Background
Natural killer (NK) cells are appealing cells for immunotherapy because they are very potent effector lymphocytes of the innate immune system that can attack and kill many different tumor target cells without prior sensitization [
1‐
3]. Studies show that some cytotoxic immune cells, including T- and NK cells, are capable of killing multiple target cells, sequentially, in a process termed serial killing [
4,
5]. Moreover, serial killer cells are often faster at delivering lytic hits and inducing target cell death than non-serial killer NK cells [
6]. Unfortunately for adoptive immunotherapies, many NK cells kill only once, with less than 30% capable of serial killing [
7‐
9]. While the number of target cells killed per effector varies by study, it is consistently noted that a minority of NK cells is responsible for the majority of killing events. Furthermore, there are challenges to obtaining sufficient numbers of functionally active NK cells from a patient’s blood because, despite being widespread throughout the body, NK cells represent just 2–18% of lymphocytes in human peripheral blood and it is challenging to obtain sufficient numbers of NK cells needed to overwhelm the number of tumor cells [
10,
11].
To readily obtain the large numbers of NK cells needed, immortalized cytotoxic cell lines have been established from patients with NK-cell cancers; however, little is known about their serial killing capacities. NK-92 are one of over eight available NK cell lines (Additional File
1: Table S1) and have reproducible cytotoxicity to a variety of tumor types [
10,
12,
13], even under hypoxic conditions [
14]. As of February 2022, NK-92 cells are the only NK cell line approved by the FDA for clinical trials and have been the subject of over 550 publications cited in PubMed USA. In addition to their innate activities, NK-92 cells can also be genetically manipulated to express receptors that recognize specific tumor antigens and that augment therapeutic monoclonal antibodies through antibody-dependent cellular cytotoxicity (ADCC). These NK-92 cells and CAR-NK-92 variants are immediately available and more affordable than current CAR-T-cell therapy [
15]. In fact, haNK (NK-92) cells, engineered to express the high affinity CD16A allele (in order to recognize tumor cell-bound monoclonal antibodies), were tested in combination with anti-PD-L1 antibody, avelumab, and have now been further modified to also express a PD-L1-specific chimeric antigen receptor [
16,
17]. NK-92 cells have been infused into patients with advanced cancers, resulting in clinical benefits with limited side effects. Additionally, NK-92 cells are being tested in several clinical trials in four different countries and for patients with a range of malignancies, including leukemia, glioblastoma, and melanoma [
18‐
21]. It is important to assess what can happen to these cells in vivo, following transfer, a challenging issue that has been addressed so far only by monitoring cells circulating in the patients’ blood [
19]. In this report, we assessed in vitro
, potential hazards to NK-92 cell serial killing that could occur in vivo after adoptive transfer, including losses of cytotoxic serial capacity following irradiation, ligation of NK-92 cell Fas by cells residing in the tumor, as well as vulnerability of the NK-92 cells to attack by blood primary NK cells.
As far as the authors are aware, we are the first to observe killing frequencies [
5] (KF) > 1 by NK-92 cells using standard release assays (presented at American Association of Immunologists annual meeting, 2021). We monitored this serial killing to predict potential losses of activity to the cells during the time that the cells remain viable after adoptive transfer. Our data demonstrate several potential complications that would result in losses to therapeutic efficacy in vivo when NK-92 cells become impaired following irradiation. It was previously reported that when NK-92 cells were irradiated with 10 Gy, NK-92 cell proliferation was prevented and cytolytic activity was substantially conserved within the live cells remaining 1 day following irradiation [
22]. We found, however, that NK-92 cell serial killing significantly decreased 1 day after irradiation. Irradiation also increased NK-92 cell susceptibility to Fas-ligation as well as to attack by lymphokine-activatable primary blood NK cells.
