Characterization of ex vivo-expanded pNK cells for 14 days
The mean number of total mononuclear cells in the leukapheresis products from four donors was 2.11 ± 0.24 × 1010 cells, and the mean percentage of NK cells was 13.2 ± 4.15% (range, 7 ~ 18.7%). After isolation, an average of 1.05 ± 0.04 × 109 cells (range, 1 ~ 1.1 × 109) with an average of 94.9 ± 2.66% CD3−CD56+cells were obtained. The isolated pNK cells were divided into 6× 107 cells/vial and cryopreserved.
Each vial of pNK cells from the four donors was expanded for 14 days with a mixture of antibodies and cytokines without feeder cells. The total cell numbers were measured on Days 0, 6, 11, and 14 during expansion. On Day 14, the expansion fold was 175 ± 43-fold (Fig. 1a) and the viability of expanded cells was 93 ± 1% (Fig. 1b). We analyzed the proportion of cells before culture(D0), after 14 days of culture(D14), and after cryopreservation (D14 cryo).
The proportion of CD3−CD56+ pNK cells was 99.5 ± 0.4%, 99.7 ± 0.2%, and 99.9± 0.1% on D0, D14, and D14 cryo, respectively. CD56+CD16+ pNK cells were 94.7 ± 3.9%, 95.2 ± 2.1%, and 96.7 ± 3% on D0, D14, and D14 cryo, respectively (Fig. 1C). On D14, CD3+ T cells, CD14+ monocytes, and CD19+ B cells were 0%, 0.8 ± 0.3%, and 0.3 ± 0.1%, respectively. After freezing, no change was observed in cell distribution (Fig. 1d). On culturing pNK cells until 21 days, the total cell number continuously increased (4.9 ± 0.2 × 1010 cells at Day 21) and a high viability was maintained throughout (Additional file: Fig. S1a, b). The pNK cells at Day 21 also retained a high proportion (99.5%) of CD3−CD56+ cells (Additional file: Fig. S1c).
Phenotypic comparisons of resting and expanded pNK cells
We evaluated various receptors of pNK cells on D0, D14, and D14 cryo using fluorescence-activated cell sorting (FACS) analysis. On D14, the expression of activating receptors such as NKG2D, NKp30, NKp44, and NKp46 was significantly increased, CD16 was maintained at a high level (95%), and NKG2C was increased in some donors as compared to those on D0. OX40 and 41BB slightly increased (Fig. 2a). Co-receptors such as CD94, DNAM-1, and 2B4 were significantly expressed from D0 to D14 (Fig. 2b). Before and after freezing, receptors such as CD16, NKG2C, NKp30, OX40, CD94, DNAM-1, and 2B4 showed an equivalent expression. NKG2D, NKp44, and 41BB showed slightly decreased expression levels; however, the changes in NKG2D and 41BB expression were not statistically significant. NKp46 expression, however, was significantly decreased. Inhibitory receptors such as CD158b and CD158e decreased on D14 than on D0, but only NKG2A slightly increased and CD158f did not change. Interestingly, the major immune check point protein, programmed cell death-1 (PD-1), did not increase in our culture medium (Fig. 2c). Additionally, some transcription factors involved in NK cell functions were analyzed. Eomes and GATA3 significantly increased during expansion while T-bet and E4BP4 were high since Day 0. The proliferation marker, Ki67, was significantly increased during expansion (Additional file: Fig. S2a). The expression of chemokine receptors such as CCR2, CCR5, and CCR6 significantly increased on D14 as compared to that on D0, CCR3 and CCR7 did not change, CCR4 increased, and CCR6 slightly decreased in some donors (Fig. 2d). CXCR3 and CXCR6 increased whereas CXCR1 and CXCR4 decreased after expansion (Additional file: Fig. S2b). Expanded pNK cells significantly expressed adhesion molecules, such as CD2, CD11a, CD18, ITGA1, and ITGB7, except CD62L (Additional file: Fig. S2c). CD69, an activation marker, was expressed shortly after NK cell activation, and it increased after expansion while CD57, a senescence marker [20, 21], decreased after expansion (Additional file: Fig. S2d). Cytolytic proteins such as perforin and granzyme B were significantly expressed from D0 to D14, and the levels were maintained after freezing. The expression of IFN-γ was slightly increased. CD107a degranulation did not change significantly, but it was increased after cryopreservation (Fig. 2e).
