The immune revolution: a case for priming, not checkpoint

Abstract

Most tumors are unresponsive to immune checkpoint blockade, especially if deep immunosuppression in the tumor develops prior to and prevents T cell immunosurveillance. Failed or frustrated T-cell priming often needs repair before successful sensitization to PD-1/PD-L1 blockade. CD40 activation plays a critical role in generating T cell immunity, by activating dendritic cells, and converting cold tumors to hot. In preclinical studies, agonistic CD40 antibodies demonstrate T cell-dependent anti-tumor activity, especially in combination with chemotherapy, checkpoint inhibitory antibodies, and other immune modulators. With the advent of multiple CD40 agonists with acceptable single-agent toxicity, clinical evaluation of CD40 combinations has accelerated.

Introduction

The Immune Revolution in cancer is upon us. Deep tumor regressions achievable in multiple cancers with checkpoint inhibitory antibodies and remissions in refractory leukemia from chimeric antigen receptor (CAR) T cells have prompted a series of FDA approvals that have begun to change the face of cancer care. Still, there is a bittersweet quality to these successes: most patients do not respond to current renditions of checkpoint or CAR T cell therapy, and many patients too quickly relapse after an initial response. The PD-1 antibody pembrolizumab, for example, is approved for use as first-line therapy for patients with metastatic non-small cell lung cancer that overexpresses PD-L1 – outperforming chemotherapy in a manner thought impossible 10 years ago. Yet, nearly 30% of such patients are found to be refractory to therapy at the first restaging studies and another 25% have tumor progression at one year (Reck et al., 2016).

Cold tumors

The extent of intratumoral T cell infiltration positively predicts overall survival across many cancer types, and is also considered a major predictor of clinical response to checkpoint therapy with PD-1 and PD-L1 monoclonal antibodies (mAb). These so-called “hot” tumors stand in contrast to “cold” tumors that do not respond to single-agent PD-1 or PD-L1 therapy (Sharma and Allison, 2015; Teng et al., 2015). Pancreatic ductal adenocarcinoma (PDA), which now accounts for more deaths in the United States than breast cancer (Siegel et al., 2018), is an example of a tumor which remains almost entirely refractory to single-agent PD-1 or CTLA-4 antibody therapy. The only exception appears to be in <1% of PDA patients with high microsatellite instability (Le et al., 2017). Studies of the adaptive immune response in PDA, based on the KPC genetically engineered mouse model of this disease KRas^LSL-G12D/+, Trp53^LSL-R172H/+, Pdx1-Cre (Hingorani et al., 2005), highlight the extent to which macrophages dominate the spontaneous tumor microenvironment (TME) and the extent to which T cells are classically excluded from the TME even at the earliest stages of the disease (Clark et al., 2007). These observations are consistent with a model of a “cold” tumor (Sharma and Allison, 2015) and “immune privilege” (acquired or not) that requires a novel approach for therapy (Vonderheide and Bayne, 2013). T cell exclusion in KPC mice is a prominent feature not only in primary tumors but also metastatic lesions (Aiello et al., 2016; Clark et al., 2007). KPC tumors exhibit a low mutational burden, near absence of classically defined neo-epitopes, scant T cell infiltration and resistance to checkpoint therapy (Beatty et al., 2011; Evans et al., 2016; Feig et al., 2013; Winograd et al., 2015). In humans, T cell exclusion in PDA – with marked surrounding desmoplasia – is also well-described; however, a subset of primary pancreatic tumors do exhibit moderate infiltration of CD8⁺ T cells and other immune cells (Bailey et al., 2016; Balachandran et al., 2017; Fukunaga et al., 2004; Ino et al., 2013b; Wormann et al., 2014), and this phenotype correlates with expression of functional cytotoxicity genes and overall survival (Balachandran et al., 2017; Balli et al., 2017; Fukunaga et al., 2004; Ino et al., 2013a). In addition, human pancreatic tumors express a moderate range of non-synonymous mutations that are predicted to function as neo-epitopes (Balachandran et al., 2017; Balli et al., 2017; Rech et al., 2018). Recently, an in-depth evaluation of neo-epitopes in human PDA revealed special qualities of neopeptides, such as high differential agretopicity index (DAI) (Duan et al., 2014; Rech et al., 2018), that predicts long-term survival (Balachandran et al., 2017). High DAI reflects a mutated peptide that binds with high affinity to MHC whereas the wild type counterpart peptide does not. Yet, even in the setting of neoepitopes, checkpoint antibody therapy nearly universally fails in PDA. Thus, there is a disconnect, at least in PDA, in applying the notion that CD8⁺ T cell infiltration and non-synonymous mutations are sufficient to confer responsiveness to checkpoint blockade.

