Predictive markers for the efficacy of CAR T-cell therapy: the interplay between CAR T-cell fitness and systemic immunity Open Access

Chimeric antigen receptor (CAR) T-cell therapy has transformed the therapeutic landscape for lymphoid malignancies, offering durable remissions in patients with relapsed or refractory disease.^1,2 However, pivotal trials show heterogeneous outcomes for CAR T-cell therapy in diffuse large B-cell lymphoma (DLBCL) and acute lymphoblastic leukemia (ALL). In second-line DLBCL, axicabtagene ciloleucel (axi-cel) and lisocabtagene maraleucel improved event-free survival over standard care, whereas tisagenlecleucel did not.^3-5 These differences were likely driven by trial design (eg, allowance of bridging chemotherapy), manufacturing logistics, and baseline risk. Toxicity profiles also diverge. Axi-cel has higher severe cytokine release syndrome (CRS) and immune effector cell–associated neurotoxicity syndrome (ICANS) than lisocabtagene maraleucel, consistent with faster expansion from its CD28 vs 4-1BB costimulation, and these adverse effects plus a substantial nonresponder fraction (∼30%-35%) constrain broader use.^6-8 In ALL, adult regimens (obecabtagene autoleucel, brexucabtagene autoleucel, and rapid off-rate CD19 CARs) and tisagenlecleucel for pediatric or young adults achieve meaningful remissions but with wide interpatient variability in depth and durability.^9-12 Collectively, these data underscore the need for robust predictive biomarkers to guide product selection, anticipate toxicity, and personalize therapy.

Recently, Levstek et al comprehensively categorized predictive biomarkers into 6 principal groups to improve CAR T-cell therapy outcomes and minimize adverse events: mitochondrial dynamics, endothelial activation, central nervous system impairment, immune system markers, extracellular vesicles, and the inhibitory tumor microenvironment (TME).¹³ They emphasize that existing biomarkers and predictive scores, such as the CAR-HEMATOTOX model, endothelial activation and stress index (EASIX), and modified cumulative illness rating scale, demonstrate limited predictive power.^14-16 Thus, developing integrated predictive models that consider diverse biomarkers can provide a deeper understanding of the complex interactions between CAR T cells and the patient’s immune system.^17,18 

This review aims to integrate these insights, highlighting the importance of CAR T-cell intrinsic functional markers and systemic immune contexture, supported by emerging multiomic approaches and advanced analytics.¹⁹ A comprehensive understanding of these factors can help refine CAR T-cell manufacturing, enhancing patient stratification, and improving therapeutic outcomes in hematologic malignancies.

The efficacy of CAR T-cell therapy is shaped by the composition of infused T-cell subsets.²⁰ Memory differentiation proceeds from naïve T cells (Tns) to effector T cells (Teffs), with a fraction persisting memory T cells (Tmems), including central memory (Tcms), effector memory T cells, and memory stem cells (Tscms).^21,22 Tcms (CCR7⁺CD27⁺SELL⁺) sustain long-term responses; effector memory T cells provide rapid effector function; and Tscms, although rare, exhibit multipotency and self-renewal.^21,23-25 

Clinical and translational research consistently demonstrates that products enriched with early memory subsets, particularly Tcms and Tscms, correlate with superior CAR T-cell expansion, persistence, and durable remision.^26-29 Premanufacturing T-cell quality, especially Tscm-like cells (CD8⁺CD45RA⁺CD27⁺), further predicts efficacy and reduced CRS.^29,30 

Approaches to enhance early Tmem subsets in CAR T-cell products include optimizing cytokine environments during manufacturing. Interleukin-7 (IL-7), IL-15, and IL-21 have been shown to support the development and maintenance of Tscms, whereas IL-2 can drive terminal differentiation and impair CAR T-cell longevity.^31,32 Pharmacological modulation of signaling pathways, such as phosphoinositide 3-kinase inhibitors or Bruton tyrosine kinase (BTK) inhibitors, has also emerged as a promising strategy to preserve less differentiated CAR T cells, resulting in improved therapeutic outcomes.^33,34 Phosphoinositide 3-kinase pathway inhibition has been tested predominantly during manufacturing to bias cells toward memory-like phenotypes and enhance persistence.^33,35 In contrast, BTK inhibition (eg, ibrutinib) has largely been evaluated in vivo as bridging or concomitant therapy to reduce tumor burden and modulate the B-cell lymphoma microenvironment.^34,36 

These insights emphasize the critical role that T-cell subsets and phenotypes play in the predictive assessment and therapeutic optimization of CAR T-cell products. They suggest that manufacturing processes prioritize enriching and preserving early Tmem populations to enhance the long-term success of CAR T-cell therapies.

