Last week, the ACIR team attended the AACR Annual Meeting 2024 in San Diego, California. This week’s extensive special feature covers select talks from the conference. We have organized the content by topics below.
AACR-Cancer Research Institute Lloyd J. Old Award
Gordon J. Freeman
Regulation of anti-tumor immunity
Sébastien Talbot
Stefani Spranger
Dmitry I. Gabrilovich
Justin Micah Balko
Sjoerd H. Van Der Burg
Nir Hacohen
Strategies to enhance cell-based therapies
Livnat Jerby
Marcela V. Maus
Aude G. Chapuis
Immune checkpoints
James P. Allison
Ira Mellman
Tumor antigens and antigen presentation
James R. Heath
Sebastian Amigorena
Robert D. Schreiber
Vinod Balachandran
T cell exhaustion
Greg M. Delgoffe
Microbiome
Marcel R.M. Van Den Brink
Bispecifics
Dimitris Skokos
AACR-Cancer Research Institute Lloyd J. Old Award
AACR-Cancer Research Institute Lloyd J. Old Award in Cancer Immunology- Gordon J. Freeman - Dana Farber Cancer Institute, Boston, MA, USA
Gordon Freeman, the recipient of the prestigious 2024 AACR-Cancer Research Institute Lloyd J. Old Award in Cancer Immunology – an award given to active researchers who have been instrumental in promoting, building, and sustaining the field of tumor immunology – presented an overview of both his groundbreaking work in the field and his recent contributions. Freeman discovered the molecule now known as PD-L1. In collaboration with Clive Wood and Tasuku Honjo, he worked out the PD-L1/PD-1 inhibitory axis, which tunes down exuberant T cell responses, preventing pathology and enabling memory. In the years that followed, PD-1/PD-L1 blockade moved to the clinic and became a staple in cancer immunotherapy, and more complex models of PD-1 and T cell dysfunction modification emerged beyond just reinvigoration of already exhausted cells. Despite the clinical success of blocking the PD-1/PD-L1 axis across a range of tumor types, there is still major room for improvement. Freeman outlined the importance of combination therapies, many of which have failed to date, but emphasized the need to develop more accurate and relevant animal models. Freeman then turned to his recent work uncovering a novel interaction between PD-L1 and CD80 in cis, which is also immunoinhibitory. However, blocking this interaction led to onset of autoimmune type I diabetes, pointing to the need to clearly understand the complex interplay of multiple elements and interactions affecting immune stimulation and inhibition. Detailed analysis of these interactions eventually identified an antibody that blocked CD80 cis interactions, which alleviated the autoimmune response, potentially uncovering a new drug target. Freeman also discussed his work uncovering RMGb/PD-L2 as a new and important axis. PD-L2 is a homologue of PD-L1, which, like PD-L1, binds to PD-1, but unlike PD-L1, also binds RMGb – a GPI-anchored membrane protein found in many tissues, particularly lung. Where this axis is important became apparent as Freeman’s lab, together with Arlene Sharpe and Dennis Kasper, was investigating why microbiota-free mice (germ-free or following antibiotic treatment) were resistant to anti-PD-1/PD-L1 therapy. Analysis of CD11c+ and CD11b+ myeloid cells in tumor draining lymph nodes of microbiota-free, anti-PD-1-resistant mice showed higher PD-L2 expression. Normal responder gut microbiota, however, could downregulate PD-L2 and RGMb expression. Combined blockade of both PD-L1 and PD-L2 or both PD-1 and RGMb resulted in effective antitumor immunity in microbiota-free mice or mice reconstituted with human non-responder microbiota, eventually demonstrating that the RGMb–PD-L2 interaction was specifically preventing response to anti-PD-1 therapy and suggesting new potential combination therapies or bispecific molecules. Finally, Freeman closed with a hopeful, data-driven outlook for cancer immunotherapy; thanks for those who have motivated, supported, and enabled his work; and a photo with a healthy patient who 10 years earlier had been scheduled for hospice before getting onto anti-PD-1 treatment.
Regulation of anti-tumor immunity
Nociceptor neurons affect cancer immunosurveillance- Sébastien Talbot - Queen’s University, Kingston, Ontario, Canada
Sensory neurons (nociceptors) in tumors can not only receive signals, but also return inflammatory signals to the periphery in the form of neuropeptides, which can then act on blood vessels, increase blood flow, and allow extravasation of immune cells. Sébastien Talbot and his team studied this crosstalk between nociceptors and immune cells in melanoma. In mice in which nociceptors were ablated, tumors grew significantly slower than in control mice, and were infiltrated with T cells that had higher cytotoxic activity and less exhaustion. RNAseq analysis of neurons cocultured with cancer cells revealed upregulation of many neuropeptides, including galanin and CGRP (aka calca), both of which were very highly expressed. CGRP is the ligand for RAMP1, which acts through a C-type lectin receptor. CGRP was found to drive exhaustion of CD8+ T cells via RAMP1 in vitro and in vivo. According to TCGA data, approximately 1% of patients with melanoma had a high number of RAMP1+ TILs, and this was associated with shorter survival time. To understand which cells express RAMP1 in tumors, melanoma-infiltrating immune cells were analyzed via scRNAseq early after tumor implantation. This revealed that dendritic cells were the main cells expressing RAMP1. A series of in vitro experiments showed that neuron-released CGRP directly regulated the activation and function of cDC1 via RAMP1 to restrain T cell activation and promote tumor progression. In DCs exposed to CGRP, co-inhibitory receptors were induced, transcription factors and chemokines required for their antitumor activity were downregulated, and the capacity to present antigen was decreased. In vivo, in the absence of nociceptors, more DCs that had engulfed mCherry-labeled tumor cells were present in the tumor, and more XCR1+CD103+ migratory CDs were found in tumor-draining lymph nodes. Correspondingly, RAMP1 deletion in DCs increased their number in the tumor, along with infiltrating CD8+ T cells with increased effector function. RAMP1- DCs were also more immunogenic. However, Talbot and his team also identified indirect ways by which CGRP–RAMP1 activity mitigated influx of cDC1s into the tumor. Nociceptors and the CGRP–RAMP1 axis were found to control stem-like memory TCF1+ CD8+ T cells and production of the chemokine XCL1 by T cells. Immunofluorescence imaging of Yumm1.7 melanoma and patient samples confirmed that immune cells were found in regions that were rich in CGRP+ neurons. Spatial transcriptomics further verified that regions high for CD11c expression were also enriched for the expression of the CGRP-RAMP1 pair. In the last part of his talk, Talbot showed that tumors can reprogram innervating nociceptors to express markers of nerve injury (including the transcription factor ATF3) and secrete neuropeptides (including galanin). Galanin seems to directly act on CD8+ T cells to drive their exhaustion, and galanin expression in patient samples correlated with shorter survival time.
