Authors: Erica L. Braverman (1Department of Immunology, University of Pittsburgh School of Medicine, Pittsburgh, PA; 2Department of Pediatrics, Division of Pediatric Hematology/Oncology, University of Pittsburgh School of Medicine, Pittsburgh, PA), Giuliana P. Mognol (3Moores Cancer Center, University of California, San Diego, La Jolla, CA), Andy J. Minn (4Institute for Immunology and Immune Health, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA; 5Abramson Family Cancer Research Institute, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA), Dario A.A. Vignali (1Department of Immunology, University of Pittsburgh School of Medicine, Pittsburgh, PA; 6Tumor Microenvironment Center, UPMC Hillman Cancer Center, Pittsburgh, PA; 7Cancer Immunology and Immunotherapy Program, UPMC Hillman Cancer Center, Pittsburgh, PA), Judith A. Varner (3Moores Cancer Center, University of California, San Diego, La Jolla, CA; 8Department of Pathology, University of California, San Diego, La Jolla, CA)
Categories: Article
Source: American Society of Clinical Oncology educational book. American Society of Clinical Oncology. Annual Meeting
Authors: Erica L. Braverman, Giuliana P. Mognol, Andy J. Minn, Dario A.A. Vignali, Judith A. Varner
Immune checkpoint inhibitors are cancer therapeutics that have shown remarkable success in extending lives in many cancers, including melanoma, MSI-high cancers, and other cancers. However, these therapeutics have not shown benefit for many patients with cancer, especially those with advanced cancer diagnoses. In addition, many patients develop resistance to these therapeutics and/or life-altering adverse events that can include cardiotoxicity, pneumonitis, thyroiditis, pancreatitis, and hepatitis. Extensive efforts to improve cancer care by uncovering mechanisms of resistance to immune therapy in solid tumors have led to identification of new sources of resistance and to the development of new approaches to activate or sustain antitumor immunity. Chronic stimulation of T cells by tumors and by checkpoint inhibitors can lead to a progressive state of T-cell exhaustion. Chronic T-cell activation by the tumor microenvironment (TME) or immune therapeutics can upregulate the expression and function of alternate checkpoints, including the T-cell protein LAG-3. Persistent interferon signaling in the TME can drive epigenetic changes in cancer cells that enable tumors to counter immune activation and disrupt tumor cell elimination. In addition, immune-suppressive macrophages can flood tumors in response to signals from dying tumor cells, further preventing effective immune responses. New clinical developments and/or approvals for therapies that target alternate immune checkpoints, such as the T-cell checkpoint LAG-3; myeloid cell proteins, such as the kinase phosphoinositide 3-kinase gamma isoform; and chronic interferon signaling, such as Jak 1 inhibitors, have been approved for cancer care or shown promise in recent clinical trials.
In 2018, the Nobel Prize for Medicine was awarded to James Allison and Tasuko Honjo for identification of novel strategies to promote antitumor immunity. These investigators made fundamental observations that proteins expressed on the surface of chronically stimulated T cells, PD-1 and CTLA-4, suppressed T-cell functions, thereby rendering T cells exhausted.^1–4^ Antibody inhibitors of PD-1 and CTLA-4, nivolumab (nivo)/pembrolizumab and ipilimumab, respectively, have shown remarkable success as cancer therapeutics and are approved widely for treatment of patients with cancer.^5,6^ A 10-year study of patients with melanoma treated with nivo plus ipilimumab versus either antibody alone recently showed that 43% of dual antibody-treated patients with melanoma survived a median of 10 years.^7^ Additional inhibitory immune checkpoints LAG-3, TIGIT, TIM3, and GITR and activating checkpoints ICOS, 4–1BB, OX40, and CD40 have since been identified and therapeutically targeted.^8^ While showing remarkable success in Hodgkin’s lymphoma, melanoma, microsatellite instability (MSI)-high cancers, and some other cancers, immune checkpoint blockade (ICB) therapeutics have not yet shown benefit for many patients with cancer, especially those with advanced cancer diagnoses.^9–11^ In addition, many patients eventually develop resistance to ICB and/or life-altering adverse events that can include cardiotoxicity, pneumonitis, thyroiditis, pancreatitis, and hepatitis.^12^
The development and commercialization of chimeric antigen receptor T (CAR-T) cells has also revolutionized cancer care. CAR-T cells are autologous T cells transduced to express engineered cell surface receptors for tumor cells with intracellular motifs designed to activate T-cell proliferation and cytotoxicity.^13^ CAR-T-cell therapy has achieved notable success in some hematologic cancers, including acute lymphoblastic leukemia, non-Hodgkin’s lymphoma, and multiple myeloma. Although a revolutionary development, CAR-T-cell therapy has not yet achieved clinical approval for treatment of solid tumors.^14^ Lack of persistence or penetrance in the tumor microenvironment (TME) and toxicity are the main causes for failure of solid tumor CAR-T therapy.^14^ Improvements in CAR-T construction include enhancing the affinity of the tumor cell–targeting receptor, adding elements to improve trafficking to the TME, and adding signaling domains to maintain T-cell activation through autocrine IL12 expression^15^ Combination of CAR-T therapy with other immune therapeutic approaches promises to sustain efficacy for the success of this novel therapeutic approach. Additional approaches to restore antitumor immunity include novel engineered bispecific antibodies (BsAbs) that bind to CD3 on T cells and a tumor cell–specific antigen BiTEs.^16^ Extensive efforts to uncover mechanisms of resistance to immune therapy in solid tumors have led to new clinical developments and/or approvals for therapies discussed in this article.
