Supplementary Figure 4. Hazard ratios for OS (in blue) and PFS (in red) derived from (A) unadjusted Cox proportional hazards models, (B) models stratified by CAR-T product, and (C) models both stratified by CAR-T product and adjusted for confounders, including pre-CAR-T age, LDH pre- lymphodepletion, and bridging. Each landmark time on the X-axis represents the time at which the Cox proportional hazards models are landmarked. For instance, the hazard ratios corresponding to landmark time 35 utilizes day 20 metrics (using the latest measurement within the pre-specified time range of days 10–35). Similarly, the hazard ratios corresponding to time 75 utilizes day 60 metrics (using the latest measurement within the pre- specified time range of days 50–75).

Supplementary Figure 7. Gating strategy for the T regulatory cells assay. (A, B) CD3+CD4+ T helper cells are passed through a CD25 vs CD127 dot plot to identify CD4+CD127-CD25+ T regulatory cells, and (C) a CD25 vs CCR4 dot plot to identify CCR4+CD25+ cells. T regulatory cells are defined as CD4+CD127-CCR4+CD25+ via a series of AND gates.

Innate immune cells respond rapidly to environmental cues through signal-regulated transcription factors (SRTFs) that sense changes in the tissue microenvironment. Signal transducer and activator of transcription (STAT) proteins are critical regulators of cytokine signaling and determine polarized immune responses. Herein, we reveal that activated type 2 innate lymphocytes (ILC2s) express STAT4, a SRTF canonically linked to type 1 immunity. STAT4 expression is induced in ILC2s upon activation by the alarmin IL-25 and linked with accumulation of lung inflammatory ILC2s (iILC2s). Despite elevated STAT4 expression, iILC2s do not acquire type 1 features, such as interferon (IFN)-γ production or T-bet expression and do not respond to IL-12 stimulation. Instead, STAT4 is activated by type I IFNs and supports the maintenance of the iILC2 pool. Transcriptomic analysis of Stat4-deficient ILC2s reveals enhanced type I IFN signaling and impaired proliferation, suggesting that STAT4 functions to antagonize IFN-driven suppression. Our data uncover a novel regulatory axis in which IL-25-induced STAT4 expression equips ILC2s to modulate interferon responses and to prevent aberrant autocrine function of type I IFNs, thus sustaining inflammatory effector populations during immune activation. These findings broaden the understanding of ILC2 activation and suggest new avenues for modulating innate lymphocytes in inflammatory diseases.

Supplementary Figure 6. Gating strategy for the T cell subsets assay. (A, B) CD3+CD4+ T helper cells are passed through a CCR7 vs CD45RA dot plot to identify CD4+CCR7+CD45RA+ naïve cells, CD4+CCR7+CD45RA- central memory (CM) cells, CD4+CCR7-CD45RA- effector memory (EM) cells, and CD4+CCR7-CD45RA+ effector cells. (C, D) Similarly, CD3+CD8+ T cytotoxic cells are passed through a CCR7 vs CD45RA dot plot to identify CD8+CCR7+CD45RA+ naïve cells, CD8+CCR7+CD45RA- CM cells, CD8+CCR7-CD45RA- EM cells, and CD8+CCR7-CD45RA+ effector (TEMRA) cells.

Supplementary Tables S1-S15 show additional analyses that support arguments and claims made in the main manuscript.

Supplementary Figure 3. Kaplan-Meier survival curves for OS and PFS. Comparisons include (A) OS and (B) PFS for CD4 <155 CD4 !155 CD4 <155 CD4 !155 CD4+ T cell counts above and below the median day 30 metric of 155 cells/μL. P-values are derived from multivariate Cox proportional hazards model stratified by CAR-T product and adjusted for confounders, including pre-CAR-T age, LDH pre-lymphodepletion, and bridging.

