When, Where, and How Much: Toward Smarter Design and Delivery of CAR T‑Cell Therapy

Abstract

Chimeric antigen receptor (CAR) T-cell therapies have fundamentally reshaped clinical oncology by achieving durable remissions in refractory hematologic malignancies. However, their broader translation remains bottlenecked by severe, treatment-associated hyperinflammatory toxicities, including cytokine release syndrome (CRS), immune effector cell–associated neurotoxicity syndrome (ICANS), and immune effector cell-associated hemophagocytic lymphohistiocytosis-like syndrome (IEC-HS). Current management strategies are reactive and rely on systemic immunosuppressants, such as corticosteroids and the anti-IL-6 receptor antibody tocilizumab. While these interventions blunt acute inflammation, they can exert systemic effects that have the possibility to impair CAR-T cell persistence, prolong cytopenias, and compromise long-term antitumor efficacy. In this review, we challenge the conventional paradigm that treats toxicity as a uniform, unavoidable consequence of systemic immune activation. Instead, we present emerging evidence demonstrating that CAR-T-associated toxicities possess discrete spatial, temporal, and kinetic organization across distinct tissue compartments and biological phases. We propose a cohesive engineering framework that leverages advanced drug delivery technologies to convert toxicity management from a reactive clinical hurdle into a systems-level design problem. By organizing future therapeutic strategies around three core pillars: temporal control (gating interventions to specific phases of immune escalation), dosing and kinetic control (flattening proliferation slopes via fractionated delivery, dynamic feedback loops, and mathematical modeling), and spatial control (anatomical targeting of inflamed vasculature, lymphoid organs, and the central nervous system), it may be possible to selectively decouple localized toxic inflammatory cascades from therapeutic CAR-T cell expansion. Ultimately, integrating these spatiotemporal and kinetic parameters provides a blueprint for engineering the next generation of safer, scalable, and highly efficacious cellular immunotherapies.

The Toxicity Bottleneck in CAR-T Cell Therapy

Chimeric antigen receptor (CAR) Tcell therapies have fundamentally reshaped the landscape of clinical oncology, achieving durable remissions in patients with otherwise refractory hematologic malignancies and showing early promise in solid tumors. − CAR T-cell therapy, in which a patient’s own T cells are genetically reprogrammed to recognize tumor antigens, demonstrates the immense power of harnessing the immune system to actively eliminate malignant cell populations. − Yet despite these transformative clinical outcomes, the broader expansion of CART therapy remains heavily constrained by severe treatment-associated toxicities. −

In clinical practice, toxicity represents a significant barrier to the wider deployment of CART therapies. − Acute complications of CAR T-cell therapy comprise a spectrum of hyperinflammatory syndromes that include cytokine release syndrome (CRS), immune effector cell-associated neurotoxicity syndrome (ICANS), and immune effector cell-associated hemophagocytic lymphohistiocytosis (HLH)-like syndrome (IECHS). − Delayed complications, including prolonged cytopenias and endothelial dysfunction, also occur and can impair immune recovery from preconditioning regimens and complicate post-treatment care. − Because these toxicities are frequently severe and require treatment in intensive care units, they can significantly increase treatment cost, limit outpatient CAR T-cell administration, and create barriers to broader patient access.

Current clinical strategies to mitigate these toxicities are fundamentally limited. Most interventions are reactive, initiated only after inflammatory syndromes have already developed, and many are associated with substantial side effects. − Corticosteroids are a cornerstone of treatment for these hyperinflammatory syndromes, particularly ICANS and IEC-HS, and interleukin (IL)-6 blockade with the anti-IL-6 receptor antibody tocilizumab is the standard first-line therapy for CRS. − Although these interventions are effective at suppressing inflammatory responses, they exert broadly immunosuppressive effects that can impair CAR-T cell expansion and persistence, prolong cytopenias, and increase susceptibility to infection. − More broadly, these interventions treat toxicity as a uniform systemic process, despite mounting evidence that inflammatory responses during CAR T-cell therapy arise from distinct cellular pathways and anatomical compartments.

Herein, we discuss evidence showing that CAR T-cell toxicities are inherently structured across time and tissue compartments and propose that therapeutic strategies must be designed to reflect this organization. Toxicity should be viewed not as an unavoidable consequence of immune activation, but as a systems-level design problem. Drug delivery technologies offer a powerful set of tools for controlling where, when, and how strongly immunomodulatory interventions act within the body. By integrating spatial targeting based on specific tissues, temporal control aligned with immune activation phases, and kinetic modulation of inflammatory signaling, it may be possible to blunt localized toxic inflammatory cascades while preserving CAR T-cell expansion and antitumor efficacy.

We propose drug delivery as a conceptual framework for engineering safer CAR T-cell therapies through precise spatiotemporal control of immune activation. We argue that there are three major facets in the design of safer and more effective immunotherapies: temporal control (Figure ), dosing and kinetic control (Figure ), and spatial control (Figure ). Spatial control defines where immune modulation occurs; temporal control governs when immune activity is permitted or restrained; and dosing determines how much intervention is delivered, how rapidly it is introduced, and how these parameters shape the kinetics and magnitude of immune activation.

1.

Pillar 1Temporal Control Framework for CAR T-Cell Optimization. Schematic representation of temporal control in engineered immunotherapy safety, illustrating how the timing of therapeutic activity can be regulated to shape immune activation. Temporal control governs when interventions are initiated, modulated, or halted, allowing treatment to be aligned with the evolving stages of immune activation and toxicity.

2.

Pillar 2Dosing Control Framework for CAR T-Cell Optimization. Schematic representation of dosing control in engineered immunotherapy safety. Dosing control regulates the magnitude, rate, and sequencing of therapeutic interventions to shape the kinetics of CAR T-cell expansion and immune activation.