Finally, we evaluated the cytotoxicity of pNK cells against K562 cells at D0, D14, and D14 cryo. Cytotoxicity against K562 was significantly increased on D14 than on D0 (14.5 ± 3.5% vs. 93.2 ± 4.1% at an E:T ratio 10:1, P < 0.05), and it was maintained after freezing at an E:T ratio of 10:1 (Fig. 2f). The cytotoxicity of 21 days-expanded pNK cells against K562 was similar to that of 14 days-cultured pNK cells at E:T ratios of 5:1, 10:1, and 20:1 (Additional file: Fig. S1d).
Gene expression signature of ex vivo-expanded pNK cells using RNA sequencing
We performed RNA sequencing to employ an unbiased approach to identify the altered gene expression enhancing the functions of ex vivo-expanded pNK cells for 14 days, and we transcriptionally compared Day 0 and Day 14 pNK cells. In total, 5471 genes were differentially expressed between Day 0 and Day 14 pNK cells with 2855 upregulated (Category I) and 2616 downregulated genes (Category II) on Day 14 as compared to those on Day 0 (Fig. 3a). Gene Ontology Biological Process (GOBP) analysis results revealed that the upregulated DEGs were enriched in DNA metabolic processes, cell cycle regulation, DNA biosynthesis, nuclear division regulation, and cytokine regulation (Fig. 3b).
The NK activation markers, NCR2 (1064-fold), NCR3 (5-fold), KIR2DS4 (2-fold), and TNFRSF9 (2-fold), were increased on Day 14 pNK cells than on Day 0 pNK cells (Fig. 3c). Concomitantly, the genes related to cytotoxicity (TNFSF10, GZMK, GZMA, PRF1, and FASLG), chemokines (CCR1, CCR2, CCR5, and CCR6), anti-apoptosis (BAX, BCL2, and BCL2L1), and cell proliferation (MKI67, PCNA, FOXM1, AURKA, and PLK1) were enriched on Day 14 (Fig. 3c). Several genes involved in immune suppression, including CD22, IL-10, TGFβ, and TGFβR3, were also decreased on Day 14. PD-1, which was only slightly expressed (1%) on Day 0 and Day 14 pNK cells using FACS analysis (Fig. 2c), was not detected in the RNA sequencing data. Among the significantly expressed transcripts, CD40L and CCR5 were increased by 12,018-fold and 49-fold on Day 14 pNK cells than on Day 0 pNK cells (Fig. 3d).
To investigate whether CD3−CD56+ selection leads to gene expression alterations, we performed RNA sequencing on Day 0 and Day 14 samples of PBMCs without CD3−CD56+ selection (non-selectively expanded pNK cells) after an ex vivo expansion of cells using the same manufacturing method (Additional file: Fig. S3a-3c). Similar to the results of selectively-expanded pNK cells (Fig. 3), NK activation markers, cytotoxicity-related genes, chemokines, anti-apoptosis, and cell proliferation were enriched in non-selectively expanded pNK cells on Day 14 than on Day 0 (Additional file: Fig. S3). However, the increase of some activating receptors, chemokines, and anti-apoptotic genes, such as NKp44 (1064-fold), TNFSF10 (24.5-fold), CCR2 (130.1-fold), CCR3 (1427.5-fold), CCR5 (49.9-fold), CCR6 (124.1-fold), and BCL2 (10.5-fold), was more significant in selectively expanded pNK cells than in non-selectively expanded pNK cells (6.3-, 9.7-, 6.6-, NA, 16.7-, NA, and 1.7-fold, respectively) (Fig. 3 and Additional file: Fig. S3).