Primary or acquired resistance to immune therapy can arise because of treatment selection pressures that lead to genetic changes in tumor cells; wound healing responses of the microenvironment to therapy-induced damage that can result in increased blood vessel development (angiogenesis), macrophage accumulation, and deposition of extracellular matrix that resembles granulomas; and upregulation of mechanisms in the immune system that prevent over-responsiveness to stimuli. Antigen presentation by tumor cell major histocompatibility complex (MHC) I is essential to enable T cells to target and kill tumor cells. A high tumor mutational burden, which can generate numerous neoantigens, has been associated with responsiveness to anti–PD-1 therapy.^17^ Heterozygosity of MHC I protein (MHC 1) loci is also associated with enhanced responsiveness to checkpoint inhibitors.^18^ Loss of antigen presentation in tumors caused by somatic loss of heterozygosity at MHC-I loci,^19^ loss of β2 macroglobulin expression on tumor cells,^20^ and loss of responsiveness to interferon gamma (IFNγ) because of inactivating or deletion mutations of JAK1, JAK2, or other IFNγ response genes can promote resistance to therapy.^20^ Therapeutic resistance can also occur by immunoediting, when activated T cells eliminate tumor cells bearing recognized neoantigens, leaving behind viable tumor cells that are no longer recognized by host T cells.^21^ In some tumors, persistent interferon production by immune cells and chronic interferon signaling in tumor cells lead to T-cell exhaustion and resistance to immune therapy.^22^ Inhibition of JAK1 signaling can prevent T-cell exhaustion and break resistance to checkpoint inhibitors in animal models of cancer.^23^ Accordingly, the JAK1 inhibitor itacitinib promoted T-cell differentiation toward nonexhausted states when added to anti-PD-1 in patients with non–small cell lung cancer (NSCLC) who had previously shown resistance to anti–PD-1 therapy.^23^ In addition, the Jak1/2 inhibitor ruxolitinib reduced neutrophil/lymphocyte ratios and immune-suppressive myeloid cell content and enhanced the efficacy of anti–PD-1 therapy in animal models of solid and hematologic tumors.^24^ Ruxolitinib combined with nivolumab also improved response rates in patients with Hodgkin lymphoma who were previously refractory to anti-PD-1.^24^ These paradoxical findings resemble those observed in studies of chronic viral infection; limited interferon signaling is required for effective immune responses, but chronic interferon signaling suppresses T-cell function because of upregulation of a host of immune-suppressive interferon–stimulated genes (ISGs), including immune checkpoints. Thus, mutations and changes in the TME that interfere with effective immune activation or persistent signaling from ineffective immune responses can each promote immunotherapy resistance.
Chronic T-cell activation by the TME or ICB therapy can upregulate the expression and function of a variety of immune checkpoint molecules, including PD-1, PD-L1, CTLA-4, LAG-3, TIM-3, TIGIT,^25^ and others. Chronic stimulation of T cells leads to a progressive state of T-cell exhaustion.^26^ Immune checkpoint inhibitors (ICIs) target coinhibitory molecules present on the surface of exhausted T cells, restoring T-cell function^27^ and/or preventing the exhaustion of new T-cell clones that are recruited to the tumor.^28^ The US Food and Drug Administration (FDA) has approved several ICIs, including monoclonal antibodies against PD-1 (nivo, pembrolizumab, and cemiplimab), CTLA-4 (ipilimumab), and LAG-3 (relatlimab), whereas additional inhibitors of TIM-3 and TIGIT are currently undergoing clinical trials (Table 1). Despite the notable clinical achievements with these inhibitors, some patients still do not exhibit a positive response, and others develop resistance mechanisms to even combinatorial ICIs.
Three key mechanisms may be most significant in generating resistance to targeted immune dysregulation of immune responses and chronic inflammatory signaling leading to expression of alternative immune checkpoints; immune-suppressive environments created by macrophages, neutrophils, or regulatory B cells^53,54^; and a profibrotic microenvironment driven by TGFβ signaling that inhibits T-cell migration, proliferation, and survival.^55^ Tumor microenvironments have been classified into three broad categories based on the presence and localization of T T-cell–inflamed, in which abundant CD8^+^ T cells are closely associated with tumor cells; T-cell–excluded, in which T cells are present in larger numbers but are unable to approach tumor cells; and T-cell deserts, in which no T cells are present.^56^ Of these subsets, T-cell–inflamed tumors are the most response to ICB therapy.^56^ Consequently, there is a need to understand key signals and cellular features that maintain suppressive TMEs to develop innovative strategies to enhance patient response and survival (Table 2).
Immunosuppressive myeloid cells, which can include macrophages, dendritic cells, and immune-suppressive monocytes and granulocytes (sometimes called myeloid-derived suppressor cells [MDSCs]), play significant roles in promoting resistance to immune therapy within the TME.^57^ Tumors are often characterized by large increases in immune-suppressive CD206+ macrophages or granulocytes when compared with normal tissues.^54^ Immunosuppressive myeloid cells can enhance the stemness, survival, and proliferation of cancer cells through the secretion of cytokines such as insulin-like growth factor, transforming growth factor β, and interleukin 6 family member; facilitate angiogenesis through the production of vascular endothelial growth factors (VEGFs), platelet derived growth factors, and other factors; stimulate fibrotic responses through direct and indirect regulation of extracellular matrix protein secretion; and suppress innate and adaptive immune responses.^54^ Tumor-associated macrophages promote resistance to immune checkpoint inhibitors as they express checkpoint ligands, such as PD-L1, PD-L2, CD80, CD86, and CD155, a TIGIT ligand.^58^ However, macrophages are versatile, exhibit tissue-specific differences, and can adopt proinflammatory phenotypes, whereupon they can stimulate adaptive immune responses and eliminate tumor cells.^59^ Attempts to modulate the trafficking and gene expression of these populations to enhance immunotherapy are currently underway.^54^ Strategies to target myeloid cell– mediated immune suppression clinically have included inhibitors of the macrophage growth factor receptor CSF1R,^42^ myeloid cell trafficking receptors such as CCR2^60^ and^39^ CXCR2,^39^ macrophage signaling receptors MERTK/AXL,^49^ regulators of lipid processing, such as TREM2,^52^ and myeloid cell–specific kinases such as phosphoinositide 3-kinase gamma isoform (PI3Kγ).^61^ While none of these approaches have yet been approved for widespread use in cancer, their examination in combination therapies with ICB and targeted therapies is actively under investigation, and current studies suggest that these novel therapeutic approaches will provide clinical benefit to a wide range of patients with cancer.