Supplementary Figure 5. Gating strategy for the TBNK assay. (A) FSC vs SSC dot plot was used to identify debris, characterized as low SSC and low FSC. Non-debris events are used for subsequent gating. (B) Using a FSC-A vs FSC-H dot plot, a singlet gate isused to include singlet and remove doublet events. (C) Singlet events are used in a CD45 vs SSC dot plot, and a gate is set around the CD45+ lymphocyte events, characterized by low SSC and bright CD45. (D) CD45+ lymphocyte events arethen passed through a CD3 vs CD19 dot plot to identify CD19+ B cells and CD3+ T cells. (E) CD3+ T cells are further identified as CD3+CD4+ T helper cells and CD3+CD8+ T cytotoxic cells in a CD4 vs CD8 dot plot. (F) Finally, CD56+CD16+ NK cells are identified in a CD3 vs CD16+56 dot plot.

Lymphocyte differentiation during infection depends on the integration of antigen and cytokine signals, yet how the timing and sequence of these cues program cell fate remains unclear. We found that interleukin-12 (IL-12) plays a context-dependent role in immune memory formation. Without prior antigen-receptor signaling, IL-12 drove cytotoxic lymphocytes toward terminal effector differentiation. In contrast, antigen signaling redirected IL-12-STAT4 activity through cooperation with AP-1 transcription factors to promote memory formation. This stepwise signal integration enabled lymphocytes to acquire memory rather than effector fates. Whereas CD8+ T cells were protected from premature IL-12 signaling by delayed receptor expression, natural killer (NK) cells, which constitutively express the IL-12 receptor, must engage their antigen receptor before cytokine signaling for efficient adaptive programming. Together, these findings define a framework in which sequential antigen and cytokine signaling coordinates effector versus memory differentiation, ensuring both robust primary responses and selective enrichment of high-avidity memory clones.

Supplementary Figure 1. Immune subset trajectories from time of CAR-T (time 0) through one year (time 365) following CAR-T with superimposed boxplots at predefined time-points. Subsets include (A) CD3+ T cells, (B) CD3-19+ B cells, (C) CD3-56+16+ NK cells, (D) CD4+ T cells, (E) CD4+CCR7-45RA- EM cells, (F) CD4+CCR7-45RA+ effector cells, (G) CD4+CCR7+45RA- CM cells, (H) CD4+CCR7+45RA+ naïve cells, (I) CD8+ T cells, (J) CD8+CCR7-45RA- EM cells, (K) CD8+CCR7-45RA+ TEMRA cells, (L) CD8+CCR7+45RA- CM cells, and (M) CD8+CCR7+45RA+ naïve cells. Trajectories depicted are estimated using LOESS or local regression modeling.

Supplementary Figure 2. Boxplots with superimposed dot plots demonstrate immune subset distributions at clinically significant timepoints, including days 30, 100, 180, and 365 post-CAR-T. Axi-cel in highlighted in blue, liso-cel in green, and tisa-cel in red. Subsets shown are (A) CD4+CCR7-45RA+ effector cells, (B) CD4+CCR7+45RA+ naïve cells, (C) CD4+CCR7+45RA- CM cells, (D) CD8+CCR7+45RA+ naïve cells, (E) CD8+CCR7-45RA+ TEMRA cells, (F) CD8+CCR7-45RA- EM cells, and (G) CD8+CCR7+45RA- CM cells. Ranges of days 20 to 45, days 80 to 120, days 150 to 210, and days 300 to 400 were used to calculate the distributions for the boxplots for days 30, 100, 180, and 365, respectively. Note differences in y-axes across panels. Asterisks above each set of bar plots represent P-values as calculated by the Kruskal- Wallis test. * <0.05, ** <0.01, *** <0.001, and **** <0.0001, with absence of any asterisks representing P > 0.05.

PD-L1 blockade benefits even PD-L1-negative tumors, suggesting that non-tumor cells contribute to PD-L1 expression. Natural killer (NK) cells, vital mediators of innate immunity, vigorously express PD-L1 upon activation. We demonstrate that the ligation of PD-L1 on circulating and tumor-infiltrating NK cells with the therapeutic anti-PD-L1 antibody atezolizumab, soluble PD-1, or PD-1+ cells enhances NK cell-mediated tumor clearance via changes in metabolism, adhesion, and migration. PD-L1 engagement increases NK cell tumor infiltration via the CXCR3 pathway and cytoskeletal remodeling, supported by a metabolic shift from glycolysis to fatty acid oxidation (FAO). Loss of a key FAO enzyme, CPT1A, in NK cells abrogates the PD-L1-mediated anti-tumor effect, supporting a critical role for FAO in enhanced NK cell killing. The PD-L1-triggered shift away from glycolysis permits NK cells to remain highly effective at tumor killing in glucose-restricted TME. Taken together, PD-L1 ligation enhances NK cell cytotoxicity and tumor infiltration and contributes to NK resilience in challenging TME conditions, resulting in a more effective anti-tumor immunity. One sentence summary PD-L1 engagement on NK cells enhances their tumor infiltration and cytotoxic activity by inducing a metabolic switch from glycolysis to fatty acid oxidation, enabling sustained function in the glucose-deprived tumor microenvironment. ### Competing Interest Statement K.C.H. is a member of the scientific advisory board for Wugen. B.S. receives research support from Genentech. First Eagle Investments Foundation

