Low-Strength Type I Interferon Signaling Promotes CAR T-Cell Treatment Efficacy

Abstract

CD19-directed chimeric antigen receptor (CAR) T-cell therapy has significantly advanced the treatment landscape for relapsed/refractory diffuse large B-cell lymphoma (r/r DLBCL). However, up to 60% of patients do not achieve a complete response. To uncover determinants of therapeutic efficacy, we analyzed the infusion products of eight r/r DLBCL patients with distinct clinical responses to axicabtagene ciloleucel using single-cell transcriptomics. Compared to patients who exhibited progressive disease, infusion products of complete responders demonstrated enriched signatures of type I interferon (IFN-I) signaling. Based on these findings, we developed a novel strategy to improve CD19-directed CAR T-cell treatment efficacy by incorporating IFN-I as an enhancer during the ex vivo manufacturing process. For both CD28- and 4-1BB-costimulated second-generation CARs, we found that low-strength IFN-I signaling enhanced CAR T-cell cytotoxicity and in vivo efficacy. On the other hand, high-strength IFN-I signaling compromised cell viability and in vivo efficacy. Our low-strength IFN-I signaling approach leverages an existing FDA-approved pharmacologic agent and is compatible with current CAR constructs and manufacturing workflows. Together, our results establish IFN-I as a potent and costimulation-independent enhancer of CAR T-cell efficacy and provide a translationally feasible approach to enhance CAR T-cell therapies for r/r DLBCL.

INTRODUCTION

Chimeric antigen receptor (CAR) T-cell therapy has advanced the standard-of-care for hematologic malignancies, achieving complete response rates ranging 39-71% across relapsed/refractory diffuse large B-cell lymphoma (r/r DLBCL)¹, acute lymphoblastic leukemia², and multiple myeloma^3,4. However, the therapy's effectiveness is constrained by significant rates of treatment failure. In r/r DLBCL, only ~40% of patients exhibit a complete response, whereas the remaining ~60% of patients exhibit a partial response, stable disease, or progressive disease, as defined by the International Working Group response criteria.¹ Patients who cannot clear their lymphoma burden have limited therapeutic options and often succumb to their disease.

Existing studies have correlated treatment failure with diminished CAR T-cell polyfunctionality⁵, increased T-cell exhaustion signatures within CAR T-cell products^6-8, inhibition by CAR Tregs⁹, as well as failure of CAR T cells to persist and expand in vivo^1,10, among others¹¹. Given the importance of CAR T-cell-intrinsic factors in determining treatment outcomes, we investigated the infusion products (IP) of CAR T-cell therapy patients with r/r DLBCL using single-cell transcriptomics. We observed that, compared to patients with progressive disease, the IPs of complete responders demonstrated enriched signatures of type I interferon (IFN-I) signaling. This observation was also validated using an independent, published dataset from Maus et al.¹²

As potent immunomodulators, IFN-I cytokines have long been of interest for their potential to modulate immunotherapy response rates.^13,14 Upon binding to type I interferon receptors (IFNAR), IFN-I cytokines trigger a phosphorylation cascade that activates transcription factors such as signal transducer and activator of transcription 1 (STAT1), STAT2, and interferon regulatory factor 9 (IRF9).¹³ These transcription factors coordinate transcription of interferon-stimulated genes, such as IRF7, MX1, ISG15, ISG20, IFITM1/2/3 which mediate antiviral immunity and effector functions.^13,15 In CD8⁺ T cells, IFN-I signaling enhances cytotoxicity¹⁶ and induces central memory differentiation¹⁷, while also controlling T-cell exhaustion¹⁸, inhibitory receptor expression¹⁹, and apoptosis^20,21. Due to their immunomodulatory effects, clinical formulations of IFN-I cytokines (e.g., Avonex, Intron A, Alferon N) have been approved by the FDA for the treatment of malignancies (e.g., hairy cell leukemia, follicular lymphoma)²², autoimmune diseases (e.g., multiple sclerosis)²³, and viral infections (e.g., hepatitis B and C)²⁴. Despite being an attractive FDA-approved cytokine for regulating immune functions in cancer, autoimmunity, and infection, IFN-I has not been utilized to modulate CAR T-cell therapy responses.

Given our bioinformatic discoveries and the literature-established roles of T-cell-intrinsic IFN-I signaling in modulating T-cell differentiation and functionality¹³, we hypothesize that IFN-I signaling enhances CAR T-cell efficacy during the ex vivo manufacturing process. In this study, we detail our discovery and development of IFN-I cytokines as an enhancer to partially address CAR T-cell therapy failure rates. Through in vitro functional assays and in vivo lymphoma xenograft mouse models, we demonstrate proof-of-principle for a novel ex vivo manufacturing method that leverages low-strength, FDA-approved IFN-I cytokines to improve the in vivo treatment efficacy of CAR T-cell therapy.

RESULTS

Enrichment of IFN-I signaling in infusion products of complete responders.

