This prospective study was conducted between March 2020 and April 2021. Anonymized healthy UCB samples were obtained from the Seoul Metropolitan Government Public Cord Blood Bank (Allcord) upon approval. This study was approved by the Institutional Review Board of Samsung Medical Center, Seoul, Korea (approval No. 2020-03-057). All experiments were conducted in accordance with relevant guidelines and regulations.

UCB mononuclear cells were isolated by density gradient centrifugation (25 minutes, 37°C, 500×g) using Ficoll-Hypaque (d=1.077, Lymphoprep; Axis-Shield, Oslo, Norway) and washed twice with phosphate-buffered saline (Welgene, Gyeongsan, Korea). All protocols were performed in accordance with relevant guidelines and regulations.

Isolated UCB mononuclear cells were cultured in RPMI 1640 (Hyclone Laboratories, Logan, UT, USA), a mixture of DMEM/Ham’s F-12 (Gibco, Grand Island, NY, USA; 2:1 v/v), and AIM V serum-free (Gibco) basal medium. RPMI 1640 medium was supplemented with 10% heat-inactivated FBS (Gibco), 100 U/mL penicillin, 100 μg/mL streptomycin (Lonza, Walkersville, MD, USA), and 4 mmol/L glutamine (Gibco) (RC). DMEM/Ham’s F-12 (2:1) was supplemented with 10% heat-inactivated human AB serum (Thermo Fisher Scientific, Waltham, MA, USA), 25 µM beta-mercaptoethanol (Thermo Fisher Scientific), 50 µmol/L ethanolamine (Sigma-Aldrich, St. Louis, MO, USA), 20 µg/L ascorbic acid (Sigma-Aldrich), and 5 ng/L sodium selenite (Sigma-Aldrich) (DS). AIM V was supplemented with 5% ICSR (Life Technologies, Grand Island, NY, USA).

K562 and Raji cells were acquired from the American Type Culture Collection (Manassas, VA, USA) and were cultured in RC medium at 37°C in a humidified 5% CO₂ incubator. K562-OX40L-mbIL18/21 feeder cells were generated by transducing parental K562 cells with lentiviral vectors encoding OX40L, IL-18, and IL-21 [13].

To eliminate donor-to-donor variation in NK cell expansion, isolated UCB mononuclear cells from each donor were split into same-size aliquots for paired expansions in each culture medium assessed. The numbers of donors used for NK cell expansion are indicated in the figure legends. The isolated UCB mononuclear cells were expanded in RC, DS, and AI. K562-OX40L-mbIL18/21 feeder cells were irradiated with 100 Gy of radiation and co-cultured with the UCB mononuclear cells in a 24-well plate. In the first week, 10 U/mL recombinant human IL-2 (PeproTech, Rocky Hill, NJ, USA) was added to each culture. After one week, the IL-2 level was increased to 100 U/mL and an additional 5 ng/mL of soluble IL-15 was added to each culture, as previously described [13]. UCB NK cells were stimulated with K562-OX40L-mbIL18/21 cells on days 0, 7, 14, and 21. The culture medium was replaced every 2-3 days, and the cells were continuously cultured until day 28.

In the culture medium shifting experiment, the culture medium was shifted on day 14. During the shift, AI was added to the DS plate, and DS was added to the AI plate, along with 100 U/mL IL-2 (PeproTech) and 5 ng/mL IL-15 (PeproTech). Expanded UCB NK cells were harvested on day 28 for further experiments. Cell expansion is presented as “fold expansion,” which was calculated by dividing the total number of CD56+/CD3– live NK cells in each week by the number of CD56+/CD3– NK cells on day 0.

