NK-cell therapy, conventional T-cell therapy, and CAR-T therapy are related forms of adoptive immunotherapy, but they are not interchangeable. They differ in how they recognize abnormal cells, how strongly they depend on human leukocyte antigen (HLA), whether the product can come from a healthy donor, how long the cells persist, and how mature the clinical evidence is. “NK therapy” may mean unmodified, activated, memory-like, antibody-supported, or CAR-engineered NK cells; “T-cell therapy” can mean tumor-infiltrating lymphocytes, virus-specific T cells, T-cell-receptor (TCR)-engineered cells, or CAR-T cells. The exact product matters more than the broad label.

For patients, the central distinction is practical as well as biological. Approved autologous CAR-T products can produce durable remissions in selected blood cancers, but require individualized manufacturing and specialist toxicity management. Donor-derived NK platforms may support banked, repeatable production with less alloreactivity, yet CAR-NK remains investigational and has a far smaller evidence base. This review explains the trade-offs without treating an early platform as a proven replacement.

What are NK cells, T cells, and CAR-engineered cells?

NK cells are innate lymphocytes that continuously integrate activating and inhibitory signals. They can attack stressed, infected, or transformed cells without first recognizing one specific peptide through a clonally rearranged receptor. Inhibitory killer-cell immunoglobulin-like receptors (KIRs) and NKG2A sample HLA class I molecules; loss of normal HLA restraint can create a “missing-self” signal. Activating receptors such as NKG2D and natural cytotoxicity receptors sense stress ligands, while CD16 enables antibody-dependent cellular cytotoxicity. The final decision reflects a balance rather than a single lock-and-key event [1].

Conventional alpha-beta T cells use a T-cell receptor to recognize peptide displayed by HLA. CD8 T cells usually recognize peptide–HLA class I, while CD4 T cells generally recognize peptide–HLA class II. This gives exquisite specificity and supports immune memory, but the target is HLA-restricted. A TCR-engineered therapy therefore works only when both the target peptide and the required HLA allele are present.

A chimeric antigen receptor changes the recognition input. A CAR typically combines an antibody-derived binding domain with transmembrane and intracellular signaling modules. CAR-T and CAR-NK cells can bind a surface antigen directly without peptide presentation by HLA. The host cell still matters: a CAR-T cell retains T-cell biology, while a CAR-NK cell retains innate receptors, CD16 activity, and NK-specific signaling constraints [4].

Scientific comparison of NK missing-self recognition, T-cell peptide-HLA recognition, and CAR surface-antigen recognition

Recognition and HLA: the key mechanistic difference

Natural TCR recognition is HLA-dependent; CAR recognition is largely HLA-independent; native NK recognition is HLA-informed but not peptide-HLA restricted. This wording matters. NK cells are not blind to HLA: inhibitory NK receptors use HLA as a safety signal. However, they do not require a particular tumor peptide to be presented by a particular HLA molecule in the same way a conventional TCR does.

CARs bypass peptide presentation by binding cell-surface molecules. That can overcome one route of tumor escape—loss of antigen processing or HLA presentation—but it introduces another: the tumor may lose or reduce the surface antigen targeted by the CAR. CAR-NK cells may preserve additional, CAR-independent killing through native activating receptors and CD16, potentially offering a second route against heterogeneous targets. Whether that theoretical redundancy improves durable outcomes remains product- and disease-specific [6].

Do not confuse three concepts: HLA restriction determines what a receptor can see; HLA matching addresses donor–recipient compatibility; and host-versus-graft rejection determines how long allogeneic cells survive. An HLA-unrestricted CAR target does not automatically make every donor T-cell product safe, and low GVHD potential does not guarantee persistence.

Side-by-side comparison

FeatureNK-cell therapyConventional / engineered T-cell therapyCAR-T therapy
Native recognitionBalance of activating and inhibitory receptors; missing-self, induced-self, and antibody-dependent killingTCR recognizes peptide presented by a compatible HLA moleculeCAR binds a chosen surface antigen; native TCR remains unless removed
HLA relationshipNot peptide-HLA restricted, although HLA–KIR/NKG2A signals regulate activityUsually HLA-restrictedCAR recognition is HLA-independent; donor TCR can still cause alloreactivity
Typical sourcing directionHealthy-donor blood, cord blood, cell lines, or stem-cell-derived banks; autologous approaches also existAutologous, matched donor, tumor-derived, or virus-specific depending on productCurrently approved products are autologous; allogeneic CAR-T remains an active development field
PersistenceOften shorter and more variable; cytokine support, memory-like programs, or repeat dosing may be neededMemory T cells can persist long termCan expand substantially and persist for months or years, varying by construct and patient
CRS / ICANSGenerally lower in early NK/CAR-NK experience, but not impossibleDepends strongly on product and activationEstablished, potentially severe risks requiring trained monitoring
GVHDLow intrinsic risk with purified allogeneic NK products; contaminating T cells and manufacturing still matterRelevant for conventional allogeneic alpha-beta T cellsNot expected from autologous products; a central engineering concern for donor-derived CAR-T
Evidence maturityUnmodified NK has decades of study; engineered and CAR-NK evidence remains mainly early-phaseVaries widely by platform and indicationMultiple regulatory approvals and pivotal trials in defined hematologic malignancies