ABSTRACT:Patients treated with chimeric antigen receptor T-cell (CAR-T) therapy are subject to profound immunosuppression. Dynamics of immune reconstitution (IR) and impacts of IR on outcomes following infusion across CAR-T products are not well understood. In this study, we profiled IR in 263 patients with relapsed/refractory large B-cell lymphoma receiving CAR-T therapy (axicabtagene ciloleucel 44.9%, lisocabtagene maraleucel 30.4%, and tisagenlecleucel 24.7%). Following infusion, patients remain persistently immunosuppressed, with 48.1% having CD4+ T-cell counts <200/µL and the median CD3-CD19+ B-cell counts remaining zero through 1 year after CAR-T therapy. IR differences exist by product, with the fastest CD4+ T-cell recovery seen for tisagenlecleucel, driven primarily by more rapid recovery of the CD4+CCR7−CD45RA− effector memory subset. NK cell, but not CD4+ T cell, recovery is significantly associated with favorable progression-free (HR, 0.65; 95% confidence interval, 0.48–0.88) and overall survival (HR, 0.64; 95% confidence interval, 0.44–0.92) and inversely correlated with inflammatory markers measured at the time of infusion. SIGNIFICANCE:This study reveals differences in IR patterns after CAR-T therapy in patients with large B-cell lymphoma, with early NK cell recovery emerging as a key predictor of survival. These findings provide potential future avenues of research for improving patient outcomes and tailoring post-therapy management strategies to mitigate relapse risk.

HLA evolutionary divergence (HED) can serve as a surrogate for the degree of immunopeptidome diversity of HLA phenotypes. Different degrees of HED in the patient-donor pair may influence the presentation of peptides relevant for alloreactive responses involved in graft-versus-host and graft-versus-leukemia (GvL) effects, potentially impacting outcomes after hematopoietic cell transplantation (HCT). This study was conducted to test whether higher HED scores (both class I and class II) correlate with improved GvL and survival after HLA-matched HCT. The study cohort comprised pediatric and adult patients (n = 9231) reported to the Center for International Blood and Marrow Transplant Research (CIBMTR) database who underwent a first HCT from 8/8 matched unrelated donors between 2008 and 2018 for the treatment of acute myeloid leukemia, acute lymphoblastic leukemia, myelodysplastic syndrome, chronic myeloid leukemia, or lymphoma were included. HED was calculated on the amino acid sequences of HLA-A, -B, -C, and -DRB1, and class I (HLA-A, -B, and -C), and HLA-DRB1 HED scores were assigned to each patient-donor pair. The association between increasing HED (top quartile versus lower 3 quartiles) and HCT outcome was evaluated with malignant disease relapse, disease-free survival (DFS), and overall survival (OS) as primary endpoints. Secondary endpoints were transplantation-related mortality (TRM), acute and chronic graft-versus-host disease (GvHD), and engraftment. Greater HLA-DRB1 HED was associated with significantly decreased malignant disease relapse (hazard ratio [HR], 0.86; 95% confidence interval [CI], 0.79 to 0.94; P = .0014), better disease-free survival (HR, 0.92; 95% CI, 0.86 to 0.98; P = .0067), and improved OS (HR, 0.91; 95% CI, 0.85 to 0.96; P = .0019) in the total population after adjustment for other significant clinical variables. There was no significant association between HLA-DRB1 HED and TRM or the risk of acute or chronic GVHD. Conversely, higher (upper quartile) HLA class I HED did not significantly impact OS, DFS, TRM, relapse, or acute and chronic GVHD compared with the lower 3 quartiles. In addition, neither HLA-class I nor HLA-DRB1 HED significantly impacted neutrophil or platelet engraftment post-transplantation. Our findings demonstrate that higher HLA-DRB1 HED scores are associated with reduced relapse and improved DFS and OS in patients undergoing matched unrelated donor transplantation for hematologic malignancies. These findings contribute to the growing evidence supporting the importance of HED in post-transplantation outcomes; however, further refinement and validation are needed before incorporating HED into clinical transplantation risk assessment.

