A systematic review of the last two decades of NK cell-based clinical trials: state of the art of AML therapy

Abstract

Background

Acute myeloid leukemia (AML) remains a significant challenge in hematologic oncology, with high relapse rates and limited treatment options. Natural killer (NK) cells, as key effectors of innate immunity, have shown strong anti-leukemic activity, making them promising candidates for immunotherapy. Despite increasing clinical interest, a comprehensive evaluation of NK cell-based therapies in AML is still needed.

Methods

This systematic review follows the PRISMA guidelines to analyze clinical trials evaluating NK cell therapy in AML, either as a standalone treatment or in combination with hematopoietic stem cell transplantation (HSCT). A literature search across five major databases identified relevant studies, with data extraction focusing on NK cell sources, isolation and expansion strategies, clinical efficacy, and safety outcomes.

Results

A total of 48 clinical trials were identified, including 27 trials specific to AML and 21 trials involving AML along with other hematologic malignancies. Peripheral blood (PB)-derived NK cells were the main source (82%), with purification methods mainly using CliniMACS-based CD3 depletion and CD56 selection. Short-term activation (≤ 24 h) and long-term expansion (> 7 days) were employed in 36% of studies each. In non-HSCT transplant settings, NK cell therapy achieved a complete remission (CR) rate of 37.1% and an event-free survival (EFS) of 71.3%, while post-HSCT overall survival (OS) reached 39.5%. Notably, graft-versus-host disease (GVHD) incidence stayed low, highlighting the favorable safety profile of NK cell therapy.

Conclusion

NK cell-based therapy represents a promising and well-tolerated immunotherapeutic approach for AML. However, optimizing NK cell expansion, persistence, and clinical applications requires further investigation through large-scale, controlled trials.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12885-025-14837-y

Introduction

Allogeneic hematopoietic stem cell transplantation (allo-HSCT) is an effective therapy for patients with acute myeloid leukemia (AML) and myelodysplastic syndromes (MDS), where the curative effect is mediated by the graft-versus-leukemia (GVL) effect exerted by donor T and natural killer (NK) cells [1, 2]. Therefore, allo-HSCT is considered the prototype of cellular immunotherapy [2]. Unfortunately, the overall survival (OS) of patients receiving HSCT has not improved because an increase in relapse-related mortality has counterbalanced the reduction in non-relapse mortality [3]. Relapse remains a significant cause of long-term mortality in transplant recipients, accounting for 30–40% of post-transplant deaths, depending on the donor type and disease status at the time of transplantation [4]. Furthermore, the majority of elderly patients and those with comorbidities are not suitable candidates for allo-HSCT, thus limiting their treatment options and long-term survival chances [5, 6]. Therefore, developing alternative immunotherapy approaches that can potentially be applied to all ages with minimal toxicity and maximum clinical efficacy is a high-priority research area [7].

NK cells are the first lymphocytes to reconstitute following allo-HSCT, and their robust recovery within the first two months post-transplant is correlated with reduced disease relapse and improved survival outcomes. In contrast to T cells, NK cells exhibit a regulatory function in graft-versus-host disease (GVHD) while preserving their GVL effect in the allo-HSCT setting [8, 9]. These findings highlight the therapeutic potential of NK cell-based immunotherapy in AML. The growing understanding of NK cell biology and advancements in isolation, expansion, engineering, differentiation from stem cells, and the discovery of memory phenotypes in NK cells have led to a widespread exploration of NK cell-based therapies in AML, both in the context of transplantation and non-transplant settings [10, 11]. Despite the increasing interest in NK cell-based therapies for AML, a comprehensive systematic review of clinical trials in this area is currently lacking.

This study aimed to systematically review clinical trials utilizing NK cell-based treatments for AML in pediatric and adult populations, focusing on the methods employed for the isolation, activation, or ex vivo expansion of NK cells, and assessing clinical outcomes and adverse events. Due to the heterogeneity of the articles and the small number of included patients, it was impossible to conduct a meta-analysis. Therefore, through a systematic review, we aimed to conclude the advantages and disadvantages of NK cell therapy in the treatment of AML.

Methods

Search strategy and selection strategy

This systematic review was conducted by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Two authors conducted a comprehensive literature search to identify all relevant studies evaluating the safety or efficacy of NK-cell therapy for AML/MDS. Our literature search included five major electronic databases: PubMed, Web of Science, Scopus, Embase, and Cochrane Library, utilizing MeSH terms and keywords such as “natural killer cell” and “acute myeloid leukemia” (Supplementary file 1). We also searched the gray literature, including references of the included studies, related reviews and systematic reviews, ClinicalTrials.gov, and ASH conference abstracts, to identify additional relevant studies. The search encompassed studies from the inception of these databases to April 25, 2024. No filters or language limits were applied to the search. All search results were imported into the EndNote (version X9.0) reference manager, and duplicates were identified and manually removed using EndNote. The most recently updated results of each clinical trial were analyzed from published articles or ASH conference proceedings.

Data extraction

Two authors independently screened titles and abstracts (n = 8,892) to exclude irrelevant articles. Disagreements were resolved through mutual consensus and consultation with a third author. The full texts and supplementary materials of the remaining articles (n = 219) were then assessed for eligibility for data extraction by two authors based on predetermined criteria. The inclusion criteria were as follows: (1) original studies (clinical trials, case-control, retrospective, and prospective cohort), (2) both adult and pediatric patients, and (3) studies that investigated the impact of NK cell infusions in combination with HSCT or without HSCT. We excluded articles with insufficient data (where efficacy or safety after NK-cell immunotherapy was not addressed), reviewed articles, pre-clinical studies, protocols, case reports, and published articles that were not in English if translations were unavailable.

Five authors completed the data extraction independently. Conflicts were resolved by discussion with senior reviewers. Data were extracted using a predetermined form that included study characteristics such as author, year, study identifier, study`s phase, country, and sponsor (Table 1), NK cell source, NK cell donor, NK cell culture protocol, and product characteristics (Tables 2 and 3), number of patients, age, gender, disease, HSCT donor (when applicable), NK cells dose, number of infusions, conditioning regimen, outcome data including complete remission (CR), OS, partial response (PR), stable disease (SD), morphologic leukemia-free state (MLFS), minimal residual disease (MRD) status, relapse, safety data including cytokine release syndrome (CRS), GVHD, infection and follow-up time (Table 4).

Table 1.

Overview of clinical trials for AML

#	Author	year	Trial identifiers	Study phase	Country	Sponsor	Publication Type
1	Ciurea et al. [13]	2024	NCT01787474	I	USA	M.D. Anderson Cancer Center	FT
2	Durán et al. [14]	2023	NCT05304754	I/II	Spain	Instituto de Investigación Hospital Universitario La Paz	Abs
3	Ahmadvand et al. [15]	2023	IRCT20200621047859N	I	Iran	NM	FT
4	Lee et al. [16]	2023	NCT02477787	II	Korea, Republic of	Asan Medical Center	FT
5	Kulkarni et al. [17]	2023	Clinical Trials Registry-India (CTRI/2019/02/017505)	II	India	Department of Biotechnology, India	FT
6	Sauter et al. [18]	2023	NCT04623944	I	USA	Nkarta Inc.	Abs
7	Rubio-Azpeitia et al. [19]	2022	NCT03669172	I/II	Spain	Martín, José Luis Díez, M.D.	FT
8	Bednarski et al. [20]	2022	NCT03068819	I/II	USA	Washington University School of Medicine	FT
9	Berrien-Elliott et al. [21]	2022	NCT02782546	II	USA	Washington University School of Medicine	FT
10	Ciurea et al. [22]	2022	NCT01904136	I/II	USA	M.D. Anderson Cancer Center	FT
11	Berrien-Elliott et al. [23]	2022	NCT03050216	II	USA	Masonic Cancer Center, University of Minnesota	FT
12	Shapiro et al. [24]	2022	NCT04024761	I	USA	Dana-Farber Cancer Institute	FT
13	Otegbeye et al. [25]	2022	NM	I	USA	NM	FT
14	Wang et al. [26]	2022	NM	NM	China	NM	Abs
15	Huang et al. [27]	2022	NCT05008575	I	China	Xinqiao Hospital of Chongqing	Abs
16	Heuser et al. [28]	2022	NCT04632316	I/II	Belgium	Glycostem Therapeutics BV	Abs
17	Devillier et al. [29]	2021	NCT01853358	I	France	Institut Paoli-Calmettes	FT
18	Gómez García et al. [30]	2021	NCT02763475	II	Spain	Instituto de Investigación Hospital Universitario La Paz	FT
19	Vasu et al. [31]	2021	NCT04220684	I	USA	Sumithira Vasu	Abs
20	Silla et al. [32]	2021	NCT02809092	I/II	Brazil	Hospital de Clinicas de Porto Alegre	FT
21	Kim et al. [33]	2021	NCT03349502	II	Korea, Republic of	Seoul National University Hospital	Abs
22	Berrien-Elliott et al. [34]	2020	NCT01898793	I/II	USA	Washington University School of Medicine	FT
23	Mani et al. [35]	2020	NCT02316964	I	USA	Sumithira Vasu	FT
24	Zhao et al. [36]	2020	NM	NM	China	NM	FT
25	Cooley et al. [37]	2019	NCT02781467	I	USA	Celularity Incorporated	Abs
26	Cooley et al. [38]	2019	NCT01385423/NCT02395822	I	USA	Masonic Cancer Center, University of Minnesota	FT
27	Nguyen et al. [39]	2019	NCT00703820	II	USA	St. Jude Children’s Research Hospital	FT
28	Wang et al. [40]	2019	NM	NM	China	NM	FT
29	Fehniger et al. [41]	2018	NCT01520558	I/II	USA	Coronado Biosciences, Inc.	FT
30	Vela et al. [42]	2018	NCT01944982/NCT02074657	I/II	Spain	Hospital Infantil Universitario Niño Jesús, Madrid, Spain	FT
31	Björklund et al. [43]	2018	EudraCT number 2011-003181-32	I/II	Sweden	NM	FT
32	Boyiadzis et al. [44]	2017	NCT00900809	I	USA	ImmunityBio, Inc.	FT
33	Dolstra et al. [45]	2017	EudraCT number 2010-018988-41	I	Netherlands	NM	FT
34	Jaiswal et al. [46]	2017	NM	I	India	NM	FT
35	Curti et al. [47]	2016	NCT00799799	I	Italy	University of Bologna	FT
36	Shaffer et al. [48]	2016	NCT00526292	II	USA	Memorial Sloan Kettering Cancer Center	FT
37	Lee et al. [49]	2016	NCT00402558/NCT01390402	I	USA	M.D. Anderson Cancer Center	FT
38	Rubnitz et al. [50]	2015	NCT00697671/NCT00187096	NM	USA	St. Jude Children’s Research Hospital	FT
39	Kottaridis et al. [7]	2015	EudracT number: 2005 006087-62	I	United Kingdom	University College London	FT
40	Miller et al. [51]	2015	NM	I	USA	NM	Abs
41	Thakar et al. [52]	2015	NM	I/II	USA	NM	Abs
42	Choi et al. [53]	2014	NCT00823524	I/II	Korea, Republic of	Asan Medical Center	FT
43	Bachanova et al. [54]	2014	NCT00274846/NCT01106950	NM	USA	Masonic Cancer Center, University of Minnesota	FT
44	Stern et al. [55]	2013	NCT01386619	I/II	Switzerland Germany	University Hospital, Basel, Switzerland	FT
45	Rubnitz et al. [56]	2010	NCT00187096	NA	USA	St. Jude Children’s Research Hospital	FT
46	Yoon et al. [57]	2010	NCT00569283	I	Korea, Republic of	Asan Medical Center	FT
47	Uharek et al. [58]	2010	NM	I/II	German	NM	Abs
48	Passweg et al. [59]	2004	NM	I	Switzerland	NM	FT

Abbreviations FT Full text, Abs Abstract, NM Not mentioned, NCT National clinical trial

Table 2.