Not all T cell responses are the same: how DC shape anti-tumor immunity- Stefani Spranger - Koch Institute for Integrative Cancer Research at MIT, Boston, MA, USA
Not all T cell responses are created equal, and to understand why, Stefani Spranger and colleagues investigated the cell types that underlie T cell activation. In prior work looking at conventional DCs in models of tumor regression (spontaneous tumor control), and progression (immune escape), Spranger uncovered an important role for an interferon-sensing gene signature (ISG)+ DC cluster that was capable of stimulating antitumor T cell responses, and found that regressing tumors produced more IFNβ than progressing tumors, supporting this cluster. Next, to understand why tumor types with similar TMB and T cell infiltration respond differently to immunotherapy, Spranger demonstrated that a low TMB tumor model implanted in the flank responded to ICB, while the same tumor implanted in the lung did not. Intratumoral T cell numbers and dynamics were different at each site, and by examining the respective draining lymph nodes, Spranger found that while both dLNs activated T cells to equal degrees, there were over a thousand differentially expressed genes between them. Notably, T cells primed in the lung dLN were more consistent with effector or effector exhausted T cells, while T cells primed in the flank dLN were more memory-like. This was found to be controlled by an expanded Th1 effector-like subset of Tregs (expressing CXCR3 and T-bet). Higher IFNγ levels were observed in lung dLNs, even in tumor-naive mice, but the effect was abrogated in germ-free mice, suggesting that the increased IFNγ is likely associated with microbiome of the lung (a mucosal barrier tissue). Next, Spranger discussed immune-desert tumors, which often display neoantigen heterogeneity and are unresponsive to checkpoint blockade. Using naturally derived neoantigens added to the low TMB KP tumor cell line to model complex antigen expression patterns, researchers found that co-expression of two strong neoantigens resulted in a pattern of clonal dominance only when both were presented on the same cancer cell (“clonal”) and on the same MHC molecule. When a strong and weak neoantigen were combined, the weak antigen response was enhanced, but again, only when the two antigens were coming from the same cancer cell. As these differences were observed in the spleen, it was likely the result of differential T cell priming in the lymph node, rather than T cell encounter with antigens in the tumor. Tracking antigen uptake by DCs in the LNs, the researchers found that DCs from subclonal tumors (each cell expressing one antigen or the other) usually only cross-presented one antigen at a time, whereas DCs from clonal tumors expressed multiple neoantigens at once. Further, DCs expressing multiple antigens were found to express more costimulatory and less inhibitory molecules, suggesting a more immune-activating phenotype. In regards to therapeutic translation, these results suggest that targeting clonal neoantigens may help to overcome immune evasion. There was even some efficacy when co-targeting weak antigens in combination with checkpoint blockade, opening up new potential vaccine targets that are less likely to result in immunoediting.
Regulation of T cells by myeloid cells in cancer- Dmitry I. Gabrilovich - AstraZeneca - MedImmune, LLC, Gaithersburg, MD, USA
Depending on their context, myeloid cells tend to split into two major functional states: under acute threat, they serve an important role in protective immunity, but under chronic inflammation, as in cancer, they become immunosuppressive. There are multiple types of suppressive myeloid cells derived from different sub-lineages (DC, monocyte/macrophage, or neutrophil), and often such cells exhibit functional redundancy in suppressive pathways. This led Dmitry Gabrilovich to hypothesize that it might be necessary to attack such cells at their root. Gabrilovich began by discussing ferroptosis – a type of programmed cell death in tumors driven by changes in the redox state. While attempts to use ferroptosis to control tumor growth have failed, suppressive neutrophils (PMN-MDSC) have gene signatures associated with ferroptosis, and induction of ferroptosis in neutrophils converts them to a suppressive state, leading to the release of factors that suppress T cell activity. Consistent with this effect and counter to the expected direct effects on tumor cells, treatment with a pharmacological inhibitor of ferroptosis enhanced antitumor efficacy (particularly when used in combination with checkpoint blockade) while induction of ferroptosis enhanced tumor growth. In human data from TCGA and other datasets, a ferroptosis signature was positively correlated with PMN-MDSCs and negatively correlated with T cell infiltration and patient survival. Next, Gabrilovich described another mechanism of immunosuppression that was identified based on clinical observations in which patients with NSCLC who progressed after anti-PD-L1 showed notable objective response rates upon treatment with anti-PD-L1 and an inhibitor of ATR (involved in DNA damage repair). In these patients, ATRi was associated with reduced monocyte signature and reduced T cell exhaustion, as well as with increased type I and II IFN genes. Modeling this in mice, researchers found that ATRi alone had antitumor activity, and enhanced the antitumor effects of anti-PD-L1 in a manner that was dependent on CD8+ T cells. Further investigation showed that ATRi suppressed proliferating T cells. In line with this, ATRi was only effective against tumors with intermittent scheduling. Upon rest and recovery, antigen-specific T cells became more polyfunctional due to changes in the TME, including an increased IFN-I signature. This was validated in a GEMM model of lung cancer, which also showed depletion of immunosuppressive TAMs and monocytic MDSCs, but not PMN-MDSCs or DCs. This was likely due to differential expression of DNA damage repair machinery between monocyte-derived suppressor cells and PMN-MDSCs. While PMN-MDSCs were not reduced by ATRi, the increased IFN-I in the TME reduced their propensity towards ferroptosis, essentially eliminating their immunosuppressive activity. The increased IFN-I also activated DCs, further supporting T cell accumulation and T cell-mediated antitumor immunity. IFN-I also induced PD-1 upregulation on T cells, supporting the observed synergy with anti-PD-L1. Overall, ATR inhibition appears to fundamentally drive remodeling of the TME myeloid cell compartment towards an antitumor state, emblematic of attacking the immunosuppressive compartment at its root.
