lower than the experience with anti–CTLA-4 antibody therapy. Inflammation of the lung, pneumonitis, is rare, but it has been a cause of treatment-related deaths in early clinical evaluation (53).


The potential for biomarkers to inform and guide the clinical use of checkpoint-blocking antibodies has been an area of intense interest and research. To date, although many potential immunological markers have been explored in retrospective studies, validation in prospective trials remains to be performed. However, a review of these studies is worthwhile nonetheless, because they have informed our understanding of the biology of checkpoint blockade and may help guide future analyses.

Early studies of ipilimumab identified the absolute lymphocyte count (ALC) as a potential biomarker of interest. Measured by different parameters including an absolute cutoff (ALC >1.0 K/μl) or measuring the rate of rise, a higher ALC after the initiation of treatment with ipilimumab has generally been linked with more favorable clinical outcomes (82, 83). More recently, higher quantity of myeloid-derived suppressor cells, an aberrantly differentiated myeloid population with immunosuppressive features, has been associated with inferior overall survival and less ALC rise after ipilimumab in retrospective series (84).

ICOS is a costimulatory molecule expressed on activated T cells. It is thought to be involved in T cell proliferation, survival, and memory (85). T cell phenotype/activation markers, including ICOS, HLA-DR (human leukocyte antigen–DR), and CD45RO, have also been explored as potential biomarkers (86). For example, in a retrospective analysis of patients with advanced melanoma treated with ipilimumab, a higher frequency of peripheral blood CD4+ICOShigh T cells, sustained over 12 weeks’ time, correlated positively with increased overall survival (87). Presumably, maintenance of activated T cells occurs at the clonal level, as suggested by analysis of T cell repertoire via TCR β chain sequencing studies in peripheral blood, which demonstrated an association between maintenance of high-frequency T cell clones and better outcomes after CTLA-4 blockade (88). Immune responses to specific antigens have also been monitored in patients treated with checkpoint blockade, most commonly focused on so-called shared tumor markers or cancer-testes antigens such as NY-ESO-1. Two published retrospective studies have observed that patients with melanoma who were treated with ipilimumab and who tested positive for serum antibodies to NY-ESO-1 were more likely to have derived benefit (89, 90). At present, it is unclear if these antigen-specific responses are causally related to clinical benefit or serve as a proxy for immune activation. Two studies have explored antigen-specific T cell responses directed against specific mutations unique to the individual tumor in single cases of patients treated with ipilimumab, an area for potential future investigation (91, 92). Last, several studies point toward the tumor microenvironment as a potentially relevant area to assess ongoing antitumor immune responses. For example, in a prospectively designed phase 2 study of melanoma patients treated with ipilimumab, expression of immune-related genes in pretreatment tumor biopsy specimens, including IFN-γ–responsive genes, was correlated positively with clinical activity (93, 94). In a retrospective study of patients with melanoma who were treated with ipilimumab, the number and character of tumor mutations, potentially reflecting tumor-specific antigens recognized by the immune system, correlated with clinical outcomes (91).

Ideally, there would also be a way to determine which patients are likely to benefit from an anti–PD-1/PD-L1 approach using tumoral or peripheral blood characteristics. Of such characteristics, expression of PD-L1 by the tumor cells or tumor-infiltrating immune cells has been the most extensively studied. Tumor cell expression of PD-L1 can be genetically driven, as occurs in Hodgkin’s lymphoma, or may occur in response to IFN-γ or other cytokines derived from infiltrating T cells (25, 60). In other instances, macrophages in the tumor stroma may express PD-L1 (95). In early studies of nivolumab, response rates were positively associated with the presence of PD-L1 on the tumor cells (53). The recently published randomized phase 3 study of nivolumab compared to dacarbazine in melanoma patients, however, showed that PD-L1–negative patients had an overall survival benefit with nivolumab compared to dacarbazine, indicating that PD-L1 expression on tumor cells does not predict improvement in overall survival with nivolumab treatment (55). The currently available data do not support using tumor expression of PD-L1 as a single biomarker to select patients to receive nivolumab or other anti–PD-1/PD-L1 antibodies. It is possible that models incorporating many immunologic factors, including the presence of infiltrating CD8+ T cells in a tumor (95), may ultimately lead to a more predictive biomarker. PD-L1 expression can also be heterogeneous and dynamic within an individual patient, stressing the complexity of this possible biomarker (96).


