Antagonist Antibodies to PD-1 and B7-H1 (PD-L1) in the Treatment of Advanced Human Cancer

Antigen-specific T-cell responses are controlled positively and negatively by costimulatory and coinhibitory molecules, respectively. Coinhibitory molecule signaling prevents inappropriately directed immunity and limits the size and duration of immune responses. Among the key coinhibitory molecules, broadly categorized as “checkpoint molecules,” are CTL antigen-4 (CTLA-4), which controls early stages of T-cell activation, and programmed death-1 (PD-1; ref. 1). PD-1 (CD279) is a member of the B7-CD28 family that regulates T-cell activation, peripheral tolerance, and the prevention of bystander tissue damage during immune responses (2–4).

PD-1, so named for its involvement in classical programmed cell death (1), is expressed on activated CD4⁺ and CD8⁺ T cells, natural killer (NK) T cells, B cells, and activated monocytes and dendritic cells (DC; ref. 4). PD-1 protein is not detectable on resting T cells but is found on the cell surface within 24 hours of T-cell activation (4). The known counter-receptors of PD-1, B7-H1 (also called PD-L1; ref. 5) and B7-DC (also called PD-L2; ref. 6)—both of which have been observed on cancer cells (7, 8)—have distinct expression profiles. Low levels of B7-H1 mRNA are found in virtually all normal tissues and cell types examined thus far (7). However, constitutive expression of B7-H1 cell-surface protein in normal tissues is rare and has been found [via immunohistochemistry (IHC)-based analysis] only in a fraction of tissue macrophages within lung, liver, tonsil, and placenta (9). These findings indicate the existence of 1 or more posttranscriptional mechanisms controlling B7-H1 cell-surface protein expression.

The biologic consequences of B7-H1 expression depend on cell membrane localization because it is presumed that B7-H1 is functional only when it ligates a counter-receptor. B7-H1 cell-surface protein can be induced by various inflammatory mediators, including IFN-α, -β, and -γ, bacterial lipopolysaccharide, granulocyte-macrophage colony-stimulating factor, VEGF, and the cytokines interleukin-4 (IL-4) and IL-10 (9–12). In particular, the IFN family of cytokines are potent inducers of B7-H1 mRNA and protein on cultured B7-H1⁻ cells. In addition to binding PD-1, B7-H1 can also bind CD80 on activated T cells, thus inhibiting T-cell activation and production of cytokines (4).

B7-DC is expressed on myeloid dendritic cells, activated T cells, and some nonhematopoietic tissues (including lung; ref. 6), although only on a minority of patient tumors (6, 8, 13–15). Further studies are required to define the role of B7-DC expression, induction, and signaling on T-cell activation and function. Results from studies of B7-DC-knockout mice and in vitro studies have been inconsistent and show either increased or decreased response to antigens (14–16). These results are consistent with an as-yet unrecognized second receptor for B7-DC. Several studies in the literature have provided evidence for a preferential inhibitory role of B7-DC on Th2 responses (17), which in addition to the known binding between B7-H1 and CD80, could explain potential differences in clinical activity and toxicity of antibodies targeted against B7-H1 versus those directed against PD-1.

In a healthy host, PD-1 signaling in T cells regulates immune responses to minimize damage to bystander tissue and prevents the development of autoimmunity by promoting tolerance to self-antigens. Ligation of PD-1 results in the formation of PD-1/T-cell receptor (TCR) inhibitory microclusters that recruit SHP2 molecules, which dephosphorylate multiple members of the TCR signaling pathway, effectively turning off T-cell activation (18). Inhibition of RAS and PI3K/AKT pathways was also shown, resulting in downstream suppression of cell-cycle progression and T-cell activation (ref. 19; Fig. 1). PD-1 ligation by B7-H1 on macrophages, other antigen-presenting cells (APC), or endothelium inhibits production of several cytokines, including IFN-γ, IL-2 (which protects against T-cell apoptosis), and TNF-α (4, 13), and promotes T-cell apoptosis via inhibition of survival factor Bcl-xL (4).