Revisiting cancer immunosurveillance

In the prevailing view of cancer immunosurveillance, T cell recognition of tumor antigens leads to tumor clearance unless the tumor undergoes immunoediting (e.g. loss of MHC or the antigen) or immunosuppressive pathways arise to dampen T cell reactivity (e.g. PD-L1) (Schreiber et al., 2011) (Figure 1). In each case, immune suppression follows as a consequence of T cell recognition. These observations were deduced from landmark studies in the highly mutated 3-methylcholanthrene (MCA) carcinogen-induced model of mouse sarcoma (Matsushita et al., 2012; Shankaran et al., 2001). For PDA and other cold tumors, however, it has been hypothesized that immunosuppression is an early (not secondary) event, and T cell reactivity fails to fully unfold during the entire natural course of the tumor (Vonderheide and Bayne, 2013) (Figure 1). In such tumors, poor T cell priming – or even immunological ignorance – may be linked to downstream mediators of oncogenes such as mutant Kras and others. Thus, in one scenario (classic MCA model), checkpoint blockade is successful; in another scenario (KPC model), checkpoint blockade alone is ineffective.

Fig. 1. Potential manifestations of cancer immunosurveillance.

In situations of an active T cell response to cancer (illustrated in the Figure by small round blue cells), a heterogenous tumor (illustrated in the upper left by a mix of red and green cells) may respond by immunoediting (top pathway) or by invoking peripheral tolerance pathways such as PD-1 and CTLA-4 (middle pathway) – each an example of immune escape. In some settings (bottom pathway), however, there is poor T cell reactivity from the start of tumor formation – potentially related to immune ignorance or immune privilege (illustrated by concentric circles around the tumor) – and thus there is no classic “escape” as noted in the first two scenarios. Rather priming or boosting of T cell responses, potentially enabled by CD40 or other means of dendritic cell activation, is required for therapeutic effect and sensitization to checkpoint blockade.

To understand immune surveillance further in cold tumors, the class MCA experiments were redone in the KPC model (Evans et al., 2016). The development of spontaneous tumors was followed in KPC mice that were either replete or depleted of T cells. There was no difference in overall survival nor time to diagnosis. Tumor cell lines harvested from T cell replete or depleted KPC mice grow equally upon reimplantation into syngeneic normal mice and are never rejected regardless of whether host mice are T cell-depleted. These observations are in contrast to those from similarly designed experiments in the MCA model.

Upon expression of the strong model antigen ovalbumin (OVA), KPC tumors are uniformly rejected unless (i) CD8⁺ T cells were depleted in host mice or (ii) host mice were made tolerant to OVA. OVA in these experiments is a neoantigen. When OVA-negative and OVA-expressing KPC tumor clones are injected together, only the OVA-negative cells grow out in T cell competent mice – a variation of immunediting in which only antigen-negative tumor cells emerge.

These observations (Evans et al., 2016), and similar findings in other genetically engineered mice (Casanovas et al., 2005; Ciampricotti et al., 2012; Ciampricotti et al., 2011; DeNardo et al., 2009), suggest that (i) the cardinal features of immunoediting are not universally observed, (ii) in such cases, this lack of immunoediting may be a consequence, and not a cause, of poor antigenicity, and (iii) antigen strength dictates the outcome of immune surveillance and can overcome even deep immmunosuppression in the TME. These conclusions are similar to other experimental models that highlight the dependency of cancer immunosurveillance on antigenic strength (DuPage et al., 2012). These observations carry an important implication for the design of novel clinical strategies: namely, in the absence of strong antigens, there may be no Darwinian-like pressure from T cells; thus, the underlying tumor cells may remain susceptible to T cells, but only if these T cells can be boosted or provoked.