Effective transduction is critical for CAR T-cell therapy, ensuring sufficient CAR expression on T cells and directly influencing efficacy and safety.³⁷ CAR density strongly affects activation, proliferation, and antitumor activity of CAR T cells, with adequate expression required for efficient antigen recognition and cytotoxicity.³⁸ 

Heterogeneous CAR expression arises from semirandom transgene integration during viral transduction.³⁹ Such heterogeneity can lead to inconsistencies in CAR T-cell responses, because high-density cells may undergo premature exhaustion, whereas low-density cells may fail to kill targets.⁴⁰ Furthermore, transduction protocols using viral vectors, such as lentiviruses or gamma-retroviruses, significantly influence the integration pattern and expression level of the CAR. Advanced methodologies, such as CRISPR-based site-specific integration, are being explored to achieve uniform CAR expression levels and improve safety by reducing the risks associated with insertional oncogenesis.^41,42 Site-specific CRISPR/CRISPR-associated protein 9 integration into safe-harbor loci (eg, T-cell receptor alpha constant) can normalize CAR density and reduce insertional mutagenesis risk; however, off-target edits remain a translational concern.^43,44 Comprehensive off-target mapping (eg, genome-wide, unbiased identification of DSBs enabled by sequencing/circularization for high-throughput analysis of nucease genome-wide effects by sequencing) and high-fidelity nucleases are required before clinical application.^45,46 

Optimizing CAR density requires a fine-tuning of viral vector dosage, multiplicity of infection, and stimulation protocols, whereas quantitative monitoring ensures balanced expression.^47,48 Achieving optimal CAR density through precise control and refinement of transduction processes is crucial for enhancing the therapeutic potential of CAR T cells.

Cytokines play a vital role in the manufacturing process of CAR T cells, because they influence T-cell activation, expansion, and differentiation into specific subsets of Tmems. In these processes, IL-2, IL-7, IL-15, and IL-21 are routinely used, each influencing quality, persistence, and antitumor activity.⁴⁹ 

Although IL-2 remains the most used cytokine for ex vivo CAR T-cell expansion, sustained IL-2 exposure preferentially expands CD25^high regulatory T cells (Tregs).³² It drives terminal effector differentiation, which can predispose products to exhaustion and blunt in vivo persistence.⁵⁰ Alternative cytokine cocktails such as IL-7/IL-15 preserve stem-like/central memory states and have been associated with improved proliferation and persistence.^30,51 These phenotypes correlate with metabolic programs favoring mitochondrial fitness and oxidative phosphorylation over chronic glycolysis.^52,53 

After infusion, CAR T-cell cytokine profiles strongly influence efficacy and safety.^54,55 Effector cytokines such as interferon gamma (IFN-γ), tumor necrosis factor α (TNF-α), granzyme B, and macrophage inflammatory protein-1α mediate direct cytotoxicity,^55,56 whereas stimulatory cytokines including granulocyte-macrophage colony-stimulating factor (GM-CSF), IL-2, and IL-8 promote T-cell proliferation and survival.⁵⁷ Regulatory cytokines, such as IL-4, IL-10, IL-13, and IL-22, can modulate immune responses, often dampening excessive inflammation.⁵⁸ Inflammatory cytokines such as IL-6 and IL-17A, although beneficial in antitumor activity, can also contribute to adverse events.^57,59 Clinically, cytokines are also involved in adverse events, notably CRS and ICANS. CRS arises from macrophage/monocyte activation, with IL-6 and IL-1β driving systemic inflammation and vascular/endothelial damage.⁶⁰ ICANS typically follows CRS, marked by cytokines penetrating the central nervous system due to blood-brain barrier disruption.⁶¹ 

Single-cell multiplexed cytokine profiling has revealed striking heterogeneity and highlighted polyfunctionality, the simultaneous secretion of multiple cytokines, as a predictor of balanced antitumor efficacy with reduced toxicity.^54,56 Polyfunctionality and related biomarkers will be further discussed in “Proteomic profiling and polyfunctionality assay.”

