Molecular Pathways and Mechanisms of LAG3 in Cancer Therapy

Coinhibitory receptors (IR), such as programed death-1 (PD1) and cytotoxic T lymphocyte antigen-4 (CTLA4), are upregulated on T cells in settings of chronic antigen stimulation or inflammation and serve to limit tissue damage and autoimmune-mediated host tissue pathology (1, 2). Importantly, such IRs play an essential role in suppressing autoimmunity, but conversely this immunoregulation can limit viral clearance and suppress antitumor immunity. Expression of these IRs on the surface of CD4⁺ and CD8⁺ T cells reflects a distinct transcriptional and epigenetic T-cell lineage known as exhausted T cells (3). Cancer immunotherapeutics have sought to antagonize the interactions between IRs and their cognate ligands to reverse this exhausted state and to reinvigorate the antitumor immune response. Blockade of PD1 with mAbs is an FDA-approved strategy that has been effective across a wide range of malignancies including melanoma, renal cell carcinoma (RCC), head and neck squamous cell carcinoma (HNSCC), non–small cell lung cancer (NSCLC), and multiple microsatellite instable (MSI) cancers, among others (4). Despite this success, not all patients benefit, and some tumor types are largely refractory to current immunotherapy. As multiple IRs are expressed on exhausted T cells, immunotherapies targeting more than one IR in combination could more effectively restore T-cell function, leading to improved clinical efficacy.

The combination of nivolumab (anti-PD1) and ipilimumab (anti-CTLA4) has been shown to induce a higher response rate and progression-free survival (PFS), recently reaching superiority in 5-year survival versus monotherapy with nivolumab (5). However, this came at the cost of considerably increased toxicity. On March 18 2022, LAG3 became the third FDA-approved target of an immune checkpoint inhibitor with relatlimab (anti-LAG3) plus nivolumab for treatment of patients with unresectable or metastatic melanoma. Nearly a dozen more agents that target LAG3, alone or in combination with anti-PD1, are currently being evaluated clinically. In this review, we discuss the current understanding of LAG3 biology, including its expression pattern, structure, function, and signaling pathways. We also summarize the current state of clinical investigation of LAG3-targeted monotherapies, combinations, and bispecifics, as well as the rationale for synergistic enhancement of antitumor immunity with LAG3-targeted combinations. A timeline for the discovery and clinical development of LAG3 through FDA approval is provided in Fig. 1.

LAG3 is not expressed on naïve T cells, but induced upon stimulation of CD4⁺ and CD8⁺ T cells to limit activation and prevent autoimmunity (6). However, as a result of persistent antigenic stimulation within the tumor, sustained expression of LAG3 along with other IRs including PD1, TIM3, and TIGIT, results in a state of T-cell dysfunction, associated with a lack of proliferation, cytokine secretion, and cytolysis (7). Coexpression of PD1 and LAG3 has been demonstrated on CD4⁺ and CD8⁺ tumor-infiltrating lymphocytes (TIL) in numerous murine transplantable tumor models (B16-F10 melanoma, MC38 colon adenocarcinoma, and SA1N fibrosarcoma; ref. 8). IHC studies identified high LAG3 in a range of tumor types including ovarian cancer, melanoma, NSCLC, and HNSCC, as correlating with poor prognosis and disease-free survival (9). Importantly, extensive immune profiling of patients with melanoma treated with immune checkpoint blockade demonstrated that LAG3 expression on CD8⁺ T cells in peripheral blood resulted in poorer outcomes after immunotherapy compared with patients with no LAG3 expression. Patients with melanoma who possessed a LAG3⁺ CD8⁺ immunotype had poorer outcomes after immune checkpoint blockade with a median survival of 22.2 months compared with 75.8 months for those with a LAG3-negative immunotype (10). This demonstrates that expression of LAG3 may act as a mechanism of resistance to PD1 blockade, and thus can be used as a biomarker to identify patients who may or may not respond to treatment. Furthermore, this also suggests that LAG3 could potentially be targeted to overcome such resistance.