3.

Pillar 3Spatial Control Framework for CAR T-Cell Optimization. Schematic representation of spatial control in engineered immunotherapy safety. Spatial control directs immune modulation to specific anatomical sites to achieve organ-level precision and minimize off-target toxicity. Spatially restricted therapeutic approaches.

Toxicity’s Time and Place

Emerging mechanistic insights suggest that CAR T-cell–associated toxicities share early systemic, myeloid-driven inflammation, but differ in their tissue localization, timing, and clinical manifestations. Temporally, toxicity evolves through sequential phases of immune activation. Following infusion, CAR-T cells rapidly engage tumor antigen and undergo clonal expansion, rapidly releasing cytokines such as interferon (IFN)­γ and GM-CSF that initiate downstream immune activation. These cytokines in turn activate monocytes and macrophages, which themselves release high levels of inflammatory mediators, generating a feedforward cytokine cascade that drives CRS. As inflammation progresses, toxicity often becomes compartmentalized within specific organs, such as the central nervous system (CNS). In ICANS, for example, endothelial activation and disruption of the blood–brain barrier (BBB) allows access of inflammatory mediators to the CNS. In IEC-HS, by contrast, inflammation arises predominantly within myeloid-rich organs such as the liver and spleen. These observations suggest that toxicity arises from coordinated but compartmentalized immune dynamics rather than a single systemic inflammatory event.

CRS is the most common CAR-associated acute toxicity, occurring in more than 80% of patients receiving CAR T cells for hematologic malignancies. , CRS is characterized by massive release of pro-inflammatory cytokines by many cell types, often described as a “cytokine storm,” which typically manifests as fever and can lead to hypotension, hypoxia, and multiorgan damage in more severe cases. Although most cases of CRS are mild-to-moderate, requiring minimal intervention that can be managed in an outpatient setting, severe CRS usually requires hospitalization, with the most severe cases treated in intensive care units. Onset of CRS occurs a median time of 2–5 days after infusion for most FDA-approved CAR T-cell therapies, , with more severe cases tending to begin earlier and last longer, highlighting the need for prophylaxis and early interventions to mitigate the most severe toxicities.

The inciting events in CRS are mediated by activated CAR T cells, which release inflammatory cytokines (IFN-g, IL-2, TNF, and GM-CSF) and trigger downstream activation of other immune cells. However, the primary drivers of hyperinflammation are myeloid cells, which release additional cytokines including IL-6 and IL-1, that are recognized as key mediators of systemic inflammation in CRS. , The incidence of CRS is determined by factors beyond tumor burden and magnitude of CAR T-cell expansion, − suggesting that the endogenous immune system is an important determinant of pathogenesis. This idea is supported by the fact that the severity of CRS does not directly correlate with better CAR T-cell function and tumor killing. Indeed, while the development of any-grade CRS, compared with no CRS, appears to be associated with improved progression-free survival at 1 year, severe CRS (grade ≥3) has been associated with inferior overall survival outcomes. , Prevention of severe toxicities is thus key for improving outcomes and broader application of CAR T-cell therapies.

ICANS occurs in roughly 35% of patients with hematologic malignancies. , Although ICANS is defined and graded according to neurological symptoms (consciousness level, seizures, and presence of cerebral edema), evidence suggests that it also results from systemic, rather than CNS-specific, toxicity. ^, The association with systemic inflammation is further supported by the fact that a majority of ICANS cases occur alongside CRS, although there have been a few reports of mild ICANS in patients without CRS. −

The neuropathology of ICANS is not completely understood, but the main mechanisms are thought to be endothelial activation and BBB disruption as a result of systemic inflammation driven by myeloid-derived cytokines. − Although CAR T cells have been found to infiltrate the CNS in severe cases of ICANS, this seems to be a result of BBB permeability rather than a cause. Indeed, in patients with CNS solid tumors, in which CAR T-cells are administered directly into or designed to traffic to the CNS, ICANS develops less frequently compared to systemic delivery of CAR T cells in patients with non-CNS malignancies. − This observation also supports the idea that neurologic toxicity is driven primarily by non-T cell sources of inflammation. A key role for systemic inflammation was supported by a meta-analysis of 84 clinical trials, which found a higher incidence and severity of CRS and neurological symptoms in patients with hematological malignancies compared to those with solid tumors. Indeed, several promising studies have suggested that localized CAR T-cell delivery to solid tumors results in fewer systemic toxicities. ,,

IEC-HS, a more recently recognized toxicity associated with CAR T-cell therapy, is primarily characterized by organ damage, particularly to the liver and spleen, and hyperferritinemia. , IEC-HS mimics HLH, which is generally categorized as either primary or secondary. Primary HLH results from genetic variants that impair cytotoxic cell function, thereby extending T-cell activation due to prolonged immunological synapse formation and subsequent release of excessive levels of pro-inflammatory cytokines, whereas secondary HLH (also known as macrophage activation syndrome [MAS]) is often triggered by immune insults such as infections. In both primary and secondary HLH, macrophages are activated as a consequence of the initial inflammation and release additional cytokines, triggering a systemic hyperinflammatory state.