CD40 and T cell priming

Increasing attention has turned toward evaluating and repairing insufficient T cell priming as a root cause of cold tumors and checkpoint unresponsiveness. Antigen presenting cells (APC), particularly BATF3-dependent type I classic dendritic cells (DC), are critical in driving T cell priming and function in tumor-bearing mice (Broz et al., 2014; Byrne and Vonderheide, 2016; Durai and Murphy, 2016; Engelhardt et al., 2012; Roberts et al., 2016; Salmon et al., 2016; Sanchez-Paulete et al., 2016; Spranger et al., 2017), reinforcing a long-standing appreciation of DC dysfunction in the TME (Aspord et al., 2007; Gabrilovich, 2004). Impaired T cell trafficking into the tumor as well as hostile TME factors that limit T cell persistence in the TME are additional likely factors that drive a T cell-poor tumor (Spranger et al., 2017; Stromnes et al., 2015; Vonderheide and Bayne, 2013).

In KPC mice, classic DCs are a rare but definite population in the TME. Notably, these cells express CD40 (Winograd et al., 2015). CD40 is a cell-surface member of the TNF receptor superfamily that is most prominently expressed on DC, B cells, and myeloid cells (van Kooten and Banchereau, 2000). It is a well-known regulator of T cell immunity including in cancer (Grewal and Flavell, 1997). CD40-ligand, primarily expressed on activated T cells, interacts with CD40 on APCs resulting in a ‘licensed’ state (Lanzavecchia, 1998). Licensed DC exhibit upregulation of cytokines (such as IL-12), antigen-presenting molecules (such as MHC), costimulatory molecules (such as CD80 and CD86), adhesion molecules (such as ICAM-1) and an array of other TNF receptor family ligands that then engage receptors of the TNF superfamily on T cells. CD40 signaling mechanisms have been summarized previously (Vonderheide, 2007). The TNF ligand-receptor orientation is uniquely reversed for CD40/CD40-ligand in the DC:T cell synapse, and as such, CD40 is as a proximal regulator of other TNF family signaling receptors on T cells. Thus, upon activation of CD40 on DCs, multiple other agonistic pathways such as OX40, GITR, and 41BB are engaged. Importantly, CD40 is also prominently expressed by B cells and CD40 activation massively upregulates APC function (Coughlin et al., 2004; Schultze et al., 1997). Thus, it is possible that CD40 activation will also influence B cell function in the TME, including B regulatory cells and tertiary lymphoid structures (Gunderson et al., 2016; Lutz et al., 2014; Poschke et al., 2016).

Both loss-of-function and gain-of-function studies in KPC model systems point to a critical role for the CD40 pathway in regulating T cell priming in tumors. OVA-expressing KPC tumor clones are rejected in wild type mice but grow progressively in CD40 knock out mice or BATF3 knock out mice that do not have CD103⁺ DCs (Byrne and Vonderheide, 2016). KPC tumor cells, which do not express CD40, are poor APC. Agonistic CD40 mAb triggers T cell-dependent tumor rejection in certain experimental models with antigenic tumors (Sandin et al., 2014; van Mierlo et al., 2002), and CD40 mAb used in combination with blocking PD-L1 or PD-1 mAb functions to promote tumor regression further (Zippelius et al., 2015). CD40 can drive an IL-12-dependent downregulation of PD-1 expression on T cells, reversing T-ccell exhaustion and permitting tumor response in otherwise PD-1 mAb refractory tumors (Ngiow et al., 2016).

In KPC mice bearing spontaneous tumors, CD40 therapy leads to non-durable tumor regression and transient involution of tumor stroma in a fraction of mice, each dependent on the presence of CD40-activated macrophages (Beatty et al., 2011). T cells are not required for this effect. Although tumor regressions were observed with agonistic CD40 mAb in a clinical trial of patients with metastatic PDA (Beatty et al., 2011; Beatty et al., 2013), neither tumor-bearing patients nor mice demonstrate durable tumor regressions from this approach.