LAG3 is also expressed on CD4⁺ Foxp3⁺ regulatory T cells (T_reg) and while it has been shown to be required for control of homeostatic T-cell expansion and suppressive activity of T_regs themselves, it has been demonstrated that LAG3 acts as a brake on T_regs to limit proliferation and function in a model of diabetes (11–13). However, LAG3 expression on T_regs in melanoma and colorectal cancer patients’ peripheral blood, tumor-associated lymph nodes and the tumor itself has been correlated with production of immunosuppressive cytokines (IL10, TGFβ) compared with LAG3-negative cells (14). Likewise, LAG3 coexpression with CD49b defines an IL10-producing CD4⁺ type 1 T regulatory (Tr1) cell population, in which LAG3 also contributes to its suppressive properties (15). This population has also been associated with disease progression in colorectal cancer.

While persistent antigen exposure that drives the exhausted T-cell program within tumors and chronic viral infection, IL2, IL7, and IL12 have also been reported to drive LAG3 expression on activated T cells (16). Other mechanisms reported to drive expression of LAG3 include glycogen synthase kinase-3 (GSK-3), as inhibition was shown to downregulate LAG3 expression in T cells (17). GSK-3 inactivation was shown to potentiate the antitumor effects of anti-LAG3 to eliminate tumor cells and may promote a more efficacious antitumor immune response.

In addition to CD4⁺ and CD8⁺ T cells, LAG3 can be found on other immune populations such as γδ T cells, natural killer T cells, plasmacytoid dendritic cells (pDC), and activated B cells, although it is unclear whether LAG3 expression on these populations plays a role with regard to antitumor immunity (18–20). LAG3 has also been found to be highly expressed by malignant cells in diffuse large B-cell lymphoma (DLBCL; ref. 21).

The Lag3 gene is comprised of eight exons and its chromosomal location is adjacent to the gene for CD4 on the distal part of chromosome 12, with a similar intron-exon organization that is likely to have evolved due to a gene duplication event (22). The Lag3 gene encodes a 70 kDa type I transmembrane protein, made up of 498 amino acids with structural homology to CD4, containing four extracellular immunoglobulin-like superfamily regions (D1–D4) with one variable (V type) and three constant (C type) Ig-like domains. The membrane distal D1 domain has an additional 30 amino acid proline-rich loop that is conserved across multiple species, but not present in CD4 (23). Biochemical analysis of LAG3 has suggested that it can be expressed as monomers, dimers, and higher-order oligomers on the cell surface, independent of ligand engagement (24).

As a result of the structural homology to CD4, the canonical ligands for LAG3 are MHC class II molecules although additional ligands have been described previously (Fig. 2). The presence of a proline-rich D1 loop in LAG3 allows it to bind to MHC class II with a much higher affinity than CD4, leading to the original hypothesis that LAG3 primarily functions by competing with CD4 for binding to MHC class II (18, 25, 26). However, mutagenesis studies as well as experiments performed with C9B7W (a rat mAb that binds to the D2 domain of LAG3), which blocks the inhibitory effects of LAG3 but does not disrupt the MHC class II:LAG3 interaction, implied that LAG3 does not primarily function by disrupting CD4:MHC class II interactions (18). While further studies have shown that LAG3 inhibits T-cell responses by binding to stable complexes of peptide and MHC class II, inhibitory signals are transmitted via the intracellular region (27). Moreover, MHC class II expression by human melanoma is associated with poor prognosis, and ligation with LAG3 has been shown to be an escape mechanism to protect against apoptosis as LAG3 mutants unable to bind to MHC class II have reduced inhibitory function (28, 29). Despite this, these studies raise the possibility that LAG3 binding to other ligands may mediate its function, particularly with respect to non-MHC class II–restricted CD8⁺ T cells expressing LAG3 within the tumor.

Other ligands have been reported to interact with glycans on LAG3 including galectin-3 (Gal-3) and liver sinusoidal endothelial cell lectin (LSECtin) that are both lectins with carbohydrate-recognition domains. Gal-3 is a soluble galactose-binding lectin secreted from tumor cells and tumor stromal cells (30). LAG3:Gal-3 engagement was suggested to reduce the frequency of IFNγ-producing CD8⁺ T cells upon T-cell receptor (TCR) stimulation (31). LSECtin is expressed in the liver, but is also found on the surface of melanoma cells where engagement with LAG3 may inhibit IFNγ production by antigen-specific effector T cells (32, 33).