T cell regulation depends on a balance of positive and negative regulators of T cell activation. Clinical development of immunostimulatory antibodies thus far has focused predominantly on the negative regulators, or checkpoints, of T cell activation. However, the positive regulators of T cell activation, a diverse collection of costimulatory molecules, are also rational targets for immunotherapeutic agents. Antibodies targeting some of these molecules have shown promise and also toxicities in the early stages of clinical development. The first costimulatory molecule targeted clinically was CD28, which was evaluated in a clinical trial of the humanized monoclonal antibody TGN1412 in 2006 (TeGenero Immuno Therapeutics, humanized IgG4, CD28 superMAB). This study highlighted the potential dangers of eliciting an overly exuberant immune response and also demonstrated flaws in clinical trial design that unnecessarily exposed healthy volunteers to excessive risk. The antibody was administered in a single dose to six healthy volunteers, and the resulting immune stimulation caused a cytokine release syndrome that led to multiorgan dysfunction in at least four patients (97). Some aspects of the TGN1412 antibody have since been highlighted, including its “superagonist” function, to explain this very strong, immediate immune activation (98). Nevertheless, lessons learned from this experience have informed the subsequent development of agonist antibodies targeting T cell costimulatory molecules, and this area has proceeded with appropriate caution.

A variety of antibodies targeting costimulatory molecules are in early clinical development, including drugs targeting CD27, CD40, OX40, GITR, and CD137 (99). These targets belong to the TNFRSF and are largely involved in regulating T cell survival, proliferation, and cytokine production. In addition, a CD40 agonist may promote APC function. The expression and function of this family of molecules are reviewed extensively elsewhere (100, 101). Here, we focus on four agonist antibodies for which clinical data are now available: CP-870,893 (Pfizer, human IgG2, anti-CD40), 9B12 (AgonOx, murine IgG1, anti-OX40), PF-05082566 (Pfizer, human IgG2, anti-CD137), and BMS-663513 (urelumab, Bristol-Myers Squibb, human IgG4, anti-CD137). Two phase 1 studies of CP-870,893 have been published: one combining the antibody with carboplatin and paclitaxel in patients with advanced solid tumors and a second combining the antibody with gemcitabine for patients with pancreatic cancer (102, 103). The primary objective for these studies was to assess the safety and tolerability of this antibody. In the first phase 1 study, an MTD of CP-870,893 was established on the basis of the observation of two dose-limiting toxicities (one grade 3 cytokine release and one transient ischemic attack). The second study confirmed the MTD for this agent and also demonstrated some promising preliminary clinical activity: of 22 patients with advanced, chemotherapy-naïve pancreatic cancer treated with the combination of CP-870,893 and gemcitabine, 4 (18%) achieved a partial response. In a phase 1 study of the murine OX40 agonist antibody 9B12, 30 patients with advanced, treatment-refractory solid tumors were treated with escalating doses of antibodies. The study established the safety and tolerability of 9B12 at the tested doses, with grade 3/4 drug-related toxicities restricted to seven patients with transient lymphopenia. Although none of the treated patients met RECIST criteria for response, cases of tumor regression/disease stability were observed in patients with melanoma, renal cell cancer, squamous cell cancer, prostate cancer, and cholangiocarcinoma (104). Last, two agonist CD137 antibodies have been evaluated in clinical trials. A phase 1 study evaluating the agonist CD137 antibody BMS-663513 was presented at the American Society of Clinical Oncology (ASCO) 2008 (105). An MTD was not defined despite treatment at the highest dose level (15 mg/kg). An 11% incidence of reversible transaminitis was observed, but unfortunately after this report, two cases of fatal hepatitis occurred at the 10 mg/kg dose, temporarily halting the development of this compound. Development has resumed at lower doses. Another phase 1 study evaluating the CD137 agonist antibody PF-05082566 was presented at ASCO 2014 (106). Of the 24 patients evaluable for response, 1 patient had a partial response, 1 had a mixed response, and 7 patients had stable disease. Dose escalation continues in this study.