To uncover determinants of therapeutic efficacy, infusion products from 8 r/r DLBCL patients treated with CD28-costimulated CD19-directed CAR T cell therapy (axicabtagene ciloleucel, “axi-cel”) at the University of Chicago Medicine were profiled using single-cell sequencing. The cohort included 5 patients who achieved a complete response (CR) at one month post-infusion and another 3 patients who exhibited progressive disease (N-CR) (Figure 1a). After removing low-quality cells, we obtained 27,879 cells distributed among 7 T-cell clusters (Figure 1b) based on gene expression (Figure 1c, figure S1a). No cluster was patient specific (Figure 1d, figure S1b). The two CD8⁺ clusters were annotated as central memory (CM, SELL⁺IL7R⁺TBX21⁻KLRG1⁻) or effector memory (EM, SELL⁻IL7R⁺TBX21⁺KLRG1⁺) T cells. The two CD4⁺ clusters were annotated as memory (Mem, FOXP3^loIL7R⁺CTLA4^lo) or regulatory (Treg, FOXP3^hiIL7R⁻CTLA4^hi) T cells. The remaining three clusters consisted of naïve-activated (SELL⁺IL7R⁺TCF7⁺CD38⁺), unconventional (TRDC⁺NCAM1⁺) as a mixture of γδ and natural killer phenotypes, and proliferating (MKI67⁺TOP2A⁺) T cells. No discernable composition differences were found between CR and N-CR patients (Figure 1b, d, figure S1c).

Figure 1. Enrichment of IFN-I signaling in infusion products of complete responders.

(a) Cartoon depicting study design of single-cell analysis on infusion products of patients who received axi-cel from discovery dataset and validation datasets. CR, complete responder, N-CR, non-complete responder. (b) UMAP demonstrating 7 T-cell clusters in infusion product from discovery dataset. (c) Violin plots depicting normalized expression levels of representative markers for each cluster. (d) Stacked bar graphs depicting T-cell cluster proportions for each patient in discovery dataset. (e) Gene set enrichment analysis comparing CR and N-CR T cells. For each gene set, direction and statistical significance of enrichment are indicated by the circle’s color and size, respectively. (f) Violin plots depicting IFNα response signature scores in CR and N-CR T cells from each T-cell cluster. Signature scores were compared by Wilcoxon Rank-Sum test. (g) Tilemap depicting scaled IFNα response signature scores in each T-cell cluster across patients. (h) UMAP demonstrating 7 T-cell clusters in infusion product from validation dataset from Maus et al. (i) Violin plots depicting normalized expression levels of representative markers for each cluster. (j) Stacked bar graphs depicting T-cell cluster proportions for each patient in validation dataset. (k) Violin plots depicting signature scores in CR and N-CR T cells from each T-cell cluster. Signature scores were compared by Wilcoxon Rank-Sum test. (l) Tilemap depicting scaled signature scores of IFNα response gene sets in each T-cell cluster across patients. No Treg was found in P17 infusion products, colored in gray.

Gene set enrichment analysis revealed distinct differences between CR and N-CR T cells. CR T cells were enriched for signatures of type I interferon signaling (“IFNα response”), cytolysis, and leukocyte degranulation, whereas N-CR T cells were enriched for signatures of cell growth, hypoxia, and glycolysis (Figure 1e). Notably, the CR-specific IFNα response signature was enriched in CR T cells across T-cell clusters (Figure 1f) and individual patients (Figure 1g).

To validate our discovery of CR-specific IFN-I signaling signatures, we tested our findings using an independent and published single-cell transcriptomic dataset from Maus et al. (n = 15 patients with lymphoma who received axi-cel, Figure 1a).¹² This cohort included 10 patients who achieved a complete response (CR) and 5 who did not (N-CR), as assessed one-month post-infusion. After filtering out low-quality cells, performing dimensionality reduction, and clustering, we uncovered the same 7 T-cell clusters identified in our own dataset (Figure 1h) based on gene expression (Figure 1i, figure S2a). No cluster was patient specific (Figure 1j). Notably, expression of ISGs (including ISG20, IFITM1/3) was upregulated in IP of CR (figure S2b). Most importantly, IFNα response signatures was enriched in IPs of CR. The CR-specific IFNα response signature was consistently observed across all T-cell clusters except Tregs (Figure 1k, figure S2c) and individual patients (Figure 1l).

Taken together, findings from the discovery and validation datasets strongly support an association between IFN-I signaling and improved CAR T-cell clinical efficacy.

IFN-I signaling increases CAR T-cell cytotoxicity.

Previous studies have demonstrated the immunomodulatory role of T cell-intrinsic IFN-I signaling.¹³ For instance, IFN-I signaling promotes cytotoxicity¹⁶, T_h1 polarization^25-27, and central memory phenotype differentiation¹⁷, all of which may be desired CAR T-cell characteristics and potentially explain the differential clinical responses observed in the two patient cohorts. Based on these findings and our bioinformatics analyses, we hypothesized that IFN-I signaling can enhance CAR T-cell efficacy. To test this hypothesis, we generated both CD28- and 4-1BB-costimulated CD19-directed CAR T cells from primary human T cells. Our CD28-costimulated CAR is based on axicabtagene ciloleucel (axi-cel, Figure 2b), whereas our 4-1BB-costimulated is based on lisocabtagene maraleucel (liso-cel, Figure 2c). To induce IFN-I signaling, we treated CAR T cells with exogenous IFN-α2, a highly potent, FDA-approved, and clinically compatible IFN-I subtype, for culturing CAR T cells (Figure 2a). Prior to addition of IFN-α2, we confirmed that CAR T cells globally expressed the type I interferon receptor in a transduction-independent and costimulation-independent manner (Figure S3a, c).