Fibrinogen-like protein 1 (FGL1), which is secreted by the hepatocytes in the liver and is highly upregulated in tumors, has also been suggested as a ligand for LAG3 with immunosuppressive activities (34). FGL1 binds to the D1 and D2 domains of LAG3, independent of MHC class II interaction, and blockade of this interaction increases IFNγ levels. FGL1 may be a biomarker of poor prognosis itself as high FGL1 levels in plasma of patients with NSCLC and melanoma were associated with resistance to anti-PD1 therapy (34).

Recently, it has been postulated that LAG3 may interact in cis with the TCR:CD3 complex in the presence or absence of MHC class II ligation in both CD4⁺ and CD8⁺ T cells. In support of this hypothesis, prior to and following TCR stimulation, multiple LAG3 molecules associate and track with the (TCR):CD3 complex to the immunological synapse and via its cytoplasmic tail allows LAG3 to promote an inhibitory effect on TCR signaling (35).

Many immune IRs (checkpoints), such as CTLA4, PD1, and TIM3, inhibit T-cell activation through cytoplasmic inhibitory domains such as an immunoreceptor tyrosine-based inhibitory motif or immunoreceptor tyrosine-based switch motif, but these are not present in the LAG3 intracellular domain. Despite structural homology to CD4, LAG3 does not have the C-X-X-C binding site for the tyrosine kinase p56^Lck that facilitates coreceptor downstream signal transduction, although curiously LAG3 function was shown some time ago to be coreceptor dependent (23, 26). Instead, LAG3 possesses three conserved regions in the cytoplasmic tail: (i) a “KIEELE” motif, (ii) an acidic glutamic acid-proline (EP) repetitive sequence, and (iii) a conserved putative serine phosphorylation site (Fig. 2; ref. 6). The “KIEELE” motif has been shown to be required for LAG3 inhibitory activity in T-cell hybridomas (26). Another group has suggested that LAG3 transduces inhibitory signals through an FXXL motif in the membrane-proximal region (36).

Recent studies with LAG3 mutants have shown that the EP motif in the cytoplasmic tail, and in particular, the highly conserved and repetitive glutamic acid residues are responsible for the dissociation of p56^lck from CD4 or CD8 coreceptors, consistent with previous studies highlighting coreceptor dependency of LAG3 (26), thereby abrogating coreceptor-mediated signaling (35). Although LAG3 lacks the di-cysteine binding motif for p56^lck, a large number of LAG3 molecules moves into the immunologic synapse via its association with the (TCR):CD3 complex and the resulting number of glutamic acid residues binds to Zn²⁺ and leads to a reduction in the local pH of the immunological synapse. This reduction in pH causes p56^lck to disassociate from CD4 or CD8 resulting in a reduction in phosphorylation of ZAP70 and the consequent downstream signaling events. Therefore, LAG3 through dissociation of p56^lck signaling via this unique mechanism can limit TCR signaling and T-cell activation (35).

While many ligands could be involved in triggering LAG3 signaling, the consequences of this signaling in T cells lead to a reduction in T-cell proliferation and function. Early studies demonstrated that LAG3 knockout mice experienced uncontrolled T-cell proliferation following administration of a superantigen and ovalbumin-specific T cells lacking LAG3 similarly experienced uncontrolled proliferation following immunization (37, 38). In addition, LAG3 was found to limit the size of the memory pool in mice following infection with Sendai virus (39). In chronic viral infection, LAG3 is coexpressed with other IRs (39) that together result in a progressive reduction in effector function due to lower expression of cytolytic effectors such as granzyme B and perforin, loss of expression of effector cytokines such as IFNγ and TNFα in CD8⁺ T cells (39) and IL2, IL4 and IFNγ in conventional CD4⁺ T cells (40). LAG3 has also been shown to regulate the metabolic profile of naïve CD4⁺ T-cell metabolism to control homeostatic expansion (41). LAG3 has also been shown to regulate both the STAT5 and AKT pathways. The enhanced STAT5 signaling in LAG3-deficient CD4⁺ T cells resulted in increased aerobic glycolysis and effector function following activation. STAT5 signaling is thus required for LAG3 blockade to affect homeostatic proliferation in vivo and thus it is inferred that LAG3 may modulate the STAT5/IL2 signaling axis during homeostatic proliferation (41). Overall, the binding of ligands to LAG3 expressed on the cell surface of T cells leads to inhibition of proliferation and loss of T-cell effector functions.