Recombinant human IL-2 and IL-15 (PeproTech) were used to expand the NK cells. Allophycocyanin (APC)-Cy7-conjugated anti-human CD3, phycoerythrin (PE)-Cy7-conjugated anti-human CD56, PE-Cy5-conjugated CD56, and fluorescein isothiocyanate (FITC)-conjugated CD3 (eBioscience, San Diego, CA, USA) were used to measure the purity of NK cells (CD56+/CD3–). Pacific blue-conjugated anti-human CD16, Pacific blue-conjugated anti-human NKp46 (eBioscience), APC-conjugated anti-human DNAM-1, APC-conjugated anti-human NKG2A, and PE-conjugated anti-human NKp30 (BD Biosciences, Franklin Lakes, NJ, USA) were used to measure UCB NK cell receptor levels.

For functional assays, expanded NK cells were harvested on days 14 and 28 to measure their cytotoxicity and antibody-dependent cellular cytotoxicity (ADCC) against target tumor cells (K562 and Raji cells) by flow cytometry using carboxyfluorescein diacetate succinimidyl ester (CFSE; Life Technologies) staining of the target cells, as previously described [12, 13]. Briefly, K562 and Raji target cells were stained with 0.5 μM CFSE in fluorescence-activated cell sorting (FACS) buffer at 37°C for 10 minutes and then washed twice with RPMI medium. For the ADCC assay, Raji cells were stained with 0.5 μM CFSE in FACS buffer at 37°C for 10 minutes and then washed twice with RC medium. The CFSE-stained Raji cells were then supplemented with 5 μg/mL rituximab. The target cells (2×10⁴) were transferred into a 96-well U-bottom plate in triplicate and mixed with various numbers of expanded NK cells (0.5:1, 1:1, and 2:1 effector-to-target ratios). The plates were centrifuged at 400×g for 3 minutes and then incubated for 4 hours for functional analysis and for 30, 90, 150, and 240 minutes for the time-course killing assay at 37°C in a 5% CO₂ incubator. After the incubation, the co-incubated cells were transferred to FACS tubes. One microliter of 1 mg/mL propidium iodide (Sigma-Aldrich) was added to each tube before data acquisition. All functional assay data were acquired on a FACSVerse instrument (BD Biosciences) and analyzed using the Kaluza software version 2.1. The numbers of donors used in the NK cell functional assays are indicated in the figure legends.

Live imaging for cytotoxicity was conducted as previously described [17]. Briefly, clean coverslips (18×18 mm) were treated with air plasma (100 W, Femto Science, Hwaseong, Korea) for 1 minute and coated with a protein solution containing fibronectin (2.27 mM, Sigma) and anti-CD44 (0.25 mg/mL, clone no. IM7, eBioscience) at 37°C for 30 minutes. K562 cells were stained with 1 mM CellTrace Far Red (Invitrogen) according to the manufacturer’s instructions. The stained K562 cells (0.1×10⁶ cells) and NK cells (0.01×10⁶ cells) were mixed and seeded on the antibody-functionalized coverslips. The samples were then centrifuged to capture the cancer cells from the surface (400×g, 3 minutes).

A modified Olympus IX 83 epi-fluorescence microscope equipped with a 40× (UPlanFL N, NA=1.30) objective lens and a Zyla 4.2 sCMOS camera (Andor) was used for imaging experiments. For fluorescence imaging, a U-LH75XEAPO Xenon lamp and 75 W (Olympus) and Cy5 (EX BP 620/60, BS 660, and EM BP 770/75) filter sets were used. The microscope was automatically controlled by a micro-manager, and the stages were equipped with a Chamlide TC incubator system (Live Cell Instrument, Seoul, Korea) to maintain cell culture conditions (37°C, 5% CO₂). The images acquired were analyzed and processed using the ImageJ software (NIH, Bethesda, MD, USA).

To measure granzyme B and perforin levels, day-28 unstimulated expanded UCB NK cells were harvested and stained with APC-Cy7-conjugated anti-human CD3 and PE-Cy5-conjugated anti-human CD56 (eBioscience/BD Biosciences) on ice for 15 minutes. After surface receptor staining, the cells were washed, permeabilized, and fixed using a Cytofix/Cytoperm kit (BD Biosciences) according to the manufacturer’s instructions. Then, the cells were stained with FITC-conjugated anti-human granzyme B and PE-conjugated anti-human perforin (BD Biosciences) on ice for 30 minutes. Finally, the cells were washed and analyzed using the FACSVerse cell analyzer.