Figure 2. IFN-I signaling increases CAR T-cell cytotoxicity.

(a) Workflow of CAR T-cell ex vivo generation, treatment, and culturing. (b-c) CD28- or 4-1BB-costimulated CAR designs based on the construct used in axicabtagene ciloleucel (b) or lisocabtagene maraleucel, respectively (c). (d-e) Representative flow plots depicting gating for CAR expression (top) and T-cell lineages (bottom) in CD28-costimulated (d) and 4-1BB-costimulated (e) CAR T cells. (f, h) Transduction percentage as measured by meGFP fluorescence among T cells from each culturing condition in CD28-costimulated (f, n=3) and 4-1BB-costimulated (h, n=3) CAR T cells. (g, i) Stacked bar plot comparing CD8 vs CD4 proportions in CD28-costimulated (g, n=3) and 4-1BB-costimulated (i, n=3) CAR T cells. (j-m) Cytotoxicity of CD28-costimulated (j-k) and 4-1BB-costimulated (l-m) CAR T cells. (j, l) direct cytotoxicity against firefly luciferase-expressing OCI-Ly8 target cells at E/T ratio of 1:1 via 6-hr bioluminescence assay (n=5). (k, m) intracellular staining for granzyme B (right) and perforin (left) after 4-hr of coculture with OCI-Ly8 at E/T ratio of 1:1 as measured by flow cytometry (n=3). (n-o) Concentration of cytokines secreted by CD28-costimulated (n) and 4-1BB-costimulated (o) CAR T cells into the supernatant after 24-hr of coculture with OCI-Ly8 at E/T ratio of 1:1 as measured by ELISA (n=5). From left to right, IL-2, TNF, and IFNγ. All data are reported as mean ± standard error. Statistical analysis were performed via one-way (f, h, j-o) or two-way (g, i) ANOVA with Tukey correction for multiple comparisons. All panels are representative data from 3 independent donors.

Since each downstream target of IFN-I signaling exhibits varying sensitivity to IFN-I concentrations²⁸, we hypothesized that the IFN-I culturing concentration may affect CAR T-cell functions. To determine the IFN-I concentrations for our study, we initially titrated IFN-α2 on both CD28- and 4-1BB-costimulated CAR T cells and measured intracellular interferon regulatory factor 7 (IRF7), a key ISG²⁹ (figure S3b, d). IFN-I signaling reached saturation at concentrations above 3000 IU/mL of IFN-α2, consistent with previous studies.^16,27 We also observed half-maximal IRF7 upregulation at approximately 100-300 IU/mL of IFN-α2. Consequently, we established three culturing conditions: IFN-α2 at saturating concentration (“IFNα-hi”, 3000 IU/mL), IFN-α2 near half-maximal effective concentration (“IFNα-lo”, 300 IU/mL), and no IFN-α2 (“IFNα-neg”, 0 IU/mL). We confirmed that these conditions resulted in concentration-dependent upregulation of ISGs, including MX1, IRF7, and ISG15 (figure S4).

CAR T cells were treated based on the three culturing conditions for downstream experimentation (Figure 2a). Since IFN-I signaling can inhibit viral transduction³⁰, we chose to add IFN-α2 after CAR transduction, which enabled us to achieve consistent transduction efficiency and expression of both CD28-costimulated (Figure 2d, f) and 4-1BB-costimulated CARs (Figure 2e, h) across the three culturing conditions. Between culturing conditions, IFN-I signaling did not affect CD4/CD8 ratios (Figure 2g, i), memory phenotypes (figure S5a-d), or IFN-I receptor expression (figure S5e-h). To test cytotoxicity, CAR T cells were co-cultured with OCI-Ly8 target cells, a CD19⁺ human DLBCL cell line. In both CD28- and 4-1BB-costimulated CAR T cells, we observed IFN-I signaling significantly increased CAR T-cell cytotoxicity (Figure 2j, l), expression of cytolytic molecules (granzyme B, perforin, Figure 2k, m), and secretion of cytokines (IL-2, TNF, IFNγ, Figure 2n-o), in a concentration-dependent manner. These findings suggest that IFN-I signaling increases CAR T-cell cytotoxicity.

High-strength IFN-I signaling increases apoptosis, whereas low-strength IFN-I signaling does not.