Summary of the clinical development of non-engineering NK cell therapies for AML

Author (year)	NK cell source	NK cell purification protocol	Culture process	Final product characteristics
Ciurea et al. (2024) [13]	PB	CD3+ cells depletion/RosetteSep system	CD3-depleted PBMCs were co-cultured in T-75 flasks with irradiated (100 cGy) K562-mbIL21-41BBL feeder cells (FC21) at a ratio of 1:2 (PBMCs: FC21) for 14 days in RPMI 1640 supplemented with 10% FBS and IL-2.	NM
Durán et al.(2023) [14]	PB	NM	NK cells stimulated ex vivo with IL-15	NM
Ahmadvand et al. (2023) [15]	PB	CD3+ cells depletion/CliniMACS system	CD3-depleted cells were expanded in X-VIVO 10 media with 5% AB serum, irradiated autologous PBMCs (2,500 rad), OKT3, and rhIL-2 for 21 days.	Mean purity: 94.8% Viability: ≥ 80%
Lee et al. (2023) [16]	PB	CD3+ cells depletion/RosetteSep system	CD3-depleted cells were cultured in α-MEM media supplemented with IL-15, IL-21, and hydrocortisone (10–6 M) for 3 weeks.	Cell viability: 70–97% Purity: 71–97% CD56+ cells, < 1% CD3+ cells
Kulkarni et al. (2023) [17]	PB	CD56+ cells selection/CliniMACS device	CD56+ cells activated overnight with autologous plasma, ATO, and rIL-2.	NM
Rubio-Azpeitia et al. (2022) [19]	PB	CD3+ cells depletion followed by CD56+ cell selection/CliniMACS device	Purified NK cells were cultured in RPMI medium supplemented with 10% FBS. The cell suspension was transferred to a cell culture bag supplemented with IL-15 and incubated overnight (12–16 h) at 37 °C and 5% CO2.	Viability: 91.5% (79-96.8%) Purity: 82.2% (76.8–97%)
Bednarski et al. (2022) [20]	PB	CD3+ cell depletion followed by CD56+ selection/CliniMACS device	NK cells were activated for 12 to 16 h with rhIL-12, rhIL-15, and rhIL-18 in good GMP conditions.	Mean purity: 94% ± 1% CD3−CD56+
Berrien-Elliott et al. (2022) [21]	PB	CD3+ cell depletion followed by CD56+ selection/CliniMACS device	Purified NK cells were stimulated with rhIL-12, rhIL-15, and rhIL-18 for 12–16 h under GMP conditions.	Purity: 87.5–100%
Ciurea et al. (2022) [22]	PB	CD3+ cells depletion/RosetteSep system	CD3-depleted PBMCs were co-cultured in T-75 flasks with irradiated (100 cGy) K562-mbIL21-41BBL feeder cells (FC21) at a ratio of 1:2 (PBMCs: FC21) for 14 days in RPMI 1640 supplemented with 10% FBS and IL-2 (50 IU/mL).	Viability: 96% Median expansion: 2995-fold Purity: 99% (Feeder cell contamination < 1%, T-cell contamination < 0.1%)
Berrien-Elliott et al. (2022) [23]	PB	CD3/CD19+ cells depletion/CliniMACS device	Cells product activated overnight with ALT-803.	Purity: 53–78% CD3+CD56+ cells
Shapiro et al. (2022) [24]	PB	NK CD3+ cells depletion followed by CD56+ cells selection/CliniMACS device	NK cells were activated for 12–16 h using IL-12, IL-15, and IL-18.	Purity: ≥70% CD3– CD56+ cells
Otegbeye et al. (2022) [25]	PB	CD3+ cells depletion/CliniMACS device	CD3-depleted cells were cultured in G-REX flasks with an irradiated (90 Gy) feeder cell line (an OCI-AML3 cell line expressed membrane-bound IL-21, namely NKF) for 2 to 3 weeks. The cell culture medium was supplemented with IL-2.	viability: >77.15% Purity: 98.03% NK, 0.05% B, and 1.06% T cells Fold expansion: 152-fold
Wang et al.(2022) [26]	PB	NM	NK cells were expanded with IL-2 and IL-15.	NM
Heuser et al. (2022) [28]	CD34+ HSPCs derived from UCB	CD34+ HSPCs enrichment from UCB/CliniMACS device	Same with Harry Dolstra et al. method.	NM
Devillier et al. (2021) [29]	PB	CD3+ cells depletion followed by CD56+ cells selection/CliniMACS device	In gas-permeable cell culture bags, NK cells were cultured for 7 days in RPMI 1640 medium supplemented with 10% FBS and IL-2.	Viability:>70% Median purity: 95%
Gómez García et al. (2021) [30]	PB	PBMCs were isolated by density gradient	PBMCs were co-cultured with K562-mb15-41BBL feeder cells at a 1:1.5 ratio, supplemented with IL-2 from days 0 to 7, and IL-2 from days 7 to 21 in RPMI 1640 medium containing 10% human AB serum.	Viability: 98.04% Purity: 87.8% NK cells and 2.3% T cells Fold expansion 58.5%-fold
Vasu et al. (2021) [31]	PB	NM	NK cells were cultured with feeder cells expressing mbIL-21 (FC21) and then cryopreserved.	NM
Silla et al. (2021) [32]	PB	CD3+ cells depletion/CliniMACS device	CD3-depleted PBMCs were cultured with irradiated mb IL-21 K562 cells for 8–21 days and then cryopreserved in 10% dimethyl sulfoxide, 20% hydroxyethyl starch plasmin, and 70% albumin.	Viability: ≥ 85 Purity: >80% CD56+; ≤ 1% CD3+; ≤ 1% CD32+; < 5% CD19+
Kim et al. (2021) [33]	PB	CD3+ cell depletion/VarioMACS	CD3-depleted PBMCs were cultured in CellGro SCGM serum-free medium with 1% plasma, OKT3, IL-2, and irradiated autologous PBMCs for 14 days.	Purity: 98.10 ± 0.88% CD3−CD56+ cells Viability: 95.2 ± 1.9%
Berrien-Elliott et al. (2020) [34]	PB	CD3+ cells depletion followed by CD56+ cell selection/CliniMACS device	Purified NK cells were activated with IL-12, IL-15, and IL-18 for 12 h under GMP conditions.	NM
Mani et al. (2020) [35]	PB	NM	NK cells were expanded for 14 days using irrK562 feeder cells displaying mbIL-21, 4-1BBL, and IL-2 in RPMI media.	NM
Zhao et al. (2020) [36]	PB	NM	PBMCs were cultured in the flask and culture bag with irr mbIL-21/41BBL-K562 feeder cells in RPMI 1640 medium supplemented with human IL-2 and autologous serum for two weeks.	NM
Cooley et al. (2019) [37]	Placenta-derived stem cells and UCB stem cells	NM	UCB stem cells or placenta-derived stem cells were cultured in the presence of thrombopoietin, SCF, Flt3 ligand, IL-7, IL-15, and IL-2 for 35 days to generate PNK-007 under the GMP standards.	Viability: ≥ 80% after thawing. Purity: >95% CD56+CD3−
Cooley et al. (2019) [38]	PB	CD3+/CD19+ cells depletion/CliniMACS device	NK cells were activated by overnight incubation with rhIL-15 in X-VIVO 15 supplemented with 10% human AB serum.	NM
Nguyen et al. (2019) [39]	PB	CD3+ cell depletion and CD56+ cell enrichment/CliniMACS device	Purified NK cells were infused without in vitro exposure to cytokines.	NM
Wang et al. (2019) [40]	PB	PBMCs were isolated by density gradient centrifugation	PBMCs were co-cultured with feeder cells in GT-T551-H3 medium containing 10% autologous plasma for 13 days.	NM
Fehniger et al. (2018) [41]	PB	CD56+ cells selection/CliniMACS device	CD56+cells were cultured with CNDO109 (CTV-1 tumor cell line) lysate for 16 h under cGMP conditions, and no cytokines were used in the incubation process. After co-incubation, the lysate was removed by discontinuous density gradient separation, and then the activated NK cells (namely CNDO-109-NK cells) were cryopreserved.	Purity: 71.8% CD56+CD3– cells Viability: 96.4%
Vela et al. (2018) [42]	PB	PBMCs were isolated by density gradient centrifugation	PBMCs were co-cultured with K562-mb15-41BBL feeder cells at a 1:1.5 ratio, supplemented with IL-2 from days 0 to 7 and IL-2 from days 7 to 21 in RPMI 1640 medium containing 10% human AB serum.	Viability: 98.04% Purity: 87.8% NK cells and 2.3% T cells
Björklund et al. (2018) [43]	PB	CD3+/CD19+ cell depletion/CliniMACS device	The NK cells were cultured overnight in X-VIVO15 supplemented with IL-2.	Median Viability: 87.2%
Boyiadzis et al. (2017) [44]	NK-92 cell line	N/A	Cryopreserved cells were thawed and cultured in X-VIVO 10 media supplemented with 5% human AB serum, IL-2, asparagine, L-glutamine, and L-serine for 10 days in T-25 flasks and then transferred into G-Rex10 units, with cell numbers ranging from 5 to 10 × 106. When the cell counts exceeded 50 × 106, the cells were moved to G-Rex100 flasks. The culture was terminated when the cell concentration reached 2 to 8 × 105 cells/mL.	Viability:>70% Purity: >90% CD56+ CD3− cells; <5% CD16+CD3+ cells
Dolstra et al. (2017) [45]	CD34+ HSPCs	CD34+ HSPCs enrichment from UCB units/CliniMACS device	CD34+ cells were cultured in GBGM media supplemented with 10% HS, a low-dose cytokine cocktail including GM-CSF, G-CSF, and IL-6, in VueLifeTM culture bags for 14 days. From days 0–9, the medium was supplemented with a high-dose cytokine cocktail containing SCF, Flt3L, TPO, and IL-7. Between days 10 and 14, TPO was replaced with IL-15. After 14 days, cells were differentiated by replacing Flt3L with IL-2. Around day 14, cells were transferred to the WAVE BioreactorTM System, and after 42 days, NK cell products were harvested and washed.	Purity: 75 ± 12% Median viability: 94%
Jaiswal et al. (2017) [46]	PB	CD56+ cells selection/CliniMACS device	CD56+ cells were infused without in vitro exposure to cytokines.	NM
Curti et al. (2016) [47]	PB	CD3+ cells depletion and CD56+cell selection/CliniMACS device	NK cells were thawed and infused without in vitro exposure to cytokines.	Viability: 95% (range, 92%−98%) Purity: 93.5% (range, 66.4%−99.2%)
Shaffer et al.(2016) [48]	PB	CD3+ cell depletion and CD56+ cell selection/CliniMACS device	NK cells were infused without in vitro exposure to cytokines.	Mean viability: 89.5 Mean purity: 98.9% CD3−CD56+ cells, < 0.1% CD3+ cells
Lee et al. (2016) [49]	PB	CD3+ cells depletion (for all patients) followed by CD56+ cell enrichment (only for three patients)/CliniMACS device	The NK cells were activated overnight in culture media supplemented with IL-2.	Median purity: 0.02% CD3+ cells, 41.77% CD14+ cells, 21.84% CD19+ cells, 14.1% CD56+CD3+
Rubnitz et al. (2015) [50]	PB	CD3+ cells depletion and CD56+cells selection/CliniMACS device	Purified NK cells were infused without in vitro exposure to cytokines.	Median purity: 98.4% CD56+ cell, 0% CD3+CD56− cell and 0.31% CD19+cell
Kottaridis et al. (2015) [7]	PB	CD56+ cells selection/CliniMACS device	The CD56+ cells were cultured overnight in X-VIVO10 with a lysate derived from CTV-1 leukemia cells. After co-incubation, the lysate was removed by discontinuous density gradient separation, and aliquots of NK cells were cryopreserved.	Mean purity: 97.17% CD 56+ cells, >80% CD56+CD3− NK cells
Miller et al. (2015) [51]	NM	NM	NK cells were activated overnight with IL-15.	NM
Thakar et al.(2015) [52]	PB	CD3+ cells depletion and CD56+ cells selection/CliniMACS device	NK cells were administered without prior culturing or expansion.	Mean purity: 92% viability: >95%
Choi et al. (2014) [53]	PB	CD3+ cells depletion/CliniMACS or RosetteSep systems	CD3-depleted cells were activated and expanded over 13 to 20 days with IL-15, human IL-21, and hydrocortisone in a-MEM culture media.	Viability: 71%−85% Purity: CD56+ CD122+ cells > 90%, CD3+CD56+ cells < 3% The median fold expansion: 3.7-fold (8–70)
Bachanova et al. (2014) [54]	PB	CD3+ cells depletion (Cohort 1)., CD3+ cells depletion followed by CD56+ cells selection (Cohort 2), CD3/CD19+ cells depletion (Cohort 3)/CliniMACS system	Isolated cells were cultured overnight in X-VIVO 15 supplemented with IL-2.	Purity: cohort 1 39 ± 9% NK cells, 0.7% T cells; cohort 2 75 ± 6% NK cells,1.3% T cells; cohort 3 54 ± 16% NK cells, 0.3% T cells
Stern et al. (2013) [55]	PB	CD3+ cells depletion and CD56+ cells selection/CliniMACS device	Fresh or thawed NK cells were administered without prior culturing or expansion.	Median viability: 84%
Rubnitz et al. (2010) [56]	PB	CD3+ cells depletion and CD56+ cell selection/CliniMACS device	NK cells were administered without prior culturing or expansion.	NM
Yoon et al. (2010) [57]	CD34+progenitor cells	N/A	CD34+ cells were cultured at a density of 2 × 106 cells/ml in T-75 flasks containing Myelocult complete media supplemented with SCF, FLT3L, IL-7, and hydrocortisone. When the cell count reached around 4 × 106 cells/ml, half of the medium was changed with new complete media containing the same cytokines. After 21 days, cells were harvested, washed, and transferred to Myelocult complete media supplemented with IL-15, IL-21, and hydrocortisone. Half of the media was replaced each week with fresh, complete media. After 21 days, cells were harvested, thoroughly rinsed, and suspended in 50 mL albumin solution before being infused into patients.	Mean viability: 88% Mean purity: CD122+CD56+ cells 64%, CD3+ cells 1.0% (range, 0–2.6%).
Uharek et al.(2010) [58]	PB	CD3+ cells depletion, and then CD56+ cells selection/CliniMACS device	NK cells were infused without in vitro exposure to cytokines.	NM
Passweg et al.(2004) [59]	PB	CD3+ cells depletion, and then CD56+ cells selection/CliniMACS device	After cell enrichment, purified NK cells were immediately infused or cryopreserved without in vitro cytokine exposure.	Mean purity: 97% CD56+CD3−cells