Expression of LAG3 is controlled at multiple levels including genetic, epigenetic, cleavage at the cell surface and via intracellular trafficking. Early studies suggested that LAG3 may be subjected to additional levels of genetic control as three splice variants have been proposed to exist, which include a version lacking the last three exons producing a soluble truncated version of the full-length protein (42, 43). There is also evidence for epigenetic regulation of LAG3 expression, via DNA methylation of the Lag3 promoter, in some tumor types including melanoma and clear-cell RCC (44, 45). This epigenetically mediated regulation of Lag3 by tumor cells has been associated with increased immune cell infiltration and improved overall survival.

LAG3 levels on the cell surface are also regulated by intracellular trafficking and cell surface cleavage. Rapid translocation of LAG3 from intracellular stores in lysosomal compartments to the cell surface occurs upon TCR stimulation (46, 47). Interestingly, unlike healthy individuals, a proportion of patients with cancer demonstrate robust intracellular, but not surface, LAG3 expression in peripheral CD8⁺ T cells (48). The divergence between intracellular and surface LAG3 is the result of rapid shedding of LAG3 from the cell surface by a disintegrin and metalloprotease domain-containing protein-10 and protein-17 (ADAM10/17)-mediated cell surface shedding following trafficking to the surface. ADAM10/17 cleave LAG3 at the connecting peptide region between the membrane proximal D4 domain and the transmembrane domain releasing a monomeric soluble form of LAG3 (sLAG3; ref. 49). While ADAM10 constitutively cleaves LAG3 under resting conditions, shedding activity is enhanced upon TCR activation and ADAM17 shedding only occurs upon protein kinase C activation (49). It has been demonstrated in vitro that the shedding of LAG3 is essential for T-cell function and proliferation by generation of a noncleavable LAG3 (LAG3^NC) mutant (49). Furthermore, generation of a conditional knock-in mouse in which LAG3^NC is only expressed in T cells was shown to be resistant to antitumor immune effects of PD1 blockade—including impact on tumor growth and functionality of LAG3^NC-expressing CD4⁺ T cells within the tumor (50). In addition, the translational relevance of LAG3 shedding was demonstrated by assessment of LAG3 and reciprocal ADAM10 expression, which were correlated to disease stage and prognosis of patients with HNSCC, and the LAG3:ADAM10 ratio in fact predicted response to nivolumab in a cohort of patients with melanoma (50). These data demonstrate that regulation of LAG3 expression and function by ADAM-mediated cell surface shedding has an important effect on patient responsiveness to immunotherapy. Overall, LAG3 levels on T cells are tightly regulated by several mechanisms.

While both in vitro and in vivo studies have not uncovered a function of sLAG3 compared with the transmembrane counterpart, sLAG3 in patient sera is prognostic of outcome in several malignancies. Serum levels of sLAG3 were significantly elevated in patients with hepatocellular carcinoma (HCC) compared with healthy donor controls (51) and in patients with hormone receptor–positive breast cancer (52). High sLAG3 expression was found in relation to better prognosis with gastric cancer (53). Patients with advanced (stage III–IV) NSCLC also had significantly lower sLAG3 expression than patients with stage I–II NSCLC (54). However, in HNSCC, high levels of sLAG3 were associated with poor prognosis (55). Although LAG3 is shed from the surface of T cells, pDCs are the major contributor to sLAG3 levels in vivo, complicating the interpretation of the biological significance of sLAG3 (19). Further work is needed to better understand the significance of the role of LAG3 on pDCs and the consequent release of sLAG3, as little is known in tumors other than a role in regulation of pDC homeostasis (19). Regardless of the origins of sLAG3 in vivo, promoting LAG3 shedding activity by enhancing ADAM10/17 activity may lead to enhanced T-cell function by abrogating LAG3-mediated intracellular signaling.

In 2006, a dimeric, recombinant, soluble form of LAG3 called eftilagimod alpha was the first-in-class agent to enter the clinic. Consisting of four LAG3 extracellular domains fused to the Fc portion of human IgG1, this agent was clinically developed as an immune adjuvant to activate antigen-presenting cells (APC) and is not to be confused with the monomeric sLAG3 released as a byproduct of metalloprotease-mediated shedding. The rationale for eftilagimod alpha-based therapy lies in the observation that a soluble LAG3-Ig fusion protein directly stimulates DCs to upregulate costimulatory molecules and produce IL12 and TNFα (56). Recent data suggest a survival benefit with the combination of eftilagimod alpha and paclitaxel as a first-line chemoimmunotherapy assessed in a phase IIb study of patients with HER2-negative, HR-positive metastatic breast cancer, but only in a prespecified subgroup of patients aged under 65 years and not the overall population. This combination was well tolerated with manageable adverse events. Clinical trials involving combination of eftilagimod alpha with pembrolizumab (anti-PD1) have been initiated in a phase IIb study in patients with PDL1-positive HNSCC and NSCLC (57), as well as in melanoma with encouraging antitumor activity (58).