The pathophysiology of IEC-HS is thought to be similar to MAS but is initiated by hyperactivated CAR T cells. , In contrast to classical cytokine release syndrome (CRS), which is driven predominantly by early systemic cytokine amplification after immune effector cell activation, IEC-HS is marked by sustained macrophage and monocyte activation, pathologic hemophagocytosis, hyperferritinemia, coagulopathy, hepatic dysfunction, and, in severe cases, multiorgan failure. Although CRS and IEC-HS may overlap temporally and likely exist along a biologic continuum, IEC-HS often develops after CRS resolution or during prolonged inflammation, suggesting partially distinct immunopathologic mechanisms. ,, Emerging evidence implicates persistent IFN-γ signaling, endothelial dysfunction, and impaired immune homeostasis as key contributors. − Clinically, this distinction is important because therapies effective for CRS, particularly IL-6 blockade alone, may be insufficient in IEC-HS, where broader suppression of macrophage-driven inflammation is often required. −

Shared cytokine profiles, clinical features, and cellular mediators suggest that CRS, ICANS, and IEC-HS exist along a spectrum of CAR T-cell-associated toxicities. ,, Shared cytokine profiles, clinical features, and cellular mediators suggest that CRS, ICANS, and IEC-HS exist along a spectrum of CAR T-cell-associated toxicities. , At the same time, important differences in dominant molecular drivers must be recognized. , Whereas IL-6 blockade is effective in most cases of CRS, it has shown limited efficacy in IEC-HS, − , highlighting the distinct contribution of macrophage-driven hyperinflammation in this syndrome.

Notably, our group and others − have suggested that baseline hyperinflammation prior to therapy may predispose patients to severe immune-related toxicities. In patients with limited marrow reserve, an impaired ability to regulate or recover from excessive inflammatory responses may influence how hyperinflammation manifests clinically, potentially contributing to whether patients develop CRS alone or progress to ICANS or IEC-HS.

These distinctions have important therapeutic implications. Management of CRS is centered on early cytokine modulation and strategies to limit excessive systemic immune activation while preserving antitumor efficacy. In ICANS, approaches targeting endothelial dysfunction, blood–brain barrier integrity, and myeloid-driven neuroinflammation may be particularly relevant (Table ). In contrast, IEC-HS may require earlier identification of high-risk patients and broader suppression of macrophage-driven inflammation, including targeting IFN-γ, IL-1, or JAK-STAT signaling pathways (Table ). Across all three toxicities, integration of predictive biomarkers, toxicity-adapted cell engineering, and mechanism-specific immunomodulation will likely be essential to improving the safety and precision of immune effector cell therapies.

1. Characteristics of iRAEs in CAR-T Therapy .

Feature	CRS	ICANS	IEC-HS
Dominant cell types	Activated T cells, monocytes/macrophages	Endothelial cells, myeloid cells, CNS-resident immune cells	Macrophages/monocytes, activated T cells ,,,
Key cytokines/mediators	IL-6, IL-1, IFN-γ, TNF-α	IL-1, IL-6, GM-CSF, endothelial activation markers ,	IFN-γ, IL-18, IL-1, CXCL9, ferritin-associated inflammatory pathways ,
Primary affected organs/systems	Systemic inflammatory response; cardiovascular and pulmonary systems	Central nervous system	Liver, bone marrow, coagulation system, spleen, multiorgan involvement
Typical clinical manifestations	Fever, hypotension, hypoxia	Encephalopathy, aphasia, seizures, cerebral edema	Hyperferritinemia, cytopenias, coagulopathy, hepatic dysfunction, hemophagocytosis
Temporal pattern	Usually early after infusion	Often concurrent with or following CRS	Frequently delayed or prolonged after CRS
Standard interventions	Tocilizumab, corticosteroids, supportive care	Corticosteroids, supportive neurologic management; limited role for tocilizumab	Corticosteroids, anakinra, IFN-γ-directed strategies, intensified immunosuppression ,
Emerging approaches	GM-CSF blockade, IL-1 inhibition, engineered CAR designs ,	Endothelial stabilization, myeloid modulation, BBB-protective strategies ,,	JAK inhibition, IFN-γ neutralization, macrophage-targeted therapies

^aFor the most up-to-date clinical parameters and grading criteria, please refer to the 2026 ASTCT consensus grading for toxicities after immune effector cell therapy.

Dose–Response Dynamics

The administered dose of CAR T-cells is a key clinical variable shaping both the therapeutic activity and toxicity risk. However, CAR T-cell therapy challenges traditional pharmacodynamic concepts of dose because the infused cells can interact dynamically with the tumor and host immune compartments after infusion. Conventional dose–response thinking assumes that increasing the administered amount should predictably increase biological effect. CAR-T therapy violates this assumption because the administered product can expand, contract, persist, and interact with tumor and host immune compartments over time. As a result, identical doses can produce different expansion curves and toxicities. In addition, dosing in CAR T-cell therapy involves not only the number of infused CAR T cells, but also patient characteristics, the infusion schedule, and conditioning intensity. Because CAR T cells are living, self-amplifying therapies, a fixed inputthe administered cell dosecan produce a nonlinear biological output once cells encounter antigen, proliferate, recruit myeloid cells, and initiate cytokine cascades. Toxicity is therefore often better explained by the trajectory of immune activation than by administered dose alone. Clinically important variables that determine whether toxicities develop include peak CAR T-cell expansion, rate of expansion, and total inflammatory exposure over time. CAR T-cell dosing could therefore be redefined as a kinetic control problem, in which dosing is harnessed upstream to shape the rise, peak, duration, and resolution of immune activation, before antitoxicity interventions are used to shape the immune activation trajectory downstream.

The main limitation of fixed-dose thinking is that it treats CAR T-cell dosing as a static quantity rather than a dynamic input. A lower infused dose may still produce severe CRS if tumor burden is high, lymphodepletion strongly supports expansion, and/or the CAR T-cell construct inherently undergoes rapid proliferation. On the other hand, a higher infused dose may be well tolerated if antigen burden is lower or expansion is slower. The core problem with dosing is not just “too much dose,” but the ambiguous turn from intervention to steep immune activation kinetics.