Prior to activation, but not after, DCs have uniquely enhanced capacity to take up antigen (Albert et al., 1998; Heath and Carbone, 2001). This paradigm explains how chemotherapy followed by CD40 activation (but not CD40 followed by chemotherapy) results in the establishment of effective, T-cell dependent immunity and memory in tumor-bearing mice for which CD40 alone is insufficient (Nowak et al., 2003). The rate-limiting step of chemotherapy/CD40 effectiveness appears to be the extent to which the chemotherapy is cytotoxic against the tumor and presumably “spills antigen” (Byrne and Vonderheide, 2016). In the KPC model, gemcitabine (Gem) followed by agonistic CD40 is effective in generating ‘hot’ T cell-inflamed tumors and tumor regression only in a subcutaneously implanted tumors that have a far less complicated TME (Beatty et al., 2015). However, the addition of nab-paclitaxel (nP) to gemcitabine, which is a combination more effective in patients than Gem alone, synergizes to trigger tumor regression, establish T cell memory, and improve survival in subcutaneous, orthotopic, and spontaneous KPC tumors (Byrne and Vonderheide, 2016). This phenotype is entirely dependent on T cells, and not macrophages. In each case, the combination of Gem/nP plus CD40 leads to a marked increase in T cell infiltration in the TME, with a skewing toward IFNγ and TNFα secreting T cells, an increase in activated DCs, and loss of M2 macrophages. In addition, Tregs are cleared from the TME with therapy. Importantly, Gem/nP/CD40 completely fails in BATF3 or CD40 knockout mice, suggesting a critical therapeutic dependence on DCs (Byrne and Vonderheide, 2016).

The anti-tumor effect of Gem/nP chemotherapy with agonistic CD40 antibody is independent of toll-like receptor pathways and other innate signaling mechanisms because tumor regressions are observed in MyD88, TLR4, TLR3, TRIF, Casp 11, IL-1R, and P2X7R knock out mice as well as STING mutant and IFNAR knockout mice (Byrne and Vonderheide, 2016). Thus, CD40 activation with chemotherapy obviates the need for STING or type I interferon activation; however, these observations also suggest that the addition of CD40 mAb with STING agonists, for example, may be synergistic, given the non-redundant aspects of these pathways on DC activation.

From an immune surveillance perspective, these data in KPC mice are important for two reasons. First, the use of chemotherapy plus CD40 agonists has the capacity to sensitize tumors to checkpoint blockade (Figure 1). Combination therapy with Gem/nP/CD40 plus PD-1 antibody further extends the activity and durability of response to chemo/CD40 alone (Winograd et al., 2015). Similarly, agonistic CD40 mAb also cooperate with radiation therapy and checkpoint blockade for tumor regression (Verbrugge et al., 2012). Second, T cell infiltration, tumor regressions, and immunological memory in response to chemotherapy/CD40 is accomplished in a murine model system that does not express classically defined nor alternatively defined (i.e. high DAI) neopeptides (Evans et al., 2016). Although further investigations are ongoing to identify potential neoantigens in the KPC model beyond those arising from non-synonymous tumor mutations, there is evidence that shared antigens are responsible for the effect of Gem/nP/CD40. Cured mice are protected against subsequent challenge with unrelated KPC tumors (Byrne and Vonderheide, 2016; Evans et al., 2016). These data suggest that that are likely other types of tumor rejection antigens beyond those derived from non-synonymous mutations in the tumor.

It is not yet clear if CD40 chemoimmunotherapy, as described above, boosts weak preexisting responses and/or truly primes new anti-tumor T cells. In KPC mice bearing spontaneous tumors, implantation of syngeneic KPC tumor cells subcutaneously (i.e. two tumors) results in CD8⁺ T cell infiltration into the implanted tumor, and chemotherapy/CD40 triggers regression of subcutaneous tumors (Beatty et al., 2015). In contrast, KPC tumor cells implanted subcutaneously in wild type mice do not trigger the same level of CD8⁺ T cell infiltration, suggesting that the presence of a spontaneous KPC tumor is associated with a prior priming event in KPC mice (Beatty et al., 2015). In other studies, T cell reactivity against KPC tumors, including against shared KPC antigens, is evident among splenic T cells isolated from of tumor-bearing KPC mice, although the level of this reactivity is very low (Feig et al., 2013). In these experiments, the combination of CXCR4 inhibitor with PD-1 and CTLA-4 mAb – delivered without chemotherapy – results in T cell infiltration and modest, transient regressions (Feig et al., 2013). These observations are consistent with a low-level pre-existing T cell immunity in KPC mice, insufficient to infiltrate or persist in the spontaneous tumor. On the other hand, TCR deep sequencing shows that Gem/nP/CD40 treatment of subcutaneous KPC tumors leads to expansion of certain, pre-existing T cell clones but also is able to recruit new T cell clones specifically to the TME (Byrne and Vonderheide, 2016).