The metabolic health of CAR T cells is crucial for antitumor efficacy, persistence, and long-term outcome.⁵² Functional capabilities depend on mitochondrial integrity, bioenergetics, and glycolytic capacity,²⁰ which determine survival in immunosuppressive TME.⁶² 

Effective CAR T cells display high mitochondrial mass and the ability to shift between oxidative phosphorylation and glycolysis.^13,63 Although Tns rely on oxidative phosphorylation, activated T cells switch to aerobic glycolysis to meet rapid energy and biosynthetic demands.^53,64 Balanced plasticity between these pathways underpins proliferation and cytotoxicity.⁵² Impaired mitochondrial biomass, by contrast, correlates with poor performance.

Mitochondrial membrane potential (ΔΨm) is another critical marker of metabolic fitness.⁶⁵ High ΔΨm, typical of Teffs, reflects glycolysis and reactive oxygen species (ROS) production, and eventual cellular exhaustion. In contrast, a lower ΔΨm in Tns and Tmems indicates great metabolic reserve and persistence.^52,62 Controlled ROS is essential: low levels support signaling, but excess causes DNA damage and dysfunction.⁶⁶ Processes such as mitophagy maintain mitochondrial quality and prevent ROS accumulation.^52,66 

A promising avenue is optimizing the metabolic fitness of CAR T cells through strategic interventions during manufacturing. For example, modulation of cytokines such as IL-7 and IL-15 has shown potential not only to enhance proliferation, persistence, and memory formation but also to directly regulate metabolic pathways.^51,67 IL-7 promotes glucose uptake and mitochondrial function, whereas IL-15 supports fatty acid oxidation and mitochondrial biogenesis, thereby fostering the development of long-lived Tmems with superior bioenergetic capacity.^68,69 Strategies aimed at enhancing metabolic fitness by maintaining mitochondrial integrity, optimizing glycolytic capacity, and controlling ROS production can significantly improve CAR T-cell therapeutic outcomes, underscoring metabolic fitness as a critical determinant of CAR T-cell therapy success.^52,62 

The systemic cytokine landscape before CAR T-cell therapy has a profound influence on therapeutic outcomes. Elevated baseline cytokine levels, notably IL-15, correlate positively with robust CAR T-cell expansion and persistence, making IL-15 an important prognostic biomarker.⁵¹ Conversely, elevated baseline IL-6 and GM-CSF levels, although predictive of CRS, might simultaneously indicate a highly activated systemic immune state capable of supporting robust CAR T-cell responses.⁷⁰ Systemic inflammation markers, such as C-reactive protein and IL-6, provide valuable predictive insights into CAR T-cell therapy outcomes. Elevated baseline levels of these markers often correlate with increased risk of CRS and ICANS.⁷¹ However, these elevated markers may also indicate a robust immune environment favorable for CAR T-cell expansion and antitumor efficacy. Recent research highlights that baseline inflammation status predicts adverse events and indirectly indicates a more activated immune milieu capable of supporting robust CAR T-cell proliferation and functionality.^71,72 Accurate interpretation of these dual roles is essential for optimal patient management, enabling preemptive measures to mitigate severe toxicities and maximize therapeutic responses.⁷³ For instance, prophylactic use of cytokine-blocking agents in patients with elevated inflammatory markers has shown promising results in reducing toxicity without impairing therapeutic efficacy.^74,75 

Suppressive immune populations, notably Tregs and myeloid-derived suppressor cells (MDSCs), significantly modulate the efficacy of CAR T-cell therapies by inhibiting antitumor immune responses.^76,77 Increased presence or enhanced suppressive function of Tregs and MDSCs before therapy can predict suboptimal CAR T-cell expansion and reduced clinical effectiveness.⁷⁶ Mechanistically, Tregs suppress Teffs cell functions through direct cell-to-cell interactions and secretion of immunosuppressive cytokines such as IL-10 and transforming growth factor β.⁷⁸ MDSCs contribute similarly by producing arginase, nitric oxide, and ROS, which impair T-cell proliferation and function.⁷⁹ Therapeutic strategies targeting these suppressive cell populations, such as depletion protocols using chemotherapeutics, immune modulators, or checkpoint blockade agents, have significantly improved CAR T-cell efficacy by enhancing immune activation.^80,81 Advanced immunophenotyping techniques further aid in precisely identifying and quantifying these suppressive cell subsets, allowing more personalized and effective intervention strategies.