B7-H1 can also serve as a receptor. It has been shown that reverse signaling through B7-H1 in T cells and dendritic cells regulates cytokine production and inhibits dendritic cell maturation and survival of activated T cells (20, 21). B7-H1 signaling also inhibits tumor cell apoptosis induced by antigen-specific CD8⁺ T cells and in response to various stimuli (22). During chronic viral infection, B7-H1 expression in hematopoietic cells reduces the induction and functionality of virus-specific T-cell responses, whereas B7-H1 expressed in the nonhematopoietic compartment regulates tissue immunopathology and viral clearance (23). Thus, the functional consequences of B7-H1 interaction with its receptors are determined by the location of its expression.

Although PD-1 and its counter-receptors primarily control the function of effector cells, this pathway may also have a role in modulating T-cell priming (24). The varied regulatory roles of PD-1 signaling are supported by preclinical studies showing spontaneous, late-onset development of autoimmune disease in PD-1–deficient mice (25, 26). B7-H1–deficient mice do not display spontaneous autoimmunity (3, 27) but have increased T-cell accumulation in select peripheral organs and impaired apoptosis of CD8⁺ T cells (27). Several reports also suggest that PD-1/B7-H1 signaling contributes to regulatory T cell (Treg) induction, maintenance, and suppressive function (28) via the downregulation of mTOR and AKT phosphorylation (29).

Multiple in vivo mouse experiments showed tumor regression or prolonged host survival after abrogation of PD-1 pathway signaling alone (9, 30, 31) or in combination with other agents, including cancer vaccines (32–37). Tumor regression was accompanied by increased effector T-cell (Teff) function and cytokine production. However, the exact sequence of events within the tumor microenvironment that culminates in tumor regression after PD-1 pathway blockade is unknown. Although it is reasonable to assume that PD-1 pathway blockade reactivates PD-1–expressing tumor-infiltrating lymphocytes (TIL; Fig. 2), studies of T-cell function in chronic viral infection suggest that the population of T cells expressing high levels of PD-1 are “exhausted” and may not be reactivated with PD-1 blockade alone (38). Exhausted T cells result from chronic antigen exposure (caused by persistent viral infections or progressive tumor), express high levels of inhibitory molecules, have poor effector functionality and a distinct transcriptional profile, and may, in fact, constitute a discrete stage of T-cell differentiation. In patients with metastatic melanoma (mMEL), CD8⁺ TIL can express several cell-surface molecules associated with exhaustion including TIM-3, LAG-3, and PD-1 (with some cells expressing all markers), which may be the cause of resistance to PD-1 pathway blockade in some patients and may provide the rationale for future combinations of checkpoint/coinhibitory molecule antagonists (39, 40). Recently, Mkrtichyan and colleagues reported that treating mice with a B7-DC-Ig fusion protein in combination with cyclophosphamide and a vaccine resulted in a significant decrease (via effects on B7-DC/PD-1 signaling and cell proliferation) in the number of PD-1^hi CD4⁺ and CD8⁺ T cells within the tumor, allowing infiltration by and/or expansion of nonexhausted PD-1^lo T cells into the tumor microenvironment (41). Whether this phenomenon occurs in human patients as a result of administration of the other PD-1 pathway-targeted immunotherapies presented later also remains to be determined.

B7-H1 is expressed by various tumor tissues and high PD-1 expression is often present on TILs. Peripheral blood CD4⁺ and CD8⁺ T cells in patients with cancer can also express PD-1, in some cases in a large percentage of cells, which makes it unlikely that PD-1⁺ T cells exclusively represent tumor-specific T cells (42). These findings are consistent with recent data showing that PD-1 can be upregulated by common γ-chain cytokines including IL-2, IL-7, IL-15, and IL-21 (43).

Table 1 summarizes PD-1, B7-H1, and B7-DC expression in several different malignancies and correlates expression with patient prognosis. Most large retrospective studies show that B7-H1 expression correlates with poor prognosis and/or more aggressive disease; however, several reports indicate a lack of association (44–46) or even that B7-H1 expression is associated with improved survival and influx of lymphocytes into the tumor microenvironment (47). Although there may be biologic reasons to explain the discrepant outcomes, reviews of the literature correlating B7-H1 expression with prognosis should nevertheless be interpreted with caution, because of the heterogeneity in expression within tumor tissue, the requirement to assess membrane B7-H1 protein rather than intracellular protein or mRNA, the lack of specificity of several commercially available antibodies, and the substantial difficulty in developing reagents and methods for detection of B7-H1 expression in formalin-fixed, paraffin-embedded tissue (FFPE). B7-H1 protein contains only 2 small linear hydrophilic regions, and therefore, a limited number of antibody-binding sites (5). Because FFPE tumor tissue is generally available in patients with metastatic cancer but frozen tissue is not, a customized protocol was subsequently developed to renature B7-H1 in FFPE tissues (47, 48) permitting reproducible and specific immunohistochemical staining of B7-H1 with the 5H1 monoclonal antibody (mAb; ref. 9).