The inhibitory effects of LAG3 on CD8⁺ T cells appear to be distinct from those of PD1 or CTLA4, providing a rationale for blockade of multiple IRs. Biochemical reconstruction of the PD1 signaling pathway has indicated that the costimulatory receptor CD28 is more highly dephosphorylated following PD1 ligation with PDL1 than the TCR (59). Thus, PD1 blockade may serve to reinvigorate T cells via rescue of cosignaling pathways as opposed to directly regulating signals that drive T-cell proliferation through the TCR. Similarly, CTLA4 serves to reduce T-cell activation by depriving T cells of signaling through the CD28 coreceptor, although CTLA4 does this through competitive binding with the CD80 and CD86 costimulatory ligands for CD28 rather than dephosphorylating CD28 following signaling through PD-L1 (60). CTLA4 may also remove CD80/CD86 directly from APCs as an alternative mechanism of limiting costimulation. Thus, although blockade of PD1 and CTLA4 may lead to expansion of different subpopulations of T cells (61, 62), they may do so through a similar mechanism of action. In addition, previous studies have suggested that LAG3 may also mediate bidirectional inhibitory signaling into APCs. Conversely, as outlined above, the mechanism of action of LAG3 may limit T-cell activation by reducing signaling through the TCR. This unique mechanism of action of LAG3 suggests the potential for synergistic enhancement of antitumor immunity with less toxicity in combination with PD1 due to their distinct mechanisms.

Following the success of targeting PD1/PDL1, clinical trials began to move beyond monotherapy to combinations of immunomodulatory agents. The rationale for this is that, in many instances, blockade of a single immunomodulatory IR may be insufficient to rescue CD8⁺ T-cell function, and that targeting multiple coinhibitory pathways will lead to a more robust reversal of immune exhaustion in the tumor microenvironment. This has been demonstrated with combination of ipilimumab with nivolumab in advanced melanoma resulting in a 52% 5-year survival compared with 44% and 26% in the nivolumab and ipilimumab groups, respectively, associated with severe toxicity (63).

Preclinical investigation of LAG3 has suggested a strong rationale for combination therapy with anti-PD1. Compared with CTLA4 blockade alone, LAG3 blockade as a monotherapy leads to a small reduction in tumor growth in several murine tumor models including MC38 (8). However, combination of anti-LAG3 with PD1 blockade dramatically limits tumor growth (∼80% of mice tumor free), compared with administration of anti-PD1 alone (∼40% of mice tumor free). Increased tumor-free survival was associated with augmented CD8⁺ T-cell infiltration and enhanced cytokine production (8). Synergistic effects with combinatorial blockade have also been demonstrated in the ID8 ovarian cancer model, EG7 lymphoma model, and the 5T33 multiple myeloma model (64, 65). This evidence suggests that PD1 and LAG3 blockade in combination leads to potent reinvigoration of antitumor immune responses. Combination of anti-LAG3 with an agonist mAb targeting the immunostimulatory 4-1BB molecule was also shown to synergistically control B16 tumor outgrowth (66). The observation that LAG3 alone does not have a substantive antitumor effect also suggests that the efficacy of LAG3 in combination is reliant on perturbing distinct pathways, which may suggest the potential for fewer adverse events. In addition in an adoptive chimeric antigen receptor T-cell model, LAG3 blockade along with anti-PD1 and TIM3 was shown to overcome dysfunction in a synergistic manner to boost effector function (67).