AbbreviationsFCS Fetal calf serum, PBMCs Peripheral blood mononuclear cells, FBS Fetal bovine serum, IL Interleukin, rIL Recombinant IL, GMP Good manufacturing practices, WB Whole blood, HS Human Serum, GM-CSF Granulocyte-macrophage Colony-stimulating factor, G-CSF Granulocyte colony-stimulating factor, SCF Stem cell factor, Flt3L Fms-like tyrosine kinase 3 ligand, TPO Thrombopoietin, NM Not mentioned, NA Not applicable, PB Peripheral blood, HSPCs Hematopoietic stem and progenitor cells, UCB Umbilical cord blood

Table 3.

Summary of the clinical development of engineering NK cell therapies for AML

Author (year)	NK Cell Source	Target	Method of co-targeting	Costimulatory Domain	Culture process
Sauter et al. (2023) [18]	PB	NKG2D Ligands	Gamma-retroviral	OX40.CD3z.IL-15	PB-derived NK cells cultured with stimulatory cells, transduced and expanded by mbIL-15.
Huang et al. (2022) [27]	UCB	CD33	NM	NM	NM

Abbreviations PB Peripheral blood, UCB umbilical cord blood, mbIL-15 Membrane-bound Interleukin-15, NM Not mentioned

Table 4.

