Antitumor Immune Responses in B2M-Deficient Cancers

The success of immune checkpoint blockade (ICB) therapies relies on the antitumor activity of CD8⁺ T cells (1, 2). However, CD8⁺ T cell–based therapies rely on functional antigen presentation by tumor cells, which opens avenues for the development of resistance mechanisms. Canonically, tumor antigens are presented to CD8⁺ T cells by surface MHC class I molecules, which are composed of a heavy chain and β2-microglobulin (B2M). B2M is necessary for the proper stabilization and folding of MHC class I molecules. In its absence, the MHC class I complex does not reach the cell surface and is degraded (3). The loss of MHC class I through B2M mutations or copy-number losses results in the inability of CD8⁺ T cells to recognize cancer cells and has long been recognized to lead to tumor immunotherapy resistance (4–8). However, the use of ICB in highly immunogenic cancers that may have already gone through immune editing has demonstrated that some patients whose cancers do not express B2M can still respond to anti–PD-1 therapies (9–15). Therefore, there is a need to study how the immune system can induce responses in B2M-null tumors by analyzing murine models and studying patient-derived biopsies. Previously, we reported that anti–PD-1 resistance due to B2M loss could be overcome with the activation of natural killer (NK) cells and CD4⁺ T cells in murine models (16). The role of CD4⁺ T cells in immune responses to B2M-knockout (KO) tumors has been corroborated in murine models and biopsy samples of patients with DNA mismatch repair deficient (MMR-d) cancers (13). Furthermore, γδ T cells have also been shown to increase in MMR-d cancers with B2M defects, and to have a cytotoxic effect upon treatment with ICB (15). Nevertheless, the exact effect of defects in MHC class I antigen presentation machinery on anti–PD-1 therapy response, and the cells capable of mediating the antitumorigenic effects of PD-1 blockade in B2M-defective human tumors, have not been fully characterized.

Here we used CRISPR/Cas9 to develop sublines through B2M-KO mutations in three murine tumor models: MC38 and YUMMER2.1, which have high immunogenicity due to carcinogen-induced high mutational load and have CD4⁺ and CD8⁺ T-cell codependency in response to anti–PD-1 therapy, and the lowly immunogenic B16 model with undetectable baseline MHC class I expression and primary CD8⁺ T-cell dependency to respond to immunotherapy (17). These murine models allowed us to study the mechanisms by which immune cells orchestrate responses to PD-1 blockade in the context of B2M inactivation. To further corroborate these findings, we analyzed sequencing data from a large cohort of pretreatment biopsies from patients with melanoma for the presence of B2M somatic alterations and correlative infiltrating immune cell subsets. Our findings may support the development of combination strategies to potentiate the CD8⁺ T-cell antitumor effects and avoid resistance to ICB therapies.

All mouse studies were approved by the University of California, Los Angeles (UCLA) Animal Research Committee under protocol no. 2004-159-43I. Female C57BL/6N mice were obtained from the UCLA Department of Radiation Oncology Animal Core. Mice were bred and kept under defined flora and pathogen-free conditions at the Association for the Assessment and Accreditation of Laboratory Animal Care–approved animal facility of the Division of Experimental Radiation Oncology, UCLA. The MC38 mouse colon adenocarcinoma cell line was initially generated at the NCI Surgery Branch (originally Colo38) and was obtained from Dr. Robert Prins, UCLA Department of Neurosurgery (2014). The B16-F10 mouse melanoma cell line was purchased from ATCC (catalog no. CRL-6475, 2017). The YUMM2.1 UV mouse melanoma cell line was obtained from Dr. Marcus Bosenberg, Yale University (2009). The MC38, B16-F10, and YUMM2.1 UV mouse cell lines and subsequently established KO cell lines (see CRISPR/Cas9-mediated gene KO) were cultured at 37°C and 5% CO₂ in DMEM (Corning, catalog no. MT10013CM) supplemented with 10% FBS (Omega, catalog no. FB-21), 100 units/mL penicillin, 100 μg/mL streptomycin, and 0.25 μg/mL amphotericin B (Omega, catalog no. AA-40). Cell lines were tested for Mycoplasma contamination using the MycoAlert Mycoplasma Detection Kit (Lonza, catalog no. LT07-218), and were regularly tested for authentication. For in vivo experiments, early passage cell lines (no more than 10 passages) were utilized.

Antibodies for in vivo experiments were as follows: anti-mouse PD-1 (clone RMP1-14, catalog no. BE0146), anti-mouse CD8 (clone YTS 169.4, BE0117 catalog no. BE0117), anti-mouse CD4 (clone GK1.5, BE0003 catalog no. BE0003), anti-mouse NK1.1 (clone PK136, BE0036 catalog no. BE0036), anti-mouse CD40L (clone MR-1, BP0017 catalog no. BP0017), anti-mouse IFNγ (clone XMG1.2, BE0055 catalog no. BE0055) and isotype control antibody (clone 2A3, BE0089 catalog no. BE0089), all from BioXCell. Bempegaldesleukin (NKTR-214; ref. 18) was provided by Nektar Therapeutics through a Materials Transfer Agreement (MTA2018-00000545, 2018) and was diluted in the recommended product formulation buffer for in vivo studies.

