Various human cancer cell lines were used in this study, including those of HNSCC and NSCLCs. SNU-1066, SNU-1041, and SNU-1076 cell lines were purchased from the Korean Cell Line Bank (Seoul, Korea). Detroit 562 (CCL-138), FaDu (HTB-43), NCI-H1975 (CRL-5908), and NCI-H1650 (CRL-5883) cell lines were purchased from the American Type Culture Collection (ATCC, Manassas, Virginia, USA). The HN31 cell line was kindly granted by John F. Ensley, Wayne State University. SNU-1066, SNU-1041, SNU-1076, HN31, NCI-H1650 and NCI-H1975 cell lines were cultured in Roswell Park Memorial Institute Medium (RPMI 1640); Gibco, Invitrogen) supplemented with 10% heat‐inactivated Fetal Bovine Serum (FBS; Gibco) and 10 µg/mL gentamycin. Detroit 562 and FaDu cell lines were maintained in Eagle’s Minimum Essential Medium (EMEM) supplemented with 10% heat‐inactivated FBS and 1% penicillin/streptomycin (Gibco).

Allogeneic NK cell lines engineered to express the high-affinity CD16 allele were used. NK-92-CD16 cells (purchased from the ATCC, PTA-8836) were maintained in alpha minimum essential medium supplemented with 25% FBS, 0.2 mM inositol, 0.1 mM 2­mercaptoethanol, and 0.02 mM folic acid. NK-92-CD16 cells were cultured with recombinant human interleukin (IL)-2 (200 U/mL).

All cell lines were cultured for less than 6 months and tested negative for mycoplasma using a detection kit (iNtRON Bio Technology, Seongnam, Korea) prior to experiments. The human cancer cell lines were authenticated by CosmoGenetceh (Seoul, Republic of Korea) using short-tandem repeat profiling.

Peripheral blood mononuclear cells (PBMCs) of healthy donors were isolated from leukoreduction system chambers using Ficoll density gradient sedimentation. PBMCs were rested overnight in RPMI 1640 medium (Gibco) and activated by recombinant human IL-15 (PeproTech, Rocky Hill, New Jersey, USA) 1 ng/mL with the RPMI1640 medium for 3 days before an in vitro ADCC assay.

To perform mAb-mediated ADCC in vitro experiments, the IgG1 isotype kappa control from human myeloma plasma was used (Sigma-Aldrich, St. Louis, Missouri, USA). The following mAbs were used for anti-PD-L1 mAb. Atezolizumab (Selleck Chem, Houston, Texas, USA) is IgG1 mAb with a modified Fc region designed to limit ADCC. IMC-001 (kindly provided by ImmuneOncia, Yongin, Korea) is a fully human PD-L1 recombinant IgG1 mAb that can induce ADCC. An anti-hPD-L1-hIgG1 (InvivoGen) is atezolizumab-based wild-type IgG1 that displays ADCC.

Cells were stained with fluorescent-labeled mAbs for the following cell-surface markers: anti-CD274-PE (clone MIH1, RRID:AB_647198), isotype control-PE (MOPC-2, RRID:AB_396091), anti-CD3-FITC (UCHT1, RRID:AB_395739), anti-CD56-APC (B159, RRID:AB_398601), anti-CD107a-PE (H4A3, RRID:AB_396135), anti-CD107a –APC (H4A3, RRID:AB_1645722) (BD Biosciences, San Jose, California, USA); anti-CD16-APC (CB16, RRID:AB_2016663), anti-CD16-PE-Cy7 (CB16, RRID:AB_10714839) (Thermo Fisher Scientific, Waltham, Massachusetts, USA); anti-CD56-PE-Vio770 (REA196, RRID:AB_2726091; Miltenyi Biotech, Bergisch Gladbach, Germany). For CD107a degranulation, a protein transport inhibitor, Golgi-Stop (#554724, BD Biosciences) was used. Fixable viability dye eFluor 506 (#65-0866-14, Thermo Fisher Scientific) was used to exclude dead cells. A purified anti-human CD16 antibody (3G8, RRID:AB_314202 BioLegend) was used for blocking assay.