After the manufacturing process, both CD28- and 4-1BB-costimulated CAR T cells cultured under the IFNα-hi condition, but not the IFNα-lo condition, consistently yielded fewer viable cells compared to the IFNα-negative condition (Figure 3a, c). Given previous reports that IFN-I signaling can inhibit cell proliferation³¹, we evaluated whether CAR T-cell proliferation was affected by IFN-I signaling through assessing the proliferation marker, Ki-67. No significant differences in Ki-67 expression were observed across the three culturing conditions, indicating that IFN-I signaling does not impair CAR T-cell proliferation during ex vivo expansion (Figure 3b, d, figure S6). Since IFN-I signaling also reportedly promotes T-cell apoptosis^20,21, we next quantified CAR T-cell apoptosis via Annexin V staining. Flow phenotyping revealed that CAR T cells in the IFNα-hi condition trended towards increased apoptosis, whereas those in the IFNα-lo condition did not (Figure 3e-f).

Figure 3. High-strength IFN-I signaling increases apoptosis, whereas low-strength IFN-I signaling does not.

(a, c) Relative counts of live CAR T cells from the three culturing conditions after CAR T-cell manufacturing in CD28-costimulated (a, n=5) and 4-1BB-costimulated (c, n=6) CAR T cells. Counts were normalized to the IFNα-neg condition. (b, d) Proliferation of CAR T cells from the three culturing conditions in CD28-costimulated (b, n=3) and 4-1BB-costimulated (d, n=3) CAR T cells. Left, representative flowplots depicting gating for Ki-67⁺ proliferating CAR T cells at the day when IFNα was removed. Right, Ki-67⁺ proliferating cell proportions among CAR T cells. Non-activated naïve T cells from human PBMCs were used as a biological control for gating. (e, f) Apoptosis of CAR T cells from the three culturing conditions after CAR T-cell manufacturing in CD28-costimulated (a, n=5) and 4-1BB-costimulated (c, n=6) CAR T cells. Left, representative flow plots depicting gating for viable, early apoptotic, and dead cells among CAR T cells. Right, staining for Annexin V positivity among live CAR T cells. (g, h) Line graphs depicting viable CAR T-cell cellularity of CD28-costimulated (g, n=3) and 4-1BB-costimulated (h, n=3) CAR T cells in vitro following coculture with OCI-Ly8 at E/T ratio of 1:5. (i, j) Apoptosis of CD28-costimulated CAR T cells (i, n=3) and 4-1BB-costimulated CAR T cells (j, n=3) from the three culturing conditions at day 6 after coculture with OCI-Ly8 cells at E/T ratio of 1:5. Left, representative flow plots depicting gating for viable, early apoptotic, and dead cells among the three culturing conditions. Right, staining for Annexin V positivity among live CAR T cells. All data are reported as mean ± standard error. Statistical analysis were performed via one-way (a-f, i-j) or two-way (g, h) ANOVA with Tukey correction for multiple comparisons. Panels a, c, e-j are representative data from 3 independent donors. Panels b, d are representative data from 2 independent donors.

To further investigate the cellularity and apoptosis of CAR T cells from the three culturing conditions following antigen-specific stimulation, we subsequently quantified cellularity and apoptosis of CAR T cells after stimulation with OCI-Ly8 target cells. Following stimulation, CAR T-cell cellularity in the IFNα-lo and IFNα-neg conditions was statistically indistinguishable (Figure 3g-h). On the other hand, CAR T-cell cellularity in the IFNα-hi condition was significantly depleted compared to the other two conditions. At the timepoint when CAR T-cell cellularity in the IFNα-hi condition significantly diverged (day 6 post-stimulation in the representative data), we observed a corresponding enrichment in apoptosis (Figure 3i-j).

Taken together, these findings suggest that high-strength, but not low-strength IFN-I signaling increases CAR T-cell apoptosis, thereby negatively affecting persistence during both ex vivo expansion and antigen-specific activation.

Low-strength IFN-I signaling promotes CAR T-cell treatment efficacy in vivo.

Given that the IFNα-lo condition uniquely increases CAR T-cell functionality (Figure 2) without compromising cell viability (Figure 3) in both CD28- and 4-1BB-costimulated CAR T cells, our data demonstrate that low-strength IFN-I signaling promotes CAR T-cell efficacy in vitro by balancing function with persistence. Based on these in vitro conclusions, we hypothesized that the IFNα-lo condition uniquely optimizes CAR T-cell treatment efficacy in vivo.

To test our hypothesis, we analyzed the efficacy of CAR T cells prepared from the three culturing conditions in a xenograft mouse model based on the OCI-Ly8 cell line (Figure 4a). To control for donor-donor heterogeneity, CAR T cells were derived from multiple independent donors (three for CD28-costimulated CAR T cells and two for 4-1BB-costimulated CAR T cells). Across donors, CAR T-cell injection prolonged survival relative to the PBS-only control. For CD28-costimulated CAR T cells with donors 1 and 2, CAR T cells from the IFNα-lo condition significantly prolonged survival and slowed down tumor progression relative to CAR T cells from the IFNα-neg and IFNα-hi conditions (Figure 4b, figure S7a). For donor 3, although the differences between CAR T cells from IFNα-lo and IFNα-hi conditions were relatively modest compared to other donors, one mouse treated with CAR-T cells from the IFNα-lo condition showed a significantly better and sustained response (Figure S7b-c). For 4-1BB-costimulated CAR T cells with donors 4 and 5, CAR T cells from the IFNα-lo condition consistently yielded superior outcomes compared to CAR T cells from the IFNα-neg and IFNα-hi conditions, leading to improved survival and tumor control (Figure 4c, figure S7d). Taken together, our findings support our hypothesis and indicate that CAR T cells generated from the IFNα-lo condition, regardless of costimulatory domain, exhibit enhanced in vivo efficacy compared to those from IFNα-neg or IFNα-hi conditions across multiple donors.