Getting T and NK cells to collaborate in an MHC-I heterogeneous microenvironment- Justin Micah Balko - Vanderbilt University Medical Center, Nashville, TN, USA
Intrigued by the important role of the MHC-I pathway in antigen presentation to cytotoxic CD8+ T cells, Justin Balko took this a step further by considering how heterogeneity in MHC-I expression in tumors – a biomarker that has not received significant attention – might impact immunity. With a focus on breast cancer (BC), Balko first showed as part of a multi-omics study of TNBC that HLA-A alleles could be impacted by both hypermethylation and by focal deletion (less common with HLA-B or -C). Taking a more granular look, he used immunofluorescence in 314 patients to quantitate HLA expression at the single-cell level. While ER+ BC showed universally low HLA expression, TNBC showed the highest HLA expression and widest variation, with 25% of cases being multimodal (a mixture of on and off phenotypes). Balko then modeled this with EMT6 tumors, which are sensitive to anti-PD-L1 treatment. Knockout of B2m (B2mKO) completely abrogated the anti-PD-L1 response, but interestingly, mixtures of WT and B2mKO with as few as 2% B2mKO cells was enough to show resistance to anti-PD-L1. Analysis of tumors from mixed inoculation showed a reduction in CD8+ T cells, but an increase in NK cells. Returning to human IHC, Balko observed that when classifying tumor regions as MHC high, low, or heterogeneous, and measuring the proximity of CD8+ T or NK cells, the close proximity of CD8+ T cells was positively correlated with MHC expression level and that NK cells were slightly closer to heterogeneous tumor areas than to MHC high or low areas; these results were similar to those observed in mice. These results suggested that addition of an antibody blocking the NK inhibitory receptor NKG2A might allow NK cells to be more active and rescue anti-PD-L1 activity, which it did. Interestingly, anti-PD-1 monotherapy led to a decrease, while anti-NKG2A monotherapy led to an increase in MHC-I. To explain why NK cells were not active when B2m was totally absent, Balko postulated that some T cell activity or product, such as IFNγ, was necessary. By making B2mKO tumor cells able to inducibly produce IFNγ, he could restore NK cell recruitment and sensitivity to anti-PD-L1. Balko closed by showing that in a trial of chemotherapy alone versus chemotherapy with anti-PD-L1 in metastatic TNBC, outcomes were better in patients with low (compared to high) MHC heterogeneity. Higher levels of NK cells were observed in the more heterogeneous tumors though, suggesting that such patients might benefit further with anti-NKG2A.
Myeloid cell mediated intrinsic immune resistance in HPV-induced cancers- Sjoerd H. Van Der Burg - Leiden University Medical Center, Leiden, Netherlands
Sjoerd Van Der Burg is interested in defining the tumor-intrinsic and -extrinsic mechanisms that lead to immune resistance, and for this presentation, he focused on myeloid cell effects. Van Der Burg and colleagues have studied HPV-induced cancers in mice and humans, showing in humans that spontaneous or vaccine-induced T cell responses correlated with improved, though still not optimal, clinical outcomes. Upon therapeutic vaccination with a synthetic long peptide containing HPV16 epitopes, TC-1 tumors were initially controlled in mice, but eventually escaped, and so this was a good system to dissect the escape mechanism(s). After demonstrating that escape was not due to loss of antigen expression, HLA expression, IFNγ/TNFα signaling, differential T cell killing, or lack of immune checkpoint inhibition, Van Der Burg took a deeper look at myeloid cells, recognizing that myeloid cells increased in numbers in peripheral blood of tumor-bearing mice and patients with cervical cancer, and inhibited T cell activation in patient PBMCs . Parallel work in patients had shown that early after initiating standard-of-care carboplatin/paclitaxel chemotherapy, the high levels of myeloid cells decreased, coinciding with increased T cell reactivity. This result was replicated in mice and led to two clinical trials in which vaccination was given with chemotherapy and timed to begin during this period of myeloid cell decrease, showing improved vaccine responses and improved survival in patients responding to the vaccine. Returning to preclinical models, tumors from vaccinated mice were examined at the start of tumor regression and when tumor volume had shrunk substantially, but would soon relapse. This revealed a shift in phenotype from M1-like to M2-like macrophages and a significant increase in expression of immune checkpoints in T cells, with evidence for enhanced markers of proliferation (Ki67). In particular, CD163high, tissue-resident, suppressive macrophages were increased in these day 19 tumors. Interestingly, CSFR1 inhibition reduced overall macrophage numbers significantly, but the small CD163high population, which was found at the periphery of the tumor, was unaffected, suggesting a possible reason for failure of CSFR1 inhibitors in the clinic to date. Targeted deletion of these macrophages just after vaccination with CD163-targeted liposomes containing doxorubicin restored CD8+ T cell phenotypes and increased survival. However, once resistant tumors were established, this treatment had no effect. Transfer of early and resistant tumors into naive mice demonstrated that tumor growth of resistant tumors was unresponsive to further vaccination (while early tumors were responsive), indicating this was a tumor-intrinsic property. Vaccination was able to induce similar T cell responses in early and resistant tumors, but CD11b+ (and CD11b+MHCII+) myeloid cell numbers were decreased in resistant tumors. Returning to the prior observations, cisplatin treatment increased T cell responses and restored myeloid CD11b+ populations to almost the levels observed in early tumors, increasing survival and indicating that effective T cell therapies may need, and appear to attract, some particular myeloid cell state(s) for effective tumor rejection. Studies in other tumor models indicate that neutrophils and eosinophils may similarly have particular tumor-rejecting states important to T cell-directed strategies.
Cancer vaccines and understanding and promoting immunogenicity- Nir Hacohen - Broad Institute, Cambridge, MA and Massachusetts General Hospital, Charlestown, MA, Charlestown, MA, USA
As tumors mutate and evolve, they often produce antigens that the immune system can recognize. However, these antigens are often either edited out in response to immune pressure, or the tumor finds ways to defend itself against immune attack, leading to ongoing, but ineffective antitumor immunity. Given that every tumor presents unique challenges, Nir Hacohen recommends taking a personalized approach, namely through personalized neoantigen-targeting vaccination. However, effective neoantigen vaccination presents numerous hurdles. For example, good targets need to be both tumor-specific and immunogenic, which requires identification of neoantigens or other source antigens in tumors (e.g., unannotated ORFs), along with effective prediction of their presentation on HLA molecules. While immunopeptidomics has advanced prediction algorithms, there is still room for improvement and additional features of prediction pipelines that need to be incorporated (transcript/protein abundance, proteolysis transport, TCR recognition, etc.). Effective vaccination also requires generating a diverse immune response, which can come from targeting multiple antigens at once, targeting an appropriate combination of MHC-I- and MHC-II-presented antigens, or inducing epitope spreading. Effective vaccines also need to be optimized for the magnitude, quality, and lifetime of the response. Perhaps the most complex challenge is optimizing the quality of the response, which includes understanding key immune cell biology, including the T cell states important to tumor elimination, the suppressive cells that dampen immune responses, the coordinated interactions between immune cells, and how tumors evolve to resist antitumor immunity. Looking more closely at how immune cells “speak” to each other, Hacohen described research that identified “immunity hubs” based around CXCL10/CXCL11+ cells. In data from clinical trials, these immunity hubs were found to be enhanced in responders versus non-responders to immune checkpoint blockade. In more recent work, Hacohen and colleagues investigated these hubs further and identified a particular subset of immunity hubs that contained TCF7+PD-1+ CD8+ T cells, and strongly delineated responders, progression-free survival, and overall survival. Using the MERFISH gene panel to dig deeper, the team was able to analyze spatial neighborhoods, and found that immunity hubs containing TCF7+PD-1+ CD8+ T cells showed interactions in which ISG+ macrophages produced CXCL10/11, inducing CD4+ T cells to produce IFNγ, which in turn supported the ISG+ macrophages and T cell recruitment in a positive feedback loop. Additionally, mregDCs and stromal cells were found to produce CCL19, supporting stem-like TCF7+ and TCF7+PD-1+ CD8+ T cells.