Given the strong rationale for antagonistic blockade of LAG3 to enhance CD8⁺ T-cell function (8), numerous mAbs have been developed and are progressing through clinical trials targeting a variety of solid and hematologic tumor types (Table 1). All these clinical trials have an anti-PD1 arm to capitalize on the synergistic effects expected with dual blockade. The clinical success of an anti-LAG3/PD1 combination therapy was recently demonstrated by RELATIVITY-047, a global, randomized, double-blinded phase II/III study of patients with metastatic or unresectable melanoma in the first-line setting, which resulted in the FDA approval of relatlimab with nivolumab. Combination of relatlimab and nivolumab versus nivolumab alone, showed superior PFS (10.2 months) comparing the combination therapy with nivolumab alone (4.6 months; ref. 68). Patients with LAG3 expression greater than 1% did show improved PFS; however, in both stratified groups, there was a benefit with the addition of relatlimab to nivolumab (69). Furthermore, PFS at 12 months was 48% in the combination arm and 36% with nivolumab alone. While there was also a clinically meaningful improvement in overall survival with the combination, this was not statistically significant. Grade 3 or 4 adverse events in the combination arm (21.1%), were manageable and this safety profile was substantially more favorable than observed with anti-PD1/CTLA4 combination therapy (where grade 3–4 toxicity was encountered in 59% of patients; ref. 70). Myocarditis occurred in 1.7% of patients in the relatlimab-nivolumab group, compared with 0.6% of patients in the nivolumab alone group, although all myocarditis events resolved completely (69). Recent work presented at the Annual Society of Clinical Oncology (ASCO) Annual Meeting revealed that addition of relatlimab to nivolumab as a second-line treatment benefited patients who have progressed previously on anti-PD1/PDL1 therapy (71). Other work at ASCO revealed that pathologic complete response correlated with improved survival in the setting of neoadjuvant nivolumab and relatlimab, supporting synergistic activity when targeting these two IRs (72). Initial clinical activity, with a manageable safety profile, has also been demonstrated with fianlimab (anti-LAG3) given in combination with cemiplimab (anti-PD1) in patients with advanced melanoma at ASCO (73). In addition, work presented at ASCO showed favezelimab (anti-LAG3) plus pembrolizumab recently demonstrated safety in a phase I trial in microsatellite stable colorectal cancer (MSS-CRC; ref. 74). Overall, these early results suggest similar efficacy with substantively reduced toxicity for anti-LAG3/PD1 therapy versus anti-CTLA4/PD1 blockade.

In addition to mAbs described here, several dual antagonist bispecific agents are in clinical development targeting LAG3 with either an anti-PD1 or PDL1 pairing. FS118 is a dual checkpoint inhibitor targeting LAG3 and PDL1 that was shown to clear MC38 in preclinical settings and has now entered phase I clinical study with a cohort of patients that have previously shown resistance to anti-PD1 as a monotherapy. Interestingly, recent data have demonstrated that the bispecific activity of FS118 drives the shedding of LAG3, which is of importance for its therapeutic activity (75). In preclinical murine studies, FS118 resulted in reduced surface expression of LAG3 and increased sLAG3 in the serum of tumor-bearing mice. This increased level of sLAG3 was FS118 dose dependent and inhibited with ADAM10 metalloprotease inhibitor (GI254023X), suggesting that FS118 may promote LAG3 shedding. Likewise, in the patient cohort, FS118 increased sLAG3 in patients’ sera and decreased LAG3 TIL expression (75). While this is the first immunotherapeutic agent to suggest LAG3 shedding as part of its mechanism of action, it is unknown whether other LAG3-targeting agents can modulate cell-surface expression. An additional PD1-LAG3 bispecific agent, called tebotelimab, has been generated using a dual-affinity retargeting (DART) platform that covalently links two polypeptide chains between the variable domains of two antibodies via a disulfide bridge and a short linker that promotes heterodimerization (76). Encouraging results with tebotelimab have demonstrated early evidence of antitumor activity with a tolerable safety profile and further clinical results are highly anticipated with these bispecific agents to determine whether these are more efficacious than combinations of antagonistic mAbs.

Combination therapy targeting PD1 and CTLA4 achieves a superior response rate and PFS advantages for patients with cancer compared with either therapy alone, but at the cost of increased toxicity. LAG3 functions by limiting T-cell proliferation and plays a role in inhibiting T-cell effector function via mechanisms distinct from PD1, making it a rational choice for combination therapy. While the FDA approval of relatlimab and nivolumab for the treatment of metastatic melanoma is exciting, there remain a number of important unanswered questions that are being addressed in current studies:

• 1. Which pharmacologic compounds will be the most efficacious? Currently, there are several monoclonal and bispecific antibodies undergoing clinical assessment, with the latter category of agents possibly being more efficacious due to dual blockade of both LAG3 and PD1/PDL1.