Thus far, only a small number of studies have examined the prognostic significance of B7-DC expression on tumor tissue and other cells within the tumor microenvironment. Tumor-cell B7-DC seems to be a negative predictor of survival for patients with esophageal carcinoma (49); other studies have identified either no relationship (50) or a correlative trend (although not significant) between B7-DC expression and reduced survival (51, 52). It should be noted that in many of these studies, B7-DC was expressed by only a minority of patient tumors (8).

In human ex vivo studies of CD4⁺ or CD8⁺ T-cell activation, addition of an anti-B7-H1 or anti-PD-1 antibody augmented T-cell expansion and proliferation, increased cytokine production, and enhanced cytolytic activity (9, 53–55; Fig. 2). These data supported the clinical development of human antibodies blocking either PD-1 or B7-H1, several of which are currently being evaluated in clinical trials.

Nivolumab (MDX-1106, BMS-936558), a fully human monoclonal immunoglobulin (Ig)G4 antibody that binds PD-1 with high affinity and blocks its interaction with both B7-H1 and B7-DC, was initially evaluated in a standard phase I dose-escalation trial in which patients received intravenous doses from 0.3 to 10 mg/kg (56). Patients could be retreated with up to 2 courses, each course consisting of 2 doses spaced 4 weeks apart, at 12-week intervals. Among 39 patients, including 21 treated at 10 mg/kg, treatment produced a complete response (CR) lasting more than 21 months in a patient with colorectal cancer, partial responses in a patient with melanoma and a patient with renal cancer, and mixed responses in a patient with lung cancer and a patient with melanoma. PD-1 receptor occupancy was maintained beyond 60 days at all dose levels and beyond the time when drug levels were no longer detectable in the blood. In a subsequent, much larger phase I trial accruing 304 patients (as of July 2012), nivolumab was administered in doses ranging from 0.1 to 10 mg/kg i.v. every 2 weeks (57, 58). Small expansion cohorts were accrued in metastatic renal cell carcinoma (mRCC), melanoma, non–small cell lung carcinoma (NSCLC), prostate, and colorectal cancers. Including all dose levels, at the time of data analysis, objective responses were observed in 33 of 106 (31%), 20 of 122 (16%), and 10 of 34 (29%) of previously treated patients with mMEL, NSCLC, and mRCC, respectively. In NSCLC, objective response rate (ORR) and progression-free survival (PFS) rate at 24 weeks appeared higher at the 3- and 10-mg/kg dose levels compared with 1 mg/kg, but there was no clear dose–response relationship for activity in other diseases or for toxicity across the trial. No objective responses were observed in patients with prostate or colorectal cancer. Objective responses were durable (>6 months) in the majority of patients and several with sufficient follow-up remain progression-free even beyond 2 years. Thirteen patients showed nonconventional patterns of tumor response (such as prolonged reduction of tumor burden in the presence of new lesions) that were suggestive of clinical benefit but were not included in the calculation of ORR.

Nivolumab was well tolerated at all dose levels. Common drug-related adverse events included pruritus, rash, diarrhea, fatigue, nausea, and decreased appetite. Only 14% of the patients developed grade 3 or 4 drug-related adverse events. Pneumonitis was observed in 3% of all patients and 3 (2 NSCLC and 1 RCC; 1%) died from grade 3 or 4 pneumonitis-related events, leading to more proactive implementation of an early identification and intervention protocol including monitoring and administration of glucocorticoids. Gastrointestinal and hepatic adverse events were managed with treatment interruption or corticosteroids as necessary and endocrine adverse events were managed with replacement therapy, similar to the algorithms developed for management of ipilimumab immune-related adverse events. In a subset of 42 patients in which pretreatment paraffin-embedded tumor tissue was assessed for B7-H1 expression using the 5H1 antibody, no patients with B7-H1⁻ tumors experienced an objective response, whereas 9 of 25 (36%) patients with B7-H1⁺ tumors (defined as ≥5% membrane expression) had an objective response (P = 0.006). Although these data support a role for tumor cell B7-H1 expression as a potential predictive biomarker of response to PD-1 pathway blockade, further confirmation will be required in larger patient populations.