Strategies to enhance cell-based therapies
Decoding and rewiring cancer-lymphocyte interactions- Livnat Jerby - Stanford University School of Medicine, Stanford, CA, USA
As part of a session on the organization and crosstalk of cells within the tumor microenvironment (TME), Livnat Jerby focused specifically on the the question of what drives the formation of the tumor cell–lymphocyte synapse, beginning with the signals impacting infiltration, and ending with tumor cell clearance. Her overall vision is to use a variety of new tools and technologies to understand these spatial and signaling questions, beginning with population-level information and proceeding through individual patients, tissues, local organization of cells, and the role of individual molecules in these processes. This information can then be used to design a new generation of drugs that improve how immune cells sense cancer. Jerby highlighted the use of CosMx as a tool that allows interrogation of the expression levels of thousands of genes at sub-cellular resolution, as well as the associated bioinformatics tools used to decipher spatial architecture. She presented the first example: studying NK and T cell recruitment in a population of patients with ovarian cancer. Infiltration is a marker for good outcomes, but still imperfect. Based on analysis of 2.4 million cells, Jerby and her team have identified a malignant cell state signature that is driven by copy number alteration, and predicts infiltrating T/NK cell levels, proximity of these cells to tumor cells, and response to immune checkpoint blockade in various cancer types. Based on the transcriptome of the intratumoral CD8+ T cells, the researchers could also infer the phenotypic STATE of CD8+ T cells and, coupled with the spatial information, decipher the LOCATION of T cells with regard to malignant cells.To study the impact of many individual genes at the cellular level, Jerby turned to a highly resolved model (cancer spheroids with engineered T cells). Using CRISPR technology to knock out many individual genes at once, and the highly resolved CosMx technology to interrogate the resulting cellular and spatial changes in gene expression allowed a new look into the question of how cells “choose” their environment and their defining cellular state. This analysis identified upregulation of PTPN1 as a pattern associated with sensitivity of tumor cells to NK/T cell killing (as independently observed by others). This information can then be used to find ways to cause a desired effect (“directed evolution”). Jerby closed with an example of this approach – using the CRISPR/dCAS9-SAM system to activate interesting target genes in NK cells and examine the impact on cell migration in a model spheroid system. Interestingly, the top hits were in genes not typically expressed in NK cells, but in various myeloid cells, suggesting that infiltrating NK cells were hijacking these signaling mechanisms via various G protein-coupled receptors and metabolite targets.
TEAMS of CAR-T cells in cancer- Marcela V. Maus - Massachusetts General Hospital, Boston, MA, USA
CAR T cells often show impressive antitumor activity, but CAR T cell-based treatments need to overcome antigen heterogeneity and immunosuppressive microenvironments in order to be more effective against solid tumors. To do this, Marcela Maus made use of CAR-TEAMs. In this unique CAR T cell strategy, a CAR T cell capable of engaging a target antigen on a tumor cell would be designed to also secrete T cell-engaging antibody molecules (TEAMs) that would target an “undruggable” molecule, such as a target that is also found on healthy tissues. Because these TEAMs would be released locally and cleared rapidly, they would be unlikely to induce systemic problems, and could be used to direct additional effects only within tumors. For glioblastoma, Maus and colleagues developed a CAR-TEAM in which the CAR recognized a mutated form of EGFR (EGFRvIII) while the TEAM would target T cells towards wild-type EGFR (TEAM-E). In mice containing mixed EGFRvIII+ and EGFRvIII- GBM tumors, CAR TEAMs induced complete tumor clearance. Interestingly, TEAM-secreting cells showed strong antitumor efficacy even without CAR activity. In a skin graft model, TEAM-E showed no evidence of toxicity, allowing Maus and colleagues to move towards clinical translation. Following the development of a manufacturing protocol, TEAM-secreting CAR T cells moved to patients in a phase 1 clinical trial in glioblastoma. In the first 3 patients in the safety run-in arm, manufacturing was successful for all patients, and all patients were treated. All patients experienced CRS and encephalopathy, but these were considered tolerable, with no dose-limiting toxicity. The first patient experienced tumor progression during the manufacturing process, but within 24 hours of treatment, was already showing signs of tumor reduction. Around 6 weeks after the initial infusion, the patient started to recur, and was given a second dose, which had a less significant effect, and a late biopsy showed reduced expression of EGFRvIII and copy number alterations. Patients 2 and 3 also saw rapid initial tumor regressions with no dose-limiting toxicities, and later loss of EGFRvIII and EGFR. Looking at the pharmacokinetics, CAR-TEAM cells persisted in the cerebrospinal fluid for about 4 weeks, and TEAM molecules could be observed by flow cytometry on the CAR surface. Some CAR-TEAM cells made their way to peripheral blood by 1-3 weeks, but did not contain TEAM on their surface. Further, TEAM molecules could not be detected in peripheral blood, in line with their intended rapid clearance. In addition to glioblastoma, Maus and colleagues are investigating CAR-TEAMs in both pancreatic cancer (targeting mesothelin on pancreatic cancer cells and FAP on inhibitory CAFs) and AML.