Chimeric antigen receptor (CAR)-engineered lymphocytes treat B cell malignancies; however, limited persistence can restrain the full therapeutic potential of this approach. FAS ligand (FAS-L)/FAS interactions govern lymphocyte homeostasis. Knowledge of which cells express FAS-L in patients with cancer and whether these sources compromise CAR persistence remains incomplete. Here, we constructed a single-cell atlas of diverse cancers to identify cellular subsets expressing FASLG, the gene encoding FAS-L. We discovered that FASLG expression is limited primarily to endogenous T cells, natural killer (NK) cells and CAR-T cells, while tumor and stromal cell expression is minimal. To establish whether CAR-T and CAR-NK cell survival is FAS-L regulated, we performed competitive fitness assays using FAS-dominant negative receptor (ΔFAS)-modified lymphocytes. Following transfer, ΔFAS-expressing CAR-T/CAR-NK cells became enriched, a phenomenon that mechanistically was reverted through FASLG knockout. By contrast, FASLG was dispensable for CAR-mediated tumor killing. In multiple models in female mice, ΔFAS coexpression enhanced antitumor efficacy. Together, these findings reveal that CAR-engineered lymphocyte persistence is governed by a FAS-L/FAS autoregulatory circuit. Klebanoff and colleagues report that survival and persistence of CAR-T and CAR-NK cells are regulated by a FAS ligand–FAS autoregulatory circuit, showing that disabling FAS signaling enhances their antitumor efficacy in preclinical models.

Natural Killer (NK) cells can recognize and kill Mycobacterium tuberculosis (Mtb)-infected cells in vitro, however their role after natural human exposure has not been well-studied. To identify Mtb-responsive NK cell populations, we analyzed the peripheral blood of healthy household contacts of active Tuberculosis (TB) cases and source community donors in an endemic region of Port-au-Prince, Haiti by flow cytometry. We observed higher CD8α expression on NK cells in putative resistors (Interferon γ release assay negative; IGRA- contacts) with a loss of CD8α surface expression during household-associated exposure and active TB disease. In vitro assays and CITE-seq analysis of CD8α+ NK cells demonstrated enhanced maturity, cytotoxic gene expression, and response to cytokine stimulation relative to CD8α- NK cells. CD8α+ NK cells also displayed dynamic surface expression dependent on MHC class I in contrast to conventional CD8+ T cells. Together, these results support a specialized role for CD8α+ NK cell populations during Mtb infection correlating with disease resistance.

Engineered T and NK cell therapies have widely been used to treat hematologic malignancies and solid tumors, with promising clinical results. Current chimeric antigen receptor (CAR) T cell therapeutics have, however, been associated with treatment-related adverse events such as cytokine release syndrome (CRS) and are prone to immunologic exhaustion. CAR-NK therapeutics, while not associated with CRS, have limited in vivo persistence. We now demonstrate that an NK-like TCRαβ + CD8 T cell subset, identified and expanded ex vivo through its expression of the activating receptor NKG2C (NKG2C + NK-like T cells), can be transduced to express a second-generation CD19 CAR (1928z), resulting in superior tumor clearance, longer persistence and decreased exhaustion compared to conventional 1928z CAR + CD8 T cells and 1928z CAR+ NK cells. Moreover, CAR-modified NKG2C + NK-like T cells resulted in significantly reduced CRS compared to conventional CAR + CD8 T cells. Similarly, NKG2C + NK-like T cells engineered with a TCR targeting the NY-ESO-1 antigen exhibit robust tumor control and minimal exhaustion compared to TCR-engineered conventional CD8 T cells. These data establish NKG2C + NK-like T cells as a robust platform for cell engineering, and offer a safer, more durable alternative to conventional CAR-T and CAR-NK therapies.