Differences between the groups in terms of the purity, fold expansion, receptor expression, and cytotoxicity of UCB NK cells were statistically analyzed using two-tailed Student’s paired t-test for two groups and one-way ANOVA for more than two groups, using GraphPad Prism 7 (GraphPad Software, San Diego, CA, USA). P≤0.05 was considered significant.

To investigate the role of the culture medium in NK cell expansion, UCB mononuclear cells were cultured with K562-OX40L-IL18/21 feeder cells in three different culture media: RC, DS, and AI. The fold expansion and purity of each culture were checked weekly until day 28 (Fig. 1A, B). Compared with NK cells cultured in RC, those cultured in DS showed a remarkable increase in fold expansion between days 21 and 28. NK cells cultured in AI also exhibited an increased proliferation rate but not as high as that of cells cultured in DS (Fig. 1A). NK cell purity was not as substantially influenced by the culture medium as fold expansion: NK cells cultured in RC and DS exhibited a slightly higher purity (percentage of CD56+/CD3– cells) than those cultured in AI (Fig. 1B). The fold expansion of NK cells cultured in DS on day 28 was nearly five times greater (38,240) than that of NK cells cultured in RC (8,124), suggesting that an appropriate culture medium is critical for cost-efficient NK cell production.

In addition to fold expansion, the anti-tumor activity of NK cells is critical for NK cell-based adoptive cell cancer immunotherapy. We compared the anti-tumor activities of NK cells expanded in the different media using direct cytotoxicity and ADCC assays on days 14 and 28. K562 (chronic myeloid leukemia) and Raji (B-cell lymphoma) cells were used as target cells, and rituximab was added to the Raji cells. Interestingly, NK cells cultured in AI exhibited significantly higher cytotoxicity on both days 14 and 28 than those cultured in RC and DS under all conditions (Fig. 1C–E for day 14 and Fig. 1F–H for day 28). ADCC was nearly twice as high in NK cells cultured in AI than in those cultured in RC and DS (Fig. 1E, H).

We tested batch-to-batch variation in the proliferation rate and cytotoxicity of NK cells cultured in two representative media. The remarkable proliferation-promoting effect of DS and the lower functionality of NK cells cultured in DS than of those cultured in AI were consistently observed for different human serum batches. Depending on the serum batch, NK cells cultured in DS showed slightly different fold expansion (Supplemental Data Fig. S1A–C), cytotoxicity (Supplemental Data Figure S1D–F), and ADCC (Supplemental Data Fig. S1G–I); however, the trends of higher fold expansion and lower killing potency than those of NK cells cultured in AI remained.

To identify the potential cause of the differences in functionality induced by the different culture media, cytotoxic effector molecules (granzyme B, perforin), activating receptors (DNAM-1, NKG2D, and CD16), natural cytotoxic receptors (NKp46 and NKp30), and an inhibitory receptor (NKG2A) were analyzed using flow cytometry (Fig. 2A–C). Although NK cells cultured in AI exhibited superior cytotoxicity to those cultured in the other media, their granzyme B and perforin expression levels were comparable with those of NK cells cultured in RC. Similar patterns were observed for the activating receptors and natural cytotoxic receptors: NK cells cultured in AI expressed comparable (DNAM-1, NKG2D, CD16, and NKp30) or significantly lower (NKp46) receptor levels than those cultured in RC or DS. NKG2A expression was the lowest in NK cells cultured in AI.