Surface molecule expression was measured by flow cytometry. Cells were incubated with fluorochrome-conjugated antibodies for 20 min at 4°C. Data were acquired by a FACS Calibur or FACS Canto II (BD Biosciences) and analyzed with FlowJo software (Tree Star, Ashland, Oregon, USA). We presented mAb-stained cells as percentage of positive cells or mean fluorescence intensity (MFI). Analyzing PD-L1 expression on the surface of tumor cell lines, the MFI values ​​of isotype control vary slightly between cell lines. Therefore, we defined the difference between control and PD-L1-stained cancer cell lines as delta mean fluorescence intensity (ΔMFI) and use high PD-L1 expression according to the ΔMFI rather than the absolute MFI.

Various cancer cell lines were used as target cells, including HNSCC and lung cancer cell lines. Standard ⁵¹Cr-release and CD107a degranulation assays were performed to evaluate the in vitro ADCC efficacy of three groups: control group, anti-PD-L1 mAb without ADCC (atezolizumab) group, and anti-PD-L1 mAbs with ADCC group (IMC-001 and anti-hPD-L1-hIgG1 (atezolizumab with wild-type FcR binding site, hPD-L1mAb)).

⁵¹Cr-release assays were performed to evaluate cytotoxicity induction of NK-92-CD16 cells via ADCC. The method is described in a previous study.17 Cancer cell lines evaluated as targets are indicated in the figures as appropriate. Target cells (1×10⁶) were labeled with 50 µCi for 1 hour at 37°C, then 2.5×10⁵ cells were washed twice using growth media. Target cells were cultured with 10 µg/mL of the indicated antibodies for 30 min at 37°C and washed again to avoid contact interference between NK and target cells. Target cells were seeded at 5×10³ cells/well into U-bottom 96-well plates and coincubated with effector cells at the indicated effector-to-target (E:T) ratios for 4 hour at 37°C in triplicate. Following co-culture, 75 µL supernatant from each well was transferred to 96-well PP 1.2 mL cluster tubes for gamma counting. A WIZARD2 Automatic Gamma Counter (PerkinElmer, Waltham, Massachusetts, USA) was used to calculate radioactivity.

CD107a expression is commonly used to determine NK cell functional and cytotoxic activity.18–20 Thus, a CD107a degranulation assay was performed using PBMCs from healthy donors with an E:T ratio of 1:1. PBMC was activated by IL-15 for 3 days. PBMC and target cell densities were adjusted to 2×10⁶ cells/mL. After washing target cells, 5×10⁵ cells were cultured with 10 µg/mL IgG1 isotype control, atezolizumab (anti-PD-L1 mAb without ADCC), or IMC-001 or anti-hPD-L1-hIgG1 (anti-PD-L1 mAbs with ADCC) for 30 min at 37°C and 5% CO₂. Cells were then washed with growth media to eliminate excess antibody. For 1:1 E:T ratio, 2×10⁵ cancer cells and PBMCs were incubated for 1 hour at 37°C in U-bottom 96-well plates. Afterwards, 10 µL CD107a antibody was added to each PBMC and incubated for 4 hours at 37°C. Target cells and PBMCs were centrifuged at 15 g for 3 min at room temperature. After incubating for 1 hour, 5 µL/well of GolgiStop solution was added to each well. Whole samples were transferred to FACS tubes then stained with fluorescently labeled mAbs for NK cell surface markers. Stained cells were maintained at 4°C and protected from light until flow cytometry acquisition.

Genotyping was performed using PCR. First, an Exgene Cell SV Kit (Geneall Biotechnology, Seoul, Korea) was used to extract genomic DNA (gDNA) from PBMCs. gDNA was amplified with specific primers (5′-ATA TTT ACA GAA TGG CAC AGG-3′, 5′- GAC TTG GTA CCC AGG TGG AA-3′) and Green 2X premix (Applied Biosystems, Waltham, Massachusetts, USA) using the GeneAmp PCR System (Applied Biosystems). The PCR program consisted of an initial 5 min denaturation step at 95°C followed by 35 cycles for 30 s at 94°C, 30 s at 56°C, 1 min at 72°C, and 7 min at 72°C. Images were captured with the Gel Logic 200 Imaging System (Kodak, Rochester, New York, USA). PCR products were purified using the PCR Purification Kit (Invitrogen, Carlsbad, California, USA) and sequenced by bidirectional Sanger sequencing using specific primers (5′-ATA TTT ACA GAA TGG CAC AGG-3′, 5'-ATG CTG CAG AGT GAA TGA CAC-3').