Outcome of clinical trials of NK cells in AML patients

Author (Year)	patients, n	Age (years), median (range)	Male gender, n (%)	Disease, n	Disease status, n	HSCT donor/NK cell donor	NK cells infusion dose (cell/Kg) (range)	Number of infusions	Timing of NK cells infusiona	Purpose of NK infusion		Conditioning for HSCT	Efficacy	Follow-up (mo), median (range)	Safety
										Prophylaxis/consolidation	treatment
A) NK cell clinical trials in a transplantation setting
Durán et al.(2023) [14]	18	8.3	9 (50%)	AML:5 T-ALL:5 B-ALL:8	NM	Haplo/KIR/HLA matched or mismatched haplo donor	2 × 107 (cohort 1) 2–5 × 10 7 (cohort2) 5–10 × 10 7 (cohort 3)	NM	+ 7	*		NM	Relapse: 2/18 (11.1%) OS: 72.22% at 1Y DFS: 72.22% at 1Y	NM	aGVHD grade I-II:33.3% (6/18), aGVHD grade III-IV:22.2% (4/18), mild cGVHD: 5.6% (1/18), moderate cGVHD: 11.1% (2/18), severe cGVHD: 5.6% (1/18)
Lee et al. (2023)[16]	36	56 (21–70)	19 (48%)	AML:38 MDS:2	CR1 (5), CR2 (5), primary ref (20), ref/rel (8)	Haplo/same HCT donors	1 × 108 (0.5-1.0) and 1.4 × 108 (0.5-4)	2	+ 13, + 20	*		Bus + Flu + ATG (−7 to −2)	OS: 35% at 30 months PFS: 33% at 30 months CR: 77% (23/30)	33.7	aGVHD grade II-IV: 51%, aGVHD grade III or IV: 35% mod to severe cGVHD: 20%, Severe cGVHD: 8%
Kulkarni et al. (2023) [17]	14	28 (15.75–31.5)	8 (57%)	AML	primary ref (6), ref/rel (8)	Haplo/KIR ligand matched or mismatched related donor	46.16 × 106 (25.06–70.36 × 106)b	1	−12		*	Flu (−6 to −3) + Mel (−2) + TBI (−1)c	OS: 28.6 ± 12.1% at 2 Y EFS: 28.6 ± 12.1% at 2 Y	24	Among eight evaluable patients: aGVHD grade II-III: 25% (2/8), aGVHD grade IV: 12.5% (1/8) Extensive cGVHD: 84% (5 out of 6 evaluable patients)
Rubio-Azpeitia et al. (2022)[19]	5	51 (23–69)	4 (80)	AML	CR (4) MRD+ (1)	Haplo/KIR matched or mismatched related haplo donor	3.5 × 106 (1-4.7 × 106)	2	+ 8, + 15	*		Clo/Mel (RIC), Flu/Bu (MAC), Flu/Bu (RIC)	CR: 80% (4/5) Relapse: 20% (1/5)	15 (6.4–31)	NM
Bednarski et al. (2022) [20]	9	9.4 (1.5–28)	3 (37.5)	AML	Rel	Haplo, MUD, MSD/same HSCT donors	11 × 106 (4–10 × 106)	1	NM		*	Myeloablatived	CR/Cri: 25% (2/8) at d100 PR: 12.5% (1/8) at d100 PD: 50% (4/8) at d100 OS: 42% at 2Y	24	Skin GVHD grade I: 12.5% (1/8)e
Berrien-Elliott et al. (2022) [21]	15	67 (19–73)	11 (73%)	AML	MRD + (7), blast > 5% (8)	Haplo/same HSCT donors	0.5–10 × 106	1	+ 7	*		Cyc (−6 to −5) + Flu (−6 to −2) + TBI (−1)	composite CR: 27 (4/15) at d100 OS: 29% at 1Y EFS: 3.2 M	7 (1.2–27.5)	CRS grade 1 : 6.6% (1/18) aGVHD grade I :27 (4/15), aGVHD grade II:40% (6/15), mild cGVGD: 10% (2/10) Mod cGVGD: 10% (2/10)
Ciurea et al. (2022) [22]	24	46 (18–65)	12 (50%)	AML: 13 CML: 7 MDS:4	CR:10 Rel: 3 First CP for CML: 4	Haplo/same HSCT donors	1 × 105 to 1 × 108	3	−2, + 8, +28	*		Flu (−6 to −3) + mel (−7) + TBI (−3)	OS: 75% at 1Y DFS: 71% at 1Y Relapse: 4% at 1Y	24 (12–51)	aGVHD grade II: 37.5% (9/24), aGVHD grade III/IV: 4% (1/24),
Shapiro et al. (2022) [24]	6	NM	NM	AML:3 MDS:1 CML:1 BPDCs:1	Rel/Ref	Haplo/same HSCT donors	5–10 × 106	1	NM		*	NMf	CR: 50% (3/6) at D28 MLFS: 17% (1/6) at D28	NM	Grade 2 CRS: 16.6% (1/6) No GVHD
Otegbeye et al. (2022) [25]	3	43 (23–73)	1 (33%)	AML: 1 MDS:2	Rel	NM/HLA-disparate unrelated donor	1-2.5 × 107	2	NM		*	NMg	CRi: 33% (1/3) Relapse: 33% (1/3)	15	No GVHD and CRs
Wang et al.(2022) [26]	12	NM	NM	AML	NM	Haplo	1 × 109 (0.8–1.8 × 109)	1	−1	*		NM	OS: 67% at 1Y Relapse: 25% (3/12)	12.5	aGVHD: 25% (3/12) No cGVHD
Devillier et al. (2021) [29]	16	59 (38–68)	8 (50%)	AML: 6 MDS:1 Lym:5 ALL: 1 MM: 2 PMF: 1	Morphological CR	MSD/same HSCT donors	DL1:1 × 106 DL2: 5 × 106 DL3: 5–50 × 106	1	+ 60 to + 120	*		Flu + bus + ATG	Relapse: 25% (4/16) CR: (69%) 11/16	37	Mild cGVHD: 6.25% (1/16) Mod cGVHD: 12.5% (2/16) Severe cGVHD: 6.25% (1/16)
Jaiswal et al. (2017) [46]	10	36 (2–65)	5 (50%)	AML	Ref	Haplo/NK-KIR ligand matched or mismatched	6.7 × 106 (1.7–17.7 × 106)h	1	+ 7	*		Treo (−6 to −4) + Flu (−5 to −2) + TBI (0)	CR: 40% (4/10) Relapse 50% (5/10) OS: 50%	23	cGVHD: 32.3% aGVHD: 0%
Shaffer et al.(2016) [48]	8	19 (1.9–55.9)	NM	AML: 6 MDS: 2	Rel/Ref	Haplo MSD or MUD/matched or mismatched related donor	10.6 × 106 (4.3–22.4 × 106)	1	6.8 months after HSCT (3.9–152)		*	NMi	OS: 50% at 1Y CR: 25% (2/8)	65	No GVHD
Lee et al. (2016) [49]	21	51(2–63)	15(71)	AML: 8 MDS:6 CML:7	active disease: 14 CR/CCR:7	Haplo MSD or MUD/KIR mismatch or mismatched related donors	2.96 × 106(0.02–8.32 × 106)	1	−8	*		Flu (−13 to −10) + Bu (−13 to −10) + ATG (−3 to −1)	Relapse: 60% (12/20) mOS: 233 D mRFS:102D	Approximately 74 M	aGVHD grade I: 15% (3/20), aGVHD grade II: 25% (5/20), aGVHD grade III: 10% (2/20), cGVHD: 30%
Thakar et al.(2015) [52]	40	45 (8–75)	NM	ALL: 11 AML: 9 MDS: 6 HL: 6 MM: 4 NHL: 3 CLL: 1	NM	Haplo donor/NM	2.5 or 5 × 106	1	+ 7	*		Flu + Cyc + TBI	OS: 73% at 1Y PFS: 62% at 1Y Relapse: 31%	18	aGVHD grade II-III: 36%, extensive cGVHD: 16%
Choi et al. (2014) [53]	41	47 (17–75)	23(56%)	AML: 32 ALL: 7 MDS: 1 DLBL: 1	Ref: 38 CR1: 3	Haplo-related donor/same HSCT donors, KIR mismatch, or mismatched	DL1: 0.2 × 108 DL2: 0.5 × 108 DL3: 1 × 108 DL4: ≥ 1 × 108	2	+ 14, + 21	*		Bus (d −7 and − 6) + flu (−7 to −2) + thy (−4 to −1)	Relapse: 46% CR: 68% EFS for AML patients: 31% OS for AML patients: 35%	31.5	aGVHD: 22% cGVHD: 24%
Stern et al. (2013) [55]	16	Center A: 23 (8–32), Center B: 10 (8–23)	NA	AML: 8 (50), ALL: 5 (31), HL: 2 (12.5), SCA: 1 (6)	Active disease: 7 rel:2	Haplo-related donor/same HSCT donors	Center A: 1.2 × 107 (0.5–3.4) Center B: 1.3 × 107 (0.3–3.8)	Center A: 2 Center B: 3	Center A: +40, + 100 Center B: +3, + 40, +100	*		ATG (center A) OKT3 (center B)	Relapse: 7 (43.8) CR: 2 (12.5%) OS: 44 ± 12% at 1Y	69.6 (63.6–81.6)	Agvhd grade ≥ II: 25% (4/16) No cGVHD
Yoon et al. (2010) [57]	14	39.5(23–65)	8 (57%)	AML:11 ALL:1 SMD:2	Ref: 6 CR1: 1 CR2: 3 CR3: 1 CR4: 1 RCMD: 2	Haplo-unrelated donor/same HSCT donors	9.28 × 106 (0.33–24.50)	1	+ 43 to + 50	*		Bus (d −7 and − 6) + flu (−7 to −2) + ATG (−4 to −1) + CSA (−1 to 0)	Relapse: 64% (9/14) CR: 29% (4/14)	18.7	aGVHD grade II: 7% (1/14) Mild cGVHD: 14% (2/14) Mod cGVHD: 7% (1/14) Severe cGVHD: 7% (1/14)
Uharek et al.(2010) [58]	25	NM	NM	AML: 16, ALL: 5 CML: 2 HL: 1 MDS: 1	Rel	Haplo donor/same HSCT donors	9.8 × 106 (1.61–32.2)	1	+ 2	*		TBI + thio + flu + OKT3	Relapse: 16% (4/25) CR: 36% (9/25) OS: 29% at 2Y OS in patients with AML: 40% at 2Y	3.9Y	cGVHD: 8% (2/25)
Passweg et al.(2004) [59]	5	16 (3–25)	2 (40)	AML: 4 CML: 1	CR1: 2 CR2: 1 Rel: 1 CP: 1	Haplo-related donor/same HSCT donors	0.93 × 107 (0.21–1.41 × 107)	1	3–12 months after HSCT (3–26)		*	Eto + cyc + ATG + TBI	CR: 80% (4/5) Relapse: 20% (1/5)	12	No GVHD
B) NK cell infusion in a non-transplantation setting
Ciurea et al. (2024) [13]	12	60	8 (67%)	AML	R/R	KIR ligand matched or mismatched related donor	106 to 107	6	0, + 2, +4, + 7, +9, + 11		*	Flu + Cyt ± G-CSF	CR: 5 (58.3%), OS: 41.7% at 1Y	52	No GVHD
Ahmadvand et al. (2023) [15]	9	50.5 (29–61)	6 (60%)	AML	R/R	KIR ligand mismatch unrelated donor	2 × 106 to10 × 106	2	0, +7		*	Flu (−7 to −4) + Cyc (−3 and − 2)	SD: 6/9 (66.6%) CR: 0% OS: 0% at 1Y	12	No DLT, grade 1 AEs: 40% (4/10)
Sauter et al. (2023) [18]	6	61.5 (27–70)	NM	AML	R/R	NM	1.5 × 109 cell/dose	3	0, + 7, +14		*	Flu + Ara-C	CR/CRi: 66.6% (4/6)	NM	No CRS, neurotoxicity, or GVHD
Berrien-Elliott et al. (2022) [23]	8	62.5 (52–77)	5 (62.5)	AML	R/R	NM	8 × 106	1	0		*	Flu (−6 to −2) + cyc (−5 and − 4)	CR: 0% SD: 14.3% (1/72) MLFS: 14.3% (1/7)	NM	NM
Huang et al. (2022) [27]	10	44.5(18–65)	NM	AML	R/R	Off-the-Shelf	6 × 108, 1.2 × 109 or 1.8 × 109 cells/dose	1 to 3	NM		*	Flu + Cyt	CR: 60% (6/10) at D28	NM	Grade 2 CRS: 10% (1/10)
Heuser et al. (2022) [28]	5	73 (69–81)	NM	AML	CR/CRi with MRD+	off-the-shelf	Up to 1 × 109 cells/dose	1 to 3	0, + 4, +8		*	Flu + cyc (−5 to −3)	At DLI, all three patients achieved MRD negativity	NM	No ICANS, CRS, or GvHD
Gómez García et al. (2021) [30]	7	7.4 (0.78–15.98)	5(71.5)	AML	CR	KIR/HLA-I receptor mismatch related donor	36.44 × 106 (6.92–193.2 × 106)	2	0, +7	*		Cyc (−7) + flu (−6 to −2)	CR: 71% (5/7) Relapse: 28% (2/7)4 OS: 83.3% at 3Y DFS:71.4% at 3Y	33	No GVHD
Vasu et al. (2021) [31]	6	18–80	NM	AML	R/R	Off-the-shelf	1 × 107	6	First dose day 0, Subsequent doses every 2–3 days for six total doses over 2 weeks		*	< 60 y: Flu + Cyta (−6 to −2) > 60 y: Flu (−5 t0 −2) + Deci (−6 to −2	NA	56 days	No GVHD and neurotoxicity
Silla et al. (2021) [32]	13	22 (11–47)	8 (26)	AML	R/R	KIR ligand matched or mismatched donor	1 × 106, 5 × 106 and 10 × 106	6	0, + 2, +4, + 7, +9, and + 11		*	Flu + cyta (−7 to −3) G-CSF (−8)	ORR1: 78.6% at D28-30 CR: 50% at D28-30 OS: 35 ± 15% at 1Y DFS: 67 ± 14% at 6 M	21	aGVHD grade I:0.7% (1/13)
Kim et al. (2021) [33]	11	NM	5(45%)	AML	R/R	NM	2–5 × 109 cell/dose	6	NM		*	Flu + cyc	PR: 28.6% (2/76) SD: 28.6% (2/7) ORR: 28.6 (3/11) OS: 3.4 (2.5–4.3) M	NM	No CRS or GVHD
Berrien-Elliott et al. (2020) [34]	15	72 (43–83)	11 (73.3)	AML: 14 MDS: 1	R/R	NM	DL1: 0.5 × 106 DL2: 1 × 106 DL3: 2–10 × 106	1	0		*	Flu (−6 to −2) + cy (−5 to −4)	CR/CRi: 47% (7/15) MLFS: 20% (3/15) Median LFS: 84 D Median OS: Approximately 276 D	Approximately 42 M	No ICANS, CRS, or GvHD
Mani et al. (2020) [35]	7	55.5(45–66)	4(57)	AML, MDS	R/R	NM	1.43 × 107 (0.5–2.76 × 107)	1	0		*	Flu + das or das alone (−5 to −1)	CR:0 PR:0 PD: 86% (6/7) mDOS: 161 D	Approximately 9 M (1–17)	No GvHD AEs grade3-4: 71% (5/7)
Zhao et al. (2020) [36]	20	59 (37–76)	9 (45%)	AML	MRD + Relapse (5) MRD + resistance (15)	Haplo-related donor	8.04 × 109 cells/dose (5.98–21.48 × 109)	1j	0		*	Flu + cyc (−4 to −2) or cyt/anthracyclines (−2 to 0)	Relapse: 45% (9/20) LFS, MRD (-): 45% (9/20) LFS: 5% (1/20) LFS, MRD (+): 5% (1/20) mDOS postdiagnosis: 887 D	Approximately 29 M	NM
Cooley et al. (2019) [38]	42	51.9 (20–71)	26 (42)	AML: 29 MDS: 8 MDS/CML:1 CML: MPN: 2 CMML:2	R/R	KIR match or mismatched related donor	1.2 × 107 (for IL-15 IV group) or 1.9 × 107 (for IL-15 SC group)	1	0		*	Cy (−5 to −4) + flu (−6 to −2)	IL-15 IV: ORR: 32% (8/25), CR: 24% (6/25), CRi:8% (2/25), OS: 19% at 1Y, PFS: 12% at 1Y IL-15 SC: ORR: 40% (6/15), CR: 6% (1/15), CRi:33% (5/15), OS: 21% at 1Y, PFS: 19% at 1Y	Approximately 45 M	9 (56%) patients in the SC cohort but not the IV cohort experienced CRS (any grade). Neurologic toxicities were observed in 6 (37%) patients who received SC but not IV dosing of IL-15.
Nguyen et al. (2019) [39]	21	6 (0.1–15.3)	11(52)	AML	CR	KIR–HLA-matched or mismatched related donor	12.5 × 106 (3.6–62.2 × 106)	1	0	* As consolidation		Cy (−7) + flu (−6 to −2)	Relapse: 38% (8/21) OS: 84.2 ± 8.5% EFS: 60.7 ± 10.9%	57 M	No GVHD
Wang et al. (2019) [40]	23	50.1 ± 15.3	11 (47.8%)	AML	Hematologic CR	Haplo-related donor	7.5 × 109/L(6.6 to 8.6 × 109)	1–4	NM	* As consolidation		Non-HD-Ara-C regimens	LFS: 65.1 ± 11.1% at 3Y OS: 78.1 ± 10.2 at 3Y Relapse: 30.4% (7/23)	35	No GVHD
Fehniger et al. (2018) [41]	12	73 (57–79)	8(67)	AML	CR1	Haplo-related donor	DL1: 3 × 105 DL1: 1 × 106 DL3: 3 × 106	1	0	*		Cyc (−5) + flu (−6 to −2)	RFS: 33% at 1Y mRFS: approximately 453 D mOS: approximately 519 D	Approximately 17 M	No GVHD, eight patients (67%) experienced grade 3 or 4 TEAEs
Vela et al. (2018) [42]	20 (HNJ:7, LYDIA:13)	12 (1–23)	11 (55)	T-ALL: 8 B-ALL: 5 AML: 6 BAL: 1	R/R	KIR matched or mismatched related donor	6.76 × 106 (0.7-34.16 × 106)	2–4	HNJ NKAES trial: 0 LYDIA trial: 0 and + 7		*	HNJ NKAES trial: VP, Ara-G and Cyc(−12 to 0) LYDIA trial: CLOVE, FLAG-Ida or Flu/Cyc (−12 to −2)	CR MRD−:6 CR MRD+: 7 No remission:2	NM	No GVHD
Björklund et al. (2018) [43]	16	64 (40–70)	10 (62.5%)	AML: 3 MDS/AML:8 MDS:5	R/R or PIF	KIR-ligand matched or mismatched related donor	6.7 × 106 (1.3–17.6 × 106)	1	0		*	Flu (−7 to −4) + cy (−3 to −2) + TLI (−1)	ORR: 37.5 (6/16), CR: 31% (5/16), PR: 6.5% (1/16), SD: 12.5% (2/16), MLFS:6.5% (1/16) OS: 19% at 3Y	More than 20 M	Grade 3 CRS: 6% (1/16) Grade 5 CRS: 6% (1/16) No GVHD
Boyiadzis et al. (2017) [44]	7	71 (56–80)	6 (85%)	AML	R/R	Off-the-shelf	DL1: 1 × 109 cells/m2 DL2: 3 × 109 cells/m2	2	+ 1 and + 2		*	NM	No patient achieved CR or PR	NM	No DLT, no grade 3–4 TRAEs
Dolstra et al. (2017) [45]	10	72 (68–76)	6 (60%)	AML	morphologic CR (in 2 patients, MRD was detectable)	NM	3–30 × 106	1	0	*		Flu + cy (−6 to −3)	Both patients with MRD in the BM before infusion became MRD negative. Relapse: 60% (6/10) OS: 80% at 1Y 4 pats alive after therapy	60	No GVHD and toxicity
Curti et al. (2016) [47]	17	64 (53–73)	9 (53%)	AML	CR:14 MRD+: 3	KIR-ligand mismatched donor	2.5 × 106 (1.29–5.53 × 106)	1	0	*		Flu (−7 to −3) + cy (−2)	Relapse: 44% (7/16) DFS: 56% (9/16)	22.5	No GVHD
Kottaridis et al. (2015) [7]	7	65 (49–73)	5 (71.4%)	AML	CR1-3: 6 PR: 1	HLA-mismatched related donor	1 × 106	1	0		*	Flu (−6 to −2) + TBI (−1)	CR at 1Y: 14% (1/7) Relapse: 71% (5/7) mOS:468.5 D	24	No GvHD
Miller et al. (2015) [51]	24	NM	NM	AML	Ref	NM	NM	1	NM		*	Flu + cy	CR: 12(3/24) OS: 32% at 1Y	NM	NM
Bachanova et al. (2014) [54]	57	44 (3–71)	32 (56%)	AML	R\R	KIR ligand matched or mismatched related donor	0.96 ± 0.3 × 107 0.34 ± 0.05 × 107 2.6 ± 1.5 × 107	1	0		*	Flu (−6 to −2) + Cy (−5 and − 2) + IL2DT8 (−1 ± −2)	CR in patients who did not receive IL2DT: 21% (9/42) at D35 (median duration: 2.3 M) CR in patients who receive IL2DT: 53% (8/15) at D35 (median duration: 11.2 M), OS: 20% at 1Y	32	No GVHD or CRS
Rubnitz et al. (2010) [56]	10	2.5 (0.2–21)	5 (50%)	AML	CR1	KIR ligand matched or mismatched related donor	29 × 106 (5–81 × 106)	1	0	*		Cy (d −7) + flu (−6 to −2)	EFS: 100 at 2Y CR: 100% (10/10)	32.1	No GVHD
Cooley et al. (2019) [37]	10	66 (30–70)	NM	AML (5 history of MDS/five history of HSCT)	R/R	NM	1 × 106, 3 × 106 or 10 × 106	1	0		*	Cy (−5 to −4) + flu (−6 to −2)	MLFS: 10% (1/10) at D14 CRp: 10% (1/10) at D14 MOS: 2.5 M	Approximately 2.5 M	No GVHD CRS: 10% (1/10)
Rubnitz et al. (2015) [50]	29	Prior HSCT < 10 10, > 10:5	9	AML:6 ALL:9	R/R	KIR-matched or mismatched donor	18.6 × 106 (3.5–103 × 106)	1	0		*	Clo + eto + cy (−6 to −2)	CR/CRi: 48% (14/29) PR: 20% (6/29) DFS: 31% OS at 1 Y: 40% in patients with no prior HCT and approximately 42% in patients with prior HCT	6Y	No GVHD, CRS, or effusion syndrome
		No prior HSCT < 10:8, > 10:7	9	AML:6 ALL:8