Despite the promise of immune checkpoint-blocking antibodies, not all patients benefit from treatment with a single agent, and combinations are being actively explored. Various chemotherapeutics can potentially synergize with immunotherapy by inducing homeostatic reconstitution, antigen release, and transient depletion of regulatory T cells or myeloid-derived suppressor cells (107109). Preclinical models have demonstrated that focal radiotherapy can activate tumor stroma to cross-present antigens, and anecdotal reports of abscopal effects (killing of tumors remote from the site of treatment) in the context of ipilimumab therapy have been reported in patients (110, 111). Finally, nonoverlapping mechanisms support a rationale for combining checkpoint-blocking agents and/or agonist antibodies (112, 113). To date, many studies have tried combining immune checkpoint-blocking antibodies with conventional cancer treatments such as chemotherapy, targeted therapy, radiotherapy, as well as other immunotherapies. However, at this time, combinatorial approaches generally remain early in clinical investigation.

Chemotherapy has been safely combined with anti–CTLA-4 therapy. In a trial of patients with non–small cell lung cancer, patients receiving ipilimumab in a phased schedule combined with chemotherapy were shown to have improved progression-free survival by irRC criteria compared to patients receiving chemotherapy alone (114). For patients with melanoma, ipilimumab has been evaluated with dacarbazine, carboplatin, and paclitaxel, with generally acceptable tolerability (115).

Extensive preclinical data suggest that combining targeted therapies against oncogenic drivers, such as the BRAF mutation, with immunotherapy may be synergistic [reviewed by Hu-Lieskovan (58)]. Nevertheless, early clinical testing of ipilimumab with the BRAF inhibitor vemurafenib (Genentech) was frequently associated with transaminitis and rash (116). It may be possible to safely combine ipilimumab with dabrafenib (GlaxoSmithKline LLC, BRAF inhibitor), but the combination of ipilimumab, dabrafenib, and trametinib [GlaxoSmithKline LLC, MEK (mitogen-activated protein kinase kinase) inhibitor] was associated with cases of severe colitis and is not recommended (117). Experience combining antiangiogenic agents, such as sunitinib (Pfizer) and pazopanib (GlaxoSmithKline LLC), with anti–PD-1/PD-L1 is also being investigated (118).

Because radiotherapy remains an important component of anticancer treatment and has been shown to result in immunologic effects in preclinical models [reviewed by Formenti and Demaria (119)], radiotherapy is an attractive combinatorial partner with anti–CTLA-4 and anti–PD-1/PD-L1 therapy. Although anecdotal evidence suggests that this approach may have provided benefit in some patients and is generally safe (111, 120123), we await prospective data to truly determine whether radiotherapy adds to the efficacy of immunotherapy.

The most clinically advanced immunotherapy combinations involve studies of multiple immunotherapy agents. One randomized phase 2 study showed that adding granulocyte-macrophage colony-stimulating factor (GM-CSF) to ipilimumab (10 mg/kg) improved overall survival and decreased toxicity compared to ipilimumab alone (124). Although promising, longer follow-up of this study is necessary to confirm the overall survival benefit. Further, it remains unknown whether GM-CSF has similar favorable effects when combined with the FDA-approved dose of ipilimumab (3 mg/kg).

Ipilimumab and nivolumab have also been directly combined in a phase 1 dose-finding study in patients with melanoma (125). A higher response rate was seen with the combination than with either ipilimumab or nivolumab alone, but whether this response rate will translate into improved overall survival compared to ipilimumab or nivolumab alone remains the subject of an ongoing phase 3 study (NCT01844505). The rate of toxicity when ipilimumab is combined with nivolumab is believed to be higher than with either antibody approach alone. Many of the reported side effects, however, have been asymptomatic changes in laboratory values of unclear significance, such as elevations in amylase and lipase. Investigations of ipilimumab and nivolumab (and tremelimumab + MEDI4736) are ongoing in many different malignancies.


Recent treatment successes with antibodies that regulate immune activation have essentially ended the debate about whether the immune system sees and regulates cancer growth. Tumor immunotherapy has entered the mainstream and is a strategy to be considered within the clinician’s toolbox of standard therapies for cancer. Although antibodies blocking PD-1 are only FDA-approved in melanoma and non–small cell squamous lung cancer, additional PD-1 and PD-L1 blocking antibodies are expected to be approved in multiple malignancies.