Two other PD-1 antagonist antibodies are currently being evaluated in clinical trials. MK-3475, a humanized IgG4 with high affinity for PD-1, was well tolerated in doses ranging from 1 to 10 mg/kg administered i.v. 4 weeks apart and subsequently every 2 weeks (59). IL-2 production from ex vivo stimulated lymphocytes was increased at all 3 dose levels. No grade ≥3 drug-related adverse events were observed, although 1 patient developed pneumonitis-requiring corticosteroids. For the 13 patients with melanoma treated in the dose-escalation phase and 10-mg/kg expansion cohort, 7 confirmed or unconfirmed objective responses were observed. One of 4 patients with NSCLC also achieved a partial response (unconfirmed at the time of the presentation). Cohorts of patients with melanoma were subsequently treated at 10 mg/kg i.v. every 2 to 3 weeks or 2 mg/kg every 3 weeks (60). Among 83 evaluable patients at the time of presentation, including 25 previously treated with ipilimumab, the ORR was 47% and 40% in patients with prior ipilimumab therapy. Although follow-up was relatively short, responses were ongoing in most patients. The latter data confirmed the high rate of activity in mMEL observed with nivolumab and also showed substantial activity of anti-PD-1 even in patients with prior exposure to another checkpoint inhibitor (anti-CTLA-4). It remains unclear if the higher reported rates of response for MK-3475 are due to differences in patient selection or the antibody characteristics.

Another anti-PD-1 mAb, CT-011, is a humanized IgG1 antibody shown to enhance human NK and T-cell function in vitro. In a phase I trial in advanced hematologic malignancies examining doses from 0.2 to 6 mg/kg, only 1 patient experienced drug-related adverse events, and antitumor activity was observed in 1 patient with follicular lymphoma and 1 patient with acute myelogenous leukemia (AML; ref. 61). Development of CT-011 has been focused primarily in hematopoietic malignancies and in combination with select chemotherapies, although a large, single-agent phase II trial is being conducted in patients with mMEL. In addition, a phase I trial of a third PD-1 antagonist, the B7-DC-Fc fusion protein AMP-224, is ongoing.

Other antibodies blocking B7-H1 are also in clinical development. BMS-936559, a high-affinity human IgG4 that blocks B7-H1 binding to PD-1 and CD80, was evaluated at doses ranging from 0.3 to 10 mg/kg administered i.v. every 2 weeks (62). Among the 207 enrolled patients with advanced cancer, only 9% developed grade 3 or 4 treatment-related adverse events, similar to those observed with nivolumab. Across all dose levels and expanded cohorts, objective responses were observed in 9 of 52 (17%) patients with mMEL, 2 of 17 (12%) with mRCC, 5 of 49 (10%) with NSCLC, and 1 of 17 (6%) with ovarian cancer. A subset of responses was durable. Although the activity of this drug seems to be less than that observed with nivolumab, sample sizes were smaller and the patient populations were heterogeneous. It is also possible that lower response rates were caused by differences in the antibodies administered or in the functions of their molecular targets. Data from the phase I trial of anti-B7-H1 antibody MPDL3280A/RG7446 are not yet publicly available. Recently, Medimmune, the Ludwig Institute for Cancer Research (New York, NY), and the Cancer Research Institute (New York, NY) announced a partnership to pursue development of multiple anticancer immunotherapies, including the anti-B7-H1 mAb MEDI4736 (63). Table 2 presents ongoing trials with PD-1 pathway-targeted agents.