MC38, B16, and YUMMER2.1 murine cell lines were subjected to CRISPR/Cas9-mediated knockout of B2M and JAK1 as described previously (16). The following single-guide RNA (sgRNAs) targeting B2M and JAK1 were used: forward: 5′-TCACGCCACCCACCGGAGAA-3′ reverse: 5′-TTCTCCGGTGGGTGGCGTGAC-3′ for B2M, and forward: 5′-CAGCGGAGAGTATACAGCCG-3′ reverse: 5′-CGGCTGTATACTCTCCGCTG-3′ for JAK1. These were cloned into the pSpCas9(BB)-2A-GFP vector (Addgene, catalog no. 48138; ref. 19) and then transformed into Stbl3 competent cells (Macro Lab Facility, University of California, Berkeley, CA, catalog no. Stbl3) and cultured overnight in Lysogeny Broth (LB; MP Biomedicals, catalog no. 113002016) plates containing ampicillin. Selected colonies were grown overnight in LB medium and DNA was isolated with the QIAprep midiprep kit (Qiagen, catalog no. 27104). To verify the sequence of the plasmids, a U6 promoter primer forward 5′-GCCTATTTCCCATGATTCCTTC-3′ was utilized. Next, cells were transfected using Lipofectamine 3000 following the manufacturer's protocol (Thermo Fisher Scientific, catalog no. L3000001), and GFP-positive cells were collected and single-cell sorted 48 to 72 hours posttransfection at the UCLA Jonsson Comprehensive Cancer Center (JCCC) Flow Cytometry Core. For each clone, genomic DNA was isolated using the NucleoSpin Tissue XS kit (Macherey-Nagel, catalog no. 740901.50), and a 700-bp region including the sgRNA was amplified via PCR using HotStarTaq Master Mix (Qiagen, catalog no. 203446). Finally, disruption was verified by Sanger sequencing utilizing the Tracking of Indels by Decomposition (TIDE; ref. 20) web tool and further confirmed via Western blot analysis.

Selected cell lines were maintained in 10-cm culture dishes and analyzed when 70% to 80% confluent. Western blot analysis was performed as described previously (16). Primary antibodies were specific for B2M (clone EP2978Y, ref 75853, Abcam) and GAPDH (clone D16H11, Cell Signaling Technology, catalog no. 5174). Immunoreactivity was analyzed with the ECL-Plus Kit (Amersham Biosciences Co.) using the ChemiDoc MP System (Bio-Rad Laboratories).

On day 1, murine cancer cell lines (B16, MC38, and YUMMER2.1) were plated at 2 × 10⁵ per 6-well plate and, when cell confluence reached 70%–80%, cells were collected for surface staining. The day after, culture media was replaced with fresh media with or without IFNγ 100 ng/mL (PeproTech, catalog no. 315-015) for 18 hours. On day 3 after incubation, cells were trypsinized and then incubated at 37°C for 2 hours with the same concentrations of IFNγ. Next, cells were centrifuged to remove the media and resuspended in 100% FBS. Cells were first stained with Zombie Green viability dye (BioLegend, catalog no. 423111) for 15 minutes, then washed and stained with PE-conjugated anti-mouse PD-L1 (clone MIH5, BD Biosciences, catalog no. 558091), APC-conjugated anti-mouse MHC I (clone 28-14-8, eBioscience, catalog no. 17-5999-82), and AF700-conjugated anti-mouse MHC II (clone M5, eBioscience, catalog no. 56-5321-82), and left on ice for 20 minutes. Cells were then washed once with 3 mL PBS and resuspended in 300 μL of PBS. Following staining, samples were analyzed using the Attune Flow Cytometer (Thermo Fisher Scientific) platform at the UCLA JCCC Flow Cytometry Core. Data were analyzed using FlowJo software (version 10.0.8r1, Tree Star Inc.). Experiments were performed at least in duplicate per cell line.

To seed subcutaneous tumors in mice, 0.3 × 10⁶ MC38, 0.3 × 10⁶ B16, or 1 × 10⁶ YUMMER2.1 wild-type (WT) or established B2M-KO cells were injected into the flanks of female C57BL/6N mice (6–10 weeks of age). Once tumors became palpable, four to six doses of 300 μg of anti–PD-1 or isotype control antibody were injected intraperitoneally every 3 days. Bempegaldesleukin was injected at 0.8 mg/kg every 9 days for one to two doses intravenously (tail vein). For depletion studies, 300 μg of anti-CD8, 300 μg of anti-CD4, 300 μg of anti-NK1.1, 200 μg of anti-CD40L, 200 μg of anti-IFNγ or various combinations were administered every 3 days starting the day before anti–PD-1 or bempegaldesleukin treatment and up until the end of the experiment. To validate depletion efficacy, splenocytes from control and depleted corresponding mice were harvested for comparison by flow cytometric analysis. Tumor growth was followed using caliper measurements two or three times per week and tumor volume was calculated using the following formula: tumor volume = ((width)² × length)/2. Mean and SE of the tumor volumes per group was calculated.

To characterize and quantify the dendritic cell (DC), T cell, and myeloid cell populations, mouse tumor samples were collected at day 9 and day 16 after mice were inoculated with MC38 WT or B2M-KO cells and treated with anti–PD-1 or isotype. Tumor samples were processed using the mouse tumor dissociation kit (Miltenyi Biotec, catalog no. 130-096-730) per the manufacturer's protocol. Samples were stained using the three antibody panels listed in Supplementary Table S1. Following staining, samples were analyzed using the LSRII (Becton, Dickinson and Company) at the UCLA JCCC Flow Cytometry Core and analyzed using FlowJo software (version 10.0.8r1, Tree Star Inc.).