The pretreatment TME significantly affects the efficacy of CAR T-cell therapy through various immunosuppressive mechanisms.⁸² Hypoxia, nutrient depletion, and programmed death ligand 1 (PD-L1) overexpression collectively create an environment hostile to effective CAR T-cell function. Hypoxia particularly upregulates immunosuppressive molecules such as adenosine, PD-L1, and vascular endothelial growth factor, all of which impede effective CAR T-cell infiltration and cytotoxicity.⁸³ Strategies combining CAR T cells with agents specifically targeting these TME factors, including immune checkpoint inhibitors (eg, programmed cell death protein 1/PD-L1 inhibitors), hypoxia modulators, and metabolic reprogramming drugs, have demonstrated substantial promise in clinical trials.^84,85 Furthermore, in early phase clinical studies, anbalcabtagene autoleucel, a novel CAR T-cell product engineered with programmed cell death protein 1 and T-cell immunoreceptor with Ig and ITIM domains silencing, showed encouraging antitumor activity.⁸⁶ These approaches help to alleviate immunosuppressive conditions, facilitate greater CAR T-cell infiltration, and enhance antitumor activity. Using multiomic technologies, comprehensive profiling of the TME enables clinicians to accurately identify immunosuppressive signatures and tailor combination therapies effectively, thereby significantly improving CAR T-cell persistence and clinical outcomes.^87,88 

Single-cell RNA sequencing (scRNA-seq) has transformed CAR T-cell biology by dissecting transcriptional profiles at the individual cell level.^89,90 Unlike bulk sequencing, scRNA-seq reveals heterogeneity within CAR T-cell populations and the TME, offering resolution to identify transcriptional states linked to activation, exhaustion, persistence, and memory formation.^91-93 Recent studies with approved CD19-specific CAR T-cell products have uncovered specific signatures of CAR T-cell subsets that maintain effector functions and proliferative capacity in DLBCL and ALL.^87,93 

In axi-cel–treated DLBCL, Rezvan et al used integrated scRNA-seq and multiomic profiling to identify a distinct CD8⁺ subset (“CD8-fit”) with enhanced migration, serial killing, and mitochondrial fitness.⁸⁷ These cells expressed high levels of cytotoxic (GZMB, GNLY, PRF1, FASLG), cytokine/chemokine (CCL3-5, IFNG), and mitochondrial biogenesis (PGC1α pathway) genes, and correlated with superior in vivo persistence and clinical response. Independent data sets confirmed the CD8-fit signature as a predictive biomarker for durable efficacy.

Bai et al analyzed a large preinfusion scRNA-seq atlas in ALL, identifying a “type 2–high” CAR T subset (IL-4, IL-5, IL-13, and GATA binding protein 3 [GATA3]) linked to ultralong persistence (median B-cell aplasia of 8.4 years).⁹³ Integrated transcriptome/epigenome profiling revealed GATA3-driven regulation and intercellular signaling that limited dysfunction. Functional and longitudinal studies confirmed enhanced cytokine production, superior expansion, and robust recall responses, suggesting strategies such as IL-4 priming during manufacturing to improve outcomes.

As demonstrated by clinical outcomes and laboratory-based findings from CAR T-cell clinical trials conducted in patients with DLBCL and ALL,^87,93 accurately evaluating the intrinsic antitumor activity of CAR T-cell products regardless of cancer type requires single-cell transcriptomics as a central approach, complemented by functional assays and additional supportive methodologies.⁹⁴ By capturing gene expression profiles at the resolution of individual cells, scRNA-seq reveals distinct cellular subsets that differ significantly in their activation status, memory differentiation patterns, and exhaustion markers.⁹⁵ Additionally, scRNA-seq can help uncover potential resistance mechanisms and identify molecular pathways associated with therapeutic failures. The integration of cellular indexing of transcriptomes and epitopes by sequencing complements scRNA-seq by simultaneously profiling protein expression on the cell surface alongside transcriptomic data at the single-cell level resolution.⁹⁶ This dual measurement enhances the characterization of CAR T cells by directly correlating phenotypic protein markers with transcriptional signatures. Consequently, researchers gain a comprehensive understanding of the CAR T-cell populations, enabling precise identification of highly functional subsets and exhausted or dysfunctional cells.