Methods
Cell lines and culture
All cell lines regularly tested negative for mycoplasma using the MycoAlert™ mycoplasma detection kit (Lonza, Walkersville MD). NK-92 cells (ATCC CRL-2407) were cultured in alpha Minimum Essential Media with l-glutamine and sodium pyruvate, no ribonucleosides or deoxyribonucleosides (Gibco, Waltham MA), with 0.2 mM inositol, 0.2 mM 2-mercaptoethanol, and 0.02 mM folic acid, 12.5% horse serum (Gibco), 12.5% fetal bovine serum (FBS) (Atlanta Biologicals, Flowery Branch GA), and 1000 U/ml Tecin, Teceleukin recombinant interleukin 2 (IL-2) (Roche, Basel Switzerland) at 5% CO2 and 37 °C. K562 cells (ATCC CCL-243) were cultured in Dulbecco’s modified Eagle Medium (DMEM) with 4.5 g/L glucose, l-glutamine, and sodium pyruvate (Corning Life Sciences) and with 10% fetal bovine serum (FBS) and 1% penicillin–streptomycin (pen-strep) solution (MilliporeSigma, Burlington MA) at 5% CO2 and 37 °C. Raji (ATCC CCL-86), Daudi (ATCC CCL-213), and Jurkat (ATCC TIB-152) cells were cultured in Roswell Park Memorial Institute (RPMI) media with l-glutamine (Corning Life Sciences, Tewksbury MA), 10% FBS and 1% pen-strep at 5% CO2 and 37 °C. PBMCs (STEMCELL Technologies, Vancouver, Canada) were either resting, unstimulated after culture in DMEM with 4.5 g/L glucose, l-glutamine, and sodium pyruvate and with 10% FBS and 1% Pen-Strep at 37 °C and 5% CO2 overnight or LAK activated, cultured in DMEM with 4.5 g/L glucose, l-glutamine, and sodium pyruvate and with 10% FBS and 1% pen-strep at 37 °C and 5% CO2 and supplemented with 1000 U/ml IL-2 for 3 days prior to assay.
Cell irradiation
NK-92 cells were gamma irradiated in 15 ml of culture media, using a cesium137 source (JL Shepherd Model I-30). A dose range of 2.5–20 Gy was tested. After radiation, cells were cultured for the times indicated (0–48 h) in 1000 U/ml IL-2. Control cells were processed in parallel without applying irradiation.
Fas-receptor ligation
For cytotoxicity assays, NK-92 cells were brought to 2.5 × 105 cells/ml and cultured overnight with either 1 ug/ml LEAF purified anti-human CD95 clone EOS9.1 or LEAF purified mouse IgM clone MM-30 as an isotype control. For flow cytometric assays of potential cell death induced after ligation of cellular Fas/CD95, NK-92 or Jurkat cells were cultured overnight at 2.0 × 106 cells/ml with 1 ug/ml LEAF purified anti-human CD95 or LEAF purified mouse IgM isotype control. When combined with irradiation, cells were first irradiated and then mAbs were added either immediately after irradiation or 1 day after irradiation, as indicated in the results. Control cells were processed in parallel without addition of antibodies. The mAbs were from BioLegend (San Diego, CA).
Cytotoxicity assays
Target cells were labeled with Na
51CrO
4 (Perkin Elmer, Waltham, MA) [
23]. Effector NK-92 cell counts were determined using a hemacytometer with routine samples of > 600 cells that excluded Trypan blue (MilliporeSigma). NK-92 effector cells were diluted twofold in quadruplicate replicas for each E:T in V-bottom plates (Costar 3894, 96 well) in 0.1 ml to create six to eight effector to target cell (E:T) ratios. NK cell effectors within PBMCs were diluted similarly. Radiolabeled ‘target’ cells (1 × 10
5/ml in 0.1 ml) were added to each well to induce cytotoxicity. Plates were centrifuged (Sorvall RC 6 +) at 1000 rpm for 3 min to bring the effector and target cells together and incubated at 5% CO
2 and 37 °C for 8 h (unless otherwise noted). After incubation, plates were centrifuged at 1200 rpm for 10 min and 0.1 ml of cell-free supernatant was removed for analysis in a Perkin-Elmer Wizard gamma counter. Spontaneous release was calculated using the average leak rate of target cells without effectors and the maximum release was the radioactivity released by target cells lysed with 1% SDS. The calculated % specific release is a measure of target cell killing, as targets release internalized
51Cr into the sampled supernatant when they die. Percent specific release was calculated using the following formula:
$$\% {\text{Specific Release }} = \, [\left( {Experimental \, counts \, {-} \, Spontaneous \, Release} \right)/(Max \, {-} \, Spontaneous \, Release)] \times { 1}00.$$
The data illustrated are representative of a minimum of three replicate experiments.
Flow cytometry analysis
For determining the presence of Fas-receptor CD95 on cells, the AF647 anti-CD95 clone DX2 was used with Zombie Aqua to eliminate dead cells from consideration. Cells were taken immediately from culture, washed once with PBS, and stained with a 1:100 dilution of Zombie Aqua for 30 min at room temperature (RT), protected from light. Then cells were quenched with FACS buffer with 1% FCS and brought to 5 × 10
6 cells/ml, aliquoted into flow tubes, stained with 10 ug/ml AF647 anti-CD95 clone DX2 for 30 min at RT, protected from light and washed twice.