Figure 4. Low-strength IFN-I signaling promotes CAR T-cell treatment efficacy in vivo.

(a) Schematic depicting timeline for NSG mouse model based on OCI-Ly8 lymphoma cell line. Seven days following OCI-Ly8 injection, mice were randomly divided into groups that were treated with either PBS or CAR T cells from the three culturing conditions. (b) Kaplan-Meier analysis of survival of OCI-Ly8-bearing mice treated with PBS or 5×10⁵ CD28-costimulated CAR T cells from various culture conditions. CAR T cells were generated from donor 1 (left, n=8,12,13,8), and donor 2 (right, n=5,5,4,4). (c) Kaplan-Meier analysis of survival of OCI-Ly8-bearing mice treated with PBS or 5×10⁵ (left) or 1×10⁶ (right) 4-1BB-costimulated CAR T cells from various culture conditions. CAR T cells were generated from donor 4 (left, n=6,6,6,6), and donor 5 (right, n=4,5,5,5). (d) Cartoon summarizing effects of IFN-I signaling on CAR T cells at different doses. Statistical analysis were performed by log-rank (Mantel-Cox) test.

In conclusion, our in vitro and in vivo studies support a model in which low-strength IFN-I signaling selectively promotes CAR T-cell treatment efficacy (Figure 4d). IFN-I signaling promotes CAR T-cell functionality in a concentration-dependent manner, leading to more effective cytotoxicity, cytokine release, and tumor control. However, at a high enough concentration, IFN-I signaling also increases T-cell apoptosis, leading to reduced yield, poorer persistence, and ineffective tumor control. Although the optimal level of IFN-I signaling may be subject to donor-donor heterogeneity, CAR T cells treated with low-strength IFN-I signaling, regardless of costimulatory domain, consistently exhibit enhanced effector functions without significantly increased apoptosis.

DISCUSSION

In this study, we leveraged single-cell transcriptomic analyses of patient IPs from two independent datasets, in vitro functional and phenotypic analyses, and in vivo xenograft mouse studies, to establish that low-strength IFN-I signaling during the ex vivo CAR T-cell manufacturing process selectively promotes CAR T-cell treatment efficacy in vivo in a costimulation-independent manner. Our strategy of harnessing IFN-I signaling to enhance CAR T-cell efficacy is grounded in clinical data and existing literature, while also highlighting IFN-I as an enhancer with significant translational potential for use during the CAR T-cell manufacturing process.

There are major attractive qualities that make IFN-I a uniquely promising and translatable enhancer for both CAR T-cell therapy and other emerging cellular therapy modalities. Unlike all other natural and engineered cytokines being developed as CAR T-cell enhancers (including IL-7, IL-10, IL-15, IL-21, Fc-IL-4)^32-36, IFN-Is are uniquely biocompatible drugs with an existing and extensive clinical track record for the treatment of malignancies²², autoimmune diseases²³, and viral infections²⁴. First discovered in the 1950s for their antiviral activities, IFN-I is one of only two cytokines ever approved by the FDA for cancer treatment (the other being IL-2).³⁷ Biologically, IFN-I elicit a known signaling pathway through the type I interferon receptor to induce antiviral interferon stimulated genes.^13,15 Their pharmacodynamics and pharmacokinetics have been thoroughly explored in pre-clinical and clinical studies.³⁸ By uniquely employing a biologically, pharmacologically, and clinically characterized pharmacophore, our low-strength IFN-I manufacturing method is straightforward and less likely to introduce uncharacterized toxicities. We also point out that our method was mechanistically inspired by bioinformatics findings from clinical biospecimens, thereby recapitulating a physiological and clinically relevant phenotype.

In terms of basic biology, we found that IFN-I signaling enhances CAR T-cell cytotoxicity in a dose-dependent manner; however, excessively strong IFN-I signaling also increases apoptosis, thereby reducing persistence. Our findings are compatible with existing literature on T-cell-intrinsic effects of IFN-I signaling. Enhancement of cytotoxicity and type I cytokine secretion among CD8⁺ and CD4⁺ T cells after culturing with exogenous IFN-I is well-established.^16,39 Moreover, the field’s current understanding is that T-cell-intrinsic IFN-I signaling leads to different consequences, depending on whether TCR stimulation temporally precedes (“in-sequence”) or follows (“out-of-sequence”) IFN-I signaling.^13,40 In-sequence signaling generally enhances T-cell functions, while out-of-sequence signaling generally does the opposite.¹³ In our studies, since T-cell activation (via CD3/CD28 Dynabeads) preceded IFN-I treatment, our CAR T cells were subject to in-sequence signaling under the current paradigm. Moreover, our findings mirror the expected consequences of in-sequence IFN-I signaling. Lastly, the concentration-dependent nature of our model is compatible with other studies demonstrating that different cellular functions exhibit varying sensitivities to IFN-I signaling.²⁸