The extensive proliferation of NK cells cultured in DS and enhanced cytotoxicity of NK cells cultured in AI led us to explore the possibility of obtaining high numbers of highly cytotoxic NK cells by media shifting. NK cells were cultured in DS medium for the first 14 days and then in AI medium for 14 days (DS-AI) or vice versa (AI-DS). Their fold expansion, purity, and cytotoxicity were analyzed on day 28. The fold expansion of NK cells cultured in AI-DS and DS-AI was significantly higher than that of NK cells cultured in AI but lower than that of NK cells cultured in DS (Fig. 3A). The purity of NK cells cultured in DS-AI was significantly higher than that of cells cultured in AI (Fig. 3B). NK cells cultured in DS-AI showed better expansion and purity than those cultured in AI-DS. Interestingly, the medium in which NK cells were cultured in the late phase of expansion was found to determine cytotoxicity (Fig. 3C–E). The cytotoxicity of AI-DS- (or DS-AI)-cultured NK cells was comparable with that of DS- (or AI)-cultured NK cells, indicating that DS-AI- and AI-cultured NK cells exhibited superior cytotoxicity to AI-DS- and DS-cultured NK cells. These results indicate that culture medium shifting is advantageous for recovering NK cytotoxicity (Fig. 3A, C–E).

As the expression of cytotoxic effector molecules and activating and inhibitory receptors did not explain the enhanced functionality of NK cells cultured in AI, we considered the culture medium composition. The RC and DS media contained 10% serum (10% FBS for RC and 10% human serum for DS), whereas the AI medium was serum-free. As cells cultured in serum-free culture conditions (AI and DS-AI) exhibited significantly higher functionality than those cultured in serum-supplemented conditions (RC, DS, and AI-DS), we assumed that serum promotes proliferation but suppresses the killing potency of NK cells. To confirm this, we added 10% human serum to AI (AI+10% human serum) and observed the proliferation rate and functionality of the cells. The fold expansion of NK cells cultured in AI+ 10% human serum was significantly higher than that of NK cells cultured in AI on day 28 (Fig. 4A). Purity was not altered (Fig. 4B) by the addition of serum. NK cells cultured in AI+10% human serum exhibited significantly lower cytotoxicity than those cultured in AI (Fig. 4C–E).

To examine the contrasting effects of serum supplementation on NK cell expansion and functionality further, we removed the human serum from DS and added 10% ICSR instead. NK cells cultured in DS + 10% ICSR demonstrated significantly reduced fold expansion (Fig. 4F–G), whereas their functionality was significantly increased compared with that of NK cells cultured in DS (Fig. 4H–J). Lastly, we reduced the human serum level from 20% to 5% in DS medium. In line with the above results, when the human serum level was decreased, the fold expansion of the NK cells decreased, whereas their functionality increased (Supplemental Data Fig. S2A–E).

To examine the potential reasons for the enhanced killing potency of NK cells cultured in serum-free medium, we observed the killing behaviors of cells cultured in DS (with serum) and AI (serum-free) using live cell imaging. K562 target cells (green) killed by NK cells (unlabeled) exhibited extensive membrane blebbing and cell shrinkage (Fig. 5A, Supplemental Videos 1A and 1B). The time required for killing, i.e., the time passed between initial NK cell contact and onset of membrane blebbing, was measured for each NK cell that successfully killed target cells [17]. NK cells cultured in AI were more than four times faster in killing than those cultured in DS (Fig. 5B). Thus, the killing capacity of NK cells (the fraction of NK cells that killed more than one target cell during 4 hours of imaging) was significantly higher for cells cultured in AI than for those cultured in DS (Fig. 5C). These results indicate that the superior cytotoxicity of NK cells cultured in serum-free AI is owing to their rapid recognition and killing capabilities. As the live imaging results demonstrated the rapid killing capability of NK cells cultured in AI, we analyzed NK cell cytotoxicity over various periods (30, 90, 150, and 240 minutes) (Fig. 5D). The cytotoxicity of NK cells cultured in AI tended to saturate after 90 minutes, whereas that of NK cells cultured in AI+10% human serum saturated only after 150 minutes. The fold change in cytotoxicity of cells cultured in AI+10% human versus those cultured in AI was the highest at 90 minutes (1.6-fold) and the lowest at 240 minutes (1.1-fold). These results indicate that serum in the culture medium delays the decision time for target cell killing, thus reducing the killing potency of NK cells.