CD40 with vaccines

It is thought that chemotherapy and radiation cooperate with CD40 largely by spilling antigen and as such, function as “vaccines”. CD40 mAb also exhibit capacity to enhance the activity of conventional vaccination in both tumor and non-tumor experimental systems in mice (Li and Ravetch et al, 2011; Vonderheide, 2007). In some studies, CD40 activation substitutes for T cell help. Thus, any reagent that activates adaptive immunity, especially while sparing systemic immune suppression, may be considered a logical partner for CD40 mAb. These possibilities include true “vaccine” approaches (such neoepitope-based vaccines) but also anti-tumor antibodies, oncolytic viruses, and targeted therapy (Vonderheide and Glennie, 2013). In nearly every case studied, the addition of CD40 activation enhances the activity of the “vaccine”. There is strong preclinical data that CD40 activation and TLR agonists may synergize for APC activation (Ahonen et al., 2004; Ahonen et al., 2008; Carpenter et al., 2009; Scarlett et al., 2009). It is also possible that certain chemotherapies or hypofractionated radiation therapy themselves have adjuvant properties and trigger inflammatory or immunogenic signals beyond simply antigen release from dying tumor cells (Demaria et al., 2015; Gandhi et al., 2015; Harding et al., 2017). Finally, STING agonists along with T cell agonists (e.g. CD137 or OX40 agonists) are additional strategies now entering clinical trials aimed at enhancing tumor vaccine activity even further (Broomfield et al., 2009; Uno et al., 2006).

Role for blocking PD-1/PD-L1 and other checkpoints

If T cell vaccination overcomes privilege or ignorance to generate an anti-tumor adaptive response, it is likely that anti-PD-1/PD-L1 will be needed to address T cell exhaustion that subsequently develops after priming. Cancer vaccines developed more than 15 years were shown to break tolerance to tumor antigens, but objective tumor responses in patients were unusual. Yet, none of these vaccines had the benefit of combination with anti-PD-1 or anti-PD-L1. In both humans and mice with PDA, the potential role of anti-PD-1/PD-L1 as an immunological assist after vaccination has been highlighted (Lutz et al., 2014; Winograd et al., 2015) as well as in other tumor models (Zamarin et al., 2014). There are of course many other negative immune “checkpoints” in the TME and stroma beyond PD-1/PD-L1 that may also need to be blocked at the cellular or molecular level to fully enable vaccine therapy (Coussens et al., 2013; Kraman et al., 2011).

Clinical translation

Activation of the CD40 pathway in cancer is therapeutically tractable, as has been previously extensively reviewed (Remer et al., 2017; Vonderheide and Glennie, 2013). Three main approaches have been: (i) recombinant multimeric CD40-ligand (Vonderheide et al., 2001), (ii) gene therapeutic delivery of CD40-ligand (Messmer and Kipps, 2005), and (iii) agonistic CD40 mAb, for which there is the largest clinical experience (Vonderheide and Glennie, 2013). Like all agonists in medicine, dose and schedule of CD40 agonists have been difficult to define, and there remains no consensus on the optimal route of administration. For at least two CD40 agents, a transient, moderate cytokine release syndrome has defined the maximum tolerated dose following intravenous infusion. Typical symptoms include fever, chills, and fatigue that resolve with supportive care over 1 hr to 24 hr in the outpatient setting (Johnson et al., 2017; Vonderheide et al., 2007). Agonistic CD40 mAb infusion has been associated with mild-to-moderate, transient liver function test abnormalities and transient decreases in platelets. There have no reports of autoimmune events involving colitis, hypophysitis, pneumonitis, or uveitis which are characteristic of checkpoint antibodies (Vonderheide and Glennie, 2013). The serum half-life of agonistic CD40 mAb is less than 24 hr, decidedly shorter than typical for human IgG, yet understood in relation to the large sink of CD40 molecules found on, for example, B cells and endothelial cells. Whether CD40 activation of B cells and endothelial cells contributes to treatment-related cytokine release syndrome, or to the mechanism of anti-tumor action for that matter, remains poorly understood.