Autologous versus allogeneic logistics

Autologous CAR-T is a personalized supply chain. A patient undergoes leukapheresis; cells are shipped, activated, genetically modified, expanded, tested, released, returned, and infused after lymphodepleting chemotherapy. Disease can progress or clinical status can change while manufacturing is under way. Starting material from heavily treated patients may be variable, and a manufacturing failure can be clinically consequential. The advantage is immunologic ownership: the product is the patient’s, so donor T cells cannot attack host tissues.

Allogeneic production aims to shift from one patient–one batch to one donor or master bank–many doses. Healthy-donor NK cells, cord-blood NK cells, induced-pluripotent-stem-cell-derived NK cells, and NK cell lines can be expanded into standardized lots. This could improve scheduling and distribute fixed manufacturing work across more doses. But “off the shelf” is an objective, not proof of interchangeability. Cryopreservation recovery, potency after thaw, feeder-cell removal, batch consistency, transport, host rejection, and release testing remain material challenges [14].

Allogeneic alpha-beta T cells face a different barrier: their native TCR can recognize recipient tissue and cause GVHD. Gene-edited universal CAR-T platforms attempt to disrupt the TCR and sometimes modify HLA-related pathways, adding engineering and regulatory complexity. Purified NK cells have much lower intrinsic GVHD potential, which makes donor-derived NK development attractive, but recipient immunity can still clear them.

Persistence: advantage, liability, and design choice

Persistence is not simply “more is better.” Long-lived CAR-T cells can provide continuing immune surveillance and contribute to durable disease control. The same persistence can prolong on-target effects such as B-cell aplasia when normal cells share the target. T-cell differentiation state, exhaustion, CAR tonic signaling, antigen burden, costimulatory domain, and lymphodepletion all affect expansion and survival [11].

NK products have historically shown more limited expansion and persistence, which may reduce prolonged toxicity but can also weaken durability. Strategies include IL-15 support, membrane-bound cytokines, memory-like programming, metabolic engineering, evasion of host rejection, and planned repeat dosing. Preclinical cord-blood work demonstrated that IL-15 can extend CAR-NK activity [10]. In the first published CD19 CAR-NK cohort, cells were detectable at low levels for prolonged follow-up, showing that “NK cells are always short-lived” is too simplistic [2]. Yet detection is not synonymous with durable tumor control.

CRS, ICANS, GVHD, and other risks

CAR-T has a well-characterized toxicity framework because it has been used broadly enough for uncommon and severe events to emerge. Cytokine release syndrome (CRS) can include fever, hypotension, hypoxia, and organ dysfunction. Immune effector cell-associated neurotoxicity syndrome (ICANS) can involve confusion, language disturbance, impaired consciousness, seizures, or cerebral edema. Cytopenias, infection, hypogammaglobulinemia for B-cell targets, tumor lysis, and prolonged immune effects also require structured follow-up. Severity varies by product, disease burden, and patient factors; modern grading and treatment pathways have improved management [12].

Early CAR-NK studies report a lower observed burden of CRS, ICANS, and GVHD, but “lower observed” is not “zero risk.” The landmark cord-blood study treated a small, selected cohort without these major toxicities [2]. Its expanded phase 1/2 analysis remained encouraging but was still single-center and non-randomized [3]. Other NK platforms, targets, cytokine armoring systems, conditioning regimens, and larger populations may produce different safety findings. Lymphodepleting chemotherapy itself can cause cytopenias and infections; infused cells can cause on-target/off-tumor injury; and engineered cytokine support can introduce new risks.

GVHD comparisons also need precision. Autologous CAR-T does not cause donor-versus-host disease because there is no donor. GVHD is primarily a concern for allogeneic alpha-beta T-cell products, whereas purified donor NK cells rarely initiate the classic TCR-driven process. This biological advantage does not remove the need to control T-cell contamination and product identity.