Unlocking new avenues to overcome efficacy challenges of TCR T-based strategies- Aude G. Chapuis - Fred Hutchinson Cancer Research Center, Seattle, WA, USA
Taking advantage of TCR T cells for use in therapeutics has both advantages and disadvantages compared to CAR T cells. For example, while TCR T cells can target intracellular antigens, they are restricted by HLA and require external costimulation, which is often lacking in the TME. In past research, Aude Chapuis and colleagues linked an intracellular activation domain to peptide/HLA molecules in order to directly compare features of CAR and TCR T cells. Later, in human Merkel cell carcinomas (MCC), which are largely driven by viral infection (MCPyV) and are often responsive to checkpoint blockade, Chaupis and team identified TCR T cells that recognized MCPyV antigens, including variants that were shared between some patients. In preclinical work with one of these TCRs, CD8+ T cells were more effective against antigen-containing cells than CD4+ T cells, and tumor cell killing could be enhanced with IFN-I, which induced upregulation of MHC-I. In a translational study, 5 patients with MCC were treated with autologous MCPyV-specific TCR T cells with low dose radiation to induce IFN-I, and one patient showed evidence of a late response, at a time point when TCR T cells were no longer present in the blood, but still persisted in tumors. Looking at the eventual escape lesion in this patient, expression of the target HLA was reduced, but not lost, and areas of the tumor that still expressed HLA also expressed an IFNγ signature. Based on this observation, the researchers showed that IFNγ could reverse this effect in vitro, and so exogenous IFNγ administration was added to the clinical protocol. One of two patients treated on this protocol had a delayed response, and evidence suggested that IFNγ transiently restored MHC-I, which was associated with increased activated T cells in the tumor. Investigating then whether T cell responses – especially CD4+ T cell responses – could be improved, Chapuis and others made modifications for expression of CD8αβ and/or CD200R/CD28 in T cell products. This enhanced the antitumor efficacy of cell products, and induced enough IFNγ production to eliminate the need for exogenous IFNγ. In mice, only the TCR T cells expressing both the CD8αβ and CD200R/CD28 modifications showed enhanced infiltration, on-site proliferation, and re-upregulation of HLAs in tumors. These combination-modified cells also controlled tumors better and faster than cells with single or no modifications. Only cells with the CD28 modification maintained subsets with stem-like (early memory) and cytotoxicity-associated gene expression.
Immune checkpoints
Immune checkpoint blockade in cancer therapy: Historical perspective, new opportunities, and prospects for cures- James P. Allison - UT MD Anderson Cancer Center, Houston, TX, USA
T cell activation is initiated upon recognition of an antigen, but the degree and duration of activation is controlled by a number of costimulatory and coinhibitory receptors. Jim Allison, 2018 nobel laureate, described how cancer cells often go undetected and get a headstart on T cells, and how the balance of activating and inhibitory signals is often skewed in cancer, allowing cancer cells to escape immune detection. Interventions like anti-CTLA-4 and anti-PD-1, which reduce the pressure of coinhibitory signaling and allow for enhanced T cell functionality, have helped to flatten the survival curve in patients around the globe, but still, only a small portion of patients benefit. In addition to CTLA-4 and PD-1, there is a long list of known molecules that could potentially be targeted to improve antitumor immunity, but despite extensive research, only one new checkpoint, Lag3, has joined the ranks of clinical approval in recent years. Additionally, while some tumor features, like high mutation burden/neoantigen load or high PD-L1 expression, have been associated with responses to therapy, the predictive value of these features is limited and inconsistent across cancer types. In order to increase the number of patients who benefit from immunotherapy, and to better predict who will or will not respond, Allison suggested two key future directions: 1. moving therapies into earlier disease settings, and 2. performing more data-seeking translational studies to identify mechanisms of response and resistance. In the move towards targeting earlier disease settings, Allison advocated for neoadjuvant trials, citing numerous examples in which neoadjuvant trials not only benefited patients, but progressed the field of cancer immunotherapy through the collection of treated tumor tissues for evaluation. In regards to identifying mechanisms of response and resistance, Allison cited studies in which rational combinations were established based on clinical observations and follow-up preclinical investigations. In one example, Allison and colleagues noted that Th1-like CD4+ T cells expanded by anti-CTLA-4 upregulate ICOS. Upon further investigation, they found that engaging the ICOS pathway with an agonist vaccine could enhance the antitumor efficacy of anti-CTLA-4 by increasing effector CD4+ T cells and effector/memory CD8+ T cells while reducing exhausted T cells and inducing a shift in myeloid cells from suppressive to inflammatory. Another potential target molecule is VISTA, which only shows up with anti-CTLA-4 treatment, so while it might not serve as a great target on its own, it could be beneficial to target in combination with or following anti-CTLA-4. Additionally, clinical studies have pointed towards checkpoint inhibitors being less effective in patients with bone metastases, which upon investigation, was found to be due to a reduced Th1 and increased Th17 response due to higher IL-6 and particularly TGFβ. This guided animal studies in which blocking TGFβ enhanced the limited activity of CTLA-4 blockade. Overall, Allison emphasized the importance of using clinical observations to guide research and rational combination therapies, with close attention to specific disease settings and treating patients earlier.
Reimagining checkpoints and cancer immunotherapy- Ira Mellman - Genentech, Inc., South San Francisco, CA, USA
Ira Mellman began by describing the central role of checkpoint blockade in the cancer immunity cycle, highlighting that inhibition of the PD-L1/PD-1 axis may be able to act both early (during initial T cell activation) as well as late (by overcoming T cell exhaustion in the TME) in the cycle, according to early dogma. That dogma had its basis in observing the state of T cells in cancer and chronic viral infection. Growing evidence has challenged that initial idea, particularly the difficulty of reversing the epigenetically defined exhausted state, and more recent work points to a critical role of the PD-L1/PD-1 axis during the early T cell commitment stage, which defines the differentiation trajectory. Mellman then summarized published work demonstrating that PD-1 activation downregulates both CD28 and TCR signaling through the phosphatase Shp2, and that the major source of the PD-1 ligand, PD-L1, was not on tumors, but instead on macrophages and dendritic cells, hinting at a role during antigen presentation and commitment. Further experiments confirmed an outsized importance of PD-L1 expressed on DCs in controlling tumor growth. TIGIT, another T cell inhibitory receptor that is often co-expressed with PD-1, has mechanistic similarities to PD-L1. TIGIT binds with high affinity to its ligand PVR (CD155). CD226 is a T cell immunostimulatory receptor that also binds PVR, but with lower affinity, so would typically be outcompeted by TIGIT. Surprisingly, functional studies revealed that, just as observed for CD28, PD-1 controlled CD226 stimulatory activity through Shp2, and TIGIT acts only by competing with CD226 for binding to PVR. Investigating the phenotype of cells in tumors and draining lymph nodes following blockade of both PD-1 and TIGIT revealed a greater overall number of antigen-specific CD8+ T cells, as well as a greater fraction of CD226+ cells and a decrease in TOX+ cells. Hence, negative regulators (such as PD-1) work by regulating the activity of activating receptors (such as CD226 and CD28). Taking advantage of a very large single-cell sequencing data set (RNA-, TCR-, tetramer+-, CITEseq) of many T cells in mice, the clonotypes and phenotypes of T cells in the tumor, draining LN, and blood were examined following different treatments. Interestingly, the combination of anti-TIGIT and anti-PD-1 (but not the monotherapies) identified tetramer+ TCRs in the blood that were significantly expanded in the lymph nodes and tumors. In the blood, these cells were in a newly identified CCL5+ cluster, suggesting this cluster comprised transiting effector cells. When phenotype trajectories in the different tissues were analyzed, the combination treatment had the most dramatic effect, particularly in expanding stem-like cells in the dLN, and shifting away from the exhausted state and toward a high-grade effector state in the tumor. Efficacy of the combination in controlling tumors depended on dendritic cells and PVR. Antibodies blocking B7.1 (ligand of CD28) and/or CD226 abrogated efficacy, indicating that dendritic cells provide the ligands for both PD-1 and TIGIT/CD226.