Clinical experience in patients with relapsed or refractory chronic lymphocytic leukemia showed that CAR-T cells can persist and induce durable remissions, but without a simple relationship between infused dose and downstream clinical activity. A trial of the CD19-targeted CAR T-cell therapy tisagenlecleucel in pediatric and young adult patients with B-cell acute lymphoblastic leukemia showed that in vivo CAR-T expansion and persistence are closely associated with both therapeutic efficacy and toxicity, including CRS. A subsequent clinical pharmacology analysis of tisagenlecleucel found that response and CRS were more closely associated with in vivo expansion and persistence than with initial infused dose, supporting the idea that downstream cellular kinetics, rather than administered dose alone, could determine clinical outcome.

The rate of CAR T-cell expansion is also important. CRS is a rapid systemic inflammatory response, and severe cases often occur when immune activation escalates faster than the host system can buffer it. Mathematical modeling demonstrates that the biological rate of antigen-mediated expansion (how quickly CAR-T reach their targets and start replicating) combined with the CAR T-cell characteristics (e.g., binding affinity, phenotypes, tonic signaling, etc.) determine the relationship between cellular kinetics and clinical outcomes, which in turn can correlate with the resulting toxicity. , Multiple kinetic parameters appear to influence toxicity severity, including the rate of CAR T-cell expansion, peak expansion magnitude (Cmax), persistence, cumulative cytokine exposure over time (AUC), and the timing and duration of inflammatory signaling. − Mathematical and systems pharmacology models suggest that rapid antigen-driven expansion kinetics may contribute to inflammatory amplification by accelerating cytokine accumulation beyond host buffering capacity. − However, these relationships remain incompletely defined and likely interact with additional variables including tumor burden, lymphodepletion intensity, CAR construct design, and host immune state.

The patient’s tumor burden influences the effective CAR T-cell dose by changing the amount of antigen available for CAR engagement. A given number of infused CAR T cells does not enter the same antigenic environment in every patient: in patients with higher antigen load, CAR T cells may encounter more target cells, undergo stronger activation, and release more inflammatory cytokines. This increased T-cell activation can then recruit and activate myeloid cells, amplifying cytokine production, and increasing the risk of CRS. Experience with other CD19-directed T-cell–engaging therapies, such as blinatumomab, supports this mechanistic link by showing that forced T-cell engagement of leukemia cells can trigger CRS characterized by abnormal macrophage activation and cytokine-directed inflammatory amplification. Thus, antigen availability may act as a multiplier of biological dose, converting the same infused CAR T-cell number into different levels of immune activation across patients.

Conditioning regimens function as hidden dosing modifiers. Lymphodepletion is often used in preparation for CAR T-cell infusion, but this pretreatment strongly affects how the CAR T cells behave once in the patient. For example, by reducing endogenous immune competition and increasing homeostatic cytokines such as IL-7 and IL-15, lymphodepletion can enhance CAR-T expansion and persistence. , Conditioning intensity thus changes how a given CAR T-cell dose translates into in vivo expansion kinetics.

New Approaches to Dosing

Because CAR T-cell therapy defies conventional dosing concepts, dosing strategies must be suitably nuanced. One practical dosing strategy is fractionated dosing, in which the total CAR T-cell dose is split across multiple infusions rather than delivered all at once. Fractionation changes the rate at which the immune system is perturbed and creates opportunities to pause, reduce, or modify later dosing based on early toxicity signals. Clinical analyses suggest that dose fractionation can reduce CRS severity while preserving therapeutic intent. In design terms, fractionated dosing regulates the rate of system perturbation, not only the total dose.

Another strategy is to engineer the CAR T-cell product so that its activation and expansion are constrained. The CAR construct itself determines in part how quickly and intensely cells respond after antigen engagement. For example, inclusion of costimulatory domains can increase expansion, persistence, and durability, making the construct design itself a form of dosing. CD28-based CARs are generally associated with more rapid expansion, earlier peak activity, and higher rates of acute toxicities such as CRS and ICANS, whereas 4–1BB–based CARs typically demonstrate slower expansion kinetics, greater persistence, and comparatively delayed inflammatory responses. − Antigen sensitivity can also be tuned so that CAR T cells preferentially respond to high-density tumor antigen while limiting activation against cells with low or physiologic antigen expression. In this approach, kinetic control is encoded directly into the re-engineered T-cells.

CAR T-cell activity can be pharmacologically modulated after infusion at multiple levels of the immune response. Some strategies target the CAR T cells directlyfor example, engineered kill switches or reversible signaling inhibitors such as dasatinib, which can transiently suppress CAR signaling and function early in the activation cascade. Others act downstream on inflammatory pathways, such as JAK inhibition, which can dampen cytokine signaling in severe or refractory CRS but may also reduce CAR T-cell expansion. Together, these approaches suggest that dosing could be managed as a dynamic, time-varying controlled process rather than a single fixed infusion.

Adaptive, feedback-driven dosing is another possibility. If early biomarkers can identify patients likely to develop severe CRS, then dosing and immunomodulation can be adjusted before the inflammatory cascade becomes clinically dangerous. Pediatric biomarker studies have identified early cytokine patternsincluding IL-2, IFN-γ, IL-6, and IL-10as predictors of severe CRS. Such biomarkers might provide a clear method to tweak dosing, in which measurement of the early immune trajectory allows one to infer toxicity risk and adjust subsequent dosing or immunomodulation accordingly.