A total of 0.3 × 10⁶ MC38 WT or established JAK1 and B2M KO tumor cells were implanted into the flanks of C57BL/6N mice. On day 13 postimplantation, tumors were harvested from mice that had received predefined treatments. Tumors were digested using the mouse tumor dissociation kit (Miltenyi Biotec, catalog no. 130-096-730). Spleens were dissociated and filtered using a 70-μm filter, followed by digestion with the ammonium-chloride-potassium (ACK) lysing buffer (Lonza, catalog no. BP10-548E). Sample staining and data acquisition were performed as previously described with the additional modifications that 3% paraformaldehyde was utilized and samples were not barcoded (16). The immune marker panel has been described previously (16). Samples were analyzed using the Fluidigm Helios mass cytometry system at the UCLA JCCC Flow Cytometry Core. Samples were manually gated for cells, singlets and double expression of the viable CD45 single-cell–positive population using FlowJo software (version 10.4.2), and data files were analyzed using the standard settings in OMIQ data analysis software (www.omiq.ai).

For human tumor biopsy analyses, a clinical dataset of patients with advanced melanoma (n = 514) that has been previously harmonized for somatic variant detection and gene expression profiling was utilized (21). This dataset contains clinical response, whole-exome sequencing (WES), and bulk RNA sequencing (RNA-seq) data across seven clinical trials of patients with advanced melanoma who received ICB therapy (anti–PD-1, anti–CTLA-4, or combination therapy). Copy-number alterations (CNA), LOH, and tumor purity estimates were determined using Sequenza (22). For gene expression analysis, the normalized log-counts per million (CPM) file that was batch-effect corrected with ComBat-seq was utilized (23).

Among the seven clinical trials, the Liu cohort (24), CheckMate 038 (9), CheckMate 064 (25), and CheckMate 067 (26) cohorts were further analyzed because these are the only trials that have both WES and RNA-seq data (Supplementary Fig. S1). For these analyses, pretreatment melanoma tumors of cutaneous, mucosal, acral, unknown, and uveal origins were utilized. Samples with best overall response categories of complete response (CR), partial response (PR), and progressive disease (PD) to ICB therapy were used for clinical response group comparisons, with CR/PR representing clinical responders and PD representing clinical nonresponders (RECIST v1.1; ref. 27). Stable disease (SD) tumors were excluded from clinical response group comparisons but included in somatic variant analysis.

Using the WES data (see Clinical dataset) of samples with greater than 10% tumor purity, baseline tumors were analyzed for B2M genetic alterations (n = 295; Supplementary Table S2). Tumors were considered B2M altered if they had nonsilent mutations, CNAs, or LOH at the B2M locus. Nonsilent mutations correspond to single-nucleotide variants and insertions and deletions with a high or moderate Ensembl VEP impact designation (https://uswest.ensembl.org/index.html). CNAs refer to somatic alterations that result in gained or lost copies of a genetic region. LOH describes events where one copy of an allele is lost, irrespective of copy-number status. For B2M group comparisons, tumors with B2M gains were excluded because they do not correspond to downregulating B2M genetic alterations. In addition, SD tumors were excluded from the dataset for clearer responder versus nonresponder comparisons (Supplementary Fig. S1).

Immune cell deconvolution of RNA-seq data (see Clinical dataset) was performed with CIBERSORTx (https://cibersortx.stanford.edu) using the LM22 signature matrix of 22 functionally defined mature human immune cell subsets (28–30), which distinguishes different cell types and cell states (naïve, memory, resting, and activated) based on 547 significantly differentially expressed genes (Supplementary Table S3). Each cohort was run separately using the uncorrected RNA-seq fragments per kilobase transcript per million mapped reads (FPKM) values to infer cell fractions and total cell numbers (absolute mode) per tumor sample (Supplementary Table S4). Each job was run following the recommended parameters using 1,000 permutations. The P values calculated by CIBERSORTx ranged from 0 to 0.857 in regular mode (median 0.065) and 0 to 0.836 (median 0.075) in absolute mode (Supplementary Fig. S2), and these values were loosely correlated with estimated tumor purity by Sequenza (WES; Supplementary Fig. S3). We performed comparisons of immune cell types across clinical groups two ways: all samples or excluding samples with P > 0.5. The significance in comparison across immune cell types was not different across these analyses, so the findings from the first comparison, including all samples, were reported. Differences in immune cell populations between groups were visualized and evaluated in R/RStudio (http://www.R-project.org/) by comparing the medians using a Wilcoxon test and were considered significantly different if the P value was less than 0.05.

Statistical analyses were performed using R/RStudio (v2022.07.1+554, http://www.R-project.org/). All plots analyzing B2M groups were generated using the R package ggplot2 (https://ggplot2.tidyverse.org/). For median group comparisons, P values less than 0.05 were considered statistically significant.

The genomics data used in this study are from a previous publication (21), with the raw sequencing data available through the Sequence Read Archive accession identifier PRJNA923698. Processed data from the harmonized biopsy sequencing dataset, including annotated variants and gene expression values, are available at https://github.com/ParkerICI/MORRISON-1-public. All other data generated in this study are available within the article and its Supplementary Data or upon request from the corresponding author.