HN31 cells were seeded on gelatin-coated coverslips and placed in a cell culture incubator for 12 hours to allow cells to adhere and spread on the surfaces. Then, the HN31 cells on coverslips were treated with various antibodies (10 µg/mL) for 30 min in the cell culture incubator, washed three times with cell culture media, and loaded in a magnetic chamber (Chamlide CF, Live Cell Instrument, Korea) for live cell imaging. The magnetic chamber was mounted on a microscope stage equipped with a Chamlide TC incubator system (Live Cell Instrument), which maintains a cell culture condition (37℃, 5% CO₂), and NK-92-CD16 cells were added to the chamber. Time-lapse imaging was initiated 15 min after the addition of NK-92-CD16 cells. Differential interference contrast (DIC) images were acquired every 5 min for 4 hour. A modified Olympus IX 83 epifluorescence microscope with a ×40 (UPlanFLN, NA=1.30) objective lens and an ANDOR Zyla 4.2 sCMOS camera was used for imaging experiments. The microscope was automatically controlled by Micro-manager. Acquired images were analyzed and processed with Image J.

Data are shown as mean±SD. Data were analyzed by GraphPad Prism software V.7.0 (GraphPad Software, La Jolla, California, USA). Statistical significance in multiple groups was compared by one-way analysis of variance. Paired groups were compared by a paired two-tailed Student’s t-test. Two-sided p<0.05 was considered significant.

To find target cells with high PD-L1 expression, we screened PD-L1 mRNA expression levels in cancer cell lines using the Cancer Cell Line Encyclopedia (https://portals.broadinstitute.org/ccle/)21 database (figure 1A). PD-L1 protein expression at the cell surface in different tumor types and human cancer cell lines was then analyzed by flow cytometry (figure 1B,C). High PD-L1 cell lines were defined by ΔMFI≥5. SNU-1076, FaDu, HN31, and NCI-H1975 cell lines were classified as high PD-L1 cell lines, and SNU-1041, SNU-1066, Detroit 562, and NCI-H1650 cell lines were classified as low PD-L1 cell lines.

We performed the a standard ⁵¹Cr-release assay to measure the ADCC-mediated cytotoxicity of NK-92-CD16 cells against HNSCC (figure 2A) or lung cancer cell lines (figure 2B) in the presence of anti-PD-L1 mAbs. To distinguish NK-92-CD16-induced cytotoxicity based on PD-L1 expression in target cells, HNSCC and NSCLC cell lines were grouped as ‘PD-L1 low’ or ‘PD-L1 high’.

NK-92-CD16-induced cytotoxicity was greater against PD-L1 high cell lines than PD-L1 low cell lines. In combination with IMC-001, an anti-PD-L1 mAb with preserved ADCC, NK-92-CD16 cytotoxicity was slightly increased in PD-L1 low cell lines, except for Detroit 562 cells, in which IMC-001 had no beneficial effect on ADCC with NK-92-CD16 cells. Interestingly, the other anti-PD-L1 mAb anti-hPD-L1-hIgG1 demonstrated significant cytotoxicity improvement for NK-92-CD16 cells against HNSCC and NSCLC cell lines, irrespective of PD-L1 expression.

To evaluate the contribution of anti-PD-L1 mAbs to ADCC-mediated NK cell cytotoxicity against HNSCC and NSCLC cell lines, we generated two groups: a control group (isotype and atezolizumab) and an anti-PD-L1 mAb with ADCC group (IMC-001 and anti-hPD-L1-hIgG1) (figure 2). For the control group, we found no significant increase in NK-92-CD16 cytotoxicity mediated by ADCC. However, NK-92-CD16-induced cytotoxicity significantly improved in the anti-PD-L1 mAbs with the ADCC group for PD-L1-positive cell lines.