One clinically feasible implementation of this framework could involve biomarker-guided adaptive intervention during early CAR T-cell expansion. Patients with high baseline tumor burden, elevated ferritin, CRP, angiopoietin-2, or early postinfusion cytokine elevation could be identified as high risk for severe CRS or ICANS before overt clinical deterioration. − These patients could undergo intensified monitoring during the expected window of peak CAR T-cell expansion, with temporally adjusted interventions including fractionated CAR T-cell dosing, delayed second infusions, preemptive IL-1 blockade, or early corticosteroid initiation. Pharmacokinetic monitoring of CAR T-cell expansion and cytokine trajectories could further guide escalation or de-escalation of therapy in real time. This example illustrates how temporal and dosing control may be implemented clinically through adaptive intervention strategies without requiring complex engineered delivery systems.

A preemptive strategy is to use systems pharmacology models to anticipate CAR-T cell expansion and toxicity before they occur. These mechanistic models leverage and can incorporate individual patient-specific variables such as baseline tumor burden, infused product phenotype (e.g., effector versus central memory), elevated baseline inflammation. This allows investigators to simulate alternative dosing strategies and predict outcomes for both CAR T-cells and cytokines, such as the peak expansion concentration and the total systemic exposure over time for both CAR-T cell populations and inflammatory cytokines. ,, By linking the biological inputs to expansion kinetics and cytokine dynamics, such models aim to identify dosing strategies most likely to produce safe and effective immune activation trajectories.

Finally, dosing must be adapted to the biological context rather than determined solely by body size. CAR T-cell kinetics vary widely across disease states, tumor burden, immune environments, and product designs, reflecting the combined influence of product-intrinsic and patient-specific factors. In this sense, dosing integrates with the spatial and temporal control axes described above: the amount and rate of intervention shape the magnitude and kinetics of immune activation, while temporal and spatial factors determine when and where toxicity ultimately manifests.

Treating Toxicity: Current Approaches and Limitations

Interleukin-6 Inhibition

IL-6 is a pleiotropic cytokine recognized to be a key mediator of CRS, and blockade of IL-6 signaling via tocilizumab is the first recommended treatment for patients experiencing CRS following CAR T-cell infusion. Tocilizumab is an IL-6 receptor (IL-6R) antagonist that has been successful for treating most cases of CRS but has little to no impact on ICANS. The lack of efficacy in ICANS might be attributable to tocilizumab’s poor ability to cross the BBB, differential IL-6 signaling modalities, and/or the fact that tocilizumab actually leads to even more elevated levels of free IL-6 in the serum and CSF. , The exact mechanisms by which IL-6 contributes to pathogenesis, and why tocilizumab is not effective in all patients, are also currently unclear. IL-6 has two primary signaling modalities: classic and trans-signaling, both of which are blocked by tocilizumab. Classic signaling, wherein IL-6 binds to membrane-bound IL-6R, is typically associated with anti-inflammatory, protective functions and is restricted to cell types expressing membrane-bound IL-6R. By contrast, trans-signaling involves binding of the cytokine to soluble IL-6R (sIL-6R), which is thought to be more pro-inflammatory. − Tocilizumab thus dampens both inflammatory IL-6 trans-signaling and beneficial classical IL-6 signaling, which may reduce its efficacy in ameliorating CRS-associated endothelial cell damage and coagulopathies. Both IL-6 pathways involve the protein gp130, which forms a complex on the cell surface with IL-6 and membrane-bound IL-6R and also binds to IL-6-sIL-6R complexes. gp130 is expressed on nearly all tissues across the body, so nearly all cells are susceptible to IL-6 trans-signaling. This may be an important mechanism by which IL-6 leads to broad inflammation and multiorgan damage. Although there is little research on whether increased IL-6 trans-signaling is linked to increased CRS severity, data from COVID-19, which has a similar inflammatory profile, shows elevated sIL-6R in severe cases and suggests that sIL-6R is predictive of critical illness. − Trans-signaling has also been found to be the primary mechanism of IL-6-associated damage in the CNS. ,

As such, targeting IL-6 blockade to trans-signaling across the body could reduce toxicity while maintaining the protective functions of IL-6 in endothelial cells immune cell types that express IL-6R. The protein sgp130Fc, which specifically blocks IL-6 trans-signaling, has been shown to reduce chronic inflammation in patients with gastrointestinal diseases , and shows promise in reducing cytokine storms in preclinical models of severe COVID-19. , However, similar to tocilizumab, sgp130Fc does not effectively cross the BBB due to its large size, lowering its potential for treating ICANS. Thus, future research could explore development of small molecule inhibitors of gp130 that would be able to inhibit IL-6 trans-signaling in the CNS, thereby potentially reducing neurotoxicity.

Corticosteroids

Patients with CRS who do not respond to tocilizumab, or who develop mild-to-moderate ICANS, are treated with glucocorticoids (a subtype of corticosteroids). Their ability to cross the BBB makes them an effective treatment for ICANS. Glucocorticoids function by binding to cytosolic glucocorticoid receptors in various cell types, which can then act as transcription factors to suppress the expression of pro-inflammatory cytokines and upregulate anti-inflammatory signals. In addition to modulating cytokine expression, glucocorticoids also reduce capillary membrane permeability, thus suppressing migration of immune cells such as lymphocytes and neutrophils. , Currently, glucocorticoids are typically used reactively in response to toxicity emergence, but optimal dosing and timing of administration is still undetermined. , Studies have begun to explore earlier or prophylactic treatment as a means to prevent toxicities from occurring or worsening. Although early trials raised concerns that steroid use could suppress CAR T-cell function, later studies have shown that lower doses administered early are able to reduce toxicities without negatively impacting CAR T-cell efficacy. ,

Although corticosteroids are effective in reducing CRS and ICANS, their broad immunosuppressive function has a number of drawbacks. By suppressing recovery of the endogenous immune system after lymphodepleting preconditioning, they can prolong cytopenias. They also increase the risk of infection, particularly when prolonged treatment or higher doses are given for more severe toxicities. ,

Fine-tuning of corticosteroid timing and dosing for treatment of CAR T-cell-associated toxicities has led to improved outcomes, but significant downsides still remain due to the nature of systemic immunosuppression. Although it is unlikely that steroid use will be completely eliminated as a treatment option, better targeting of the specific mediators of toxicities will reduce the need for these drugs.