A number of challenges remain in understanding which patients are susceptible to immune modulating therapies. Biomarkers such as ALC, markers of T cell activation, myeloid-derived suppressor cells, features of the tumor microenvironment, and others discussed above require prospective validation both individually and in composite models to enable better selection of patients for currently available therapies. In addition, these biomarkers need to be tested in the context of new costimulatory antibody targets and in combination studies. A practical challenge in understanding the biology of response is the need for serial sampling of the tumor microenvironment in patients responding, progressing, or achieving an initial response and subsequently progressing after therapy. These types of studies are critical to both investigating the mechanisms of adaptive resistance and developing strategies to optimally combine or sequence therapies.

Whole-genome and whole-exome sequencing efforts made possible by advances in next-generation sequencing have identified somatic mutations that lead to new epitopes (neoantigens) in human tumors that are distinct from self proteins and can be presented by common HLA alleles (126). Recently, a sequencing study of tumors from melanoma patients treated with ipilimumab identified a neoantigen signature with homology to shared epitopes from infectious agents that strongly correlated with durable clinical benefit (91). Further study is needed to evaluate if patients responding to other T cell costimulatory molecules or alternative checkpoint blockade share similar neoantigen signatures. In addition, it remains unknown to what extent neoantigens play a role in cancer immunosurveillance among tumors with a lower frequency of mutations or those not associated with known carcinogen exposure. Perhaps these tumors will not be as sensitive to T cell modulation with checkpoint blockade and will require alternate strategies to facilitate immune responses. In addition, immune editing and neoantigen loss could be a potential mechanism of adaptive resistance identified through sequencing of serial tumor biopsies.

The area of neoantigen discovery and epitope prediction also opens new possibilities for tumor immunotherapy and combination strategies with T cell costimulation. Preclinical models have defined properties of neoantigens that can lead to tumor rejection, including higher-order differences from the wild-type protein that lead to conformational differences in the peptide-binding groove and stability of the MHC/peptide interaction (127). In addition, CTLA-4 blockade has been shown to unmask immunity to neoantigens in methylcholanthrene-induced tumors in mice, and long-peptide vaccines to these neoantigens demonstrated similar clinical activity as checkpoint blockade (128). Similar high-throughput epitope discovery approaches have successfully identified neoantigen-reactive CD8+ and CD4+ T cells in patients with melanoma (92, 129). Together, these studies suggest a potential for massive parallel sequencing in combination with epitope prediction and validation strategies to identify somatic mutations that can serve as vaccination targets at the individual patient level. Studies have recently been initiated in melanoma and glioblastoma to evaluate these types of personalized vaccine approaches (NCT01970358 and NCT02287428).

The multiple molecules involved in T cell costimulation raise hopes that more patients will ultimately derive benefit from continued basic research and clinical study of this family of therapeutic targets. Because additional targets are clinically tested for safety and efficacy, one could envision an era where high-resolution immunophenotyping made possible by mass cytometry (130) and massive parallel sequencing lead to highly personalized immune therapy strategies combining neoantigen vaccination, T cell costimulation, and conventional therapeutic treatment strategies in a patient-specific manner. All of this is enough to make one wonder if perhaps there is a limit to the reign of the “Emperor of All Maladies.”


Acknowledgments: Funding was provided by the Ludwig Collaborative Laboratory and Swim Across America. A.M.L. also wishes to acknowledge funding from the American Cancer Society (MRSG-11-054-01-LIB) and Cycle for Survival. Competing interests: A.M.L. reports consulting for Bristol-Myers Squibb, Janssen, and Efranat and receiving research support from Bristol-Myers Squibb. M.A.P. reports consulting for Bristol-Myers Squibb and Amgen and receiving research support from Bristol-Myers Squibb. M.C. reports consulting for Bristol-Myers Squibb, GlaxoSmithKline, and Kyowa and receiving research support from Bristol-Myers Squibb. M.C. has a family member employed by Bristol-Myers Squibb. J.D.W. reports consulting for Bristol-Myers Squibb, Genentech, MedImmune, and Merck and receiving research support from Bristol-Myers Squibb.

Stay Connected to Science Translational Medicine

Navigate This Article