Alongside transcriptomics, complementary functional assays conducted in vitro or in vivo are essential. Traditional chromium release assays measure the release of radioactive chromium from lysed tumor cells, providing quantitative assessments of direct cell killing capability.⁹⁷ Bioluminescence-based assays use target cells expressing luciferase, enabling the sensitive and rapid detection of cytotoxicity through emitted luminescence, which allows real-time monitoring and quantification of cell killing. Similarly, impedance-based assays track real-time electrical changes as CAR T cells lyse target cells, providing kinetic data on cytotoxic interactions.⁹⁸ Flow cytometry–based viability assays further complement these approaches by providing precise and multiparametric analyses of target cell viability after exposure to CAR T cells, revealing specific killing mechanisms and the efficiency of tumor clearance.⁹⁹ Furthermore, cytokine release profiling assays, including enzyme-linked immunosorbent assays, enzyme-linked immunospot assays, and multiplex cytokine platforms, are pivotal in evaluating CAR T-cell activation and functional capacity.¹⁰⁰ These assays quantify critical cytokines, such as IFN-γ, IL-2, TNF-α, and GM-CSF, indicative of potent antitumor responses and CAR T-cell activity. Advanced single-cell secretome profiling platforms, such as IsoLight, further enhance the resolution of these assessments by quantifying cytokine secretion at the single-cell level, enabling the identification of CAR T cells with highly polyfunctional profiles associated with improved clinical outcomes.^54,101 

Comprehensive proteomic analyses provide a framework for predicting CAR T-cell efficacy and toxicity through the precise quantification of protein expression patterns, including cytokines, chemokines, and other soluble mediators.¹⁰² Multiplex cytokine bead assays and advanced single-cell secretomics have proven valuable in evaluating CAR T-cell products.¹⁰¹ Recent studies demonstrated that polyfunctional cytokine secretion profiles significantly correlate with successful CAR T-cell expansion, potency, and toxicity profiles.^88,102 Cytokine evaluation by optimized single-cell platform revealed qualitative differences in cytokine secretion patterns between CAR T cells targeting different antigens, B-cell–activating factor receptor (BAFF-R) and CD19, underscoring the importance of antigen specificity and stimulation conditions in shaping CAR T-cell responses.⁸⁸ Serial proteomic profiling in ALL provides essential insights into the mechanisms underlying CRS and ICANS.¹⁰² Elevated levels of inflammatory cytokines such as IL-6, IL-1β, and GM-CSF detected via proteomic analyses can serve as predictive biomarkers for severe adverse events, enabling preemptive therapeutic interventions.^61,103 Incorporating proteomic profiling into routine practice could improve patient stratification and therapeutic optimization.

Polyfunctionality assays, which assess the ability of single T cells to secrete multiple cytokines simultaneously, are emerging as robust predictors of CAR T-cell functionality.^54,104 This capability reflects the broader spectrum of immune responses, providing a more comprehensive insight into CAR T-cell quality and therapeutic potential compared with measuring single cytokines alone. Polyfunctional subsets demonstrate superior proliferation, persistence, and cytotoxicity, outperforming single-cytokine readouts.^55,105 Rossi et al used a single-cell multiplexed platform to measure 32 cytokines simultaneously, identifying polyfunctional subsets (eg, IFN-γ, TNF-α, IL-17A, IL-8, and macrophage inflammatory protein-1α) correlated with clinical response.⁵⁶ The polyfunctionality strength index, combining frequency and intensity of cytokine secretion, outperformed expansion metrics as a predictive biomarker. Michelozzi et al further showed that lower-affinity CAR T cells (CAT CAR T cell) displayed greater polyfunctionality and less exhaustion than higher-affinity CAR T cells (FMC63 CAR T cell).⁵⁵ This highlights the biological significance of polyfunctional cytokine responses not only as markers of potent effector function but also as indicators of sustained cellular fitness and durability, ultimately translating into better clinical efficacy and reduced toxicity profiles.

Integration of polyfunctionality assays with scRNA-seq provides a comprehensive map of CAR T-cell states.^106-108 Moreover, it helps pinpoint key regulatory genes and pathways crucial for polyfunctional activity, informing the optimization of CAR T-cell design.^109,110 Additionally, profiling polyfunctionality alongside molecular characteristics may enable clinicians to predict clinical outcomes before CAR T-cell infusion.^56,88 For example, a recent study published in Science used pooled functional screening of a combinatorial signaling domain library with scRNA-seq to elucidate the impact of CAR signaling architectures on T-cell activation and longevity.¹¹¹ These data correlated in vitro CAR T-cell phenotypes with clinical outcomes, underscoring the role of a clinically informed screening approach.