Interleukin-1 Inhibition

Mouse models have identified monocytes and macrophages as the primary producers of IL-1 early in the inflammatory cascade, upstream of IL-6, in multiple models of CRS and ICANS. , These models also suggest that IL-1 signaling is particularly relevant to neurotoxicity, supported by studies showing that IL-1 drives endothelial activation and cerebrovascular inflammation. ,

Early clinical trials of the IL1 receptor antagonist anakinra have shown promise in reducing CRS and ICANS, and anakinra is recommended for patients with severe ICANS refractory to corticosteroids. − Its smaller size enables partial BBB penetration, which may contribute to improved activity in neurotoxicity relative to larger cytokine-targeting agents strategies such as tocilizumab. , Notably, anakinra and other IL-1 blocking strategies have been established as standard and effective treatments in other inflammatory settings such as rheumatoid arthritis. Studies of prophylactic anakinra for prevention of CRS or ICANS have not yet yielded concrete results, but generally suggest a trend toward reduced severity and shorter duration of toxicities. − Though IL-1 blockade shows promise, dosage and timing of administration for optimal efficacy is still under study. Studies of IL-1 inhibitors for treatment of MAS suggest that specific dosing is a key part of inflammatory control, emphasizing the importance of careful temporal and dose modulation in drug treatment.

Spatial Control Dynamics

CAR T-cell–associated toxicities share a common foundation in systemic, myeloid-driven inflammation. This being said, their downstream manifestations are not anatomically uniform. CRS, ICANS, and IEC-HS reflect overlapping inflammatory biology, but their clinical consequences localize to distinct vascular, myeloid, and central nervous system compartments. Current mitigation strategies tend to ignore this compartmentalization and apply broad immunosuppression across the entire immune system. This mismatch suggests that toxicity control should instead incorporate anatomical specificity as a core design parameter. Spatially restricting immunomodulation may allow local inflammatory damage to be alleviated while preserving larger-scale antitumor immunity.

One of the clearest examples of spatial heterogeneity lies within the vascular and endothelial compartment. Endothelial activation has emerged as a central driver of severe CRS and ICANS, with endothelial dysfunction correlating strongly with toxicity severity. Clinical and mechanistic studies have demonstrated that elevated angiopoietin-2 and other markers of endothelial injury are associated with increased neurotoxicity, highlighting the importance of vascular integrity in disease progression. During inflammatory activation, endothelial cells upregulate adhesion molecules such as ICAM-1 and VCAM-1, facilitating leukocyte adhesion and trafficking, which further amplifies local inflammation. , In the context of ICANS, disruption of the BBB either precedes or occurs simultaneously with neurologic symptoms, allowing inflammatory mediators to access the central nervous system. These findings suggest that the vascular endothelium may function as an active participant rather than solely a passive target of inflammation.

Activated endothelial cells upregulate adhesion molecules such as ICAM-1 and VCAM-1, providing molecular entry points for targeted therapies. Nanoparticle- or antibody-based systems directed toward these markers have been shown to selectively localize therapeutic payloads to inflamed vascular regions. , By focusing on the inflamed vasculature, these strategies aim to reduce vascular leak and BBB disruption without broadly suppressing CAR T-cell function.

A second layer of spatial organization is seen in myeloid-rich lymphoid organs such as the liver, spleen, and bone marrow. Within these compartments, monocytes and macrophages serve as major producers of IL-6 during CRS, positioning them as key amplifiers of systemic inflammation. CAR-T cells themselves contribute to this process through the secretion of GM-CSF, which recruits and activates myeloid cells, further propagating cytokine production. ,,,, Importantly, preclinical models have shown that depletion of myeloid cells or neutralization of GM-CSF can significantly reduce CRS severity without impairing the cytotoxic function of CAR T cells. These findings have motivated early clinical testing of GM-CSF neutralization strategies such as lenzilumab, but evidence for clinical benefit remains preliminary; ongoing and early phase studies are evaluating whether GM-CSF blockade can reduce CRS or neurotoxicity while preserving CAR T-cell efficacy.

Lymphoid-organ targeting showcases the importance of intervening at the earliest stages of immune activation. Early CAR-T activation and expansion are closely linked to lymphoid tissue interactions, including lymph nodes and bone marrow, whereas early activation signals shape downstream cytokine production. , Strategies that alter CAR T-cell expansion and persistence within lymphoid compartments, including lymphoid-directed antigen boosting and memory-enriching manufacturing conditions, suggest a potential route for tuning early activation kinetics; however, their ability to reduce systemic inflammatory toxicity remains to be established. This approach reflects the principle of controlling the inflammatory cascade at its point of origin.

The central nervous system represents a third spatial compartment, particularly in the context of ICANS. Neurotoxicity is characterized by BBB permeability changes and activation of CNS-resident immune processes, creating a localized inflammatory environment that differs from systemic circulation. Whereas IL-6 blockade has limited efficacy in ICANS, IL-1 blockade with anakinra has shown promise in preclinical studies in reducing CNS inflammation and improving neurotoxicity outcomes. , These findings support the idea that ICANS is partially insulated from systemic cytokine blockade due to BBB dynamics, creating a compartment in which inflammatory signaling can occur despite peripheral immune modulation.