Allogeneic hematopoetic cell transplantation harnesses donor T cell alloreactivity against leukemic blasts and is curative in a subset of patients with AML. Aside from transplant, however, T cell based immunotherapies have been unsuccessful in AML and in patients with AML there is evidence for impaired endogenous immune responses. Despite these observations, mechanisms underpinning ineffective anti-leukemic T cell immunity are not fully known. Here we hypothesized that leukemic blasts drive impaired T cell immunity leading to distinct T cell compositions during different disease states. Using multi-modal approaches to study T cell phenotype and T cell receptor (TCR) repertoire across AML disease states we identified dominant clonally expanded terminal effector memory CD45RA+ (TEMRA) CD8 T cells in the marrow of AML patients with active disease along with abundant immunosuppressive CD4 T regulatory cells (Tregs). CD8 TEMRA clones maintain over time in patients with persistent AML and exhibited numerous interactions with malignant blasts suggestive of ongoing immune modulation by antigen producing leukemic cells. A subset of these CD8 effectors (expressing CX3CR1 and other NK-like markers) exert anti-tumor cytotoxic activity ex vivo but are suppressed by interactions with marrow Tregs. Consistent with this, Treg depletion rescued CD8 effector activity and promoted AML eradication. To study leukemic blasts and T cell immunity in the AML tumor microenvironment (TME), we performed integrative analysis of protein (CITE-seq and 31-color spectral flow), transcript, and TCRs in individual lymphoid and myeloid cells from longitudinal marrows. We applied this to 179 patient samples from 91 subjects (71 AML, 20 controls), sequencing >670K cells total and >190K T cells. We found that patients harbor a highly abundant CD8 TEMRA population at AML diagnosis that persisted over time in patients who did not achieve remission. These cells express clonally expanded TCRs, a subset of which were marked by CX3CR1, TIGIT (but not PD-1 or TIM3) and attributes of cytotoxicity including granzyme B, perforin, and NK-like markers. To interrogate the function of expanded effector CD8 T cells, we investigated marrow T cells in an unirradiated syngeneic AML mouse model (C1498 cells into C57B6 mice). As AML accumulated in the marrow, TIGIT+ effector CD8s increased in frequency, as in human AML. A subset of these effectors expressed CX3CR1, which we hypothesized might have tumor killing capability given their cytotoxic profile in patient data. Indeed, in vitro functional analysis revealed increased killing of endogenous tumor by marrow CD8 effectors including CX3CR1+ effectors compared to naïve CD8s harvested 18-20 days post-tumor injection. Importantly in patients with AML we harnessed the TCR CDR3 sequence as a barcode to track phenotypes of CD8 clonotypes over time and found that certain CX3CR1+ TEMRAs transition to a CX3CR1- state in ongoing disease, suggesting loss of cytotoxicity. Analyses of cell-cell interactions from CITE-seq suggested altered myeloid-T cell interactions in AML compared to control and remission samples, including increased signaling between myeloid cells, Tregs, and memory CD8s in patients with active AML. AML blasts also had increased expression of T cell inhibitory molecules including TIGIT ligands, CD244, and VISTA compared to healthy myeloid cells. Notably, in addition to expanded CD8 TEMRAs, we found increased Tregs in marrow from patients with AML and C1498 engrafted mice. Tregs in both human and mouse AML expressed high levels of TIGIT, CD39, ICOS, and CCR4 but were not clonally expanded. Ex vivo AML marrow Tregs suppressed CD8 effector function. Importantly depletion of Tregs in vivo through transgenic FoxP3 diphtheria toxin receptor mice prolonged host survival, promoted tumor clearance, and led to an increase in marrow CX3CR1+ effector CD8 T cells. These data demonstrate the active immunologic landscape of the AML bone marrow in both patient samples and mouse models of the disease. We find that although CD8 T cells have the potential for anti-leukemic immunity, their efficacy is impaired by leukemic blasts and suppressive Tregs. These studies suggest Treg-targeting interventions as a therapeutic avenue to overcome the immunosuppressive TME in AML and nominate a host of potentially targetable T cell and blast cell surface proteins that restrain T cell anti-tumor immunity in AML.