Our findings may also reconcile the currently obscure relationship between IFN-I signaling and CAR T-cell function. On one hand, Jung et al. and Evgin et al. reported that IFN-I signaling hinders CAR T-cell efficacy.^21,41 Both groups subjected their CAR T cells to IFN-β, which has up to 30-fold higher affinity for the type I interferon receptor compared to IFN-α2.⁴² Taking into account their high-affinity IFN-I subtype and culturing concentrations, their conditions are similar to our IFNα-hi condition, which accounts for their negative findings. On the other hand, Zhao et al. reported that IFN-I signaling enhances CAR T-cell efficacy.⁴³ We reason their positive findings are due to their CAR T cells receiving a weak IFN-I stimulus, since they did not add exogenous IFN-I in their in vitro system. In conclusion, our model demonstrates a non-linear relationship between IFN-I signaling and CAR T-cell efficacy, thereby potentially reconciling these intriguing prior studies.

Our study advances the understanding of IFN-I signaling in T cells and lays the foundation for translational applications in CAR-T cell therapies, though several avenues remain for exploration. While our experiments generated CAR-T cells from healthy donors for mechanistic studies, future research should focus on using T cells from cancer patients. Since cancers often suppress IFN-I signaling to evade immune detection and weaken IFN-I-driven immune responses^44-46, modulating IFN-I signaling in cancer patient-derived CAR-T cells may yield significantly improved therapeutic effects. Furthermore, we tested only two IFN-I concentrations, emphasizing the need for more rigorous dosing studies to determine the optimal therapeutic window. Finally, although our study focused on IFN-α2 due to its clinical relevance, it underscores the broader potential of the IFN-I family, with many subtypes still underexplored and potentially capable of further enhancing CAR-T cell functionality through cytokine modulation.

METHODS

Collection of patient biospecimens

Deidentified patient biospecimens were gathered in accordance with ethical guidelines under the institutional review board at the University of Chicago Medicine. Residual infusion product cells were collected from the patient’s infusion product bag and stored in CELLBANKER 1 (Amsbio, 11888). Cells were cryopreserved in liquid-phase nitrogen.

Single-cell RNA-seq data generation.

Cryopreserved biospecimens were thawed and washed with cold FACS buffer (PBS, 2% BSA, 0.05% sodium azide). Antibody-stained cells were conjugated with LIVE/DEAD Fixable Near-IR viability dye (Invitrogen, L34975) at 1:1000 dilution in PBS for 5 minutes at room temperature. Finally, cells were washed 3x in cold cell media (RPMI, 10% FBS) before sorting (BD Biosciences, FACSAria Fusion) for live cells. Sorted cells were partitioned into droplets for single-cell RNA-seq via Chromium Next GEM Single-Cell 5’Kit v2 (10x Genomics, 1000263). RNA-seq libraries were prepared according to manufacturer protocols. Libraries were quantified via the Qubit dsDNA HS Assay Kit (Invitrogen, Q32851), quality-checked for fragment sizes via high-sensitivity D5000 screentapes (Agilent, 5067-5592), pooled, and sequenced (Illumina, Novaseq-6000 and Nextseq-550).

Single-cell RNA-seq data analysis.

Alignment, filtering, counting, and demultiplexing of RNA-seq reads were performed using the Cell Ranger count (10x Genomics, version 7.1.0). For alignment, the GRCh38 reference genome was appended with the CAR transgene sequence of axicabtagene ciloleucel.

Downstream analysis was performed using Seurat⁴⁷ (version 4.3.0). After converting count matrices and merging all Seurat objects, we filtered cells with >15% mitochondrial reads and 40,000 UMIs to remove debris, doublets, and dead cells. The matrix was normalized using NormalizeData with default options. The top 5,000 variable genes found by FindVariableFeatures with the default “vst” method were taken as variable genes. Data were centered and scaled using ScaleData, with additional regression against (1) the percent of mitochondrial RNA content and (2) difference in cell cycle S-phase and G2M-phase scores. Scaled data were then used for principal component analysis (PCA) using RunPCA. Data harmonization to remove patient-specific effects was performed on the principal components using Harmony⁴⁸ (version 0.1.1) through RunHarmony. Top 50 harmony components were used for Uniform Manifold and Projection (UMAP) dimensionality reduction via RunUMAP. The same harmony components were used to construct the shared nearest neighbor (SNN) graph using FindNeighbors, which was then partitioned to identify clusters using FindClusters with the “original Louvain” algorithm.