One therapeutic approach that has shown promise in animal models of cancer and in early clinical trials is the use of selective inhibitors of PI3Kγ. The PI3K family of lipid kinases phosphorylates phosphatidylinositol in response to growth factor or G protein–coupled receptor (GPCR) signaling, thereby playing crucial roles in regulating receptor signaling leading to cell proliferation, migration, vascular permeability, and metabolic changes and immune responses.^62^ The class 1A PI3K isoform alpha (PIK3CA) has been identified as a frequently mutated driver of cancer progression that also regulates glucose metabolism.^62^ For this reason, inhibitors of this PI3K isoform have proven to be too toxic for broad use in cancer therapy.^62^ However, among the four class I PI3Ks, the class 1B PI3Kγ isoform emerged from studies at the University of California, San Diego, as a promising target for myeloid cell–based cancer immunotherapy, with evaluations in several clinical trials demonstrating safety and survival benefits in phase II clinical trials of this therapeutic approach.^44–48,63–66^ PI3Kγ is predominantly expressed in myeloid cells and endothelial cells, where it facilitates the trafficking and infiltration of monocytes and granulocytes into tissues and promotes vascular leak.^48,65^ PI3Kγ is activated by cytokines in the TME that are ligands for myeloid cell surface receptors, such as CSF1R, VEGFR, IL6 receptor family members, toll-like receptors, and GPCRs.^48^ By controlling a cascade of signal transduction events that include rapid activation of BTK, PLCγ, Rap1a, RIAM, talin, paxillin, and myosin light chain kinase, PI3Kγ orchestrates changes in the ordered structure of integrin cytoplasmic tails to promote integrin unfolding and activation, leading to myeloid cell adhesion to endothelium and extravasation of myeloid cells, including MDSCs, into tumors.^48,66^ Inhibition of PI3Kγ and other components of this signaling cascade blocks immune-suppressive myeloid cell trafficking to tumors and thereby reduces tumor growth.^48,66^ Thus, PI3Kγ plays an essential role in tumor immune suppression by promoting myeloid cell trafficking to tumor tissues.
Tumor-associated macrophages are characterized by fatty acid oxidation, which, in turn, elevates PI3Kγ expression, thereby fostering an immune-suppressive macrophage phenotype that promotes weight gain, insulin resistance, and immune suppression.^67^ PI3Kγ controls immune-suppressive myeloid cell polarization by inhibiting NFκB activation and promoting mammalian target of rapamycin/S6 kinase activation, leading to C/CAAT enC/CAAT enhancer binding protein beta-dependent transcription and inhibition of NFκB-dependent transcription, thereby regulating expression of immune-suppressive factors, such as arginase,^63^ and preventing expression of IL12, tumor necrosis factor alpha (TNFα), and other proinflammatory factors.^63^ However, PI3Kγ inhibition reprograms myeloid cells to a proinflammatory phenotype that recruits and activates CD8^+^ T cells and NK cells and promotes tumor regression mediated by cytotoxic T cells in mouse models of cancer.^63^ Inhibition of this kinase reduces myeloid cell trafficking to tumors and changes the microenvironments of some tumors from cold to hot.
PI3Kγ inhibition promoted tumor regression and long-term survival and suppressed resistance to ICB in mouse models of head and neck cancer, pancreatic cancer, and melanoma.^44,63,64^ PI3Kγ inhibitors were most effective when administered concurrently with checkpoint inhibitors in mouse models. The clinical PI3Kγ inhibitor, eganelisib, enhanced responses to nivo in patients with nivo-resistant melanoma and head and neck carcinoma in the phase I trial MARIO1 and enhanced survival when combined with nivo in patients with platinum-refractory urothelial cancer in the phase II trial MARIO275.^45–47^ When combined with atezolizumab and nab-paclitaxel in patients with triple-negative breast cancer in the phase II trial MARIO-3, eganelisib decreased circulating monocytic MDSCs and increased circulating Ki67+ memory T cells. Treatment with eganelisib, atezolizumab, and nab-paclitaxel increased human leukocyte antigen-DR+ macrophages and granzyme+ CD8^+^ T cells in tumors. This therapy also altered the TME by decreasing expression of epithelial mesenchymal transformation genes, such as SPP1, PDGFs, integrins, and matrix metalloproteins while increasing expression of MHC class I and MHC class II antigen-presentation genes, IFNγ signaling genes, and biomarkers of T-cell activation.^45–47^ Thus, PI3Kγ inhibitors show promise as effective immune-oncology agents that overcome resistance to ICB in cancer therapy.
Spatial transcriptomics and immune profiling of squamous cell carcinomas revealed that tumors are characterized by the gradual development of an extensive tumor cell–cancer associated fibroblast (CAF) barrier characterized by SPP1+ macrophage-CAF interactions, the presence of a dense fibrillar collagen network, and isolation of T cells distant from tumor cells (T-cell exclusion).^68^ The presence of this dense collagen network is associated with a poor prognosis. Macrophages participate in the formation of this dense matrix by expressing proteins that crosslink collagens, fibronectin, and other extracellular matrix components into dense fibrils that appear to entrap T cells. Interestingly, PI3Kγ inhibition strongly suppressed tumor desmoplasia/fibrosis and prevented T-cell exclusion in mouse models of cancer.^63^ Notably, PI3Kγ inhibition reduced expression of the gene for osteopontin, SPP1 (secreted phosphoprotein 1), and upregulated expression of the chemokines CXCL9 and CXCL10 in triple-negative breast cancer tumors during recent clinical trials. A low SPP1/CXCL9 ratio has been shown to be prognostic of good outcomes from ICB therapy.^69^
Some novel cell therapeutic approaches take advantage of unique properties of myeloid chimeric antigen receptor macrophages (CAR-Ms), which are presently undergoing clinical evaluation, are genetically engineered autologous monocyte/macrophage constructs that express CAR with phagocytic and proinflammatory signaling capabilities.^70^ First-generation CAR-Ms were derived from monocytes or macrophages and expressed chimeric antigen constructs targeting tumor cell antigens. These constructs consisted of a single-chain variable region fragment derived from an antibody targeting a tumor cell antigen (eg, EGFR), a hinge domain, a transmembrane domain, and a cytosolic signaling domain such as CD3ζ or Fc gamma receptor.^70^ These CAR-Ms were capable of promoting tumor cell phagocytosis but were susceptible to reprogramming by the tumor microenvironment and did not have durable impacts on tumor growth.^70^ Second-generation CAR-Ms derived from inducible pluripotent stem cells and transduced to express CAR constructs containing a toll/like interleukin receptor signaling domain of toll-like receptor could sustainably activate NFκB and proinflammatory transcription and immune stimulation.^71^ Sustained inhibition of nuclear factor-erythroid factor 2-related factor 2 also maintained CAR-M activity.^72^ These approaches have shown success in animal studies and have proven to be safe in early-phase clinical trials. CAR-M therapy has also demonstrated synergy with anti-PD-1 and anti-CD47 therapy.^70^ It is interesting to speculate that CAR-Ms and CAR-T cells might 1 day be used together, or their functionalities merged for improved cellular therapeutic approaches. Combinatorial therapies targeting myeloid cells and T cells offer the potential to enhance and synergize with current therapies (Fig 1).