^aThe timing of NK cell infusion is reported relative to transplant day (day 0), where negative and positive values indicate days before and after transplantation, respectively

^bThe infused cell product contained a heterogeneous population of CD56 + lymphocytes, with NK cells (CD56⁺ CD3^-) predominating over NKT cells (CD56⁺ CD3⁺) at a median ratio of 1.97

^cThe NK conditioning regimen comprised Flu + Ara-C + G-CSF ± idarubicin (n = 7), mitoxantrone + etoposide (n = 6) and G-CSF + cladribine + cytarabine (n = 1)

^dThe NK conditioning regimen included a 5-day FLAG (Flu/cytarabine/filgrastim) course, followed 2–4 weeks later by sequential infusion of donor T-cell-based DLI (day − 1) and NK cells (day 0)

^eThis patient, who had persistent grade 1 skin GVHD following haplo-HCT, continued to exhibit skin involvement after NK cell infusion, which was accompanied by elevated liver enzyme levels

 ^fThis study administered NK cell therapy (day 0) to post-transplant relapse patients following fludarabine/cyclophosphamide conditioning (day − 1)

 ^gPatients received NK cells lymphodepletion (cyclophosphamide day − 7 plus fludarabine days − 6 to −2), followed by NK cell infusion on day 0 and + 14

 ^hPatients received a single dose of a CD56-enriched cellular product containing a mean dose of 6.7 × 10⁶ cells/kg (1.7–17.7) NK cells and 1.3 × 10⁶/kg cells/kg (0.13–5.3) NKT cells

 ⁱCytoreductive chemotherapy typically involved cyclophosphamide (day − 3 to −2), although one patient received cyclophosphamide (day − 6 to −5) combined with fludarabine (day − 6 to −2)

 ^jPatients who achieved an effective clinical response received 1–2 additional cycles of NK cell infusion

Abbreviations AE Adverse event, aGVHD Acute graft versus host disease, ALL Acute lymphoblastic leukemia, AML Acute myeloid leukemia, Ara-G Nelarabine, ARC-c Cytosine arabinoside, ATG Anti-thymocyte globulin, BAL Biphenotypic Acute Leukemia, B-ALL B cell acute lymphoblastic leukemia, BPDC Blastic plasmacytoid dendritic cell neoplasm, BUS Busulfan, cGVHD Chronic graft versus host disease, CLL Chronic lymphocytic leukemia, Clo clofarabine, CLOVE Clofarabine, etoposide and cyclophosphamide, CML Chronic myeloid leukemia CMML Chronic myelomonocytic leukemia, CP Chronic phase, CR1 Complete remission 1, CRi Complete remission with incomplete count recovery, CRS Cytokine release syndrome, CSA Cyclosporin A, Cy Cyclophosphamide, Cyt Cytoxan, Cyta Cytarabine DeciDecitabine, DFS Disease-free survival, DLBL Diffuse large b-cell lymphoma, DLI Donor lymphocyte infusion, DLT Dose limiting toxicity, EFS Event free survival, ETO Etoposide, FLAG Fludarabine, cytarabine (Ara-C), granulocyte-colony stimulating factor (G-CSF), and idarubicin, FLAG-Ida Fludarabine, idarubicin, cytarabine and G-CSF, FLU Fludarabine, G-CSF Granulocyte- colony-stimulating factor, HL Hodgkin lymphoma, HSCT Hemopoietic stem cell transplantation, ICANS Immune effector cell-associated neurotoxicity syndrome, IL-2DT IL-2-diphtheria fusion protein, ITT Intrathecal triple therapy, IV Intravenous, LFS leukemia-free survival, LYM Lymphoma, Mdos Median duration of survival, MDS Myelodysplastic syndrome, Mel Melphalan, MLFS Morphological leukemia-free state, MM Multiple Myeloma, mOS Median overall survival, MPN Myeloproliferative neoplasm, MRD Minimal residual disease, mRFS Median relapse free survival, MSD Matched sibling donor, MUD Matched unrelated donor, NA Not available or not applicable, NHL Non-Hodgkin Lymphoma, ORR Overall response rate, OS Overall survival, PD Progressive disease, PFS Progression-free survival, PIF Primary induction failure, PMF Primary myelofibrosis, PR Partial response, PTCy Post-transplantation cyclophosphmid, R\R Relapse\Refractory, SC Subcutaneous, SD Stable disease, T-ALL T Acute lymphoblastic leukemia, TBI Total body irradiation, TLI Total lymphatic irradiation, VP Etoposide, RIC Reduced-intensity conditioning, MAC Myeloablative conditioning, Treo treosulfan, das; Decitabine,

Quality assessment

Two authors independently evaluated the quality of each included study using the National Institutes of Health (NIH) quality assessment tool for before-after (pre-post) studies with no control group. The tool consists of 12 criteria for rating study quality (good, fair, or poor) (See Supplementary file 2 for the quality assessment of studies) [12]. Disagreements were solved by mutual consensus.