Differentially expressed genes were determined using FindMarkers in Seurat package with default parameters. Gene set enrichment analysis and pathway enrichment analysis were carried out using clusterProfiler⁴⁹ (version 4.6.2) based on pathways from the msigdbr database’s built-in msigdbr⁵⁰ package (version 7.5.1). Gene module scores were calculated for each cell using AddModuleScore in Seurat for msigdbr or customized gene sets.

Lentiviral production and transduction.

CD28- or 4-1BB-costimulated anti-CD19 human CARs were designed based on the sequences of axi-cel or liso-cel. The axi-cel sequence was obtained from a previously published sequencing study,⁵¹ while the liso-cel sequence was retrieved from the Global Substance Registration System database (UNII: 7K2YOJ14X0). A GFP-tagged, CD28-costimulated or 4-1BB-costimulated, anti-CD19 human CAR was cloned onto the pHR vector backbone to constitute the transfer plasmid. Plasmids were transfected into the Lenti-X 293T packaging cell line (Takara, 632180) for lentiviral production. In brief, Lenti-X 293T cells were seeded and grown overnight to ~90% confluence. Subsequently, transfer, packaging (psPAX2), and envelope (pMD2.G) plasmids were co-transfected into Lenti-X 293T cells via Lipofectamine 3000 (Invitrogen, L3000001). After 72 hours, the viral supernatant was collected and stored at −80°C until transduction. During transduction, lentiviruses were added to human primary T cells along with protamine sulfate (Millipore Sigma, P3369-10G) to a final concentration of 10 μg/mL. Cells were spinoculated at 800 x g for 60 minutes at room temperature.

CAR T-cell production and culturing.

PBMC samples were collected from healthy donors under an Institutional Review Board approved protocol. Cryopreserved human PBMCs were thawed in warm T-cell media (RPMI supplemented with 10% FBS, 1% penicillin/streptomycin, 2 mM L-glutamine, and 50 μM 2-mercaptoethanol). Subsequently, PBMCs were cultured for 3 hours in complete T-cell media (T-cell media supplemented with 50 units/mL of IL-2) in order to allow monocytes to adhere to the culture flask. The non-adhered cells were transferred to a new culture flask. Cells were counted via the LUNA-FL cell counter. T-cell percentage (% CD3⁺) was calculated by flow cytometry.

T cells were seeded in complete T-cell media at 0.75 x 10⁶/mL. CD3/CD28 Dynabeads (Gibco, 11161D) were washed and added at 1:1 ratio. After overnight activation, T cells were transduced with CAR lentiviruses (MOI of 5). The media was refreshed after 24 hours with complete T-cell media dosed with recombinant human IFN-α2 (BioLegend, 592702). T cells were subsequently expanded, and media was refreshed every 2 days.

Quantitative reverse transcription PCR.

CAR T cells were treated under IFN-I culturing conditions for 24 hours. Then, cells were collected, and stained for viability (Invitrogen, L34975). Live CAR T cells were sorted (BD Biosciences, FACSAria Fusion). Total RNA was extracted using the Monarch Total RNA Miniprep Kit (NEB, T2010S), according to manufacturer protocols. RNA was quantified via the Qubit RNA HS Assay Kit (Invitrogen, Q32852). cDNA was synthesized using the GoScript Reverse Transcription System (Promega, A5000).

Quantitative PCR was performed using Taqman Gene Expression Assay (ThermoFisher Scientifc, 4331182) and Applied Biosystems TaqMan Universal PCR Master Mix (ThermoFisher Scientifc, 4304437), on LightCycler^® 96 Instrument (Roche). The cycling conditions were as follows: UNG incubation at 50°C for 2 minutes, initial denaturation at 95°C for 10 minutes, followed by 40 cycles of denaturation at 95°C for 15 seconds and annealing/extension at 60°C for 60 seconds. Each sample was run in triplicate.

The specific Taqman Gene Expression Assay used were as follows: ACTB (Hs01060665_g1), IRF7 (Hs01014809_g1), MX1 (Hs00895608_m1), and ISG15 (Hs01921425_s1). Relative gene expression levels were calculated using the 2^-ΔΔCt method, with ACTB serving as the internal control. Statistical analysis was performed using one-way ANOVA followed by a Tukey's post hoc test for significance.

In vitro co-culture immunoassays.

For all co-culture immunoassays, CAR T cells (E) and luciferized OCI-Ly8 cells (T) were seeded under a fixed E/T ratio. To measure in vitro cytotoxicity, co-cultures were established with E/T ratio of 1:1. After 6 hours, One-Glo Luciferase substrate (Promega, E6110) was added to each well. Residual tumor cells were quantified via luciferase activity. As a positive control, target cells were directly lysed with 1% NP-40 (Abcam, ab142227). To measure cytotoxic molecule expression, co-cultures were established with E/T ratio of 1:1 with 1x brefeldin A (BioLegend, 420601). After 4 hours, cells were collected for intracellular staining and flow cytometry. To measure cytokine secretion, co-cultures were established with E/T ratio of 1:1. After 24 hours, the culture supernatant was collected and stored. Secreted cytokines were quantified using LEGENDplex immunoassays per manufacturer’s protocol (BioLegend). To measure in vitro persistence, co-cultures were established with E/T ratio of 1:5. Every three days, cells were counted using Precision Count Beads (BioLegend, 424902) and measured for apoptosis on flow cytometry. Cell media was then refreshed.