Tumor antigens and antigen presentation
High throughput experimental discovery and validation of tumor antigens- James R. Heath - Institute for Systems Biology, Pasadena, CA, USA
With the perspective of using or inducing T cells for immunotherapy of cancer, Jim Heath posited that we need very large libraries of antigen-specific TCRs, due to the diversity of targets and HLA molecules, the importance of having multiple TCRs for a given patient to counter loss of individual HLAs or tumor heterogeneity, and the need for T cells representative of a range of phenotypes, not just cytolytic, and not just CD8+. Advances in technology have driven the magnitude of TCRs that can be screened for inclusion in such a library, and Heath previewed the rest of his talk indicating that now we can imagine screening 1,000 different peptide:MHC “targets” across 10 HLA alleles with T cells from 100 patients all in one experiment, concomitantly collecting cellular transcriptomes, patient-identifying SNPs, and more. A key advance toward this goal was the development of approaches to prepare stable peptide:MHCs by expressing the entire peptide, MHCα, and B2m as a single-chain trimeric (SCT) molecule or as a precursor to that complex that can be cleaved and exchanged with the target peptide (single-chain dimer [SCD]). With this approach, large libraries of barcoded peptide:MHC complexes can be prepared and used to identify TCRs in reactive T cells. This high-throughput approach allows a liberal selection of peptides to be tested. In fact, restricting peptide selection to only moderately tight binders (rank < 2%) eliminates the opportunity to find TCRs for weaker binders, which do exist and are effective, indicating there is a lot of biology between biochemical binding and actual immunogenicity. The broad range of TCRs recovered demonstrated a range of cytotoxicity, speed of cytotoxicity, and functionality when transfected into Jurkat cells. The wealth of recovered TCRs though makes finding one with the desired characteristics a challenge, so Heath proposed to increase the information content by conducting 10X single-cell transcriptome analysis, including TCR sequencing and CITEseq on a large collection of patient T cells from multiple timepoints in the patient’s history, captured by a large collection of barcoded tetramers from multiple alleles. Each single cell then contains patient, tetramer, TCR, and cell phenotypic information, and dimensional reduction can be used to cluster the cells based on relevant parameters, and relate back to individual TCRs and their cognate immunogenic antigen. Applying this process to a set of patients vaccinated with an HPV E6/E7 vaccine, liberally choosing what potential epitope to interrogate resulted in detection of known and previously unknown HLA-restricted epitopes. Strikingly, particular clonotypes were usually restricted to particular cellular phenotypes, raising the possibility that the phenotype, and potentially the trajectory of the stimulated T cell was an intrinsic property of the combination of TCR and peptide:MHC. If such rules could be uncovered, this would aid in selecting clonotypes with the properties desired. As a first step to uncovering such “rules”, Heath conducted a high-information content screen of nearly 1,000 tested antigens from regions of SarsCov2, focusing on HLA-A0201. Armed with the biochemical and physical information of each peptide antigen and the sequence properties of the CDR3 region of each TCR, clustering of the properties of each reactant and analysis of combinations with respect to the observed phenotype of each clonotype led to clear pairings that tracked with cell phenotypes.
A new family of immunogenic antigens- Sebastian Amigorena - Institute Curie, Paris, France
To identify novel sources of antigens for vaccination or other immunotherapy approaches, Sebastian Amigorena described how such antigens can be identified from the non-coding, or “dark” genome. One source described by Amigorena involves non-canonical splicing in coding genes that result in the translation of intronic regions (often transposable elements) and presentation of unique peptides on MHC. While these aberrant splicing events result in junctions between exons and transposable elements (JETs) that can be identified through analysis of RNAseq data, their expression is often low and there is a lot of “noise” in the data, making it difficult to sort and classify. However, tumors often have defects in splicing, so identifying non-canonical splicing events could potentially serve to dramatically extend the pool of targetable tumor-specific neoantigens. Evaluating JETs in tumors and healthy tissues, Amigorena and colleagues studied 3 tumor cell lines in mice and found that JET expression effectively distinguished tumors from one another and from healthy tissues, suggesting that non-canonical splicing events can be tumor-specific. Further, immunopeptidomics analysis was used to validate processing and presentation, and the researchers found that JETs resulted in antigens that were immunogenic and elicited antitumor immune responses in mice. Vaccination with these antigens further enhanced antitumor immune responses and delayed tumor growth. Similar results were observed in patients, where some splicing events were found to be recurrent across patients and even across tumor types. Extending the pipeline to other types of non-canonical splicing events, the researchers developed two peptide identification and validation platforms to identify tumor-specific peptides derived from splicing events, allowing work on a larger scale to identify additional candidates. ScFvs that target these novel peptides can be utilized for therapeutic use in any of several formats (IgGs, CARs, etc.). A CAR T cell example showed evidence of activation and antitumor efficacy in vitro and in vivo, validating the identification of new druggable targets. T cell engagers targeting dark genome-derived peptide:MHC molecules were also effective in vitro. Currently, the researchers are looking at the clinical potential of targeting these peptides through the development of vaccine panels. The antigen MNO-P7 is of particular interest, as it has been shown to be highly expressed in numerous cancers across TCGA, with little to no expression in healthy tissues or blood cells. Overall, these results show that targeting non-canonical splicing-derived antigens extends the range of possible druggable targets that are both tumor-specific and shared across patients.