Results

Clinical trials overview

Our study identified 48 clinical trials investigating NK cell therapy, of which 27 studies focused exclusively on AML, while 21 included AML with other hematological disorders, including MDS, lymphoma, acute lymphoblastic leukemia (ALL), chronic myelogenous leukemia (CML), etc. (Fig. 1). The temporal analysis of these studies reveals a significant acceleration in research activity over the past two decades, particularly in trials specific to AML. Only two trials (2/48; 7.4%) were reported between 2010 and 2015, followed by a substantial increase to nine trials (9/48; 33.3%) from 2015 to 2020. The most dramatic increase occurred from 2020 onwards, with 16 studies (16/48; 59.3%) conducted. Given the relatively nascent nature of NK cell therapy, most of these trials were early-phase studies. Among the 27 AML-specific studies, 11 (11/27; 40.8%) were phase I trials, 4 (4/27; 14.8%) were combined phase I/II studies, and Phase II trials accounted for six studies (6/27; 22.2%), while six studies (6/27, 22.2%) did not specify their study phase (Table 1). The study phases in other hematological malignancies trials followed a similar pattern to AML-specific studies, with eight (8/21; 38%) phase I trials, seven (7/21; 33.3%) phase I/II trials, four (4/21; 19%) phase II trials, and two (2/21; 9.5%) studies where the phase was not specified. Details are presented in Table 1.

Fig. 1.

PRISMA diagram of included and excluded articles

Geographical distribution

The geographical distribution of AML-specific studies reveals a concentration of research in developed countries, particularly in North America. The United States led the research with 12 studies (12/27; 44.4%), followed by China with four studies (4/27; 14.8%). India and Spain each contributed two studies (2/27), accounting for 7.4% each. The remaining seven studies (7/27; approximately 26%) were distributed across various countries, including Belgium, Brazil, Iran, the Netherlands, Italy, the United Kingdom, and the Republic of Korea, with each country conducting one study. In contrast, the 21 studies that included other hematological disorders exhibited a similar geographical pattern, albeit with broader international representation. The United States maintained its leading position with 11 studies (11/21; 52.3%), while European participation was notable with contributions from Spain (2/21;9.5%), Switzerland/Germany (1/21; 4.7%), Netherlands (1/21; 4.7%), Switzerland (1/21; 4.7%), Germany (1/21; 4.7%), and France (1/21; 4.7%). Asian representation in these mixed-population studies included contributions from the Republic of Korea (3/21; 14.2%).

NK cell sources and manufacturing

NK cell sources

Analysis of NK cell sources in AML-focused studies revealed that PB was the predominant source, utilized in 22 out of 27 studies (82%). Alternative NK cell sources included CD34⁺ hematopoietic stem/progenitor cells (HSPCs) in two studies (7.4%) [28, 45], while the NK-92 cell line was used in one study (3.7%) (44). In one study (3.7%), the NK cell source was not explicitly mentioned [51] (See Table 2).

Two clinical trials specifically investigated engineered NK cell therapies in AML patients [18, 27]. The first study by Sauter et al. (2023) utilized PB-derived NK cells engineered to target NKG2D ligands using gamma-retroviral vectors with OX40.CD3z.IL-15 costimulatory domains [18]. The second study by Huang et al. (2022) employed umbilical cord blood (UCB)-derived NK cells targeting CD33; however, the specific engineering methodology and costimulatory domains were not detailed in the report [27] (See Table 3).

Among the other hematological studies, a similar preference for PB-derived NK cells was observed (See Table 2).

NK cell purification protocols

Among the AML-specific studies, of the 25 non-engineered NK cell studies, 18 (18/25; 72%) reported their NK cell purification protocol, while seven (7/25; 28%) did not specify or did not use a purification protocol. Of the 18 studies that detailed their purification protocols, four (4/18; 22.2%) employed CD3 depletion alone [13, 15, 32, 33], and four (4/18; 22.2%) utilized CD56 positive selection exclusively [7, 17, 41, 46]. The combination of CD3 depletion followed by CD56 selection was the most prevalent approach, implemented in seven studies (7/18; 38.8%) [19–21, 39, 47, 54, 56]. CD3/CD19 depletion was employed in two studies (2/18; ~11%) (23, 54), while another two studies (2/18; ~11%) utilized CD34⁺ enrichment from UCB [28, 45]. The Miltenyi CliniMACS system was the predominant purification platform, used in 15 studies (15/18; 83.3%) (See Table 2 for more details on each study).

Various purification protocols were reported for the remaining 21 studies that included other hematological malignancies. The distribution of purification methods in these studies followed similar patterns to the AML-specific studies, with CD3 depletion and CD56 selection being the most commonly employed approaches. The RosetteSep system was used in four of the 21 studies (4/21; 19%) involving other hematologic disorders. Notable examples include Thakar et al. (2015), who treated a diverse cohort of 40 patients with various hematological malignancies using CD3 depletion and CD56 selection [52], and Choi et al. (2014), who employed either CliniMACS or RosetteSep systems for NK cell enrichment in a mixed cohort of 41 patients [53]. Notably, Bachanova et al. (2014) demonstrated a unique approach by comparing three different purification strategies in their study: CD3 depletion alone (Cohort 1), CD3 depletion followed by CD56 selection (Cohort 2), and CD3/CD19 depletion (Cohort 3) [54]. Details are presented in Table 2.

NK cell culture process

Among the 25 non-engineered NK cell studies that focused explicitly on AML patients, nine studies (9/25; 36%) employed short-term activation of NK cells, lasting up to 24 h. The most commonly used cytokines for NK cell activation were IL-2 (2/9, 22.2%) [17, 54] and a combination of IL-12, IL-15, and IL-18 (2/9, 22.2%) [21, 24]. Additionally, IL-15 was used in two studies (2/9, 22.2%) [19, 51], and one study (1/9; ~11%) utilized ALT-803, an IL-15 superagonist complex [23].

Moreover, nine studies (9/25; 36%) employed longer-term expansion protocols (lasting 7 days or more). The most commonly used durations were 21 days (3/9, 33.3%) [15, 30, 32], 14 days (2/9, 22.2%) [13, 33], and up to 42 days in one study (1/9; ~11%) [45]. Cytokines used for expansion included IL-2 (6/9; 66.3%) [13, 15, 30, 33, 45, 57], IL-15 (4/9; 44.4%) [16, 37, 45, 57], and IL-21 (3/9; 33.3%) [16, 45, 57], either alone or in combination with other cytokines.

Three studies (3/9; 33.3%) (37, 45, 57) employed cytokine cocktails. Among these, two studies used a similar cytokine cocktail, including granulocyte-macrophage colony-stimulating factor (GM-CSF), granulocyte colony-stimulating factor (G-CSF), and interleukin (IL)−6 for NK cell differentiation and expansion in two weeks (37, 45). However, another study used SCF, Flt3L, and IL-7 for maturation and differentiation, while IL-15 and IL-21 are used for expansion of NK cells over 21 days (57).

Among the studies, ten (10/25; 40%) employed feeder cells, predominantly using irradiated K562-derived cell lines. Specifically, four studies utilized K562 cells expressing membrane-bound IL-21 and 4-1BBL (K562-mbIL21-41BBL), two studies employed K562 cells co-expressing membrane-bound IL-15 and 4-1BBL (K562-mb15-41BBL) (30, 42), and one study used K562 cells engineered to express membrane-bound IL-21 (32). Additionally, one study utilized an OCI-AML3-derived feeder cell line expressing membrane-bound IL-21 (NKF) (25), while two studies employed irradiated autologous PBMCs as feeder cells (15, 33) (See Table 2 for more details).

Safety and efficacy outcomes

NK cell therapy in transplantation settings

Table 4 divides the studied patients into two groups: patients who received HSCT (Table 4a) and those who did not (Table 4b). Twenty studies involving 342 patients evaluated NK cell infusion in a transplantation setting. A shared characteristic among all 20 studies was the inclusion of AML patients, while 14 also involved patients with other hematological disorders. Notably, most of the studied patients (n = 225, 66%) presented with AML, while only 117 patients (34%) had other disorders, including MDS, ALL, CML, etc. Six studies involving 65 patients specifically addressed AML, with an average age of 29.6 years [17, 19–21, 26, 46]. In gender-specified articles, more than half of the cases were male (59.5%). Unfortunately, not all studies provided details regarding the disease phase or remission status of their AML patients before transplantation. In studies that provided this information, only four patients (7.6%) were in CR, while 49 patients (92.4%) were in a relapsed/refractory (R/R) state. In 6 of 20 included studies, 61 patients underwent haploidentical transplantation (haplo-HSCT) (93.8%). In addition, two patients received stem cells from a matched related donor (MSD), while 2 received stem cells from matched unrelated donors (MUD) [20]. Of the 225 patients diagnosed with AML, 204 (90%) received a conditioning regimen. The maximum injected dose of NK cells was 1 × 10⁹ cells/kg. All patients received at least one NK cell injection, with a maximum of three injections administered. In two studies, 23 patients (35.4% of the total 65 AML patients) received NK cells as a therapeutic intervention [17, 20], whereas the four remaining articles, involving 42 patients (64.6%), used NK cells for prophylactic purposes [19, 21, 26, 46]. At a median follow-up of 17.6 months (7–24 months), the average OS was 39.5% (n = 92). CR was reported in 38 patients, with an average rate of 37%, and the relapse rate was 35.6% (n = 23). The acute and chronic GVHD rates were 37.3% and 23.7%, respectively (See Table 4a).