The promising clinical data available for agents blocking PD-1 and B7-H1 validate the PD-1/B7-H1 pathway as a critical anticancer and immunotherapy target and extend the potential for immunotherapy activity beyond melanoma and RCC to other solid tumors. ORRs in melanoma, RCC, and NSCLC for the anti-PD-1 antibody nivolumab and in melanoma for MK-3475 were sufficiently high to pursue registration trials as a single agent and phase I to III trials of more traditional empiric combinations with other approved chemotherapy or targeted agents. In the absence of definitive predictive biomarkers or clear understanding of tumor–host immune relationships in most malignancies (including the role of B7-DC), the activity observed to date may merit empiric phase II exploration of PD-1 pathway blockade in multiple types of solid tumors. Preclinical data showing the expression of PD-1 on B- and T-cell lymphomas, the role of B7-H1/B7-DC in blunting graft-versus-leukemia effect in a murine model of chronic phase chronic myeloid leukemia (64), and the role of the pathway in suppressing CTL activity in a murine leukemia model (65) also support exploration of PD-1 pathway-targeted agents in hematologic malignancies, and indeed, preliminary evidence of activity was observed with the IgG1 CT-011.

There are valid reasons to expect that differences will emerge in activity or toxicity of the different PD-1/B7-H1 antagonists, because of differences in binding affinity, target, or antibody isotype. Antibody isotype can influence the ability to mediate antibody- and/or complement-dependent cell-mediated cytotoxicity [antibody-dependent cell-mediated cytotoxicity (ADCC), complement-dependent cellular cytotoxicity (CDCC)]. An IgG1 anti-PD-1 antibody that mediates strong ADCC or CDCC could result in the death of PD-1–expressing T cells, thus eliminating the cell population responsible for mediating antitumor effects. Conversely, ADCC and CDCC by an anti-PD-1 might be advantageous in eliminating intratumoral Treg or malignant hematopoietic cells expressing PD-1, and could also be desirable functions in anti-B7-H1 antibodies, by predominantly targeting cancer cells (although T cells can also express B7-H1). Human IgG4 isotype antibodies have a reduced capacity to mediate ADCC and CDCC in comparison with IgG1 isotype antibodies. Agents with lower binding affinity to the target may affect the adequacy or duration of pathway blockade and blocking PD-1 versus B7-H1 may lead to differing outcomes because of other unblocked coreceptor interactions in this pathway as discussed previously. Predicted differences in potential toxicity profiles (i.e., lower pulmonary toxicity) by blocking B7-H1 versus PD-1 because of unblocked expression of B7-DC in lung have not been verified yet in clinical trials. Additional data, including possibly head-to-head trials, will be required.

The key challenge for further clinical development of PD-1 pathway blockade is to understand the tumor–host immune relationship(s) which is/are permissive for clinical activity without other interventions, and the tumor–host immune relationships which require other therapeutic interventions in addition to PD-1 pathway blockade to produce optimal antitumor effects. A recent study showed that B7-H1 was expressed on a proportion (57 of 150) of various melanocytic lesions including primary, mMEL, and various types of nevi, most commonly only at the edges of the tumor bed or in opposition to TILs (47). Four distinct groups of tumors were identified and described as having the presence of both B7-H1 and TILs, presence of TILs without B7-H1, B7-H1 expression without TILs, or absence of both TILs and B7-H1 expression (Fig. 3). As analyzed by laser-capture microdissection of the tumor edge, selective expression of IFN-γ mRNA, a potent inducer of membrane B7-H1, was shown to correlate with tumor B7-H1 expression. These data suggest that a subset of patients have an ongoing immune response within the tumor microenvironment and that B7-H1 expression is an adaptive method of tumor resistance to cytokine-producing TILs. Another patient subset likely upregulates B7-H1 expression through aberrant cell signaling, for example, via PTEN loss (66) or via activation of the MEK/ERK or MyD88 signaling pathways, which participate in the upregulation of B7-H1 in response to IFN-γ or Toll-like receptor signaling (67–69). The data also raise important questions about the function and specificity of TIL in B7-H1⁻ tumors and the biologic basis for tumors lacking TIL. In the latter group, it is appropriate to ask whether the lack of T-cell infiltration reflects the absence of tumor antigen-specific responses or an as-yet unidentified process that excludes TIL from the microenvironment.