Cancer cell lines have different levels of constitutive surface MHC class I expression and frequently do not express MHC class II, but both can be increased by exposure to IFNγ. Thus, using flow cytometry, we characterized the baseline and inducible surface expression of MHC class I and II on three murine cell lines, MC38, B16, and YUMMER 2.1. We also analyzed expression of PD-L1 as it is readily induced by IFNγ (31). To determine the role of B2M in modulating the expression of these surface markers, we generated corresponding B2M-KO sublines using CRISPR/Cas9 gene editing (16). Constitutive MHC class I surface expression was highest in MC38, intermediate in YUMMER2.1 and lowest in B16 (Fig. 1). The same trend was observed when staining for the constitutive surface expression of MHC class II. Exposure to IFNγ resulted in increased MHC class I and II expression in the three models. In all three models, B2M KO led to the loss of surface expression of both MHC class I and II, but maintained the upregulation of surface expression of PD-L1 in response to IFNγ (Fig. 1). We reasoned that these different levels of MHC class I expression may result in different sensitivities to effector immune cells in vivo, as MHC class I is required for tumor recognition by CD8⁺ T cells, and loss of MHC class I sensitizes target cells to NK-cell cytotoxicity.

To analyze the role of effector immune cell subsets in antitumor responses to MC38 murine colon adenocarcinoma tumors, we inoculated C57BL/6N mice with MC38 WT or MC38 B2M-KO tumors, followed by antibody-mediated CD4⁺ T-cell and/or CD8⁺ T-cell depletion in both groups, and antibody-mediated NK1.1⁺-cell depletion in B2M-KO tumors. In the MC38 WT control group, depletion of CD4⁺ T cells, CD8⁺ T cells, or both, abrogated the anti–PD-1 therapy antitumor response, while for WT untreated tumors, there were no significant effects with depletion of either immune cell subset (Fig. 2A). On the other hand, MC38 B2M-KO tumors did not respond to anti–PD-1 therapy. Untreated MC38 B2M-KO tumors grew with similar kinetics as the WT control group, but CD4⁺ T-cell depletion led to significant tumor overgrowth of MC38 B2M-KO tumors for both untreated and treated with anti–PD-1 therapy groups (Fig. 2B). NK-cell depletion in MC38 B2M-KO tumors had a lower impact, both in tumor growth and in response to anti–PD-1 therapy. Therefore, CD4⁺ T cells are critical for controlling the growth of MC38 B2M-KO tumors regardless of antitumor treatment.

To analyze the immune cells infiltrating MC38 WT and MC38 B2M-KO tumors, mice were inoculated with MC38 WT or MC38 B2M-KO lines and treated with isotype control or anti–PD-1 antibody (n = 3 mice per group). We harvested spleens and tumors 9 and 16 days after tumor inoculation and studied the cell populations by multiplex flow cytometry (Supplementary Fig. S4). Regardless of the anti–PD-1 treatment status, no differences were observed in the infiltration by CD4⁺ T cells, regulatory T cells (Treg), CD8⁺ T cells, or NK cells (Supplementary Fig. S5A). The percentage of CD62L⁺CD44⁺ central memory CD4⁺ T cells was significantly higher in MC38 B2M-KO tumors compared with MC38 WT tumors at day 9 after tumor-cell inoculation. The percentages of CD8⁺PD1⁺, CD8⁺Ki67⁺, and CD62L^–CD44⁺ effector CD8⁺ T cells were significantly decreased in MC38 B2M-KO tumors compared with MC38 WT tumors at day 16 after tumor-cell inoculation (Supplementary Fig. S5A), indicating that T-cell proliferation is occurring in favor of CD4⁺ T cells (without Tregs) over CD8⁺ T cells in MC38 B2M-KO tumors.

No significant differences in the percentages of tumor-associated macrophages (TAM), M1 macrophages, M2 macrophages, CD4⁺ or CD8⁺ T cells per gram were observed at day 9 (Supplementary Fig. S5B and S5C). However, M1 macrophage infiltration and the M1/M2 ratio were significantly decreased at day 16 in the MC38 B2M-KO tumors (anti–PD-1–resistant tumors) compared with MC38 WT tumors (Supplementary Fig. S5B). In addition, the number of NK cells per gram at day 9 in MC38 B2M-KO tumors was significantly lower relative to WT tumors (Supplementary Fig. S5C). In the MC38 B2M-KO tumors, we observed a sharp decrease in the percentage of type 1 conventional DCs (cDC1; migratory CD11c^intMHC-II^hiCD103⁺DCs and resident CD11c^hiMHC-II^intCD8a⁺DCs), a subset of DCs that are specialized in priming CD8⁺ T-cell responses through cross-presentation (32). In contrast, MC38 B2M-KO tumors had an increased percentage of type 2 conventional DCs (cDC2), resident CD11c^hiMHC-II^intCD11b⁺ DCs (Fig. 3; Supplementary Fig. S5B). These data suggest that loss of B2M in MC38 tumors abrogates the activation and migration of cDC1 regardless of anti–PD-1 treatment and promotes attraction of cDC2, which are more potent activators of CD4⁺ T cells, into the tumor microenvironment (32).