Evidence maturity and regulatory approvals

CAR-T is the mature platform in this comparison. Regulators in the United States and other jurisdictions have approved multiple autologous CAR-T products for specific relapsed or refractory B-cell leukemias, lymphomas, and multiple myeloma settings. Those approvals are indication-, target-, age-, and prior-treatment-specific; “CAR-T is approved” never means every CAR design or every cancer is approved. Pivotal and post-marketing evidence now support clinical guidelines, manufacturing standards, and toxicity programs.

CAR-NK therapy is not currently an approved standard therapy. Published clinical evidence includes encouraging early studies, especially in CD19-positive B-cell malignancies, but it remains much smaller and less mature. The initial 11-patient study reported activity without CRS, neurotoxicity, or GVHD [2]; the later 37-patient analysis examined response determinants and longer outcomes [3]. These findings justify controlled development—not equivalence claims. Solid-tumor evidence for both CAR-T and CAR-NK is generally less established than the evidence for approved hematologic indications because trafficking, antigen heterogeneity, and the immunosuppressive tumor microenvironment remain formidable barriers [13].

Unmodified or cytokine-activated NK infusions should also not be conflated with CAR-NK. A clinic offering “NK-cell therapy” may be using a product with different sourcing, expansion, targeting, potency, and evidence. Patients should ask for the exact product name, regulatory status, protocol, indication, trial identifier, cell source, genetic modification, and adverse-event plan.

How costs differ—without misleading sticker prices

Cost follows the care pathway, not only the cell bag. Autologous CAR-T costs reflect patient-specific collection, vector-based engineering, release testing, logistics, lymphodepletion, infusion, specialist monitoring, and possible inpatient or intensive management of toxicity. Coverage and total out-of-pocket burden vary sharply by country, indication, payer, and treatment center. A list price cannot describe the complete episode.

Allogeneic NK and CAR-NK manufacturing may eventually reduce per-dose production costs by spreading one validated batch across multiple recipients and shortening scheduling. That economic promise has not yet been proven across commercial-scale approved CAR-NK products, because there are no approved CAR-NK products to price against routine CAR-T care. Repeat dosing, enhanced cytokine support, additional conditioning, cryogenic distribution, potency testing, and treatment in a clinical trial can offset apparent manufacturing savings. A responsible comparison therefore asks what is included, what is investigational, who pays for complications, and what happens if a batch or response fails—not which brochure displays the lower number.

Which platform may fit which clinical question?

For an eligible patient with a disease and treatment line covered by an approved CAR-T indication, CAR-T has the strongest evidence of these options. Suitability still depends on organ function, infection status, disease tempo, prior therapies, target expression, caregiver and proximity requirements, and the center’s ability to manage toxicities. Bridging treatment may be needed during manufacture.

NK-cell or CAR-NK studies may be rational when the patient meets a well-designed clinical-trial protocol, particularly where an allogeneic, rapidly available, or repeat-dose strategy addresses a defined limitation. Trial participation should include transparent consent about uncertainty, alternatives, genetic engineering, conditioning, monitoring, and long-term follow-up. Conventional T-cell products may be relevant where intracellular antigens can be targeted through a defined peptide-HLA complex, or where tumor-infiltrating lymphocyte approaches have indication-specific evidence.

What remains uncertain?

No head-to-head randomized trial establishes that CAR-NK is safer, more effective, or more cost-effective than approved CAR-T across comparable patients. Cross-trial response percentages are particularly misleading because disease mix, tumor burden, prior therapy, endpoint timing, conditioning, dose, and follow-up differ. Small early-phase cohorts are designed primarily to establish feasibility and safety, not definitive comparative effectiveness.

Open questions include optimal NK source, freeze–thaw resilience, manufacturing consistency, dose and repeat schedule, host rejection, persistence needed for each cancer, cytokine-armoring safety, antigen escape, and performance in solid tumors [15]. CAR-T development likewise continues toward faster manufacture, allogeneic products, multi-antigen targeting, controllable activation, and improved solid-tumor trafficking [5]. The likely future is not a single winner: distinct immune-cell platforms may be selected or combined according to disease biology.

Frequently asked questions

Is NK-cell therapy the same as CAR-NK therapy?

No. NK therapy may use activated or expanded NK cells without a CAR. CAR-NK adds a genetically encoded receptor for a chosen surface antigen while retaining native NK-cell functions.

Does CAR-T require HLA matching?

CAR binding itself does not require peptide-HLA matching. Approved autologous CAR-T products use the patient’s own cells. Donor-derived CAR-T needs strategies to control native TCR-mediated alloreactivity and host rejection.

Can donor NK cells cause GVHD?