For analysis of cell death following irradiation and/or anti-Fas ligation, the following fluorescent probes were used: FITC annexin V (BioLegend, San Diego CA) and 7-aminoactinomycin D (7-AAD) (MilliporeSigma). Irradiated or anti-Fas treated, and control cells were washed with annexin V binding buffer and stained with 4 ug/ml 7-AAD and 4.5 ug/ml FITC annexin V for 20 min at RT, protected from light, then washed twice and brought up with annexin V buffer containing 20 ug/ml actinomycin D (AD) (MilliporeSigma), and fixed with 0.5% formaldehyde.
For determining NK counts within PBMCs, TruCOUNT™ Beads (BD Biosciences) and the following fluorescent antibody panel was used: PacBlue anti-CD45 clone HI30, BV711 anti-CD56 clone HCD56, BV711 anti-CD16A clone 3G8, FITC anti-CD3 clone OKT3, AF647 anti-CD244 [
24] clone C1.7, together with 7-AAD to identify necrotic cells,
without washing the cells (in order to prevent cell loss). Gating sequence available in Additional file
2: Fig. S1. Resting or LAK activated PBMCs (as previously described) were taken from culture, spun down, and resuspended in FACS buffer, then aliquoted into flow tubes. Cells were stained for 30 min at RT, protected from light. After staining 20 ug/ml AD was added and the cells fixed by addition of formaldehyde (MilliporeSigma, Boston MA) to a final concentration of 0.5%. All mAbs were from BioLegend, San Diego CA, and titrated for the concentrations suitable for no-wash conditions. The samples were analyzed within 1 day, using a BD Biosciences Special Order Research Product LSR II analytical flow cytometer with a high throughput sampler (HTS) unit.
Cytometric data were analyzed with FlowJo software (BD Biosciences) to determine cell counts, %positive cells, median fluorescence intensity (MFI), and statistical comparisons between samples. The data illustrated are representative of a minimum of two replicate experiments.
Statistical analyses
Cytotoxicity assay data were calculated and graphed with Microsoft Excel and evaluated using SPSS Statistics (IBM,
version 28, Armonk, NY) using linear regression analysis or difference-in-difference comparisons. LU
50’s (the number of effector cells needed to cause 50% lysis) were calculated by linear regression equations of cytotoxicity (y = % specific
51Cr release, x = log
10 of the E:T cell ratios) to determine the number of cells needed to kill 50% of the ‘target’ cells. Then the lytic activity was expressed as LU
50/1.0 × 10
6 effector cells [
25,
26]. The slopes of this linear regression of cytotoxicity provide information that is useful to detect differences in cellular cooperativity or multiple ‘hits’ needed to kill ‘target’ cells. FlowJo “compare population” tool was used to calculate Overton subtraction [
27] and chi-squared statistics to analyze flow cytometric populations.
Discussion
In this study we demonstrated the remarkable serial killing potential of NK-92 cells towards several ‘target’ cell types and the potential for gamma-irradiation to affect this killing. We also tested NK-92 cell functionality remaining after Fas-ligation and documented NK-92 susceptibility to attack by circulating bNK cells, which are two potential limitations of adoptive cell therapy. These findings suggest that NK-92 cells have immense potential in adoptive cancer immunotherapy but should be carefully optimized before infusion into patients to ensure greatest therapeutic efficacy.
Serial killing by NK-92 cells has been documented before; however, we directed our attention to potential in vivo inhibitory effects on serial killing that are absent from standard in vitro NK assays (e.g., FasL on myeloid cells). A previous study focused on the development of a droplet-based cytotoxicity assay that utilized a lowest E:T of 1:3 and showed that ~ 50% of observed NK-92 cells are able to serially kill two or more K562 targets in 12 h [
37]. Serial cytotoxicity has also been observed with time-lapse cinematography using
genetically modified, IL-2 producing NK92-MI cells. This study indicated that one NK-92 cell could kill as many as 14 target HeLa cells over 6 h [
30]. Many studies have used standard radioactive release assays to characterize NK-92 killing towards various cell types, but, without an excess of ‘targets’, these assays were unable to address NK-92 serial cytotoxicity. Our results confirm that NK-92 cells are serial killers and show that serial killing is target cell type dependent, with Raji and Daudi targets reaching KFs > 10, while K562 targets are far less susceptible to serial killing with KFs < 3.0 (Fig.