We demonstrated previously that treatment with the CD122-preferential IL2 agonist bempegaldesleukin alone or in combination with anti–PD-1 overcame therapeutic resistance to anti–PD-1 therapy in MC38 B2M-KO tumors (16). In the prior study, depletion analyses suggested that NK and CD4⁺ T cells, which are not restricted by MHC class I, played a key role in this antitumor immunity. We also demonstrated that despite tumor-intrinsic IFNγ defects, MC38 JAK1-KO tumors with sufficient basal expression of MHC class I could overcome anti–PD-1 therapy resistance with Toll-like receptor 9 (TLR9) agonist coadministration, an effect mediated mainly by CD8⁺ T cells. Both loss and downregulation of MHC class I antigen presentation are considered major immune escape mechanisms (6–8), but they have varying effects on effector immune cells. Thus, we here used mass cytometry by time of flight (CyTOF) analysis of the tumor microenvironment to characterize the effects of bempegaldesleukin and anti–PD-1 plus bempegaldesleukin on the immunologic response changes in MC38 WT, MC38 B2M-KO, and MC38 JAK1-KO tumors, as these readily respond to bempegaldesleukin combined with anti–PD-1 (Supplementary Fig. S6). At 13 days after tumor inoculation, the combination of bempegaldesleukin with anti–PD-1 led to increased CD4⁺ T cells and NK cells and reduced CD8⁺ T cells in MC38 B2M-KO tumors (Fig. 4). In contrast, CD8⁺ T, CD4⁺ T, and NK cells all increased in bempegaldesleukin and anti–PD-1–treated MC38 JAK1-KO tumors (Fig. 4). This is consistent with our previous findings in which CD8⁺ T cells were the main immune effectors in MC38 JAK1-KO tumors with sufficient basal MHC class I expression (16). CD103⁺ murine DCs decreased in MC38 B2M-KO tumors regardless of treatment (Figs. 3A and 4C; Supplementary Fig. S7A). On the other hand, bempegaldesleukin significantly expanded CD103⁺ murine DCs, which are required for cross-presentation, in both MC38 WT and JAK1-KO tumors. Altogether, our results support that the infiltration of cDC1 CD103⁺ murine DCs decreases in MC38 B2M-KO tumors regardless of treatment, and that anti–PD-1 therapy resistance can be overcome with an IL2 pathway agonist by activating CD4⁺ T-cell and NK-cell responses.

Because our data suggested that CD4⁺ T cells played a key role in MC38 B2M-KO tumors, we further explored the role that tumor-infiltrating CD4⁺ T cells may play in determining macrophage differentiation. While bempegaldesleukin treatment led to a minimal expansion of TAMs, isotype control and bempegaldesleukin with CD4⁺ T-cell depletion (where tumors grew faster) led to a dramatic increase in macrophages (Supplementary Fig. S7B). In addition, immunosuppressive M2 macrophages were significantly increased in bempegaldesleukin with CD4⁺ T-cell depletion, with a decrease in the M1/M2 index at day 13 after tumor inoculation. These data suggest that CD4⁺ T cells limit macrophage differentiation and/or proliferation toward an immunosuppressive M2 state.

B16 murine melanoma has low basal expression of MHC class I, which can be increased with IFNγ exposure (33), and is mainly dependent on CD8⁺ T cells for the antitumor response in the setting of anti–PD-L1 therapy (34). Our prior work showed that in B16 B2M-KO tumors, bempegaldesleukin or bempegaldesleukin in combination with anti–PD-1 overcame therapeutic resistance to anti–PD-1 (16). Herein, we investigated the cell types responsible for the antitumor activity of bempegaldesleukin in B16 B2M-KO tumors. Depletion of NK cells abolished the therapeutic effect of the IL2 agonist (Fig. 5). In contrast, depletion of CD4⁺ or CD8⁺ T cells had no effect. We further studied the molecules involved in the immune cell interactions by administering blocking antibodies against IFNγ or CD40L. We observed that the antitumor activity of bempegaldesleukin was lost when blocking IFNγ, but not when blocking CD40L (Fig. 5), which is canonically expressed on CD4⁺ T cells and plays an important role in the T cell–mediated activation of DCs (35, 36). These data suggest that NK cells and IFNγ are essential for overcoming anti–PD-1 therapy resistance in B16 B2M-KO tumors. In addition, these results show that CD4⁺ T cells and T-cell help through CD40L were not crucial for developing an effective antitumor immune response in this model.

The YUMM2.1 murine melanoma cell line provides a BRAF^V600E/PTEN-null–driven model that has been previously shown to be responsive to anti–PD-1 therapy (37), with CD4⁺ T-cell depletion completely abrogating the antitumor effect and CD8⁺ T-cell depletion only having a partial effect (17). To generate a model that not only included melanoma driver oncogenes but also had increased UV light–induced mutational burden, as it is frequently found in human cutaneous melanomas (38), we obtained a polyclonal YUMM2.1 by UV exposure and expanded single-cell clones to generate 24 UV-homogeneous sublines. This resulted in sublines with truncal UV-induced mutations able to respond to anti–PD-1 therapies (39). We evaluated the in vitro and in vivo growth curves of the 24 UV-homogeneous sublines to identify the subline that behaved most similarly to the parental UV-heterogeneous cell line to select for subsequent in vivo studies (Supplementary Fig. S8). The resulting YUMM2.1 UV-clone 2, named YUMMER2.1, is a model that more closely resembles human melanomas with BRAF^V600E and PTEN^−/− oncogenic driver alterations, a high mutational load induced by UV carcinogenesis, and anti–PD-1 therapy responsiveness (Fig. 6A and B). Next, we generated a B2M-KO subline of YUMMER2.1 using CRISPR/Cas9 gene editing and validated the phenotype by protein analysis (Supplementary Fig. S8). To model the in vivo response to PD-1 blockade in YUMMER2.1 B2M-deficient tumors, we injected YUMMER2.1 B2M-KO tumors subcutaneously into the lower flank of C57BL/6N mice. When tumors became palpable, mice received the first out of six injections of anti–PD-1 therapy or isotype control. Even with B2M loss, YUMMER2.1 B2M-KO tumors responded to anti–PD-1 therapy, showing that this model is largely sensitive to PD-1 blockade treatment (Fig. 6C).