Flow cytometry.

Cells were washed with cold FACS buffer (PBS, 2% BSA, 0.05% sodium azide). Fc receptors were blocked by incubation with Human TruStain FcX (BioLegend, 422301) at 1:50 dilution for 5 minutes at 4°C. Then, cells were incubated for 30 minutes at 4°C in the dark with an antibody staining solution for surface antigen staining . Antibodies were used according to manufacturer recommendations. After staining, cells were incubated briefly either with live/dead staining solution or apoptosis staining solution for 10 minutes at room temperature. Then, cells were either washed and resuspended in FACS buffer or resuspended in Annexin V Binding Buffer (BioLegend, 422201). Cells were ready for analysis by flow cytometry or intracellular staining.

For intracellular staining, cells stained for surface antigens were subject to the BD Fixation/Permeabilization Kit (BD Biosciences, 554714) for cytotoxic molecule staining or True-Nuclear^™ Transcription Factor Buffer Set (BioLegend, 424401) for IRF7 staining, following manufacturer recommendations. In brief, Cells were fixed in fixation buffer, washed 3x in permeabilization buffer, and incubated for 60 minutes at 4°C in the dark with a staining solution containing antibodies and permeabilization buffer. After staining, cells were washed 2x in permeabilization buffer, 1x in FACS buffer, and resuspended in FACS buffer for analysis by flow cytometry.

For Ki-67 staining, cells stained for surface antigens were washed 2x in PBS, fixed in prechilled 70% ethanol overnight, washed 2x in FACS buffer, and incubated for 30 minutes at 4°C in the dark with Ki-67 antibody staining solution. After staining, cells were washed 2x in FACS buffer, and resuspended in FACS buffer for analysis by flow cytometry.

Flow cytometry was performed on LSRFortessa (BD Biosciences) or Aurora (Cytek Biosciences). Data were analyzed using FlowJo software (version 10.10.0).

Antibody panels.

All antibodies were used according to manufacturer recommendations. APC-anti-IFNAR2 (10359-MM07T-A) was from Sino Biological. All other antibodies originate from BioLegend. For surface staining, AF647-anti-CD3 (clone: UCHT1, 300416), BV421-anti-CD3 (clone: UCHT1, 300434), BV510-anti-CD4 (clone SK3, 344633), PerCP/Cy5.5-anti-CD8α (clone SK1, 344709), PE-anti-CD45RA (clone HI100, 304107), BV421-anti-CD45RO (clone UCHL1, 304223), and AF647-anti-CCR7 (clone G043H7, 353217) were used. For intracellular staining, PE/Cy7-anti-granzyme B (clone QA16A02, 372213), PE-anti-perforin (clone B-D48, 353303), and PE-anti-IRF7 (clone 12G9A36, 656003) were used. For live/dead staining, Zombie NIR Fixable Viability Kit (423106) was used in PBS. For apoptosis staining, BV421-Annexin V (640923) or PE/Cy7-Annexin V (640950) was diluted in Annexin V Binding Buffer (422201). For Ki-67 staining, PE-anti-Ki-67 (clone Ki-67, 350503) was used.

Lymphoma xenograft mouse model.

Immunocompromised NOD scid IL2Rgamma^null (NSG, Strain #005557) mice were purchased from JAX and maintained at the animal resource center (ARC) of the University of Chicago in accordance with institutional guidelines. Mice were monitored daily. Care and experimental procedures were carried out in line with the Institutional Animal Care and Use Committee (IACUC) approved protocols of the Animal Care Center at the University of Chicago.

5×10⁶ OCI-Ly8-Chili-luc cells in 100 μL PBS were intravenously injected into 8-12-week-old NSG mice to build a xenograft lymphoma model. 5×10⁵ or 1×10⁶ CAR T cells were intravenously injected 1 week later. Lymphoma progression was monitored using either IVIS Spectrum (PerkinElmer) or Lago X (Spectral Instruments Imaging) following the Intraperitoneal injection of D-luciferin potassium (PerkinElmer, 122799). Animals were anesthetized at 2% isoflurane throughout acquisition. Bioluminescence was quantified with Aura software (Spectral Instruments Imaging). Mice were euthanized upon manifestation of paralysis, impaired mobility, or poor body condition (BCS<2).

Statistical analysis.

Statistical tests were performed on R (version 4.2.1) or GraphPad Prism 10 (version 10.1.1). Comparisons between two groups were performed using Wilcoxon Rank-Sum test. Comparisons between more than two groups were performed by either one-way or two-way ANOVA followed by Tukey correction for multiple comparisons. Survival data were analyzed using the Log-Rank (Mantel-Cox) test. To assess correlation, Spearman’s rank correlation coefficient was calculated, as indicated.

Data availability.

All data are available from the authors upon reasonable request.

Code availability.

Codes supporting the analyses performed in this study are available at https://github.com/ertingtang/IFN-I-CART.