Exhausted T cells in the TME chronically express multiple inhibitory receptors (IRs) on their surface, including PD-1, LAG-3, TIM3, CTLA-4, and TIGIT.^73^ Combinatorial strategies to reinvigorate T cells via blockade of multiple IRs have shown increased benefits in the clinic, with the most common being the combination of nivo (anti-PD-1) and ipilimumab (anti-CTLA4).^29^ While this regimen leads to enhanced OS in metastatic melanoma, this combination also leads to a higher incidence of severe immune-related adverse events (irAEs). Therefore, investigators are actively searching for IR combinations to enhance antitumor immunity without causing intolerable side effects (Fig 2).
LAG-3 was originally characterized in the 1990s by Frederic Triebel as a marker of T- and NK cell activation closely related in structure to CD4.^74^ Interest in LAG-3 rose when tumor-infiltrating CD4^+^ and CD8^+^ T cells were noted to coexpress both PD-1 and LAG-3 across multiple tumor models. When these IRs were targeted together, they elicited greater tumor clearance in a preclinical model than either IR alone,^75^ creating the rationale to evaluate anti-LAG-3 in the clinic both alone and in combination with other therapies. There are currently 22 monospecific and bispecific LAG-3-targeted antibody drug candidates in clinical trials (Fig 3).^31^ LAG-3 is a type I cell surface molecule induced on TCR stimulation. When LAG-3 binds its ligands, it inhibits proximal TCR signaling, leading to reduced expansion and effector function of the expressing T cells. The intracellular EP motif of LAG-3 is known to lower the pH near CD4/CD8, sequestering Zn^2+^ and leading to Lck dissociation,^76^ which can occur even in the absence of ligand binding, termed tonic LAG-3 signaling. Recently, the intracellular KIEELE and FSALE motifs were implicated in increasing LAG-3 inhibitory function in the setting of ligand binding.^77^ LAG-3 undergoes rapid ubiquitination after ligand binding at the KIEELE motif, releasing the FSALE motif from the cell membrane and enabling its inhibitory function. The roles of other intracellular domains are an ongoing focus of current research.^32^
Two different strategies to target LAG-3 function have been identified although only one has made its way to the clinic. The first is to block the LAG-3 ligand-binding domain to prevent interaction between LAG-3 and its most well-studied ligand, major histocompatibility complex II (MHCII).^78^ LAG-3 binds to MHCII at the same sites as CD4 but with approximately 1,000-fold greater affinity, suggesting that part of LAG-3 function may be competitive MHCII binding with CD4.^79^ Disrupting this interaction improves tumor clearance in murine models and humans.^33,80^ There are other LAG-3 ligands although no antibodies to block these interactions are in the clinic. These include fibrinogen-like protein 1 (FLG1), which is expressed by certain cancers including hepatocellular carcinoma (HCC) and NSCLC.^81–83^ Blocking the FLG1-LAG-3 interaction promoted antitumor immunity in a murine colorectal cancer model.^81^ There are other proposed though less well-characterized ligands for LAG-3, including the TCR-CD3 complex, LSECtin, and galectin-3.^76^
The second strategy to block LAG-3 function, which has not yet been tested in the clinic, targets the need for LAG-3 to homodimerize to bind its ligands and signal downstream.^84^ Mutations in the LAG-3 domain required for homodimerization prevented tracking to the TCR/CD3 complex within the immune synapse in murine CD8^+^ T cells, blocking LAG-3 signaling and improving antitumor function in a murine melanoma model.^85^ While these are promising preclinical data, as noted above, the efficacy of mAbs directed at blocking LAG-3 homodimerization has not yet been assessed in clinical trials.