The findings in the study by Taube and colleagues may have important implications for B7-H1 as a prognostic and predictive biomarker as well as a target for therapy. The subgroup with both B7-H1 expression and TILs may have improved prognosis compared with the group with B7-H1 expression without TILs; indeed, multiple literature studies have shown improved prognosis associated with the presence of TIL (70, 71). Consistent with this hypothesis, patients with B7-H1⁺ mMEL, which is more commonly associated with TIL, had significantly longer survival than those with B7-H1⁻ mMEL (P = 0.032; ref. 47). Current data also raise concern for using tumor B7-H1 expression as a predictive biomarker to select individual patients for treatment with PD-1/B7-H1 blockade or to select certain tumor types for development. Designation of positive B7-H1 expression remains confounded by tumor heterogeneity, variability in the assays, location of intratumoral expression, and clear cutoff values for positive versus negative expression. Moreover, the association of TIL with B7-H1 expression, or perhaps simply the presence of TIL, may be more important for predicting response than B7-H1 expression alone. Intriguingly, immune cells that infiltrate colorectal cancer have been observed to organize into intratumoral structures, which resemble lymph nodes; this associates with a particular chemokine signature, which may be useful in identifying immunotherapy-responsive patients, especially if similar phenomena are observed in other tumor types (72). The association of intratumoral lymphocytes and chemokine expression with TIL PD-1 expression and/or tumor B7-H1 expression has not yet been defined. Ascierto and colleagues discuss biomarkers for immunostimulatory antibody therapy in this issue of Clinical Cancer Research (73).

Substantial therapeutic advances for the B7-H1⁺ TIL⁺ subgroup may follow further characterization of other coinhibitory ligand–receptor interactions or immune suppressive factors in the tumor microenvironment. Recent murine studies suggest that concurrent blockade of other exhaustion/coinhibitory molecules (TIM3 and LAG3) and PD-1 can result in synergistic antitumor activity, and 1 preclinical study suggested incremental activity by blocking both PD-1 and B7-H1 (74–76). TIL and/or peripheral blood lymphocyte CTLA-4 expression increases in response to PD-1 blockade in both animal tumors and in patients, and assuming that CTLA-4 ligands are present in the tumor microenvironment, these data provide a rationale for the ongoing phase I trial of anti-PD-1 and anti-CTLA-4 (J. Weber, personal communication). For noninflamed tumors, strategies that induce and expand tumor-antigen–specific T cells and drive them into the tumor microenvironment may be necessary before or concurrent with PD-1 pathway blockade. Potential strategies to drive T cells into the tumor microenvironment include anti-CTLA-4, IFNs (which also induce surface B7-H1 expression), and signaling antagonists such as B-raf inhibitors (e.g., vemurafenib; refs. 71, 77, 78). Compared with pretreatment tumor biopsies, TIL populations (especially CD8⁺ T cells) increased in biopsies taken after patients with mMEL were treated with vemurafenib or dabrafenib, suggesting that the combination of B-raf inhibitors and immunotherapies such as PD-1-pathway inhibitors might complement one another and lead to improved patient outcomes (78). As mentioned earlier, preclinical studies have also supported combinations of anti-PD-1 antibodies with a variety of agents including chemotherapy, cytokines, and costimulatory agents such as anti-CD137 (32); CD40 is another promising immunostimulatory target (79). Although combinations with chemotherapy will be preferred in diseases where chemotherapy is standard of care, it will also be important to assess the activity of PD-1 blockade alone and in combination with other immune therapies. If additional data confirm that tumor B7-H1 expression at a predefined level is a predictive biomarker for activity of PD-1 blockade alone, the lack of expression should not preclude testing combinations of PD-1 with other agents that could drive T cells into the tumor microenvironment. Finally, combinations of PD-1 blockade with other agents such as anti-CTLA-4 antibodies would be predicted to produce greater levels of immune-related adverse events. In the setting of potentially increased activity of such combinations, the observation of unique and increased toxicities should not preclude continued development, because of the capacity to develop effective management algorithms as already shown in clinical studies of anti-CTLA-4 antibodies.

M. Sznol is a consultant/advisory board member of Bristol-Myers Squibb. L. Chen has a commercial research grant from Amplimmune Inc. and is a consultant/advisory board member of Amplimmune.

Conception and design: M. Sznol, L. Chen

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Sznol, L. Chen

Writing, review, and/or revision of the manuscript: M. Sznol, L. Chen

The authors thank Cailin Moira Wilke, PhD (StemScientific, funded by Bristol-Myers Squibb) for professional medical writing and editorial assistance.

This study is partially supported by Melanoma Research Alliance and NIH grants CA142779 and CA121974.