Intrathecal administration or BBB-penetrant systems can achieve localized drug exposure within the CNS, offering a means to address ICANS while sparing systemic immune responses. , However, invasive CNS-directed delivery approaches also introduce important procedural and safety considerations, including infection risk, hemorrhage, catheter-associated complications, and the potential for localized neuroinflammation or neurologic injury. , These risks may limit feasibility in critically ill patients and underscore the need to carefully balance compartment-specific drug delivery against procedural burden and clinical stability. Mechanistic studies further support this approach, demonstrating that compartment-specific anti-inflammatory signaling within the CNS can mitigate neurotoxicity progression. ,

A potential future implementation of this framework could involve biomarker-guided spatial intervention for ICANS. In a high-risk patient demonstrating early endothelial activation, rising angiopoietin-2 levels, or early BBB dysfunction following CAR T-cell infusion, CNS-directed intervention could be initiated using intrathecal anakinra or BBB-penetrant anti-inflammatory delivery systems while preserving systemic CAR T-cell activity. ,,, Endothelial-targeted nanoparticles or antibody-based delivery systems directed toward ICAM-1 or VCAM-1 expressed on activated cerebral vasculature may further enable selective localization of anti-inflammatory payloads within inflamed CNS-associated vascular compartments. , Serial cytokine profiling, endothelial injury markers, and neurocognitive monitoring could then guide treatment timing and duration. ,, Although many of these approaches remain investigational, they illustrate how spatial targeting may complement temporal dosing strategies to reduce localized inflammatory injury while minimizing systemic immunosuppression. ,

Target microenvironment–directed immunomodulation represents another avenue for spatially precise intervention. Layered modulation of immune effector state within the target microenvironment may improve target engagement, persistence, and functional control of immune-targeting therapies with the goal of limiting systemic inflammatory amplification. Approaches integrating immune-targeting agents, including CAR-T, with immunomodulatory strategies, including checkpoint inhibition or other locally acting immune modifiers, are being explored as methods to tune effector activity directly within the target tissue microenvironment and overcome local resistance mechanisms. , More broadly, these strategies are unified by a shared design principle: preserve systemic immune competence while spatially constraining excessive inflammatory signaling to the site of immune engagement, thereby balancing antitumor efficacy with treatment-associated toxicity. However, because localized immune modulation may also attenuate desired antitumor responses or increase on-target, off-tumor toxicities in normal tissues expressing the target antigen, these strategies require careful optimization of antigen selection, payload design, and spatial control of effector function.

In summary, CAR-T-associated toxicities are spatially structured, yet current therapeutic strategies are usually spatially unstructured. Advances in oncology and inflammatory disease have already demonstrated the feasibility of anatomical targeting, providing a foundation for more precise immunomodulation. Integrating spatially restricted delivery into CAR T-cell workflows has the potential to reduce severe complications, preserve CAR T-cell persistence, and expand access to higher-risk patient populations. However, spatial control alone is not enough to minimize toxicity. Toxicity also changes over time, requiring a focus on temporal control.

The applicability of this spatiotemporal framework to solid tumors warrants particular consideration. Compared with hematologic malignancies, the toxicity landscape of CAR-T therapies in solid tumors remains less well-defined, in part because clinical efficacy has generally been more limited and clinical experience remains comparatively early. − Although several solid tumor studies have reported lower rates of severe CRS and ICANS relative to hematologic CAR-T therapies, direct comparisons remain challenging given differences in CAR-T expansion, persistence, tumor burden, patient selection, and toxicity reporting practices. − Importantly, systemic toxicities, including CRS, ICANS, and IEC-HS, have also been observed in solid tumor settings. − In addition, solid tumors may present distinct risks related to compartmental inflammation and on-target/off-tumor injury in normal tissues expressing shared antigens. −

As a result, spatial considerations may become particularly important in solid tumor settings, where locoregional delivery strategies, including intratumoral, intracavitary, intraperitoneal, intrapleural, and intracranial administration, are increasingly being explored to maximize local efficacy while limiting systemic inflammatory toxicity. − Temporal dosing strategies may also require adaptation, including repeated low-dose administration, transient CAR expression, or controllable activation systems designed to limit cumulative tissue injury in sensitive organs. Furthermore, stromal barriers, suppressive immune microenvironments, and heterogeneous antigen distribution may alter CAR-T kinetics, persistence, and toxicity profiles, necessitating tumor-specific optimization of both spatial and temporal control strategies within the proposed spatiotemporal framework. − For example, intracranial or intrapleural CAR T-cell administration may increase local tumor exposure while limiting systemic cytokine dissemination and reducing systemic inflammatory toxicity. , Such approaches may additionally be paired with transient CAR expression systems, repeated low-dose administration schedules, or inducible activation platforms to reduce cumulative on-target/off-tumor tissue injury within sensitive organs.

Temporal Control Strategies

CAR T-cell–associated toxicities develop along a predictable but dynamic time course, emerging as waves of immune activation rather than a single inflammatory event. , Early T-cell activation and cytokine release initiate a myeloid-driven inflammatory amplification loop that produces systemic cytokine elevations characteristic of CRS, and as these cascades intensify they can become self-sustaining and progress toward organ-specific injury, particularly within the central nervous system. Despite this structured temporal progression, clinical management remains largely reactive, typically initiated after symptoms appear, highlighting the need for temporally aligned strategies that intervene earlier in the inflammatory cascade and adapt treatment to the stage of immune activation.