Early studies in multiple murine tumor models demonstrated synergy between anti-PD-1 and anti-LAG-3.^78^ Importantly, recent work in mice has shown that CD8^+^ T cells lacking both PD-1 and LAG-3 demonstrated significantly improved antitumor activity compared with either IR deleted alone.^86^ These double deficient cells were found to be transcriptionally distinct from the cells with single deletions, again highlighting synergistic effects of dual blockade.^78^ PD-1 deletion alone increased proliferation but led to more exhausted T cells within the TME, whereas LAG-3 deletion alone led to reduced expression of TOX, a transcription factor that is critical for the development of T-cell exhaustion. The knockout (KO) of both ligands together in tumor antigen-specific T cells increased the proportion of self-renewing stem-like CD8^+^ T cells and created more polyfunctional effectors. Dual KO CD8^+^ T cells were able to secrete more IFNγ and TNFα than either single KO or wild-type CD8^+^ T cells. Importantly, while IFNγ can have significant antitumor activity on its own, the increased IFNγ produced by the dual KO T cells was critical for boosting the antitumor T-cell response cell intrinsically. Deleting the IFNγ receptor from the tumor or the host did not limit the efficacy of dual KO CD8^+^ T cells, but loss of the IFNγ receptor on the CD8^+^ T cells abrogated the improved antitumor efficacy of the dual-KO cells. Therefore, the combination of PD-1 and LAG-3 engagement on antitumor T cells hinders autocrine IFNγ-dependent antitumor activity of CD8^+^ T cells in mice, shedding light on the increased benefit of dual blockade. Indeed, these data are in line with observations from human studies demonstrating increased IFNγ signature in T-cell clones expanded after combination therapy.^87^
PD-1/LAG-3 synergy has been additionally studied in chronic viral infection models, in which persistent viral antigen mimics the T-cell exhaustion seen within the TME.^85,88^ In this setting, PD-1 was again found to be primarily responsible for limiting proliferative capacity, whereas LAG-3 was more involved with hampering cytotoxic function. LAG-3 was similarly found to sustain the expression of TOX in chronically stimulated cells, while also repressing expression of NKG2C, an activating receptor on CD8^+^ T cells. While PD-1 KO alone led to higher expression of additional IRs, the addition of LAG-3 KO reduced overall IR expression, again suggesting a functional benefit to the dual blockade. Furthermore, higher expression of activating NK receptors improved functionality of dual-blockade cells. Importantly, the alteration in activating/inhibitory NK receptors seen in the murine model was consistent with studies performed on the peripheral blood of patients with melanoma after dual anti-PD-1/anti-LAG-3 therapy.
To assess the role of PD-1/LAG-3 synergy in humans, a randomized phase II trial was designed in which patients with advanced melanoma received nivo (anti-PD-1) alone, relatlimab (rela; anti-LAG-3) alone, or both rela and nivo therapy for a 4-week lead-in phase, followed by combination therapy with rela and nivo for all patients (ClinicalTrials.gov NCT03743766).^87^ Single-cell transcriptomic analysis of CD8^+^ T cells from patients initially treated with rela/nivo demonstrated increased T cell receptor signaling, cytotoxic and effector gene modules, and, surprisingly, retention of exhaustion gene modules, in contrast to the transcriptional signature of blockade of either IR alone. Collectively, these observations highlight parallels between the mechanisms that lead to increased CD8^+^ T-cell activation after dual PD-1/LAG-3 deletion in murine models compared with dual blockade in humans.^86,89^
There are currently over 40 clinical trials evaluating 22 anti-LAG-3 therapeutics. Relatlimab from Bristol Myers Squibb (BMS) was the first anti-LAG-3 therapeutic to be approved by the FDA following the benefits observed in the RELATIVITY-047 trial.^33^ In this study, patients received the fixed-dose combination opdualag (480 mg of nivo and 160 mg of rela) once every 4 weeks. The median progression-free survival in patients with previously untreated metastatic or unresectable melanoma was 10.1 months (95% CI, 6.4 to 15.7), which was more than double that seen with nivo alone at 4.6 months (95% CI, 3.4 to 5.6; hazard ratio, 0.75 [95% CI, 0.62 to 0.92]; P = .0055).^33^ Impressively, this improvement in PFS was not associated with a worsening safety profile, as was seen with the ipilimumab/nivo combination. This highlighted that combinatorial strategies could be synergistic without amplifying off-target effects and has led to additional trials testing this combination in other tumor types, including Hodgkin and non-Hodgkin lymphoma (ClinicalTrials.gov NCT05255601), HCC (ClinicalTrials.gov NCT04658147), and glioblastoma (ClinicalTrials.gov NCT06325683).
Another phase I trial by Regeneron Pharmaceuticals and Sanofi (ClinicalTrials.gov NCT03005782) is testing their anti-LAG-3 mAB fianlimab in combination with the anti-PD-1 mAB cemiplimab. Importantly, this trial significantly increased the dose of the fianlimab (1600 mg once every 3 weeks [20 mg/kg per 80 kg person]) over the dose of rela used in the RELATIVITY-047 trial (160 mg once every 4 weeks).^33,36^ This difference in formulation reflects the ongoing debate on the optimal dosing of LAG-3 therapeutics. There is concern that the interaction of LAG-3 antibodies with soluble LAG-3 (sLAG-3) can limit their efficacy by acting as sink that limits on-target blockade. LAG-3 is generated when LAG-3 is cleaved from the surface of immune cells, including T cells and dendritic cells, by a disintegrin and metalloproteinase domain-containing (ADAM) proteins, ADAM10 and ADAM17.^90^ There is currently no known function of sLAG-3—it is believed, rather, that cleavage of LAG-3 from the surface of T cells is a rapid mechanism to limit LAG-3 signaling. In line with this, a high LAG-3/ADAM10 ratio on conventional CD4^+^ T cells from melanoma and head and neck squamous cell carcinoma (HNSCC) correlated with poor prognosis.^91^ Regardless of its function, there remains concern that sLAG-3 interferes with anti-LAG-3 therapeutics, prompting the tenfold increase in dose seen in this clinical trial. In combination with cemiplimab (350 mg) administered once every 3 weeks over 12 months, therapy elicited an overall response rate (ORR) of 61.2%, with 12 complete responses (CRs) and 48 partial responses (PRs) of a total of 98 patients.^36^ The Kaplan-Meier estimation of median progression free survival was 15.3 (95% CI, 9.4 to not estimated) months.^36^ While the safety profile of combination therapy was similar to that of cemiplimab alone, they did note higher incidences of adrenal insufficiency (12% of patients), which were grade ≥2 in 64% of cases. There were additionally adverse events reported in 94% of patients, with 44% being grade ≥3. This combination is currently in phase III clinical trial (ClinicalTrials.gov NCT05352672), wherein two doses of fianlimab (1,600 mg v 400 mg once every 3 weeks) are being tested. Furthermore, another clinical trial (ClinicalTrials.gov NCT06246916) will test opdualag against the fixed-dose combination of fianlimab and cemiplimab head-to-head.