Improvement of tumor neoantigen detection by high field asymmetric waveform ion mobility mass spectrometry (FAIMS)- Robert D. Schreiber - Washington University School of Medicine, St. Louis, MO, USA
In a typical immunopeptidomics workflow, mass-spectrometry (MS)-based detection of neoantigens in native tissue requires the expansion of tumor cells, from which HLA complexes can be purified and neoantigens can be identified through data analysis. Because this process requires high numbers of tumor cells, it is often difficult to implement in a clinical setting. While newer, more sensitive MS technologies are becoming available, they are expensive, and it may not be practical for labs to upgrade their entire system. Instead, Bob Schreiber described how using high field asymmetric waveform ion mobility mass spectrometry (FAIMS) could be implemented as an attachment to improve the depth of peptidome profiling on common MS instruments. FAIMS is a gas-phase separation technique using ion charge state and size, and an empirically determined compensation voltage to select which population of ions are captured. To demonstrate the neoantigen detection power of FAIMS, Schreiber and team utilized their well characterized T3 MCA sarcoma, for which they have validated a full profile of MHC-I and MHC-II neoantigens. Use of FAIMS identified the major neoantigen, Lama4, from these cells, reducing the “junk” hits and noise from MS, and allowing for clearer identification of neoantigen peptides. FAIMs also improved detection of the minor neoantigen from T3 (mAlg8) as well as MHC-II neoantigens from tumors with enforced expression of CIITA. To see “how low they could go” with FAIMS, Schreiber and team performed a dilution series to establish the lower limit of mLama4 detection with or without FAIMS. Across all cell counts tested, FAIMS showed improved sensitivity, using either data-dependent acquisition or targeted approaches. Further, FAIMS was able to identify mLama4 from 10mm-diameter tumor samples (about 13 million cells, around half of which are tumor cells). Importantly this metric would allow for the detection of neoantigens in individual tumors, without the need for tumor cell expansion. Similarly, the major class I tumor rejection antigens have been detected in the F244 and 1946 sarcoma lines, further validating these results.
Personalized RNA Vaccines for Pancreatic Cancer- Vinod Balachandran - Memorial Sloan Kettering, New Your, NY, USA
Vinod Balachandran began by outlining why cancer vaccines are on the verge of becoming the next promising cancer immunotherapy. Key requirements for such a breakthrough include meeting the challenges of strength, specificity, functionality, and durability, which Balachandrann posited are being met based on their initial and updated results in a trial of a personalized mRNA vaccine. These mRNA vaccines, comprising up to 20 neoantigens per patient in lipoplex nanoparticles, is being delivered in the adjuvant setting to patients with pancreatic cancer. The previously published results demonstrated (1) strong ex vivo ELISpot responses in 50% of the patients, (2) expansion and longitudinal tracking of TCRs, with some clonotypes reaching up to 10% of all peripheral blood T cells and (3) primarily de novo CD8+ T cell responses, which did not overlap with clonotypes expanded after a single dose of anti-PD-L1 three weeks prior to vaccination. Critically, at median follow-up of 1.5 years, none of the vaccine-responding patients (n=8) showed evidence of recurrence (median RFS was not reached) compared to non-responders (median RFS of 13.4 months). Balachandran updated these encouraging results now that patients have reached a median follow-up of 3.2 years. Extensive longitudinal tracking of 79 different TCR clonotypes from the 8 responder patients at 24 time points before, during, and after vaccination revealed durable responses that followed the classical pattern of T cells induced by infection (priming, contraction, and long-term memory) with durable responses at 3 years (mean of ~0.01% of all blood T cells per clonotype; max of 1%). TCR kinetics could be examined to determine estimated lifetime above background for 29 of these clonotypes, which indicated a median lifetime of approximately one year after priming and 6 years after a single boost, despite intervening chemotherapy treatment. Analysis of these clonotypes not only in blood, but also in other tissues (tumor, tumor-adjacent tissue, draining lymph nodes), indicated that greater than 97.5% were not detectable anywhere prior to vaccination. The single dose of anti-PD-L1 mainly expanded pre-existing T cells. At the ~3 year time point, 75% of CD8+ T cell responses were still polyfunctional upon ex vivo stimulation, and clinically, median RFS was still not reached. Two responders eventually recurred during the 3 year window. The first responder to recur demonstrated a rapid aggregate T cell clonotype contraction, even after vaccine boosting, and had the lowest estimated clonotype lifetime among all responders. The second recurrence occurred in a patient who showed the longest time after vaccine initiation to a detectable response. A randomized Phase II study is currently underway.
T cell exhaustion
Metabolic features of T cell exhaustion- Greg M. Delgoffe - University of Pittsburgh, Pittsburgh, PA, USA
T cells infiltrating a tumor encounter not only a plethora of cell types that can influence T cell activity, but also a very metabolically diverse and active environment with decreased glucose and increased lactic acid, hypoxia, and reactive oxygen species. At the same time, T cell activity is metabolically demanding. Greg Delgoffe has been looking at infiltrating T cells (TILs) through such a metabolic lens. TILs are characterized by loss of functional mitochondria across a range of tumor types, and loss of functioning mitochondria correlates with poor response to anti-PD-1 therapy. Loss of functional mitochondria also correlates with increased exhaustion, an epigenetic end-stage state of T cells. Exhausted T cells are, however, characterized by high-avidity TCRs, adaptation to the TME, and escape from apoptosis. Delgoffe asked whether such cells could be ideal soldiers in the T cell army if they could be properly stimulated, and looked to understand how metabolites influence T cell phenotype. Metabolites are transported into cells by (mostly) the SLC family of solute carriers, and by comparing expression levels in T cells from lymph nodes and tumors sorted by phenotype, Delgoffe found that a particular subfamily of SLC transporters was highly upregulated in exhausted CD8+ TILs. MCT11 (Slc16a11), a relatively uncharacterized monocarboxylate (e.g., lactic acid) transporter, was among the very top most differentially expressed genes in murine exhausted T cells, and was also expressed in human exhausted cells. Of note, lactic acid was the most abundant metabolite in the TME and was immunosuppressive at high levels. Delgoffe demonstrated enhanced transport of lactic acid into exhausted cells, but importantly, overexpression of MCT11 accelerated exhaustion of tumor-specific (OT-1) T cells in vivo. Conditional knockout (and concomitant expression of GFP) showed that MCT11-deficient cells (GFP+) no longer took up lactate, and more efficiently controlled transplanted tumor growth. Although the number of “exhausted’ cells did not decrease, these cells were now functional, as measured by cytokine production and target cell killing. As the next step, an antibody that blocks MCT11 function was developed, and in two tumor models (B16 melanoma and MEER HNSC), it showed antitumor activity as monotherapy, dependent on adaptive immunity, and synergized with anti-PD-1 in MC38. Further, anti-MCT11 has shown preliminary activity with CD19-CAR T cells in a humanized solid tumor mouse model. Addressing whether activity was due to blocking the function of MCT11 or depleting MCT11+ cells, an Fc-LALAPG variant of anti-MCT11 (which blocks binding to FcγR) was constructed and was shown to have even increased antitumor activity, indicating that activity was due to blocking the function of MCT11. Cured mice showed resistance to tumor rechallenge. Overall, these results suggest that modulation of the activity of metabolite transporters may restore the functional activity of exhausted T cells and more broadly modulate T cell activity across a spectrum of diseases.