These results indicated that the NK cell culture method from PBMCs with or without CD3−CD56+ selection increased the expression of various activating NK receptors and cytotoxicity-related genes and when pNK cells were cultured with CD3−CD56+ selection, CD40L, NKp44, and migration-related genes were more significantly expressed.
Cytotoxicity of expanded pNK cells against various ovarian and breast cancer cell lines
Next, the cytotoxicity of expanded pNK cells was evaluated in vitro against human ovarian cancer cell lines. Various tumor cell lines displayed different levels of susceptibility to cytolysis to expanded pNK cells (Fig. 4a). To understand this varied susceptibility, expression of different ligands for NK receptors was analyzed on cancer cell lines. In line with the cytotoxicity data, the pNK-sensitive A2780cis cells showed high expression levels of NKG2D ligands, such as MICA, ULBP1, ULBP4, and NKp30 ligand (B7-H6), as compared to the pNK-non-sensitive SKOV3 cells.
In contrast, the expression levels of inhibitory KIR ligand (HLA-ABC) were significantly lower in A2780cis cells than in pNK-non-sensitive cancer cells (A2780 and SKOV3 cells) (Fig. 4b, c). As CD158b (KIR2DL2/3) was increased in expanded pNK cells (Fig. 2c), we investigated the effect of HLA-C1 or C2 expression on the cytotoxicity of pNK cells. Gene expression levels of both HLA-C1 and HLA-C2 were significantly decreased in pNK-sensitive A2780cis cancer cells (approximately 0.17-fold of the level in A2780 cells) (Fig. 4d) although both A2780 and its cisplatin-resistant A2780cis subline possessed HLA-C1/C2 genotypes (Additional file: Table S1). To elucidate the contribution of HLA-C1 and -C2 expression in NK alloreactivity, we introduced siRNAs for HLA-C in A2780 cells. After confirming that mRNA and protein expression of both HLA-C1 and HLA-C2 significantly decreased after HLA-C siRNA transfection (Fig. 4e, f), we verified that the cytotoxicity of pNK cells was significantly increased in A2780 cells after HLA-C siRNA transfection (20% vs. 40% at an E:T ratio of 10:1, P < 0.001, Fig. 4g). Additionally, we also found that pNK cytotoxicity was increased after KIR2DL3 blockade rather than other inhibitory KIR blockade (Fig. 4h). Collectively, these results suggested that inhibitory KIR-KIR ligand interaction, especially KIR2DL3-HLAC1, effectively contributed to pNK cell alloreactivity. When we examined HLA-G expression in these ovarian cancer cell lines to assess its possible inhibitory function on our pNK cells, we did not find the increase of HLA-G expression (Additional file: Fig. S4).
To evaluate the role of activating NK receptors, a cytotoxic assay was performed with expanded pNK cells in the presence of blocking antibodies specific to NKG2D, NKp30, NKp44, NKp46, and DNAM-1. Although blocking a single receptor alone slightly affected the cytotoxicity, NKG2D blockade mostly inhibited NK killing as compared to any other single blocking. In contrast, blocking multiple receptors led to a substantial reduction in cytotoxicity. Particularly, blocking all four receptors led to approximately 90% blocking in A2780 and 60% blocking in A2780cis cells than in IgG control (Fig. 4i).
We also evaluated the cytotoxicity and IFN-γ secretion of expanded pNK cells against human breast cancer cell lines (Additional file: Fig. S5a & S5b), revealing that MCF-7 showed the highest sensitivity to our pNK. Unlike ovarian cancer cell lines, expression levels of NKG2D ligands were similar in both pNK-sensitive target cells (MCF-7) and pNK-non-sensitive target cells (MDA-MB-231) (Additional file: Fig. S5c). However, pNK-non-sensitive cancer cells (MDA-MB-231and MDA-MB-468) expressed high levels of HLA-C2 (Additional file: Fig. S5d) whereas pNK-sensitive cancer cells (MCF-7) expressed low levels of both HLA-C1 and HLA-C2, indicating the contribution of inhibitory KIR-KIR ligand mismatch in pNK cytotoxicity. In the blocking analysis of MCF7 cancer cells, blocking a single receptor alone except for NKp46 affected the cytotoxicity as much as blocking multiple receptors (Additional file: Fig. S5e).