Together, these results imply that PD-L1 expression in cancer cell lines correlated with NK cell cytotoxicity against tumor cells and that even low PD-L1 expression could be targeted for NK-92-CD16 cytotoxicity with anti-PD-L1 mAbs.

We confirmed that NK cell-induced cytotoxicity could be enhanced by ADCC using anti-PD-L1 mAbs in PD-L1-positive cancer cell lines. Next, we aimed to extend these findings using primary NK cells in the PBMCs. PBMCs consist of subsets of immune cells such as monocytes, immature neutrophils, macrophages, T cells and B cells, in addition to NK cells. Some of these immune cells can mediate antibody dependent cytotoxicity, such as antibody-dependent cellular phagocytosis (ADCP). ADCP initiated by macrophages must be differentiated from monocytes with macrophage colony-stimulating factor for a week.22 Unlike ADCP, it is known that ADCC classically requires NK cells. Therefore, we used PBMCs but not purified NK cells. The reason why we used PBMCs but not purified NK cells for ADCC assay was that the proportion of NK cells in PBMC is 5-10%, not much. In addition, it is only necessary to confirm the reaction within the same donor-derived NK cells. For NK cells, CD3 and CD56 molecular markers were stained to confirm NK cell-mediated ADCC analysis of NK cells. The gating strategy employed for this study is described in online supplementary figure S2, and the percentage of NK cells in the PBMCs is stated in online supplementary table 1.

Similar to previous experiments, we evaluated ADCC in vitro using NK cells in combination with anti-PD-L1 mAbs against human cancer cell lines classified by PD-L1 expression level. Figure 3A illustrates NK cell cytotoxicity against PD-L1 low cell lines. For the PD-L1 low cell line NCI-H1650, ADCC sensitivity using IMC-001 was not observed in combination with resting or activated NK cells. However, other PD-L1 low cells, including SNU-1041 and Detroit 562, showed increased ADCC in both resting and activated NK cells induced by IMC-001 and anti-hPD-L1-hIgG1 (figure 3A). Moreover, cytotoxicity induced by NK cells was increased significantly when PD-L1 high cancer cell lines were opsonized with anti-PD-L1 mAbs with preserved ADCC (figure 3B). Interestingly, FaDu cells (high PD-L1) treated with IMC-001 showed no statistical difference in NK cell-induced cytotoxicity when cocultured with resting cells, likely because of biological variation between NK cell donors.

NK cell-induced cytotoxicity in PD-L1-positive human cancer cell lines treated with anti-hPD-L1-hIgG1 improved regardless of PD-L1 expression and to similar degrees, except for NCI-H1650 (low PD-L1) with resting primary NK cells. Taken together, these findings provide crucial insights into ADCC of activated NK cells and support the previous results. NK cells could enhanced the cytotoxicity through ADCC induced by anti-PD-L1 mAb in PD-L1-positive human cancer cell lines treated with anti-PD-L1-preserved ADCC. The enhanced ADCC of NK cells occurs in a CD16-dependent manner (online supplementary figure S4).

To confirm the effects of NK cells FCGR3A gene polymorphism on ADCC induced by IMC-001, one of the anti-PD-L1 mAb groups, we revisited the results of the CD107a degranulation assays (figure 3). NK cells from six or more healthy donors were genotyped for FCGR3A polymorphism by PCR-based Sanger sequencing. We showed that IMC-001 mediated ADCC with NK cells of FCGR3A genotyped donors (online supplementary figure S3).

We compared the cytotoxicity of NK cells with V/V or V/F donor and F/F genotyped donor toward the HNSCC (figure 4A) or NSCLC cell lines (figure 4B). For the comparison, we defined the difference between control and IMC-001 percent of CD107a positive (CD107a⁺) NK cells in value as delta CD107a positive (ΔCD107a⁺ (% of NK cells)). The one-tailed unpaired t-tests were used to determine the significant differences. The figure 4 shows that the IMC-001-mediated ADCC of primary NK cells appeared to be affected by the presence of the FCGR3A high-affinity allele (V/V or V/F genotype; valine (V) phenylalanine (F)).