In addition to combination with PD-1, anti-LAG-3 is also being tested in combination with other IR blockade, including CTLA-4, TIM3, and TIGIT, and in combination with anti-VEGF and anti-ICOS monoclonal antibodies. While successes have been observed in melanoma, BMS recently terminated its phase III RELATIVITY-123 trial of opdualag in patients with previously treated, microsatellite-stable metastatic colorectal cancer because of low likelihood of meeting the primary OS end point. This result further highlights the importance of understanding how anti-PD-1/anti-LAG-3 combination therapy works to better identify those cancers that would benefit.
Bispecific antibody therapy, wherein antibodies are generated to recognize more than one IR, is also gaining popularity to more specifically target exhausted T cells by recognizing IRs together on the same cell or between T cells and target cells. There has also been suggestion that BsAbs can sterically exclude IRs from the immune synapse or accelerate their shedding, thereby blocking their inhibitory function.^31^ This mechanism has been proposed for the first BsAb to enter the clinic, FS118. FS118 is a first-in-class tetravalent BsAb against LAG-3 and PD-L1. To create this BsAb, a LAG-3 binding site was introduced into the antibody constant region of a full-length anti–PD-L1 human mAb, with two additional mutations to abrogate Fc receptor binding.^37^ The phase I trial observed no dose-limiting toxicities and demonstrated a disease control rate of 54.8% in patients with advanced cancer and PD-L1 resistance. Phase II/III trials are currently ongoing. There is also a trispecific antibody GB266T targeting PD-L1, TIGIT, and LAG-3 alongside a competent Fc domain. The goal of this combination is to increase interactions between tumor cells, T cells, and APCs. Preclinical testing in humanized non-obese diabetic scid gamma mice implanted with an RKO OKT3 human colon cancer cell line demonstrated significant tumor growth suppression with GB266T treatment,^92^ suggesting the utility of this strategy.
There are now a total of five BsAbs currently under evaluation. One of these, tebotelimab (MGD013), is a tetravalent anti-LAG-3/PD-1 BsAb developed by MacroGenics. As monotherapy against solid tumors in a phase I trial (ClinicalTrials.gov NCT03219268), irAEs occurred in 68.4% of patients (184 of 269), with 22.3% (60 of 269) being ≥grade 3. However, when tested in combination with an Fc-optimized anti-HER2 antibody, margetuximab, in patients with HER2+ tumors, only 16.7% of patients (14 of 84) developed irAEs of grade ≥3, with an ORR of 19.4% and a median duration of response of 16.7 months (95% CI, 11.04 to NE), suggesting that combination with tumor-directed therapy could reduce off-target effects of IR blockade. In addition, they observed one CR (1.4%), 13 PRs (18.1%), 25 stable diseases (34.7%), and 33 (45.8%) progressive diseases.^38^ With these promising results, the phase II/III Mahogany trial (ClinicalTrials.gov NCT04082364) is now ongoing.
Many questions remain to be answered about LAG-3 targeting, including optimal dosing, optimal combination strategies, and, perhaps most importantly, how to determine which patients/cancers will be most responsive to LAG-3 blockade. There are currently no definitive biomarkers associated with response to LAG-3 treatment, limiting our ability to choose those patients/cancers who would benefit most from targeting this IR. Indeed, the RELATIVITY-047 trial saw no correlation between either LAG-3 expression or PD-L1 expression and response to therapy.^33^ There has been some research looking at levels of sLAG-3 in the setting of checkpoint blockade, finding a negative correlation between sLAG-3 serum levels and therapy response, but this analysis did not include patients who had received anti-LAG-3 antibodies.^93^ Interestingly, a negative correlation between ICB response and the ratio of LAG-3 to ADAM10 on the surface of CD4^+^ T cells has been observed, but again, this was in the setting of anti-PD-1 alone.^86^ Finding correlates with response to anti-LAG-3 blockade is an important area of active research. There were data presented at ASCO 2024 from ClinicalTrials.gov (identifier: NCT02519322), a phase II trial testing checkpoint blockade in the neoadjuvant setting before surgery in patients with stage III, surgically resectable melanoma, evaluating 30 longitudinal patient samples before and after treatment with rela and nivo using RNAseq. Higher gene expression signatures (GES) for B cells, CD45, CD-8+ T cells, TIGIT, indoleamine 2,3-dioxygenase 1, and IFNγ were found in baseline samples from patients with major pathologic response (MPR), whereas there was higher GES for the immune checkpoint B7-H3 in non-MPR.^34^ In addition, an on-therapy rela + nivo gene signature that persisted in CD8^+^ T cells 1 month after therapy and correlated with improved responses was identified in ClinicalTrials.gov (identifier: NCT03743766).^87^ This signature also correlated with an increased percentage of CD38+Tim3+CD8^+^ T cells in peripheral blood, suggesting that coexpression of these markers could be used as an on-therapy biomarker of combination therapy. Recently, data were published from ClinicalTrials.gov (identifier: NCT04080804), a phase II neoadjuvant trial comparing nivo alone, nivo + ipilimumab, and nivo + rela in patients with HNSCC.^94^ This analysis demonstrated that patients with combined LAG-3 and PD-L1 positivity (>1%) from baseline tissue were more likely to have a MPR to the combination of nivo + rela, with no correlation found in either of the other arms. This is in contrast to data from patients with melanoma, where there was no association between LAG-3 and response to combination therapy.^33^ Furthermore, in HNSCC, the expression of LAG3 mRNA in baseline CD8^+^ TIL positively correlated with pathologic response. These seemingly conflicting results may reflect differences in the TME found in different cancer types. Analyses from the abovementioned trials and currently enrolling trials are ongoing, with further validation of these findings required.
