T cell-directed immunotherapies have changed outcomes in hematological malignancies, but durable benefit remains limited by relapse, antigen escape, poor cellular persistence, delayed immune reconstitution and infection. The key issue is not only whether a target antigen can be engaged, but whether the recruited or engineered T cells can expand, repeatedly kill, persist and recover after treatment. The Review defines T cell fitness as a multidimensional capacity that includes cellular availability, memory reserve, proliferative competence, cytotoxic function, metabolic resilience, resistance to chronic stimulation and persistence. It separates overlapping but distinct states—exhaustion, senescence and terminal differentiation—and argues for a multimodal, modality-specific assessment rather than reliance on single markers.
Exhaustion arises from persistent antigen exposure and is characterized by progressive loss of cytokine production, proliferation and cytolytic function coupled with co-expression of multiple inhibitory receptors. PD-1 alone is insufficient to define exhaustion because it also marks recent activation. More definitive exhaustion identification integrates co-inhibitory receptors (for example PD-1 with TIM-3, LAG-3, TIGIT), functional impairment, clonotype behavior and transcriptional or chromatin features.
Importantly, exhausted T cells are heterogeneous. Progenitor-like exhausted cells (TCF1-positive) retain self-renewal and can expand after PD-1 blockade, whereas terminally exhausted cells have limited proliferative recovery. Exhaustion has an epigenetic dimension: transient reinvigoration by checkpoint blockade does not necessarily erase chromatin features that constrain durable memory differentiation. Thus, a prior history of chronic stimulation can leave an ‘‘epigenetic scar’’ that affects later responses to CAR-T manufacturing, bispecific antibodies or immune surveillance even after tumor debulking.
Senescence reflects cumulative cell division, aging, telomere erosion, DNA-damage responses and impaired cell-cycle re-entry. Senescent-like T cells may retain cytotoxic molecules yet fail to expand sufficiently for manufacturing or sustained serial killing. Age reduces naïve T cell output and repertoire diversity while shifting the compartment toward differentiated memory and effector populations.
Practical senescence-associated features include CD27 or CD28 loss, CD57 and KLRG1 expression, shortened telomeres and poor proliferative responses, but no single marker is specific. The clinical significance is most evident for autologous cellular therapies: a leukapheresis product enriched for CD57+ KLRG1+ CD28– or terminal effector cells may yield a CAR-T product with limited proliferative reserve and persistence. The Review emphasizes that both patient-intrinsic cell quality and engineered signaling (for example CAR tonic signaling) jointly determine therapeutic fitness.
Terminal differentiation denotes an effector end state—often TEMRA-like cells—that can express granzyme B, perforin and CX3CR1, retain immediate cytotoxicity, but have reduced proliferative reserve, plasticity and long-term persistence. Cytotoxicity and persistence are separable properties: terminal effectors may mediate rapid tumor killing but fail to sustain prolonged immune pressure.
This distinction matters clinically. Less differentiated central-memory and stem-like memory populations generally support durable adoptive therapy, whereas a repertoire dominated by terminal effectors or senescent cells may respond initially to bispecific antibodies or other redirection strategies but then fail under repeated engagement or high antigen burden.
T cell fitness integrates measurable domains needed for a specific therapeutic platform. For CAR-T therapy the substrate must allow activation, ex vivo engineering, expansion and long-term in vivo persistence. Bispecific antibodies depend on the quantity, localization and serial responsiveness of endogenous T cells. Checkpoint blockade requires tumor-reactive cells present in a reversible inhibitory state. Consequently, the same patient may have adequate fitness for one platform but not for another.
The Review recommends moving beyond single-marker labels. Checkpoint receptors, senescence-associated markers, differentiation markers and cytotoxic molecules should be interpreted as domains within a multidimensional profile. The clinical aim is to identify which functional capacity is limiting and whether it is potentially modifiable.
No single validated clinical score currently captures T cell fitness across hematological malignancies or treatment platforms. The authors propose a domain-based fitness profile for prospective studies. At minimum, studies should measure:
Each domain should be standardized within assays and modeled against modality-specific endpoints: manufacturing success and in vivo expansion for CAR-T, serial response and immune attrition for bispecifics, and proliferative reinvigoration for checkpoint blockade. The Review cautions against deriving composite indices until training, calibration and external validation are completed; until then, domain values should be reported separately to avoid false precision.
Different hematological malignancies shape the T cell compartment through distinct mechanisms. Multiple myeloma exemplifies a convergence of older patient age, a suppressive bone marrow niche, chronic antigen exposure and cumulative therapy pressure; during myeloma evolution, clonal T cell expansions with multiple checkpoint expressions and contraction of TCR diversity have been reported. Lymphoma is influenced more by lymph-node architecture, regulatory T cells, macrophages and spatial immune contexture. Acute lymphoblastic leukemia outcomes are often dominated by product persistence and antigen escape, while acute myeloid leukemia couples marrow suppression with overlap of targets on normal hematopoiesis. These disease-specific landscapes argue against extrapolating a single exhaustion signature across malignancies.
Evidence linking T cell attributes to outcomes comes from CAR-T cell trials, bispecific antibody studies and checkpoint blockade cohorts, but the body of data has limitations. Many candidate biomarkers originate from small translational cohorts, retrospective analyses, selected trial populations or preclinical models; sampling time points and marker definitions vary, and associations in one disease or platform may not generalize to others.
For example, CAR-T response has been associated with memory-related transcriptional programs and proliferative competence, whereas nonresponse correlated with effector differentiation, glycolytic programs, exhaustion signatures and apoptosis in some studies. Bispecific antibodies recruit the endogenous T cell pool repeatedly; continuous stimulation can impair function while treatment-free intervals can permit partial recovery in experimental systems. Checkpoint blockade effects vary by disease context: PD-1 blockade performs well in classical Hodgkin lymphoma but has an unfavorable benefit–risk balance in multiple myeloma combinations, illustrating that checkpoint expression alone does not identify a universally reversible state.
The fitness framework can inform practical clinical questions. These include when to collect cells for autologous manufacture, how to select bridging therapy that preserves T cell function, which biomarkers merit prospective testing, and how treatment duration and sequencing might preserve immune competence. The Review emphasizes that modality-specific and disease-aware thresholds are needed before making definitive clinical recommendations.
The authors discuss strategies to mitigate reliance on compromised autologous T cells, including gene-edited products, CAR-NK cells and metabolic interventions designed to enhance cellular resilience. These approaches may provide alternatives when patient-derived T cells show limited fitness, but prospective testing and modality-specific benchmarks are required.
Key limitations of the current evidence base include retrospective designs, small cohorts, heterogeneous sampling and marker definitions, and disease- or platform-specific findings that may not generalize. The Review calls for prospective, standardized studies that report domain-level fitness measures, develop modality-specific endpoints, and train and validate composite indices before clinical use.
A fitness-based approach reframes biomarker development and treatment selection around the functional capacities required by each immunotherapy modality. Distinguishing exhaustion, senescence and terminal differentiation and assessing multiple domains of T cell fitness can better predict manufacturing success, in vivo expansion, durability of response and risk of immune attrition. However, the framework requires prospective validation and modality-specific thresholds before it can be adopted as a routine clinical tool.