To investigate the cell types responsible for the anti–PD-1 response in YUMMER2.1 B2M-KO tumors, we depleted CD4⁺ T cells, CD8⁺ T cells, NK cells, CD40L, and IFNγ. In agreement with previous data with YUMM2.1 (17), CD8⁺ T-cell depletion only partially abrogated the response, whereas CD4⁺ T-cell depletion and CD40L blockade completely abrogated the anti–PD-1 therapy response in YUMMER2.1 tumors (Fig. 6D and E). Depletion of CD4⁺ T cells, NK cells, and CD40L, but not CD8⁺ T cells or IFNγ, significantly curbed tumor growth inhibition by anti–PD-1 therapy (Fig. 6F and G). These results suggest that in YUMMER2.1 B2M-KO tumors, CD4⁺ T cells and NK cells are the main cells responsible and necessary for anti–PD-1 antitumor activity. In addition, we further identified the involvement of CD40L in the antitumor response in this B2M-deficient murine model.

We next wanted to study the immune cells infiltrating human melanoma biopsies with and without B2M downregulation or complete loss. To determine the B2M genetic status in biopsies of patients with advanced melanoma who received ICB therapy, we analyzed data from a harmonized clinical cohort (21). Baseline WES samples (n = 295; Supplementary Table S2; Supplementary Fig. S1) were analyzed for the presence of B2M somatic alterations, specifically nonsilent mutations, CNAs, or LOH. We found that none of the patients in this cohort had tumors with nonsilent mutations in B2M (Supplementary Fig. S9). However, 49.5% of the cohort (n = 146) had tumors with CNAs in B2M. Of the 295 patients analyzed, 0.7% of patients had biopsies with B2M homozygous loss (n = 2), 19% had B2M LOH with total copy numbers ranging from 1–5 (n = 56), and 30% had B2M copy-number gains with total copy numbers ranging from 3 to 7 (n = 88; Supplementary Fig. S9).

To elucidate the effects of downregulating B2M genetic alterations in the context of clinical response, we compared the clinical outcomes of patients whose tumors had B2M homozygous loss or LOH and unaltered status using the response RECIST v1.1 (27), separating patients with CR/PR from patients with disease progression (PD) following ICB therapy. This criterion, however, eliminated the two cases with SD and B2M homozygous loss (Supplementary Fig. S9), leaving only B2M LOH (n = 45) and unaltered tumors (n = 119) for analysis (Supplementary Fig. S1). We found that the presence of B2M LOH events in PD tumors (n = 32/95) was significantly higher compared with responsive tumors (n = 13/69), occurring approximately twice as often in biopsies of patients with no response to therapy (34% vs. 19%; χ²_Pearson test, P = 0.04; Fig. 7A). Next, to determine whether B2M LOH was associated with reduced B2M expression, bulk tumor RNA-seq data (n = 100) was analyzed for B2M expression differences (Supplementary Fig. S1). We found that B2M expression was lower in B2M LOH tumors compared with unaltered tumors (Wilcoxon test, P = 0.05; Fig. 7B). These results agree with prior studies that describe B2M LOH events as potential precursors to B2M loss and thus, reduced B2M expression (7).

To determine the immune cell subtypes present in B2M LOH (n = 25) versus unaltered tumors (n = 75), we performed immune cell deconvolution analysis on samples with paired WES and RNA-seq data using CIBERSORTx (https://cibersortx.stanford.edu). We found that B2M LOH tumors had significantly higher fractions (Wilcoxon test, P = 0.032) and quantified amounts (Wilcoxon test, P = 0.021) of activated NK cells (Fig. 7C; Supplementary Fig. S10A). It is also worth noting that compared with unaltered tumors, B2M LOH tumors had trends supporting greater fractions and numbers of M1 macrophages, γδ T cells, memory resting CD4⁺ T cells, plasma cells, memory B cells, CD8⁺ T cells, activated mast cells, and follicular Th cells (Fig. 7C; Supplementary Fig. S10A).

Furthermore, when examining the immune cell composition of B2M LOH tumors across ICB response groups, the fractions of monocytes were significantly higher in B2M LOH responsive tumors (n = 6, Wilcoxon test, P = 0.026), while the fractions of memory activated CD4⁺ T cells were significantly higher in B2M LOH progressive tumors (n = 19, Wilcoxon test, P = 0.014; Fig. 7D). These patterns were also observed with the absolute scores, where only monocytes and memory-activated CD4⁺ T cells were significantly different between both groups (Supplementary Fig. S10B). Notably, B2M LOH progressive tumors also had greater fraction and absolute score values for activated NK cells, γδ T cells, and CD8⁺ T cells; and a greater absolute score value for M1 macrophages (Fig. 7D; Supplementary Fig. S10B).