Future CAR T-cell therapies and clinical evaluations would significantly benefit from systematically incorporating polyfunctional cytokine profiling of the infusion product alongside scRNA-seq analyses to optimize treatment effectiveness and mitigate adverse events. Available single-cell polyfunctionality assays are likely robust across multiple targets. For example, Dong et al recently adapted and optimized a single-cell technical platform for characterizing cytokine polyfunctionality of T cells recognizing BAFF-R.⁸⁸ This assay is being used to support an ongoing clinical trial of BAFF-R CAR T-cell therapy in relapsed or refractory B-cell non-Hodgkin lymphomas, which is showing early, potentially transformative, results with a complete response rate of 100%, durability beyond ≥2 years, and only grade 0/1 CRS.¹¹² 

Recent advancements in CAR T-cell therapy have been accelerated by integrating immunophenotyping with machine learning (ML), enabling precise patient stratification and optimized therapeutic outcomes.¹¹³ Automated immunophenotyping and artificial intelligence (AI)–driven manufacturing are emerging tools to address product variability, standardize quality, and ensure consistent efficacy.¹¹⁴ A recent study demonstrated the utility of AI in controlling key manufacturing parameters such as cell selection, expansion, and phenotypic characterization, thereby improving batch consistency and overall efficacy.¹¹⁵ 

ML is also powerful for analyzing complex immunological data sets. Daniels et al applied combinatorial signaling motif libraries with ML algorithms to decode CAR T-cell phenotypes and predict responses based on motif composition, providing a predictive framework for potency and persistence.¹¹⁶ Similarly, ML-assisted imaging has advanced quantification of the immunological synapse, a determinant of CAR T-cell function. Convolutional neural networks can extract morphological and molecular features from synapse imaging, allowing reproducible assessments directly correlating with clinical outcomes.¹¹⁷ Naghizadeh et al further validated that ML-based quantification of synapse quality strongly predicted patient responses, underscoring the value of high-dimensional phenotypic data in personalized CAR T-cell therapy.¹¹⁸ 

Innovative combinations of AI and real-time monitoring technologies are also being explored. Boretti proposed integration of AI with Internet of things platforms to enable continuous monitoring of CAR T-cell functionality, allowing proactive adjustments and personalized dosing that improve safety and efficacy.¹¹⁹ Deep learning models are further extending clinical applications. For example, the ResNet18 with Convolutional Block Attention Module and Multi-Head Self-Attention model integrates clinical and biological parameters to predict leukemia treatment outcomes, exemplifying how computational frameworks can guide personalized regimens and mitigate risks.¹²⁰ 

Tumor antigen or epitope spreading refers to the expansion of immune responses from the initial target antigen to additional nontargeted antigens released during tumor cell destruction.¹²¹ This mechanism potentially amplifies antitumor responses by broadening immune recognition of cancer cells, thereby improving clinical outcomes and reducing relapse rates.¹²² Epitope spreading was initially observed in immune-mediated diseases and conventional cancer therapies, including radiation, chemotherapy, and novel cancer vaccines.^123,124 It has recently been identified after CAR T-cell therapies, particularly with CD19-targeted products.¹²⁵ CAR T-cell cytotoxicity induces apoptosis or necrosis, releasing neoantigens crosspresented by antigen-presenting cells, thereby priming endogenous T-cell responses against untargeted tumor epitopes.^18,121 Clinical evidence shows that such responses correlate with sustained remission and longer progression-free survival, whereas also mitigating antigen escape, a major cause of relapse.^13,55,126 

Monitoring epitope spreading can serve as an early and dynamic biomarker of CAR T-cell efficacy.^127,128 Single-cell sequencing studies have linked durable clinical responses to diverse endogenous T-cell clonotypes, whereas multiplex cytokine profiling revealed polyfunctional cytokine secretion, underscoring their role in sustained immunity.^56,88,89,93 Additionally, antigen spreading has implications for mitigating antigen-loss relapse. By promoting endogenous T-cell surveillance, antigen spreading reduces selective pressure for tumor antigen loss, thereby countering resistance.^121,122 Mechanistically, CAR T-cell–mediated tumor killing generates an inflammatory milieu favorable to crosspresentation, whereas combination strategies such as checkpoint inhibition can further enhance spreading by restoring T-cell function.^83,85 