Advances in drug delivery further enable temporal precision through programmable and responsive systems. Biomaterials capable of releasing therapeutic payloads in response to cytokine signals such as IL-6 or IFN-γ enable the potential for feedback-controlled immunomodulation. Similarly, sustained-release nanoparticles can maintain therapeutic concentrations during anticipated CRS windows, while stimulus-responsive systems can activate only when inflammatory thresholds are reached [41]. These approaches aim to align drug delivery kinetics with the dynamic progression of immune activation. It is important to note that temporal control must be integrated with CAR T-cell treatment kinetics. Toxicity severity is closely linked to CAR T-cell expansion magnitude, tumor burden, and cytokine peak levels, while lymphodepletion regimens such as cyclophosphamide and fludarabine influence expansion dynamics. As a result, temporal intervention strategies may need to scale with CAR T-cell dose, disease burden, and patient-specific immune kinetics. Incorporating pharmacokinetic (PK) modeling into CAR T-cell workflows could enable predictive timing of interventions, allowing clinicians to anticipate toxicity before it develops.

Overall, CAR-T toxicity emerges through temporally distinct immune activation phases, but current clinical strategies often intervene only after inflammatory amplification has occurred. Aligning therapeutic interventions with the timing of immune activation represents an opportunity to improve outcomes. Drug delivery technologies capable of programmable or responsive release further enhance this approach, enabling proactive rather than reactive control of inflammatory cascades. These findings suggest that the timing of immune modulation may be as important as target selection in reducing CAR-T-associated toxicities.

Conclusion

In conclusion, CAR T-cell-associated toxicity is not merely an unavoidable consequence of powerful cancer immunotherapies, but rather the product of unrestrained immune activation propagating across distinct temporal phases and anatomical compartments. The major clinical syndromes observed following CAR T infusionincluding cytokine release syndrome (CRS), immune effector cell–associated neurotoxicity syndrome (ICANS), and immune effector cell–associated hemophagocytic lymphohistiocytosis-like syndrome (IEC-HS)do not represent a uniform systemic failure. Instead, they arise from structured immune dynamics driven by the nonlinear expansion of living cellular therapies and the cascading interactions between engineered T cells, host vasculature, and endogenous myeloid populations.

Current clinical management strategies remain poorly aligned with this biological complexity. In practice, toxicity mitigation relies largely on reactive administration of systemic immunosuppressive agents such as corticosteroids or broad cytokine blockade after clinical symptoms have already developed. While these interventions can dampen inflammatory responses, their lack of spatial and temporal specificity introduces substantial clinical trade-offs. Systemic immunosuppression can impair CAR-T cell expansion and persistence, delay immune recovery, and increase susceptibility to infection, potentially compromising long-term therapeutic benefit.

Addressing this toxicity bottleneck therefore requires a shift from reactive suppression toward proactive design. Drug delivery technologies offer a potentially useful framework for achieving this transition by enabling precision control of immune modulation across three key dimensions (Figure ).

• Temporal control aligns therapeutic intervention with the kinetic progression of immune activation. By intercepting inflammatory signaling during early activation phases, before cytokine cascades undergo irreversible amplification, it may be possible to prevent severe toxicities while maintaining productive CART expansion.

• Dosing control reframes CAR-T therapy as a dynamic kinetic system rather than a fixed input quantity. Because CAR-T cells proliferate and interact with host immune networks in vivo, shaping expansion rate, peak magnitude, and overall inflammatory exposure becomes a critical determinant of toxicity.

• Spatial control focuses on restricting immunomodulatory interventions to the anatomical sites where pathology emerges. Targeting inflamed vascular endothelium, the central nervous system, or myeloid-rich reticuloendothelial organs such as the liver and spleen allows localized suppression of inflammatory damage while preserving systemic antitumor immunity.

4.

Decision flowchart illustrating a proposed framework for escalation from temporal and dosing-based toxicity mitigation strategies to spatially targeted approaches.

Looking forward, translating this framework into clinical practice will require integrating drug delivery principles directly into CAR-T engineering and treatment workflows. Biomarker-guided or adaptive interventions that respond to early cytokine trajectories could enable personalized timing of immunomodulation. Similarly, cytokine-responsive or programmable delivery systemsincluding nanoparticle or antibody-based vectors targeting activated endothelial markers such as ICAM-1 or VCAM-1may allow localized suppression of inflammatory pathways without systemic immune inhibition. These approaches illustrate how spatiotemporal control of immune signaling can be embedded directly into next-generation immunotherapy design.

Importantly, the proposed spatiotemporal framework may become increasingly relevant as next-generation cellular immunotherapy platforms continue to evolve. Emerging modalities including allogeneic CAR-T therapies, in vivo CAR engineering approaches, and mRNA-based CAR systems may fundamentally alter the kinetics, persistence, biodistribution, and toxicity profiles of engineered immune cells. ,,, For example, allogeneic CAR-T products may enable more standardized dosing, earlier intervention windows, and repeat administration strategies because of their off-the-shelf availability, but may also introduce distinct risks related to host-versus-graft and graft-versus-host immune interactions. In vivo CAR engineering strategies, including nanoparticle- or viral vector-mediated immune cell programming, may further integrate delivery technology directly into CAR generation itself, creating opportunities for tissue-selective targeting, transient expression, or anatomically restricted CAR induction. Similarly, mRNA-based CAR approaches may offer intrinsic temporal control through transient CAR expression, potentially reducing prolonged inflammatory activation and cumulative on-target/off-tumor toxicity. These emerging technologies reinforce the broader concept that future immunotherapy design will likely depend not only on antigen specificity, but also on precise control over when, where, and for how long engineered immune activity occurs.

Ultimately, the next generation of safer immunotherapies will emerge not from broadly suppressing immune activity, but from engineering precise control over immune dynamics. Drug delivery strategies capable of localizing, timing, and tuning immunomodulation offer a rational path toward expanding the therapeutic window of CAR-T therapy and enabling its broader, safer application across diverse cancers.

The authors declare no competing financial interest.

Published as part of ACS Bio & Med Chem Au special issue “2026 Rising Stars in Biological, Medicinal, and Pharmaceutical Chemistry”.