Next, to assess which subsets may be playing a more prominent role in mediating response to ICB therapy, we focused exclusively on responsive (CR/PR) melanoma tumors, comparing the estimated immune cell populations between B2M LOH tumors (n = 6) and B2M unaltered tumors (n = 35). It was observed that compared with unaltered samples, only monocytes were significantly higher in B2M LOH responsive tumors for both fractions (Wilcoxon test, P = 0.018) and absolute values (Wilcoxon test, P = 0.03; Supplementary Fig. S11). However, it is also important to note that B2M LOH responsive tumors had higher fractions and total numbers of activated NK cells and memory resting CD4⁺ T cells relative to B2M unaltered tumors (Supplementary Fig. S11). Our results indicate that in human melanoma biopsies with B2M dysregulation, activated NK cells are the most significantly elevated immune cell type infiltrating the tumors, with potential roles for other immune cells, such as CD4⁺ T cells, monocytes/M1 macrophages, and γδ T cells.

It has been well documented that cancers with B2M homozygous loss could still respond to PD-1 blockade–based therapies (9–15), which challenges the conventional thinking on the mode of action of anti–PD-1 by reinvigorating cytotoxic CD8⁺ T cells that recognize tumor antigens presented by MHC class I (2). The effector arm of a cellular antitumor immune response has two main cytotoxic cells, CD8⁺ T cells recognizing MHC class I antigen complexes and NK cells recognizing the absence of or a mismatched MHC class I. The need to avoid NK-cell killing seems to be a major driver for cancers to not lose MHC class I expression completely (40), even if this would allow the cancer to acquire resistance to ICB therapies (6–8). In addition, CD4⁺ T cells can have effector functions that induce cytotoxic cancer cell death, either directly or indirectly through the secretion of cytokines that promote bystander cytotoxicity. Furthermore, the helper function of CD4⁺ T cells, which involves cytokines such as IFNγ and CD40L engagement through DCs, results in increased cytotoxic activity of CD8⁺ T cells and NK cells (41).

We reasoned that identifying the immune cell mediators capable of exerting the antitumorigenic effects of anti–PD-1 therapy response in the absence of B2M is of critical importance to not only address antigen presentation-mediated therapeutic resistance, but also to better understand the mechanism of action of PD-1 blockade therapies. In this study, we used a combination of three different murine models with varying immune cell dependencies and human melanoma biopsy samples to interrogate the dominant immune cell subtypes and molecular markers associated with the immune response of B2M-deficient tumors. The YUMMER2.1 model is highly immunogenic and is able to respond to anti–PD-1 therapy, even in the absence of MHC class I presentation due to B2M loss, through CD4⁺ T cell and T-cell helper functions. The MC38 model has intermediate immunogenicity and does not respond to anti–PD-1 single-agent therapy when there is no MHC class I presentation. However, it can respond to therapy adding an IL2 pathway agonist, and in this case, the antitumor response is mediated by CD4⁺ T cells and NK cells, with a critical requirement for T-cell help and intratumoral DCs. The low immunogenicity B16 model is resistant to anti–PD-1 with or without B2M loss, but can respond with an IL2 pathway agonist and the response is completely dependent on NK cells with no added role for CD4⁺ T cells or T-cell helper functions. Our work in murine models has shown that CD4⁺ T cells are critical for the antitumor immunity in B2M-KO models. This is opposite of what has been reported in prior studies where it was observed that CD4⁺ T-cell depletion led to increased tumor reactivity and increased intratumoral infiltration of CD8⁺ T cells (42, 43). However, in our models, Tregs are not expected to play a significant role in tumor reactivity and response, as demonstrated by the MC38 line where WT and B2M-KO tumors had similar quantities of Tregs; thus, depleting CD4⁺ T cells in this scenario would not decrease the amount of immunosuppressive CD4⁺ T cells in the tumor microenvironment. Furthermore, because these are B2M-KO lines with a compromised MHC class I–CD8⁺ T cell immunity axis, an increase in tumor-specific CD8⁺ T cells due to CD4⁺ T-cell depletion would not lead to a therapeutic benefit. Our findings in these mouse models highlight the complexity of investigating immune responses in the context of B2M loss, stressing that tumor-intrinsic immune cell dependencies as well tumor microenvironmental factors can impact which effector cells control B2M-null tumor growth. Nonetheless, our work suggests that NK and CD4⁺ T cells are the dominant cell types curbing MHC class I–defective tumor expansion.

In our analysis of pretreatment human melanoma biopsies, we identified B2M LOH due to CNAs as the most prevalent downregulating genetic alteration and found no nonsilent mutations in B2M, which is in accordance with prior series (7, 24). We also observed that B2M LOH occurred approximately twice as often in biopsies from patients who did not respond to ICB therapy compared with those who had a clinical response. This falls in line with what was found in two independent melanoma clinical cohorts, where the frequency of B2M LOH events at baseline was roughly three times higher in biopsies of patients with PD (7). Our immune cell deconvolution analysis of B2M LOH melanoma biopsies showed that, among the immune cell types analyzed, only activated NK cells were significantly higher in B2M LOH tumors when compared with B2M unaltered tumors. In addition, we also observed higher amounts of activated NK cells in B2M LOH responsive tumors compared with B2M unaltered responsive tumors. This is consistent with the notion that MHC class I loss leaves tumor cells susceptible to NK cell–mediated killing (40, 44). Furthermore, this is consistent with similar work analyzing the transcriptional profiles of two datasets of patients with non–small cell lung cancer comparing tumors with low B2M expression against tumors with high B2M expression, where the fraction of activated NK cells was higher in B2M-low tumors (45). These results lend support toward NK cells potentially being the main mediators of MHC class I–negative tumor removal. We hypothesize that low or lost MHC class I expression in tumors leads to NK-cell activation. However, despite the greater numbers and proportions of activated NK cells found in B2M LOH melanomas, these biopsies were still associated with PD, and it was the B2M LOH progressive tumors that had the greatest amount of activated NK cells and other cytotoxic immune cell types. This suggests that there may be inhibitory molecules present in the tumor microenvironment that are impeding NK-cell cytotoxic activity, stressing the need for more in-depth studies into the tumor microenvironment of B2M-dysregulated tumors. In addition, in our samples, we found that M1 macrophages, memory resting CD4⁺ T cells, and γδ T cells were higher in B2M LOH tumors, with responsive tumors having greater amounts of monocytes and memory resting CD4⁺ T cells. Another striking observation because, as demonstrated by our murine models and previous studies, M1 macrophages that are no longer inhibited by the interaction of B2M with LILRB1 (46), and CD4⁺ T cells that are not dependent on surface MHC class I expression (13), can also play a role in eradicating B2M-null tumors. Similarly, it has recently been shown that γδ T cells, which are also not restricted by surface MHC class I expression, had high infiltration rates in B2M-inactivated MMR-d cancers and were associated with enhanced reactivity and cytotoxic activity against MHC class I–negative tumors (15).