Except for FaDu and NCI-H1650 that did not experience increased ADCC, the cytotoxicity of resting primary NK cells with the FCGR3A high-affinity genotype demonstrated significantly increased ADCC-mediated cytotoxicity. We next assessed the cytotoxicity of NK cells activated by IL-15. The cytotoxicity of FCGR3A high-affinity genotype NK cells which was activated by IL-15 revealed much higher anti-PD-L1 mAb-mediated ADCC lysis of tumor cells than donors with the F/F genotype, except for NCI-H1650 (figure 4A,B).

In summary, these results indicate that primary NK cells with FCGR3A high-affinity allele (V/V or V/F genotype) affected the sensitivity of ADCC through IMC-001. Primary NK cells with the FCGR3A high-affinity genotype showed a more pronounced ADCC increase in the presence of IMC-001.

Live cell imaging was performed to directly observe ADCC of NK cells and quantitatively evaluate cytotoxicity in a single-cell level.23 24 NK-92-CD16 cells were added to HN31 cells treated with various antibodies and observed by time-lapse imaging (figure 5A and online supplementary videos S1–4). HN31 cells (marked with white boundaries in figure 5A) and NK-92-CD16 cells (marked with yellow boundaries in figure 5A) exhibited distinct behaviors: HN31 cells spread on the substrates with minimal translocation, whereas NK-92-CD16 cells dynamically migrated and attached on HN31 cells. Some HN31 cells contacted by NK-92-CD16 cells detached from the substrates and rounded, meaning the HN31 cells were killed by the NK-92-CD16 cells.24 Only a fraction of NK cells contacting isotype-treated or atezolizumab-treated cancer cells exerted cytotoxicity, whereas the majority of NK cells contacting IMC-001- or anti-hPD-L1-hIgG1-treated cancer cells did. Of note, NK cells contacting atezolizumab-treated cancer cells undergo extensive morphology changes compared with NK cells contacting isotype-treated cancer cells, indicating NK cells recognized Fc region of atezolizumab but showed no cytotoxicity.

Next, we quantitatively analyzed cytotoxicity of NK cells in each case. Overall cytotoxicity was measured by evaluating the percentage of dying/dead HN31 cells throughout the time-lapse imaging (figure 5B). For cancer cells undergoing apoptosis, membrane blebbing proceed detachment from the substrates; thus, cancer cells exhibiting membrane blebbing were counted as a dying cells.23 25 The overall cytotoxicity of atezolizumab group was comparable to that of isotype group. In contrast, the overall cytotoxicity of IMC-001 group was ~5 times higher than that of isotype group, and anti-hPD-L1-hIgG1 group exhibited the highest overall cytotoxicity. These results agree well with the ⁵¹Cr-release assay that measures bulk population-level cytotoxicity. To assess cytotoxicity in a single-cell level, time for killing, the duration between NK cell engagement to cancer cells and the onset of cancer cell blebbing, was measured for each NK cell successfully killed cancer cells (figure 5C). For isotype and atezolizumab groups, time for killing was ~150 min, whereas for IMC-001 and anti-hPD-L1-hIgG1 groups, time for killing was <40 min (figure 5C). These results mean ADCC-mediated by anti-PD-L1 accelerate cancer cell killing rate of NK cells. Accelerated tumor cell killing mediated by ADCC frequently lead to serial killing of cancer cells, which means one NK cell kills many cancer cells. To assess serial killing capability of NK cells, the number of cancer cells killed by an NK cell was measured (figure 5D). More than 80% of NK cells in isotype or atezolizumab groups failed to kill any cancer cells or killed 0 cancer cell. In contrast, the majority of NK cells in IMC-001 or anti-hPD-L1-hIgG1 groups killed more than two cancer cells. Of note, some NK cells in an anti-hPD-L1-hIgG1 group killed up to six cancer cells. Taken together, single-cell level cytotoxicity assay revealed that anti-PD-L1-mediated ADCC of NK cells substantially accelerated cancer cell killing and enhanced serial killing.