1).
This variation in killing for different ‘target’ cells, as determined by KFs, may be due to differences in ligands on target cells that engage the diverse activation receptors of NK-92 cells. One possible explanation for low killing towards K562 cells is that NK-92 cells poorly express the receptor NKG2D, in contrast to high NKG2D-expressing KHYG-1 cell line that kills K562 cells much more effectively [
38]. Furthermore, K562s produce the granzyme B inhibitor PI-9, making them less susceptible to killing via granzyme B, which is predominately used by NK-92 cells but overcome by granzyme M used by KHGY-1 [
39]. It is possible that cleavage of NKG2D ligands by metalloproteinases such as ADAM10 [
40] combined with NKG2D downregulation during cytotoxic activation [
41] affects NK-92 killing of K562s. This loss of NKG2D may also explain the plateau in killing that we observed towards K562 after 6 h, and which has been observed for up to 12 h elsewhere [
37]. A limitation of the KF method [
5] to detect serial killing is that it measures the simple average number of targets killed per effector, rather than identifying the fraction of effectors engaged in killing within the NK cell population. Regardless, the KF method is still optimal (in terms of statistical validity, time, labor, and cost savings) to screen treatments used prior to adoptive cell transfer for their effects on serial killing.
In this report, we shed new light on potential limitations to NK-92 cell-mediated serial killing and therapeutic efficacy, specifically following irradiation, a current clinical practice preceding adoptive transfer. Prior studies have investigated the effects of irradiation on NK-92 cell-mediated killing; however, these studies used shorter cytotoxicity assays with different target cells and reported that NK-92 cytotoxic capacity is mostly retained for at least 24 h following irradiation [
22,
42]. Another study, using haNK cells 24-h post irradiation saw an increase in cytotoxicity toward multiple carcinoma cell lines compared to haNK cell killing immediately post irradiation [
43]. These findings are in contrast to our findings with unmodified NK-92 cells, which show decreases in killing toward Raji cells 1 day after irradiation (Fig.
2) as well as K562 cells (
unpublished). Possible explanations for the differences between these findings include haNK cell endogenous expression of IL-2, while our cells were supplemented with 1000 U/ml IL-2, lymphoid versus carcinoma target cells, as well as the 18-h release assays used for the haNK cells, which may allow for detection of longer-term killing potential than our 8-h release assays. It should be recognized that for definitive assessment of serial killing and its losses, two other approaches, time-lapse cinematography and microchip assays with single effector cells with multiple target cells in individual wells [
8,
30,
44,
45], have advantages over KF frequencies. These alternative approaches can discriminate between slower killing by all effector cells vs. killing by a combination of both totally inactive and fully active cells.
In this report, we have also extended the effects of irradiation to effects on cytotoxicity after Fas-ligation and to NK-92 cellular susceptibility to potential attack by patient NK cells. First, we observed a decrease in NK-92 cell viability and viable cell recovery (Table
1), as well as a consistent decrease in serial cytotoxicity 1 day after irradiation, even at lower doses of 2.5–5 Gy (Fig.
2). Notably, this decrease in killing was absent when NK-92 cells were assayed immediately following up to 20 Gy irradiation. This initial retention of activity indicates that therapeutic cell lines should be used immediately post irradiation to maximize cytotoxicity in vivo
, as in the design of one phase I trial [
21]. Ideally, an alternative anti-proliferative approach for cell lines used in adoptive therapies would be used.
The profound effects of irradiation on NK-92 cytotoxic capacity indicate that radiation effects extend beyond DNA damage and likely include direct damage to proteins [
46,
47]. Ionizing radiation produces radiolysis products, such as reactive oxygen species that inactivate proteome functions including those involved in killing and DNA repair [
42]. Low energy electron irradiation, as an alternative to gamma irradiation, inhibits NK-92 cell proliferation while maintaining higher cytotoxic capacity and for longer periods of time and could therefore be considered for clinical applications [
42]. This report also indicates that 2-h after 10 Gy gamma irradiation, there is lower expression of genes encoding multiple pathways that are critical to cell-mediated cytotoxicity [
42]. Considering the additional impact of direct proteome damage by irradiation, alternative treatments such as induction of genetically introduced type II restriction enzymes and pretreatment of cells with certain topoisomerase inhibitors (Hudig et al., unpublished results) that only inflict damage to DNA could be used prior to adoptive transfer [
47].