NK cell therapy in Non-Transplantation settings

Twenty-eight additional studies investigated the application of NK cell therapy in 456 patients diagnosed with AML, MDS, ALL, and myeloproliferative disorders (MPDs). Nearly 90% of these patients were AML patients (n = 406). The average age of the participants was 50.2 years. In gender-specific studies, it was reported that over half of the cases were male, accounting for 57.8%. Among these studies, 22 articles [7, 13, 15, 18, 21, 27, 28, 30–33, 36, 39–41, 44, 45, 47, 51, 54, 56] concentrated exclusively on AML, while the remaining publications encompassed other hematological conditions. Most cases (70.7%, n = 287) were classified as being in the relapsed or refractory phase of AML. Conversely, 26.1% of the patients, comprising 106 individuals, were in CR before commencing NK cell therapy. Most articles, 68.1% (n = 15), focused on utilizing NK cells as a treatment for AML [7, 13, 15, 18, 21, 27, 28, 31, 32, 36, 44, 54]. On the other hand, 32% (n = 7) of the articles explored the use of NK cells as prophylactic agents [30, 41, 45, 47, 56] or consolidative agents [39, 40]. This approach is particularly noteworthy as it involves administering NK cells to prevent disease onset or recurrence rather than treating an existing condition. All patients received 1 to 6 infusions, with each administration consisting of up to 21.48 × 10⁹ cells per dose. With a median follow-up of 28.7 months, the outcomes of the AML patients were as follows: 37.1% of cases achieved CR, and 28.6% reached PR. Furthermore, 71.3% of the patients reported achieving event-free survival (EFS). However, it is essential to note that 42.2% of patients experienced a relapse following NK cell treatment (More details are presented in Table 4b).

Adverse events and Treatment-Related toxicities

In this comprehensive analysis, which involved 406 patients diagnosed with AML, the occurrence of GVHD was notably low, with only one patient exhibiting signs of this condition [32]. Additionally, the studies reported that 13 patients experienced some form of adverse event (e.g., CRS, neurotoxicity) during treatment. Furthermore, it was encouraging to note that 208 patients showed no signs of GVHD, indicating a favorable response to the treatment without any complications. In addition, 49 patients reported no side effects, including critical conditions such as GVHD, CRS, or neurotoxicity. The absence of these side effects is particularly significant, as CRS and neurotoxicity are common concerns in immunotherapy and can impact patient quality of life and treatment outcomes (See Table 4).

Discussion

Our systematic analysis of 48 clinical trials on NK cell-based immunotherapy for AML revealed substantial progress in the development of NK cell manufacturing and clinical application strategies. The results underscore a sharp increase in NK cell-based clinical trials over recent years, signaling growing interest and investment in this area. Among the NK cell sources reported in the included studies, the predominant choice was allogeneic NK cells derived from PB, utilized in approximately 68% of the trials. Other sources, including HSPCs from CD34⁺ cells, NK-92 cell lines, and UCB, were used in fewer studies, indicating their more experimental status or limited scalability. PB-derived NK cells offer accessibility, established isolation protocols, and readily available mature cells with proven cytotoxic activity [10]. Furthermore, the capacity to repeatedly source cells from the same donor for subsequent product manufacturing ensures process consistency and facilitates the sustained development and optimization of NK cell-based therapeutic strategies [10]. However, PB-derived NK cells present challenges, including donor-to-donor heterogeneity, limited expansion capacity, and finite cell yields per apheresis [10]. Engineering these cells is also more complex, often requiring virus-based gene transfer methods, which can lead to variable outcomes [60]. UCB-derived NK cells were used in one study with engineered NK cells [27], presenting an alternative source with distinct benefits. Although generally displaying reduced cytotoxicity, UCB contains approximately 30% NK cells, with comparable NK cell subsets and cytokine secretion capacities to those of PB-derived NK cells [61]. UCB NK cells can be significantly expanded ex vivo, achieving up to 1000- to 2000-fold increases with exposure to cytokines and feeder cells. Their advantages include ease of expansion, high purity (median 75%) with minimal T cell contamination (< 1%), and no induction of GVHD in HLA-mismatched settings [45, 62]. The ability to select optimal donors based on killer immunoglobulin-like receptor (KIR) haplotypes enhances their therapeutic potential [45]. Nonetheless, challenges remain, including a limited starting cell source, reliance on ex vivo differentiation, and the inability to collect repeated donations from the same donor [10]. CD34⁺ HSPCs-derived NK cells, derived from cord blood, were used in 4.3% of studies [28, 45]. HSPC-derived NK cells are valued for their early sourcing, as they contain fewer mutations. They can be freshly isolated or sourced as off-the-shelf frozen products, eliminating risk to the newborn or mother [63, 64]. As demonstrated in clinical trials, HSPCs are differentiated into NK cells following a well-defined protocol. According to the data, HSPCs are initially cultured in media supplemented with a low-dose cytokine cocktail (GM-CSF, G-SCF, and IL-6), followed by early-stage differentiation cytokines (SCF, Flt3L, TPO, and IL-7) for the first 9 days. Between days 10 and 14, TPO was replaced by IL-15 to promote NK cell lineage commitment. Finally, Flt3L is substituted with IL-2 to support the terminal differentiation and maturation of NK cells [45]. These cells can differentiate into large numbers of functional NK cells within four weeks, allowing for the generation of extensive, quality-controlled NK cell batches sufficient for treating multiple patients. Despite their theoretically unlimited supply, HSPCs faced differentiation efficiency limitations, making it challenging to produce enough homogeneous NK cell populations from a single batch [64]. The NK-92 cell line was utilized in one study [44] and offers several benefits, including potent cytotoxicity against various cancer cell types and the convenience of rapid ex vivo expansion, with a doubling time of 24–36 h [65, 66]. NK-92 cells can be genetically modified through plasmid electroporation to express high-affinity Fc receptors (CD16), chimeric antigen receptors (CARs), or other molecules influencing the tumor microenvironment [67]. However, NK-92 cells are derived from a malignant cell line; they require γ-irradiation before patient administration for safety reasons. This irradiation has a significant impact on their in vivo persistence. While NK-92 cells maintain high cytotoxic activity immediately after receiving 10 Gy irradiation, they lose more than 50% of their activity within 24 h of receiving even 7 Gy [68, 69]. This reduced persistence limits their ability to sustain tumor-targeting activity in vivo [66, 70].

Among the analyzed trials, two studies (4%) investigated engineered CAR-NK cells, highlighting the emerging interest in this approach [18, 27]. Sauter et al. developed NKG2D-targeting CAR-NK cells from PB with an OX40.CD3z.IL-15 costimulatory domain [18], while Huang et al. focused on CD33-targeting CAR-NK cells derived from UCB [27]. CAR-NK cells offer several advantages over CAR-T cells for the treatment of hematologic malignancies. Unlike CAR-T cells, which require patient-specific manufacturing and carry risks of GVHD, CAR-NK cells can be used as an “off-the-shelf” allogeneic product without causing GVHD, making them more cost-effective and accessible to patients [10, 71]. Additionally, CAR-NK cells demonstrate a superior safety profile, with clinical trials showing minimal risk of severe CRS or neurotoxicity, two major concerns associated with CAR-T therapy [72, 73]. Furthermore, CAR-NK cells possess multiple tumor recognition mechanisms, combining both CAR-mediated targeting and their natural ability to recognize and kill tumor cells through various receptors and antibody-dependent cellular cytotoxicity (ADCC), making them potentially more effective against heterogeneous tumors that might escape single-antigen targeted CAR-T therapy [74]. As demonstrated in Sauter’s study, including IL-15 in the construct is an important advancement as it promotes CAR-NK persistence and function [18]. Despite these advantages, challenges remain, including optimal CAR design for NK cells, enhancement of in vivo persistence, and scalable manufacturing processes. The ongoing trials will provide crucial insights into the efficacy and safety of CAR-NK cell therapy in AML patients.

Regarding NK cell purification, our analysis revealed that about two-thirds of studies reported their purification protocol (~ 66%) (Table 2). While CD3 depletion combined with CD56 selection was the most common approach used in 48.3% of studies [7, 13, 15, 17, 32, 33, 41, 46], single-step CD56⁺ enrichment presents a potentially more advantageous strategy. Unlike the two-step process that eliminates CD3⁺ cells, CD56⁺ selection retains CD56⁺CD3⁺ cells (NKT cells) and NK cells. This approach has demonstrated several benefits. First, NKT cells have been shown to preserve GVL effects without increasing GVHD [75]. Studies have found that NKT cells, particularly those expressing CD8, can produce IL-4, which promotes regulatory T cells (Tregs) expansion, potentially explaining the lower GVHD rates after HSCT [76, 77]. Second, the retention of NKT cells through single-step CD56⁺ selection allows the deployment of this immunologically important population that would otherwise be lost with CD3⁺ depletion [78]. The study by Jaiswal et al. evaluated CD56-enriched donor cell infusion in 10 patients undergoing haploidentical transplantation and found a significantly lower incidence of grade 2–4 acute GVHD (0% vs. 50% in the control group; P = 0.01). The intervention group also showed early and robust reconstitution of mature NK cells, CD4⁺ T cells, and Tregs by day 30, without any reported viral or fungal infections. These findings suggest that CD56⁺ donor cell infusion after PTCy is not only feasible but may contribute to improved immune reconstitution and reduced GVHD, with comparable relapse rates and non-relapse mortality relative to controls [46].

The Miltenyi CliniMACS device was used as the primary tool for NK cell isolation, appearing in 63.8% of the studies (Table 2). Its widespread use suggests it is a trusted platform for clinical-grade cell separation due to its effectiveness in producing NK cell populations with high purity and minimal T cell contamination. Though less common, other methods included alternative selection and depletion techniques, each selected based on specific study goals and cell source characteristics.

The culture process of NK cells has demonstrated significant heterogeneity across studies, particularly in terms of activation and expansion protocols. Short-term activation protocols (≤ 24 h) were employed in 13 studies (27.6%), primarily utilizing IL-2 (22.2% of activation studies) [17, 54], IL-15 (22.2%) [19, 51], or a combination of IL-12/IL-15/IL-18 (22.2%) [21, 24]. However, most studies (n = 22; 46.8%) focused on longer-term expansion protocols, predominantly spanning 14–21 days (Table 2). The cytokine combinations used during expansion varied, with IL-2 being the most common (66.6% of expansion studies) [13, 15, 30, 33, 36, 44]. While IL-2 remained the most frequently used single cytokine in activation and expansion protocols, a notable trend emerged in the development of memory-like NK cells through combined stimulation with IL-12, IL-15, and IL-18. Memory-like NK cells demonstrate increased effector capabilities in response to subsequent stimuli, including CD16 engagement, tumor cell targets, and cytokine exposure [79, 80]. These cells also exhibit enhanced metabolic functions and improved infiltration into lymphoma tissues [81]. Importantly, memory-like NK cells maintain their enhanced functionality even after cell division, exhibit distinctive epigenetic and transcriptomic characteristics, and can survive for long durations in favorable immune environments, making them promising candidates for clinical research [21, 82].