Fasting is associated with improved outcomes in cancer. Here, we investigated the impact of fasting on natural killer (NK) cell anti-tumor immunity. Cyclic fasting improved immunity against solid and metastatic tumors in an NK cell-dependent manner. During fasting, NK cells underwent redistribution from peripheral tissues to the bone marrow (BM). In humans, fasting also reduced circulating NK cell numbers. NK cells in the spleen of fasted mice were metabolically rewired by elevated concentrations of fatty acids and glucocorticoids, augmenting fatty acid metabolism via increased expression of the enzyme CPT1A, and Cpt1a deletion impaired NK cell survival and function in this setting. In parallel, redistribution of NK cells to the BM during fasting required the trafficking mediators S1PR5 and CXCR4. These cells were primed by an increased pool of interleukin (IL)-12-expressing BM myeloid cells, which improved IFN-γ production. Our findings identify a link between dietary restriction and optimized innate immune responses, with the potential to enhance immunotherapy strategies.

Abstract We have recently characterized a unique CD8 T cell subset that expresses NKG2C, as well as multiple other natural killer cell markers. These NKG2C+ CD8 T cells demonstrate NK-like effector functions against various tumor targets, can be efficiently expanded and engineered using GMP-compatible techniques, and most importantly, display beneficial qualities unique to T and NK cells while minimizing the therapeutic deficits of each. NKG2C+ CD8 cells can be easily expanded using a genetically modified K562 feeder line and can be efficiently engineered via retroviral transduction. We genetically engineered NKG2C+ CD8 T cells to express an anti-CD19-28-z CAR construct to target B cell malignancies and compared effector functions of the CAR-engineered NKG2C+ CD8 T cells against CAR+NKG2C- (conventional) CD8 T cells and CAR+ NK cells in vitro and in vivo. We examined expansion, engineering efficiency, tumor control, and extent of cytokine release syndrome (CRS). We also engineered NKG2C+ CD8 T cells with a transgenic NY-ESO-1 TCR to target melanoma and compared the functional responses and efficacy of TCR-engineered NKG2C+ CD8 T cells against TCR-engineered conventional CD8 T cells. CAR+NKG2C+ CD8 T cells demonstrated superior in vitro effector function when compared to CAR+ conventional CD8 T cells and CAR+NK cells and did not upregulate characteristic T cell exhaustion markers, such as PD-1, TIM-3, and LAG-3. These CAR+NKG2C+ CD8 T cells demonstrated superior control of a CD19+ NALM6 tumor model in NRG mice, which was further enhanced with administration of IL-15. CAR+NKG2C+ CD8 T cells with IL-15 demonstrate dramatic and enhanced persistence, as well as superior cytotoxicity and activation in vivo when compared to CAR+ conventional CD8 T cells and CAR+ NK cells. Using a SCID-beige CRS mouse model, we observed reduced CRS in mice treated with CAR+ NKG2C+ CD8 T cells than in mice treated with CAR+ conventional CD8 T cells, with lower levels of proinflammatory cytokines and minimal expansion of murine myeloid subsets. Similarly, TCR-engineered NKG2C+ CD8 T cells demonstrated enhanced tumor control and in vivo persistence in an A375 melanoma-bearing NRG mouse model compared to TCR-engineered conventional CD8 T cells. We have successfully demonstrated that CAR- or TCR-engineered NKG2C+ CD8 T cells have superior anti-tumor efficacy, in vitro and in vivo, to conventional CD8 T cell and NK cell counterparts due to their innate tumor killing and greater in vivo persistence. These cells mediate less CRS and provide a safer alternative to traditional engineered T cell therapeutics. Ultimately, our NKG2C+ CD8 T cells represent a unique platform for engineered T cell therapies to target both solid tumors and hematologic malignancies, that is easily expanded ex vivo, does not require CD4 T cells for in vivo persistence, and is resistant to exhaustion. Citation Format: Kyle Lupo, M. Kazim Panjwani, Anthony Daniyan, Christopher Klebanoff, Katharine Hsu. Engineered NK-like NKG2C+ CD8 T cells mediate superior anti-tumor efficacy over conventional CD8 T cells and NK cells [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 5233.