It is also still not fully understood how LAG-3 synergizes with PD-1 and if LAG-3 could similarly synergize with other IRs. Indeed, analysis of additional data from ClinicalTrials.gov (identifier: NCT04080804) revealed distinct transcriptional states associated with response to the nivo + ipilimumab and the nivo + rela arms, wherein the former expanded effector CD8^+^ T cells, whereas the latter preferentially unleashed exhausted CD8^+^ T cells.^94^ The fact that these combinations seemingly target different CD8^+^ T-cell populations suggests that further combining these three treatments into triple therapy may lead to even greater responses across a wider variety of patients although toxicity profiles would need to be closely monitored in this setting. Either way, understanding more about the mechanism underlying combination therapy synergy and the changes that occur within the TME in patients who either respond or are resistant to anti-LAG-3 will help to elucidate drivers of ICB escape. With multiple mechanisms driving immunosuppression and reducing T-cell infiltration and function in the TME, it is likely that in cases where ICB fails, other immunosuppressive mechanisms have overwhelmed the ICB-mediated release of antitumor T cells. Identifying these inhibitory adaptations will permit the design of rational combination therapies to simultaneously target multiple immune-limiting mechanisms at once, creating treatment protocols that will make ICB more broadly effective for a higher number of patients. Ongoing research is seeking to answer these important questions, the results of which could help to expand the utility of anti-LAG-3 blockade in the clinic.
Several suppressive cellular features (eg, immunosuppressive myeloid cells, exhausted CD8 T cells), signaling pathways (eg, interferon, granulocyte colony-stimulating factor [G-CSF]), and immune checkpoint pathways (eg, PD-1/PDL1 and LAG-3) observed in tumors poorly responsive to immunotherapy likely evolved as a consequence of persistent, yet ineffective, antitumor immune responses. In normal tissues, persistent activation of the immune response can activate immune-suppressive pathways to avoid immune-mediated pathology.^95,96^ For example, in mice with chronic virus infection, deletion of PD-1 unleashes an aggressive immune response that results in tissue damage, immunopathology, and early mortality.^97,98^ Acting as wounds that never heal, tumors exploit these normal immunoregulatory mechanisms to similarly antagonize immune activation and interfere with immunotherapy efficacy.^99^ Moreover, key cytokine signaling pathways that help to initiate or maintain these immunomodulatory effects of chronic inflammation seen in normal tissue and cancers can play very different roles during the early stages of an immune response, making context-dependent effects important to understand.^100^ One such context-dependent cytokine pathway is interferon.^101^ Although type I interferon (IFN-I) and type II interferon (IFNG) are critical for immune activation, both interferons have important functions in maintaining immune suppression to protect tissues from chronic inflammation. However, because of the opposing roles of interferons in immune activation versus immune suppression, the extent to which interferon is responsible for cancer immunotherapy resistance and the feasibility of targeting it has been unclear.
Predictive GESs that contain ISGs are robustly associated with a higher likelihood of response to anti-PD-1 across many human cancer types.^102^ Tumors lacking high expression of these ISG-containing gene signatures likely have inadequate infiltration of T cells, an important producer of interferon gamma, explaining why the absence of the signature predicts a poor response. A closer examination of cell types in tumors that express IFNG-related gene signatures confirmed that T cells and myeloid cells are enriched for interferon response gene expression.^103^ However, high expression of IFNG-related signatures does not ensure immunotherapy response as many patients with signature-positive tumors fail to benefit from anti-PD-1.^102^ When gene signatures related to IFN-I signaling are considered, these signatures (called the interferon-I response gene signature, for ISG resistance signature) are found to be enriched in cancer cells compared with T cells and are associated with a lower likelihood of response to ICB.^103,104^ A similar pattern has been reported in patients with lymphoma treated with CAR-T cells, where IFNG-related signatures predict durable response, whereas the ISG.RS predicts the nondurable response.^105^ Thus, IFNG-related genes enriched in immune cells are associated with T-cell infiltration and favorable response, whereas IFN-I–related genes, such as the ISG.RS, can be enriched in cancer cells, thereby antagonizing immunotherapy response.
Clinical and preclinical evidence suggests that IFN-I–related gene signatures (ISG.RS) in cancer cells are initiated by exposure to chronic IFNG and maintained by persistent IFN-I, thereby leading to immunotherapy resistance. In a variety of mouse cancer cell lines, chronic IFNG stimulation is sufficient to confer immunotherapy resistance.^106^ Chronic IFNG stimulation and acquired resistance result in epigenetic changes and higher levels of autocrine and paracrine IFN-I production by cancer cells.^103,104^ This increased IFN-I signaling by cancer cells contributes to high ISG.RS expression, which promotes exhaustion and reduces IFNG production by intratumoral CD8 T cells. Accordingly, tumors from patients with NSCLC who relapsed after anti-PD-1 treatment exhibited a pattern of high resistance-associated ISGs and features of CD8 T-cell exhaustion in half of the relapsed patients.^22^ Moreover, in patients with NSCLC who were treated with nivo and ipilimumab in the CHECKMATE-012 trial, pathogenic somatic mutations in IFN pathway genes were associated with better progression-free survival.^103^ Similar findings have been observed in other cancers as well.^107^ Together, these observations in mice and humans point toward the therapeutic potential of targeting persistent IFN signaling to improve response to checkpoint blockade.
Because interferons are essential for initial immune activation but chronic signaling can lead to immune suppression, these opposing context-dependent functions make targeting interferon and other similar pathways a challenge. In mouse tumor models, JAK inhibitors and IFN-I receptor (IFNAR1)–blocking antibodies can be properly dosed and sequenced to prevent persistent interferon and/or JAK-dependent cytokine signaling yet preserve interferon signaling critical for immune activation.^106^ Consequently, combining a JAK inhibitor or anti–IFNAR1-blocking antibody with ICB can improve response or resensitize immunotherapy-resistant tumors across multiple mouse tumor models.^23,24^ This improved tumor response can be accompanied by improvements in CD8 T-cell differentiation, resulting in less T-cell exhaustion, and favorable changes to myeloid populations, including upregulation of antigen-presentation molecules.