Regarding the expansion protocols, recognizing that activated NK cells alone may not meet clinical dose requirements, various expansion protocols have been developed. These protocols are broadly divided into feeder-free and feeder-based approaches. Feeder-free methods primarily relied on cytokine combinations, with IL-2 monotherapy being the most common. While these approaches minimize regulatory complexity, they typically achieve modest expansion rates. However, feeder-based systems demonstrated superior expansion capabilities with several innovative approaches. In this line, four studies (44.4%) utilized K562 cells modified to express mb IL-21 and 4-1BBL, achieving up to 100-fold expansion over 21 days [13, 30, 32, 36]. K562 cells, an erythroleukemia cell line deficient in HLA class I, naturally provide a contact-dependent stimulus that drives NK cell expansion [83]. This expansion potential has been greatly enhanced through the genetic engineering of K562 cells to express additional NK cell stimulatory molecules, particularly the 4-1BBL co-stimulatory ligand and mb cytokines such as IL-15 or IL-21 [84]. Studies have shown that K562 cells engineered to express mbIL-21 can achieve high NK cell expansion, up to 47,697-fold, while maintaining increased telomere length and sustaining their proliferation capacity [85]. K562 feeder cells induced a similar phenotype in the expanded NK cells, characterized by high expression of natural cytotoxicity receptors (NCR) and NKG2D. The NK cells also exhibited prominent activation and proliferation features, characterized by elevated levels of granzyme B and perforin [85, 86]. Although alternative feeder cells have been investigated, K562-based systems remain the most widely used due to their superior and reproducible expansion efficiency. Otegbeye et al. demonstrated the efficacy of their novel OCI-AML3-based system (NKF) platform, which utilizes engineered OCI-AML3 cells expressing membrane-bound IL-21. This system achieved remarkable expansion rates, increasing 36-fold in 2 weeks and up to 152-fold over 3 weeks while maintaining high product purity (98.03% NK cells) and minimal T cell contamination (1.06%). The NKF platform represents a substantial evolution in feeder-based expansion technology, offering robust expansion capabilities and compatibility with good manufacturing practices [25]. Another interesting approach is the use of tumor cell line lysates for activating NK cells. Fehniger et al. reported using a CNDO109 (CTV-1 tumor cell line) lysate to activate purified NK cells for 16 h under GMP conditions without adding any cytokines. This resulted in an NK cell product with 71.8% CD56⁺CD3⁻ purity and 96.4% viability [41]. This multicenter Phase 1 clinical trial demonstrated the feasibility and safety of generating and administering CNDO-109-NK cells, which were primed with a tumor cell line lysate and showed promising clinical activity in high-risk AML patients who were treated in CR [41].

In this article, we conducted a systematic review to evaluate the efficacy and safety of NK cell therapy for AML. Given that this is a newer treatment, all the studies we reviewed were in phases 1 or 2 of clinical trials, featuring a limited number of patients. Additionally, the heterogeneity of the available studies and the lack of focus on a specific hematological disease, since different types of leukemia result in varying OS, EFS, and risks of relapse or CR, limited our ability to present the data. Consequently, we were unable to perform advanced analyses such as meta-analyses.

The effects of NK cell therapy vary between patients with and without stem cell transplantation due to the interplay of the patient’s immune condition, the nature of the leukemia, and their treatment history [87–89]. Therefore, we categorized NK cell recipients into two groups: patients who underwent HSCT and those who did not. This distinction was necessary due to the differing effects observed in patients who underwent HSCT compared to those who did not. For instance, recipients may exhibit elevated levels of specific cytokines after stem cell transplantation, which can either promote NK cell activation or inhibit their function [90, 91]. In HSCT, particularly from unrelated donors, HLA mismatches may lead to an enhanced GVHD, where the overall immune system, including NK cells, targets residual cancer cells. Additionally, post-transplant cyclophosphamide(PTCy), administered for GVHD prophylaxis, can lead to the release of IL-15 in the body, resulting in increased proliferation and persistence of NK cells [52]. Conversely, the immunosuppressive therapy recipients receive after transplantation may dampen their overall immune response, including NK cell activity, potentially affecting the effectiveness of NK cell therapy. There is an ongoing debate among researchers about disease burden [92]. Previously, positive MRD, even at low levels, was linked to unsuccessful transplantation [93]. However, some researchers argued that HSCT was the preferred treatment for patients who did not achieve complete remission with pre-transplant chemotherapy [94]. In our study, 92% of AML patients were in relapse or refractory phases before transplantation, with 37% achieving complete remission and 35.6% relapsing afterward. Notably, these patients received at least one dose of NK cells alongside transplantation, distinguishing our data from previous studies. Of course, in immunotherapy (for instance, CAR-T cell therapy), some studies reported significantly enhanced clinical outcomes for AML patients who were MRD-negative before transplantation [95]. The lack of adequate research in this area complicates the drawing of accurate conclusions. Similar results were observed in patients who were not candidates for transplantation but received NK cell therapy for treatment or as prophylaxis. Approximately 70% of patients were in a relapse or refractory phase before receiving NK cells; however, the final rates of CR and PR were 37% and 28%, respectively. A relatively high number of patients also achieved disease-free survival (DFS). In a similar study involving 16 patients with AML and MDS (only 3 were primary AML cases), the researchers highlighted the effectiveness of NK cells in patient outcomes. Aligned with our findings, the study reported a 37.5% complete or partial remission rate after NK cell infusion in patients in relapse or refractory phases [43]. Studies have shown that high levels of NK cells in patients with AML are associated with improved outcomes and increased OS [96]. Among individuals who received transplantation or not, over half (52.4%) of those who received NK cell prophylaxis achieved complete remission. This rate of CR is higher than the overall CR observed in our study. These findings align with a meta-analysis by Mushtaq et al., which investigated NK cell therapy in patients with hematological malignancies (1785 patients from 21 studies) following transplantation [97]. Notably, in the studies we considered, this relationship was observed in patients who received prophylactically high doses of NK cells, resulting in higher rates of CR and improved OS. Additionally, a study by Curti et al. examined 17 AML patients treated with an average dose of 3.8 × 10⁶ NK cells, finding that higher doses of reactive NK cells significantly enhanced treatment responses in these patients [47]. GVHD does not have a high incidence rate in patients who received NK cells without transplantation. This current systematic review only occurred in one case, which involved a patient with a history of previous transplantation with a presentation of diarrhea. Other patients experienced side effects from receiving NK cells, including neurotoxicity and CRS. Studies indicated that the cause of neurotoxicity is the passage of NK cells across the blood-brain barrier (BBB) [32]. The observed GVHD was not notably high. Given the results outlined earlier, it appears that NK cells by themselves do not cause GVHD. Consequently, in studies reporting GVHD, the condition has been attributed solely to inadequate donor matching, with NK cells deemed irrelevant [34].

Ultimately, we conducted a systematic review of the role of NK cells in AML treatment, comparing data from patients who received HSCT with those who did not. While we made significant strides in understanding this area, our research faced several limitations, including a limited number of patients and the heterogeneity of the included studies, which precluded a meta-analysis. Nevertheless, comparing these results can be valuable and may bring a bright outlook for the future of NK cell therapy. In this article, we have concentrated exclusively on studies that focused on AML cases; however, as noted in Table 4, other studies also encompassed additional hematological diseases. Due to the lack of separate reporting of their results, these findings were not incorporated into our discussion and conclusions.

Conclusion

In conclusion, this systematic review aimed to highlight the benefits of NK cell therapy for the treatment of AML, and the findings suggest a remarkable increase in NK cell-based clinical trials in recent years, reflecting a growing interest and investment in this promising area of immunotherapy. In this article, we conclude that no single method has been identified for extracting NK cells that is superior to others; each approach has its advantages and disadvantages. By understanding the unique benefits and limitations of each method, anyone can make an informed decision to use PB, UCB, CD34⁺ HSPCs, or NK-92 cell lines that align with their therapeutic goals. According to our data, the dual benefit of preserving GVL with reduced GVHD introduced single-step CD56 + selection as a suitable method for NK purification; however, our review indicated that fewer studies have employed this technique. Most of our studies have used IL-2 to increase the number of NK cells; however, research suggests that feeder cell-based methods are more effective for large-scale NK cell production. In AML patients, achieving CR before transplantation is ideal for better outcomes. However, those who relapse or become refractory after transplantation may benefit from NK cell therapy. Most studies in our systematic review focused on patients not in CR, yet they still achieved a CR and OS of more than 30%. Although we did not conduct a meta-analysis to determine the statistical significance of this data, it is still valuable. Notably, the low incidence of GVHD among patients receiving NK cells suggests that these cells may be safely administered without exacerbating this condition. Overall, as the landscape of AML treatment continues to evolve, our findings suggest a more nuanced understanding of NK cell therapy’s role as a standalone treatment and as a complementary approach alongside HSCT. Future research should address the current knowledge gaps, particularly concerning patient selection for NK cell therapies, to improve clinical outcomes for AML patients.

Abbreviations

HSCT

Hematopoietic stem cell transplantation

AML

Acute myeloid leukemia

MDS

Myelodysplastic syndromes

GVL

Graft-versus-leukemia

NK cell

Natural killer cell

GVHD

Graft-versus-host disease

PRISMA

Preferred reporting items for systematic reviews and meta-analyses

CR

Complete remission

OS

Overall survival

MLFS

Morphologic leukemia-free state

MRD

Minimal residual disease

CRS

Cytokine release syndrome

NIH

National institute of health

ALL

Acute lymphoblastic leukemia

CML

Chronic myeloid leukemia

PB

Peripheral blood

HSPCs

Hematopoietic stem/progenitor cells

UCB

Umbilical cord blood

FBS

Fetal bovine serum

GM-CSF

Granulocyte-macrophage colony-stimulating factor

G-SCF

Granulocyte colony-stimulating factor

IL

Interleukin

SCF

Stem cell factor

Flt3L

FMS-like tyrosine kinase 3 ligand

TPO

Thrombopoietin

MB

Membrane-bound

PBMC

Peripheral blood mononuclear cells

R/R

Relapsed/Refractory

MSD

Matched sibling donor

MUD

Matched unrelated donors

MPDs

Myeloproliferative disorder

KIR

Killer immunoglobulin-like receptor

CAR

Chimeric antigen receptor

ADCC

Antibody-dependent cellular cytotoxicity

HLA

Human leukocyte antigens

Treg

Regulatory T cell

EFS

Event-free survival

BBB

Blood-brain barrier

Authors’ contributions

MBD is considered the principal investigator and senior author, and participated in the literature search, data collection, writing of the original draft, and editing of the final draft.SK and AR were involved in quality assessment, data collection, data interpretation, and manuscript drafting.SMS contributed to the original draft writing and reviewed and edited the final version.SE, FF, and NS assisted in article screening and data collection.HR and DH participated in data collection and writing the original draft.SY contributed to the original draft writing and reviewed and edited the final version.AI, SP, ER, and AH reviewed and critically revised the manuscript.AG is considered the corresponding author, having proposed the project and contributed to the conception, review, and revision of the manuscript.

Funding

No funding was used for this study.

Data availability

No datasets were generated or analysed during the current study.

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.