Purified donor NK products have low intrinsic GVHD risk because NK cells do not use an alpha-beta TCR to attack recipient HLA. Manufacturing must still control contaminating T cells, and allogeneic cells can be rejected by the host.

Is CAR-NK safer than CAR-T?

Early CAR-NK cohorts have shown fewer severe CRS, ICANS, and GVHD events than commonly observed with CAR-T, but datasets are much smaller and products differ. It is premature to claim universal superiority.

Is CAR-NK approved?

No. CAR-NK remains investigational as of this article’s publication date. CAR-T has multiple approvals for defined blood-cancer indications.

Limitations and medical disclaimer

This is an educational comparison, not a treatment recommendation. The evidence spans heterogeneous products and cannot predict an individual outcome. Regulatory status changes, so patients and clinicians should verify the current approval label or trial registry with the relevant national authority. Never substitute a commercially described “immune-cell program” for an oncologist’s disease-specific plan or an authorized clinical trial.

References

  1. Kärre K. NK cells, MHC class I molecules and the missing self. Scandinavian Journal of Immunology. 2002;55:221–228. doi:10.1046/j.1365-3083.2002.01053.x ↩
  2. Liu E, Marín D, Banerjee P, et al. Use of CAR-transduced natural killer cells in CD19-positive lymphoid tumors. New England Journal of Medicine. 2020;382:545–553. doi:10.1056/NEJMoa1910607 ↩
  3. Marín D, Li Y, Başar R, et al. Safety, efficacy and determinants of response of allogeneic CD19-specific CAR-NK cells in CD19+ B cell tumors. Nature Medicine. 2024;30:772–784. doi:10.1038/s41591-023-02785-8 ↩
  4. Dagher OK, Posey AD Jr. Forks in the road for CAR T and CAR NK cell cancer therapies. Nature Immunology. 2023;24:1994–2007. doi:10.1038/s41590-023-01659-y ↩
  5. Sterner RC, Sterner RM. CAR-T cell therapy: current limitations and potential strategies. Blood Cancer Journal. 2021;11:69. doi:10.1038/s41408-021-00459-7 ↩
  6. Gong Y, Klein Wolterink RGJ, Wang J, Bos GMJ, Germeraad WTV. CAR-NK cell design and engineering for cancer therapy. Journal of Hematology & Oncology. 2021;14:73. doi:10.1186/s13045-021-01083-5 ↩
  7. Ebrahimiyan H, et al. Novel insights in CAR-NK cells beyond CAR-T cell technology; promising advantages. International Immunopharmacology. 2022;106:108587. doi:10.1016/j.intimp.2022.108587 ↩
  8. Rafei H, Daher M, Rezvani K. CAR natural killer-cell therapy: leveraging the power of innate immunity. British Journal of Haematology. 2021;193:216–230. doi:10.1111/bjh.17186 ↩
  9. Daher M, Rezvani K. Outlook for new CAR-based therapies with a focus on CAR NK cells. Cancer Discovery. 2021;11:45–58. doi:10.1158/2159-8290.CD-20-0556 ↩
  10. Liu E, Tong Y, Dotti G, et al. Cord blood NK cells engineered to express IL-15 and a CD19-targeted CAR show long-term persistence and potent antitumor activity. Leukemia. 2018;32:520–531. doi:10.1038/leu.2017.226 ↩
  11. Zhao Z, Condomines M, van der Stegen SJC, et al. Structural design of engineered costimulation determines tumor rejection kinetics and persistence of CAR T cells. Cancer Cell. 2015;28:415–428. doi:10.1016/j.ccell.2015.09.004 ↩
  12. Davila ML, Rivière I, Wang X, et al. Efficacy and toxicity management of 19-28z CAR T cell therapy in B-cell acute lymphoblastic leukemia. Science Translational Medicine. 2014;6:224ra25. doi:10.1126/scitranslmed.3008226 ↩
  13. Peng L, Sferruzza G, Yang L, Zhou L, Chen S. CAR-T and CAR-NK as cellular cancer immunotherapy for solid tumors. Cellular & Molecular Immunology. 2024;21:1089–1108. doi:10.1038/s41423-024-01207-0 ↩
  14. Heipertz EL, Zynda ER, Stav-Noraas TE, et al. Current perspectives on off-the-shelf allogeneic NK and CAR-NK cell therapies. Frontiers in Immunology. 2021;12:732135. doi:10.3389/fimmu.2021.732135 ↩
  15. Li T, Niu M, Zhang W, et al. CAR-NK cells for cancer immunotherapy: recent advances and future directions. Frontiers in Immunology. 2024;15:1361194. doi:10.3389/fimmu.2024.1361194 ↩