• 2. When should this combination be used? Whether the optimal setting for this combination is front line, neoadjuvant, or adjuvant remains unknown. While the combination showed improved PFS in the front-line setting in advanced melanoma it remains yet to be evaluated in the adjuvant setting.

• 3. Will anti-LAG3/PD1 combinations prove to be a viable treatment option for poorly immunogenic tumors? It remains to be seen whether the combination of LAG3 and PD1 blockade will be beneficial for tumors with a low TIL infiltrate such as MSS-CRC and clinical trials have been initiated (74). LAG3 expression is also being evaluated as a biomarker in MSS tumors (77).

• 4. Is there room for LAG3 targeting agents in hematologic malignancies? Lag3 mRNA has been demonstrated to be a prognostic marker in chronic lymphocyte leukemia and DLBCL, associated with reduced tumor-free survival (21, 78). Numerous clinical trials with LAG3 targeting agents are ongoing for hematologic malignancies, and a recent phase I study with tebotelimab achieved a clinical response in 4 of 11 patients with relapsed or refractory disease (79).

• 5. What is the mechanism of action for this combination? Understanding the cell types (i.e., conventional CD4⁺ T cells, CD4⁺ T_regs, and CD8⁺ T cells) that respond to this therapeutic combination and the activation of downstream pathways following anti-PD1/LAG3 blockade has important implications for T-cell biology and cancer immunology.

• 6. Is LAG3 expression a prognostic or predictive biomarker for response? In RELATIVITY-047, there was a clear benefit for patients receiving combined relatlimab and nivolumab regardless of LAG3 tumor expression; however, patients stratified as >1% LAG3 expression did have superior responses (69).

Overall, LAG3 as a target is an exciting new addition to the repertoire of cancer immunotherapies and addressing these unanswered questions will further our understanding of LAG3 biology and allow us to optimize its utility in the clinic.

J.M. Kirkwood reports grants and personal fees from Amgen, Bristol Myers Squibb, Checkmate Pharmaceuticals, Harbour BioMed, Immunocore LLC, Iovance Biotherapeutics, Merck, and Novartis Pharmaceuticals; personal fees from Ankyra Therapeutics, Axio Research/Instil Bio, Becker Pharmaceutical Consulting, DermTech, Fenix Group International, Intellisphere LLC/Cancer Network, IQVIA, Istari Oncology, Millennium Pharmaceuticals/Takeda Pharmaceutical, Natera Inc., OncoCyte Corporation, Pfizer, Replimune, Scopus BioPharma, and SR One Capital Management; and grants from Castle Biosciences Inc., Immvira Pharma Co., Schering-Plough, Takeda, and Verastem Inc. outside the submitted work. C.J. Workman reports a patent (8551481) issued, licensed, and with royalties paid from BMS. D.A.A. Vignali reports grants, personal fees, and other support from BMS and personal fees from Incyte and F-star during the conduct of the study as well as grants, personal fees, and other support from Tizona, Potenza, Astellas, and Novasenta; other support from Trishula and Oncurus; personal fees from Bicara, G1 Therapeutics, T7/Imreg Bio, Almirall, and Inzen Therapeutics and personal fees and other support from Werewolf and Apeximmune outside the submitted work; in addition, D.A.A. Vignali has patents in the United States (8551481, issued 10/8/2013; 9005629, issued 04/14/2015; 10787513, issued 09/29/2020; 10934354, issued 03/02/2021), Australia (2004217526, issued 8/12/2010), Europe (1897548, issued 08/14/2013), Japan (6758259, issued 09/03/2020), and Hong Kong (1114339, issued 11/22/2013) issued, licensed, and with royalties paid from BMS. No disclosures were reported by the other authors.

We wish to thank all the current and former members in the Vignali Lab (Vignali-lab.com; @Vignali_Lab) for all their constructive discussions. The authors are supported by the NIH (R35 CA263850, P01 AI108545, to D.A.A. Vignali; R01 AI144422, to C.J. Workman and D.A.A. Vignali; P50 CA254865, to J.M. Kirkwood and D.A.A. Vignali), and the NCI Comprehensive Cancer Center Support CORE grant (CA047904, to J.M. Kirkwood and D.A.A. Vignali). Figures were created with BioRender.com.