Taken together, these results demonstrated that expanded pNK cells have the ability to kill a wide range of tumor cells, and the process is not only regulated by activating receptor-ligand interaction but also inhibitory receptor-ligand interaction.
In vivo distribution of ex vivo-expanded pNK cells
To investigate the in vivo distribution of cryopreserved pNK cells, we intravenously administered pNK cells (1 × 107 cells/mouse) to BALB/c nude mice, and the mice were sacrificed at each post-injection time point until 168 h (Fig. 5a). Intravenously-injected pNK cells first appeared in the blood, lungs, spleen, kidney, liver, heart, and ovary at 2 h, where they resided for 24 h and then gradually disappeared until 48 h (Fig. 5b).
To evaluate pNK cell distribution in tumor-bearing immune incompetent mice, CFSE-stained pNK cells (1 × 107 cells/mouse) were intravenously injected into A2780cis tumor-bearing NSG mice after 40 days of tumor inoculation, and mice were sacrificed at 2, 24, and 72 h after pNK cell injection (Fig. 5c). Intravenously-injected pNK cells first appeared in the blood, lungs, spleen, kidney, liver, heart, ovary, and tumor at 2 h, and they were detected in the blood, lungs, spleen, and tumor until 72 h after injection (Fig. 5d, e). At 72 h, we observed 997 ± 343, 9657 ± 3634, and 7108 ± 3580 pNK cells in the blood, lungs, and spleen using qPCR. We also found several pNK cells (6.22 ± 2.28, 7.96 ± 1.39, and 2.92 ± 0.27 cells/cm2 at 2, 24, and 72 h, respectively) in the tumor sections via counting of CFSE-stained pNK cells.
In vivo therapeutic effects of ex vivo-expanded day 14 pNK cells against human ovarian cancer cell line xenograft tumors
To confirm the therapeutic efficacy of cryopreserved pNK cells in vivo, we established A2780cis xenograft on NSG mice. On Day 3, following subcutaneous tumor inoculation, mice were intravenously injected with 1 × 107 cells/mouse pNK cells twice a week or intraperitoneally injected with cisplatin (1 mg/kg) once a week (Fig. 6a). In line with in vitro results, on Day 40, pNK cells induced a significant inhibition of A2780cis tumor progression (76%) as compared to the control whereas cisplatin treatment did not show a significant effect (16% reduction) (Fig. 6b, c). When we sacrificed the mice, the serum levels of IFN-γ were elevated, and they were 1.6-fold higher in mice treated with expanded pNK cells than in the control group (Fig. 6d). To evaluate the effect of HLA-C expression in cancer cells on the anti-tumor effects of pNK cells in vivo, we performed in vivo efficacy tests and compared the results with those of subcutaneous (s.c.) xenograft models in NSG mice using an HLA-C-high ovarian cell line (A2780) and its cisplatin-resistant subline (A2780cis), in which HLA-C expression was notably lower than that in its parental cell line (A2780) (Additional file: Fig. S6). To compare the models under similar conditions as far as possible, we injected pNK cells (1 × 107 cells, weekly) after the tumors reached a size of 90–100 mm3 in both models, and tumor size was monitored until the mice were sacrificed after 15 days. The tumor growth in HLA-C-high A2780 xenografts was inhibited compared to that in the control (50% reduction compared to that in the control); however, the inhibition was notably lesser than that observed in HLA-C-low A2780cis (80% reduction compared to that in the control; P * < 0.05). These results were consistent with the in vitro results, in which the cytotoxicity of our pNK cells to A2780cis was markedly higher than that to A2780 (79.13 ± 2.79% vs 27.01 ± 6.95% at E:T ratio = 10:1). Taken together, these data suggest that the growth of ovarian cancer cells resistant to cisplatin was significantly inhibited by pNK treatment and this effect could be related to the low HLA-C expression in the cisplatin-resistant cells.