Building on preclinical studies aimed at blocking the immune-suppressive effects of persistent interferon and cytokine signaling, two recent early phase clinical trials have examined the combination of anti-PD-1 with a JAK1 inhibitor (itacitinib) or a JAK1/2 inhibitor (ruxolitinib).^23,24^ In a phase two study for patients with first-line metastatic NSCLC with PDL1 ≥50%, itacitinib was administered for 6 weeks after two cycles of pembrolizumab, a sequential dosing strategy that starts JAK inhibition after ICB similar to an approach used in mice. The best overall response rate was 67%, which was higher than the expected response rate for this NSCLC patient population treated with pembrolizumab alone. Among the responders were many patients who initially failed to respond to the first two cycles of pembrolizumab but responded after receiving itacitinib. Longitudinal immune profiling studies revealed that itacitinib altered CD8 T-cell differentiation, pivoting early progenitor or precursor populations away from exhaustion and toward nonexhausted CD8 T-cell subsets. By contrast, nonresponders appeared refractory to JAK inhibition, characterized by persistently elevated levels of serum inflammatory proteins and cytokine signaling along with continuous development of CD8 T-cell exhaustion.^23^ In a separate phase one clinical trial, patients with Hodgkin lymphoma previously refractory to anti-PD-1 (nivo) were treated with ruxolitinib in combination with nivo. The best overall response rate was 53%, with 6 of 19 patients achieving a complete metabolic response. A reduction in neutrophil-to-lymphocyte ratios along with a decrease in myeloid cells with suppressive features was among the favorable immune changes that correlated with response. Consistent with improvements in myeloid cells, transcriptomic signatures for G-CSF signaling, a JAK-dependent cytokine, were significantly downregulated by ruxolitinib.^24^
In total, these early-phase clinical trials using JAK inhibitors corroborate numerous immune-related changes expected from mouse models examining the impact of blocking chronic interferon and cytokine signaling on immune function. Moreover, the studies also highlight the therapeutic potential of targeting chronic inflammation and cytokine signaling to mitigate acquired immunotherapy resistance across different cancer types.
Immune checkpoint blockade relies on interfering with the function of naturally expressed inhibitory ligands on the surface of antitumor T cells. While this strategy effectively releases exhausted T cells in the tumors of treated patients, it also has the unintended side effect of causing auto-inflammation within healthy organ systems. The natural purpose of inhibitory ligands is to preserve tolerance to self-antigens, limiting the activity of autoreactive cells while similarly downregulating the immune response to avoid bystander tissue damage.^108^ By interrupting this complex system of checks and balances, autoinflammatory responses can be triggered, which have been given the name irAEs. In any given patient, irAEs can present as single-organ inflammation or demonstrate multisystem involvement, and while most present within the first few months of treatment, inflammation can occur throughout the ICB course.^109^ Low-grade irAEs (grade 1–2) occur in nearly 90% of treated patients, whereas higher-grade irAEs (grades 3–5) can range from 20% to 60%.^110^ The sites most often affected include barrier tissues like the skin, GI tract/liver, and respiratory epithelium, particularly when PD-1 is used in combination with CTLA-4.^111^ Endocrine tissues can also be affected,^112^ which can lead to the need for lifelong endocrine supplementation. Similarly, joint toxicities can also be long-lasting, with ongoing pain and mobility issues long after treatment cessation.^113^ While cardiac and neurologic toxicities are more rare, they develop rapidly and are more often fatal.^114^ Currently, the mainstay of treatment of high-grade irAEs is high-dose steroid therapy with discontinuation of the ICBs although newer therapies are actively being developed.^115^ The mechanisms driving irAEs are still an area of active study. There is likely a combination of inappropriate activation of tissue-resident memory cells, particularly at epithelial barriers,^116^ alongside the concept of epitope spreading, whereby inflammation leads to shedding of otherwise self-antigens, which, when coupled with the inflammatory microenvironment and blockade of normal checkpoints, leads to activation of autoreactive T cells.^108^ Outside of conventional T cells, checkpoints are also expressed by Tregs (particularly CTLA-4), B cells, and myeloid cells, which could also contribute to irAE-mediated pathology.^108^ Because of these varied and often significant side effects, it is the goal of newer combination therapies to avoid further worsening ICB toxicity. Indeed, data reviewing trials completed thus far demonstrate that grade 3/4 irAE rates are considerably lower with rela-nivo (19%) compared with ipilimumab-nivo (59%), likely because of the exclusion of CTLA-4.^117^ A better understanding of how irAEs are triggered will be required to effectively design better tolerated therapeutic options.
A key goal of cancer immunotherapy is to eradicate tumor cells without harming normal cells. In general, immune therapy relies on recruitment and activation of CD8^+^ T cells, which can eliminate tumor cells. However, a number of intrinsic resistance mechanisms can develop to prevent checkpoint inhibitors, CAR-T cells, or other therapeutic approaches from achieving tumor control. Here, we have identified several key means by which tumors develop resistance to a TME that is rich in immune-suppressive macrophages and granulocytes; a stromal cell–/collagen-rich microenvironment that precludes the close contact between T cells and tumor cells; mutations in antigen-presentation machinery that reduces T-cell recognition of target cells; persistent expression of type II interferon (IFNγ) and type I interferons (IFNα, IFNβ), which can upregulate expression of interferon response genes (ISGs); and expression of additional immune checkpoints, including LAG-3, TIM-3, and TIGIT. New therapies designed to circumvent these resistance mechanisms offer promise to improve cancer outcomes. We have described here the foundational studies that led to development of novel approaches to prevent resistance to checkpoint inhibitors. These studies of the myeloid cell protein PI3Kγ, the T-cell checkpoint LAG-3, and the signaling protein JAK1 have led to development of novel therapeutic approaches that have shown benefit in preventing resistance to therapy in cancer clinical trials and have led to FDA approval of the LAG-3 antibody rela in conjunction with nivo for cancer therapy. PI3Kγ and LAG-3 inhibitors to date have been found to be effective if administered concurrent with checkpoint blockade, whereas JAK1 inhibitors were found to be most effective if administered after two cycles of immune checkpoint blockers. Future clinical trials should determine optimal strategies and biomarkers of response and patient selection criteria for these immune therapies. The field of cancer immunotherapy has advanced new strategies aimed at overcoming tumor resistance. In addition to the development of new agents such as those described here, the future of cancer immune therapy will rely on personalized approaches to treatment, where identification of biomarkers such as specific mutations, presence and patterns of T cells, macrophages, granulocytes, and stroma will guide the selection of specific treatments for each patient. Continuous research and refinement of therapeutic approach efforts such as those presented here will bring the promise of immune therapy closer to fruition.