Future work should validate antigen spreading as a predictive biomarker across malignancies. Integration of immunophenotyping, single-cell multiomics, and functional assays may clarify its dynamics and clinical significance.^94,95 Harnessing this mechanism could enable personalized strategies combining CAR T cells with adjunctive immunotherapies.^116,121,122 

Figure 1 provides a schematic overview of predictive biomarkers influencing CAR T-cell therapeutic outcomes, encompassing intrinsic functional attributes, systemic immune contexture, and emerging multiomic strategies. Incorporating biomarkers discovered through advanced multiomics and computational analysis into clinical practice can improve patient selection, predict responses, and reduce adverse events. Before infusion, comprehensive immune profiling through high-dimensional flow cytometry and multiplex cytokine assays could be beneficial. Assessing baseline cytokines such as IL-6, IL-15, GM-CSF, and C-reactive protein helps identify patients at higher risk for CRS and neurotoxicity, allowing timely prophylactic interventions.

CAR T-cell products should undergo rigorous evaluation using scRNA-seq to detect signatures predictive of efficacy, such as CD8-fit subsets with cytotoxic markers and mitochondrial fitness or "type 2" subsets with robust GATA3-mediated signaling. Products enriched with beneficial T-cell subsets should be prioritized, whereas those dominated by exhausted T cells may require manufacturing optimizations or adjunctive therapies. Proteomic profiling through single-cell cytokine assays provides another key biomarker: higher polyfunctionality indices, reflecting superior expansion, persistence, and antitumor activity. In particular, IL-7 and IL-15 supplementation can enhance metabolic fitness by promoting mitochondrial biogenesis, fatty acid oxidation, and glucose use, thereby supporting long-lived memory phenotypes with greater therapeutic efficacy.

Throughout therapy, real-time monitoring with ML algorithms analyzing cytokine profiles and immunophenotyping can enable early adverse events detection and outcome predictions. Additionally, monitoring for epitope spreading after infusion through serial blood analyses using single-cell sequencing or multiplex immunophenotyping offers a provisional biomarker for sustained antitumor immunity and reduced relapse. Detecting significant epitope spreading may guide clinical decisions on monitoring, consolidation, and complementary therapies.

Beyond biomarker discovery, several practical clinical considerations are critical for successful translation into patient care. Lymphodepletion not only eliminates regulatory and suppressive immune populations but also elevates homeostatic cytokines such as IL-7 and IL-15, thereby creating a favorable microenvironment that supports CAR T-cell expansion and persistence.¹²⁹ Likewise, reducing tumor burden through bridging therapy or cytoreduction can mitigate baseline inflammatory tone, lowering the risk of severe toxicity and enabling more reliable CAR T-cell expansion.¹³⁰ Looking forward, future strategies to overcome poor-risk biomarker profiles may include manufacturing approaches that enforce metabolic fitness and memory programming, the design of armored or logic-gated CAR T-cell products resistant to inhibitory cues, rational drug combinations such as checkpoint blockade or BTK inhibition in B-cell malignancies, and adaptive algorithms that integrate baseline systemic markers with early on-treatment kinetics to guide preemptive interventions.¹³¹ 

Systematically integrating biomarker-driven strategies into clinical practice can personalize treatments by improving patient stratification and optimizing therapeutic regimens. The continued refinement of CAR T-cell manufacturing protocols informed by these predictive insights holds immense promise for improving therapeutic efficacy, minimizing adverse events, and achieving sustained remission in patients with hematologic malignancies.

This work was supported by the Gachon University research fund of 2024 (GCU-202410480001).

Contribution: K.H.Y. and L.W.K. conceptualized the review and wrote the original draft; K.H.Y., S.S., Z.D., A.K., S.-c.C., and L.W.K. critically revised and edited the manuscript; and all authors reviewed and approved the final version of the manuscript.

Conflict-of-interest disclosure: L.W.K. reports a consulting role with, and research funding from, PeproMene Bio, Inc; and equity ownership in InnoLifes and PeproMene Bio, Inc. S.-c.C. reports a consulting role with PeproMene Bio, Inc. The remaining authors declare no competing financial interests.

Correspondence: Larry W. Kwak, Toni Stephenson Lymphoma Center, Beckman Research Institute, City of Hope National Medical Center, 1500 E Duarte Rd, Duarte, CA 91010; email: lkwak@coh.org.