Objective tumor responses with single-agent agonistic CD40 mAb therapy have been observed, but the rate has been <20% in advanced, metastatic patients with solid tumors (Vonderheide and Glennie, 2013). In the absence of chemotherapy, radiation therapy or another immune combination partner, preclinical studies predict a low single-agent response rate in cold tumors. One patient with refractory metastatic melanoma was treated with repeated doses of CD40 for one year and remains in complete remission for more than a decade without other therapy (Bajor et al., 2014). The adaptive immune response in this patient has been extensively documented (Bajor et al., 2014). Response rates are higher in clinical trials of agonistic CD40 mAb combined with chemotherapy in solid tumors (Beatty et al., 2011; Nowak et al., 2015; Vonderheide et al., 2013), but the CD40 contribution to tumor regressions over and above chemotherapy has not been definitively discerned in randomized studies. The response rate of Gem/CD40 was 23.8% in treatment-naive patients with metastatic pancreatic cancer, with a progression free survival of 5.6 months (Beatty et al., 2011), higher than typically reported for Gem. In patients with metastatic melanoma previously untreated with PD-1 or PD-L1 mAb, the combination of CD40 and CTLA-4 mAb produced a response rate of 27.3% and a 1-year overall survival of 26.1 months (Bajor et al., 2015). Clinical trials are now underway with at least five different CD40 mAb across multiple cancers and in various combinations, including with PD-1, PD-L1, and CSF1R mAb.

Immune pharmacodynamics studies have provided evidence of CD40-induced activation of B cells and DC (Johnson et al., 2015; McDonnell et al., 2017; Ruter et al., 2010; Vonderheide et al., 2007). These effects are dose dependent and transient. From these studies, it seems unlikely that the maximum tolerated dose of agonistic CD40 mAb is the maximum biological dose.

Role of CD40 mAb crosslinking

The bulk of preclinical investigations using agonistic CD40 mAb utilize reagents that require Fc crosslinking by host Fc receptors. Activity of these CD40 mAb is minimal in vitro and lost in vivo in FcR deficient mice (Li and Ravetch, 2011; Richman and Vonderheide, 2014; White et al., 2013; White et al., 2011). Nevertheless, the CD40 therapeutic agent for which there is the largest clinical experience – selicrelumab (formally known as CP-870,893 or RO7009789) – is a fully human IgG2 for which crosslinking is not necessary for activity in vitro (Richman and Vonderheide, 2014). Although selective enhancement of FcγRIIB-binding remains possible and can increase in vivo activity (Dahan et al., 2016), FcR-independent activity of human IgG2 CD40 mAb is also provided by a conformationally distinct subfraction characterized by a unique arrangement of hinge disulfide bonds (White et al., 2015). CP-870,893 also does not compete with the CD40-ligand site on CD40. Each of these features (i.e., lack of required crosslinking, distinction from the CD40-ligand binding site) contrasts those of the anti-CD40 mAb used in the vast majority of murine studies, which absolutely require crosslinking and bind the CD40-ligand binding site. Considering that not all CD40 antibodies are “the same” – and to explore whether this translates into clinical reagents with important distinguishing clinical manifestations – newer CD40 clinical reagents have been designed to mimic the pharmacology of murine reagents on which the large body of preclinical data is based (Remer et al., 2017). APX005M, for example, is an Fc-mutated, humanized IgG1 that requires crosslinking for activity and competes with the CD40-ligand binding site (Johnson et al., 2017). In the first-in-human study, infusional side effects of APX005M were dose dependent and manageable, and immune pharmacodynamic studies revealed strong activation of APC, increased systemic levels of IL-12 and T cell activation after treatment (Johnson et al., 2017). Multiple combination studies with APX005M are underway, including one for treatment-naive patients with metastatic PDA who receive Gem/nP/APX005M with or without PD-1 mAb nivolumab (NCT03214250).

Conclusions and Perspectives

Immune privilege – manifesting with tumor T cell exclusion – is a biology without immunoediting. It may be especially notable in oncogene-driven carcinomas in which tumor-driven immunosuppression establishes in the earliest stages of the disease. Addressing deficient T cell priming, therefore, represents a large opportunity in cancer immunotherapy particularly for PD-1/PD-L1 refractory cancers. CD40 activation is one therapeutically tractable approach with multiple new reagents available to exploit this. Mechanistically, CD40 activation is a proximal event in T cell priming and thus, CD40 mAb may be critical in converting cold tumors to hot and generating effective T cell immunity. In the clinic, concerns that the therapeutic index of CD40 mAb is too narrow to permit clinical activity at tolerable doses have fortunately not been realized. Current efforts are aimed at using CD40 mAb in combination with non-redundant immune modulators, which is likely the best route to success.