We discovered an Achille’s heel for irradiated NK-92 cells, Fas/CD95, which has previously been noted on the majority of activated NK cells [
48,
49] and on NK-92 cells [
50]. Despite high expression of CD95, anti-Fas antibodies alone failed to affect proliferation or to initiate death of non-irradiated NK-92 cells within 1 day, even though the non-irradiated cells did respond to Fas-ligation by shrinking in size. One possible explanation for the NK-92 cell’s low sensitivity to death after anti-Fas ligation is that there are two pathways of Fas-mediated death, one of which relies on mitochondrial signal amplification. This type II, intrinsic pathway is slow and readily inhibited by expression of the Bcl-2 family of apoptotic proteins [
32,
51]. Another possible explanation is that NK-92 cells may express wild-type PI-9, which inhibits the caspase-dependent Fas/FasL-mediated death pathways [
52]. Intrinsic resistance to Fas-ligation is also indicated by evidence that NK-92 cells constitutively produce soluble Fas ligand [
42].
Even though the NK-92 cells resisted death by Fas, they did respond with decreased cytotoxic activity. Fas-ligation alone could decrease NK-92 cytotoxicity to Raji cells, but these effects were always two-fold or less for non-irradiated effector cells. However, for irradiated NK-92 cells the anti-Fas effect was remarkably stronger, with just 10% or less of control killing remaining. In synergy, Fas-ligation and irradiation profoundly reduced cytotoxicity (Fig.
4). One possible explanation for this synergistic effect, seen with cytotoxicity but not with viability, is that the cell shrinkage that occurred with Fas-ligated irradiated, non-necrotic NK-92 cells impaired their activity. This shrinkage, that was absent from Fas-ligated non-irradiated cells, is related to dehydration and has been reported as an early indicator of cell death [
53].
These findings suggest a serious risk for engagement of CD95 as a mechanism to hamper NK-92 cell therapeutic efficacy in vivo, especially if a patient’s tumor cells express the counter Fas-ligand (Fas-L/CD178). After irradiation, NK-92 cells appear to have normal viability in the face of Fas-mediated death receptor ligation but are considerably less-effective killers. A logical next step could be to remove CD95 from NK-92 cells in order to reduce their susceptibility to rapid death via the Fas pathway. Recent advances in CRISPR/Cas-9 have made the methodology a more efficient way to genetically engineer NK-92 cells, including the implementation of multiple genetic changes at one time [
54].
Having discovered that irradiation affects NK-92 cell susceptibility to Fas-ligation, we queried if irradiation would also make NK-92 cells more vulnerable to attack when encountered by patient NK cells. We found that irradiated NK-92 cells are susceptible to attack by both unstimulated and IL-2 LAK bNK, whereas non-irradiated NK-92 cells were more resistant to killing (Table
2). These results contrast with previous reports in which substantial killing to non-irradiated NK-92 cells (comparable to K562) was observed [
34,
35]. A technical consideration may contribute to these differences: the IL-2 concentration used to maintain the susceptible NK-92 cells was 20 U/ml, while we used 1000 U/ml IL-2. Our results are preliminary due to a limited number of NK cell donors but do indicate that, after irradiation, NK-92 cells may become more sensitive to attack by circulating NK cells. This NK -mediated attack could potentially be further promoted by antibody-dependent cell-mediated cytotoxicity (ADCC) supported by IgG antibodies that patients develop to NK-92 cell MHC class I proteins [
21]. We suggest that increased sensitivity to host cell attack be monitored whenever NK-92 cells are genetically modified or are treated before adoptive transfer.
Our research was limited to the cytotoxic NK line NK-92, which is only one of several lines that are available for immunotherapies (Additional file
1: Table S1). To the best of our knowledge, these other immortalized cell lines and induced pluripotent NK cells have yet to be characterized for serial killing and for the effects of irradiation combined with Fas ligation. Our study is also limited in that all assays were conducted in vitro. Nonetheless, we were able to underscore the importance of serial killing as a critical variable that may be compromised by pretreatments such as irradiation and by in vivo conditions such as intratumor Fas ligand and bNK attack. The research indicates that other cells lines should be similarly evaluated for potential effects on serial killing. Tumor counter-ligands other than Fas that stimulate NK inhibitory receptors may also profoundly compromise serial killing, a possibility that is yet to be explored. A broad implication is that it may become clinically worthwhile to genetically profile ligands of a tumor environment that affect NK serial killing to select the best NK cell line for immunotherapy.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.