Microbiome
The role of the intestinal microbiome in cancer immunotherapy- Marcel R.M. Van Den Brink - Memorial Sloan Kettering Cancer Center, New York, NY, USA
Marcel van den Brink began by noting the heightened attention the microbiome has received recently, contributing to the view of the “superorganism” comprised of 10 trillion human cells with 23,000 genes and an approximately 10x higher number of microorganisms with millions of genes in total, with each system impacting the other. Van den Brink’s research focus is on immune modulation of CAR T cell therapy and allogeneic hematopoietic stem cell transplant (HSCT), where gut toxicity is a major symptom of serious graft-vs-host disease (GvHD). Multiple features of the gut microbiota enhance the appearance of GvHD, including lower diversity, presence of enterococcus species early after transplant, and the use of broad-spectrum antibiotics. In contrast, the presence of particular anaerobes (Blutia) has been associated with reduced lethal GvHD. A very large and well controlled in-hospital diet study in patients who have received HSCT showed that sufficient calories and high fiber were positively associated with the higher diversity of the gut microbiome and presence of blutia, while being negatively correlated with presence of enterococcus. Looking at the “taxonomy” of food groups revealed that simple sugars (like those in energy drinks, which are recommended to HSCT patients with difficulties eating and drinking), especially together with antibiotics, led to low gut microbiome diversity and the presence of enterococcus. Cellulose is the most abundant fiber in food, which has multiple effects. Many microbes digest cellulose and release short-chain fatty acids, such as butyrate, which is known to ameliorate GvHD. In mice, a “sweet spot” of 12% dietary cellulose maximized survival in a GvHD model, raised butyrate levels, reduced gut barrier damage, enhanced Treg/Tconv ratios in gut mucosa, and decreased enterococcus levels. Analysis of dietary fiber in patients with and without GvHD supported the positive and negative association with bacterial taxa observed in mice. Similarly, butyrate levels in patient fecal samples in the first week after transplant – the period of initial T cell alloreactivity – showed an association between higher butyrate and protection from GvHD. Van den Brink then turned to the role of bile acids and their unexpected effect on T cells. Analysis of how various drugs used in the treatment of transplant patients affected the microbial patterns in the gut revealed a significant effect of the drug ursodiol, which is used to prevent the liver syndrome SOS, and consists of the secondary bile acid UDCA. UDCA and other secondary bile acids antagonize the activity of farnesoid X receptor (FXR; a ligand-activated transcription factor active in murine T cells) and thereby block T cell proliferation, stimulation of inflammatory molecules, and GvHD – all effects that were phenocopied in analysis of patient samples. Analysis of more than 700 patients showed that treatment with ursodiol was associated with reduced enterococcus levels and lower GvHD. Finally, van den Brink briefly summarized the important effects of the microbiome in CAR T cell therapy and provided data indicating that commercially available probiotics may contain an “anti-immune” composition, warning physicians to gather information about patient use of probiotics.
Bispecifics
Tumor-targeted costimulation via CD28 bispecific antibodies- Dimitris Skokos - Regeneron, New York, NY, USA
Dimitri Skokos and his colleagues at Regeneron propose the use of costimulatory bispecifics to convert tumor cells into antigen-presenting cells that can provide “signal 1” (activation of the TCR/CD3 complex) and “signal 2” (costimulation) by engaging CD28 receptors to T cells within the tumor microenvironment. In contrast to the CD28 superagonist (Tegenero) that demonstrated extreme toxicity in clinical studies in 2006, Regeneron’s CD28 costimulatory bispecific antibodies are only active in the presence of tumor-presented peptide:MHC (signal 1) and of a tumor-associated antigen (TAA) to serve as a target for the CD28 bispecific (signal 2). As a first example, Skokos presented data for a PSMAxCD28 antibody, Regn5678, which was combined with anti-PD-1 on the basis of findings of synergy in the induction of CD8-dependent antitumor immunity by tumor cells overexpressing the CD28 ligand CD86 and anti-PD-1 treatment. Further, the researchers found that PD-1 can exclude CD28 from the immune synapse, and that PD-1 inhibition circumvents this effect, resulting in enhanced accumulation of CD28 in the synapse. In mice, PSMAxCD28 combined with anti-PD-1 activated T cells, induced proinflammatory cytokines within the tumor, and promoted stronger tumor rejection than the corresponding monotherapies, without systemic cytokine release or activation of peripheral T cells. Anti-PD-1 alone expanded effector CD8+ T cells with high expression of coinhibitory receptors in the tumor, while the combination therapy expanded CD8+ TILs with a memory-like and less exhausted phenotype. Preliminary results from a phase 1/2 study of PSMAxCD28 and anti-PD-1 in patients with metastatic castration-resistant prostate cancer demonstrated a decline in PSA levels in 44% of patients at higher dose levels (30 to 300 mg), including 25% who showed deep or complete responses. For some responders, an association with immune adverse effects was observed, and the researchers are currently working on how to decouple these two effects. While PSMAxCD28 depends on pre-existing antitumor immunity for signal 1 via natural T cell recognition, the combination of a CD3 bispecific and a CD28 bispecific, each targeted to a distinct TAA, can induce de novo antitumor responses. The CD20XCD3 bispecific, odronextamab, has yielded clinical responses in phase 2 studies, but improvements can be made. Biopsy samples from a phase 1 trial with odronextamab showed that T cells infiltrated the tumor, that CD28 was preferentially expressed on CD8+ T cells at baseline, and that increases in CD8+ T cells after treatment were mostly driven by the CD28-expressing subpopulation. Combining increasing doses of CD20XCD3 with a CD22XCD28 bispecific in in vitro cytotoxicity assays showed enhanced cytotoxicity, proliferation, and cytokine secretion. In a DLBCL tumor xenograft model, the combination enhanced antitumor efficacy and survival, resulting in a durable survival rate of over 90%, while CD20XCD3 monotherapy led to a 0% survival rate. Finally, the combination was tested in cynomolgus monkeys, where it efficiently depleted peripheral B cells, and expanded and activated peripheral T cells, with no adverse events reported after repeated dosing. Various CD28 costimulatory bispecific antibodies as monotherapy or in combination with a CD3 bispecific antibody and/or anti-PD-1 are currently in clinical trials.
Write-up by Ute Burkhardt, Ed Fritsch, and Lauren Hitchings