In conclusion, cancers may induce different immune cell subset responses depending on their inherent immunogenicity and the level of surface MHC class I expression. The main immune cell subset mediating antitumor responses induced by PD-1 blockade in humans are CD8⁺ T cells (1), with the need for costimulation and help from CD4⁺ T cells when using PD-1 blockade therapies in implantable murine models where the immune system needs to be primed to induce tumor regression (17, 47, 48). As cancers attempt to evade the immune system by decreasing their immunogenicity through decreased tumor antigen presentation by MHC class I downregulation or loss, it triggers an alternate immune surveillance process mediated by NK cells, and potentially γδ T cells, with differing roles for CD4⁺ T cells depending on the model system.

D.Y. Torrejon reports employee of Ascendis Pharma Inc. K.M. Campbell reports personal fees and other support from Geneoscopy LLC; personal fees from Flagship Labs 81 LLC, PACT Pharma, Tango Therapeutics, and Rare Cancer Research Foundation outside the submitted work. A. Kalbasi reports research funding from Highlight Therapeutics. A. Ribas reports personal fees from Amgen, Chugai, Genentech, Merck, Novartis, Roche, Sanofi, Vedanta, 4C Biomed, Appia, Apricity, Arcus, Highlight, Compugen, ImaginAb, Kalthera-Immpact Bio, MapKure, Merus, Rgenix, Lutris, PACT Pharma, Synthekine, Tango, Advaxis, CytomX, Five Prime, RAPT, Isoplexis, and Kite-Gilead; and grants and personal fees from Agilent, Bristol Myers Squibb outside the submitted work. No disclosures were reported by the other authors.

D.Y. Torrejon: Conceptualization, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. M. Galvez: Conceptualization, data curation, software, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. G. Abril-Rodriguez: Formal analysis, investigation, methodology, writing–review and editing. K.M. Campbell: Data curation, software, formal analysis, supervision, investigation, visualization, methodology, writing–review and editing. E. Medina: Software, investigation, methodology, writing–review and editing. A. Vega-Crespo: Investigation, methodology, project administration, writing–review and editing. A. Kalbasi: Resources, investigation, methodology, writing–review and editing. B. Comin-Anduix: Formal analysis, visualization, methodology, project administration, writing–review and editing. A. Ribas: Conceptualization, resources, data curation, formal analysis, supervision, funding acquisition, investigation, methodology, writing–original draft, project administration, writing–review and editing.

This study was funded in part by the Parker Institute for Cancer Immunotherapy (PICI), NIH grants R35 CA197633 and P01 CA168585, the Ressler Family Fund, and the support from Ken and Donna Schultz, Todd and Donna Jones, Karen and James Witemyre, and Thomas Stutz through the Jonsson Cancer Center Foundation, and Jonathan Isaacson through the Melanoma Research Foundation (to A. Ribas). D.Y. Torrejon was supported by a Young Investigator Award from American Society of Clinical Oncology (ASCO), a grant from the Spanish Society of Medical Oncology for Translational Research in Reference Centers and the V Foundation-Gil Nickel Family Endowed Fellowship in Melanoma Research. M. Galvez is a predoctoral fellow supported by the UCLA Tumor Cell Biology Training Program Ruth L. Kirschstein Institutional National Research Service Award T32 CA009056 and the UCLA Medical Scientist Training Program (MSTP) NIH NIGMS Training Grant T32 GM008042. G. Abril-Rodriguez was supported by the Isabel & Harvey Kibel Fellowship Award and the Alan Ghitis Fellowship Award for Melanoma Research. K.M. Campbell is supported by the Cancer Research Institute Irvington Postdoctoral Fellowship Program, the V Foundation-Gil Nickel Family Endowed Fellowship in Melanoma Research, and the Parker Institute for Cancer Immunotherapy and V Foundation Bridge Fellows Program. Flow and mass cytometry were performed in the UCLA Jonsson Comprehensive Cancer Center (JCCC) and Center for AIDS Research Flow Cytometry Core Facility that is supported by NIH awards P30 CA016042 and 5P30 AI028697, and by the JCCC, the UCLA AIDS Institute, and the David Geffen School of Medicine at UCLA.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734.