Comparison of characteristics and tumor targeting properties of extracellular vesicles derived from primary NK cells or NK-cell lines stimulated with IL-15 or IL-12/15/18

NK cell-based therapies have shown promise for hematological cancer forms, but their use against solid tumors is hampered by their poor ability to infiltrate the tumor. NK cells release extracellular vesicles (EVs) containing cytolytic proteins, indicating that NK-cell derived EVs may have therapeutic potential. In this study, we compared the tumor-targeting potential of EVs derived from either primary NK cells or the NK cell lines NK-92 and KHYG-1 cultured in IL-15 alone or in combination with IL-12 and IL-18. Primary NK cells were also stimulated through the activating receptor CD16. Tumor cell apoptosis was measured using a panel of human colon, melanoma, glioblastoma, prostate, breast, and ovarian tumor cell line spheroids. NK cells or NK-92 cells stimulated with IL-12, IL-15, and IL-18 generated EVs with higher efficiency than EVs from resting cells, although similar amounts of EVs were produced under both conditions. Proteomic analysis indicated similar distribution of cytolytic proteins in EVs from primary NK cells and NK-92, but lower levels in KHYG-1 EVs that translated into poor capacity for KHYG-1 EVs at targeting tumor cell lines. Further, we show that CD16-stimulated NK cells release low amounts of EVs devoid of cytolytic proteins. Importantly, EVs from cytokine-stimulated NK cells penetrate into the spheroid core, and tumor spheroid susceptibility to NK-cell derived EVs was linked to differential expression of the NKG2D ligands MICA/B, which was blocked with an anti-NKG2D antibody. We conclude that EVs from activated primary NK cells or NK-92 cells has the best potential to infiltrate and target solid tumors. Supplementary Information The online version contains supplementary material available at 10.1007/s00262-022-03161-0.


Introduction
NK cells are innate lymphocytes with cytolytic activity against virally infected or malignant cells. Their activity is controlled by activating and inhibitory receptors [1]. NK cells target stress-induced ligands expressed by transformed cells via a diverse repertoire of activating receptors such as NKG2D, NKp30, NKp46, and DNAM-1. Target cell death is induced by cytolytic proteins secreted from cytolytic granules from within NK cells, or directly by death receptor-induced apoptosis [2]. NK cell activity is further modulated by cytokines. The cytokine IL-15 is necessary for NK cell survival and proliferation, and combining IL-15 with IL-12 and IL-18 induces potent IFN-γ production [3]. Short-term IL-12/15/18-stimulated NK cells differentiate into memory-like NK cells with heightened ability for cytokine secretion [4], and also enhanced capacity to target cancer cells as shown in both pre-clinical and clinical studies [5][6][7][8]. NK cells are now exploited clinically in several immunotherapeutic strategies for cancer, 1 3 with particular promising results for myeloid malignancies [9,10]. Adoptive NK cell transfer to patients with solid tumors have shown less success, despite a number of clinical trials [11][12][13]. The microenvironment of solid tumors poses a challenge to NK cell-based therapies due to suppressive signals, resulting in poor infiltration of NK cells, and suppression of NK cell effector functions [14]. Based on these challenges, strategies exploiting NK-cell derived extracellular vesicles (EVs) could represent a promising approach.
EVs have gained increasing attention due to their unique role in intercellular communication [15][16][17]. EVs are membrane bound vesicles carrying proteins, lipids, small nuclei acids, and metabolites from the secreting cell, and originate from either inside the cell in the form of exosomes or bud off from the plasma membrane as microvesicles [18,19]. A variety of therapeutic applications of EVs have been reported [20]. With regard to anti-tumor strategies, EVs isolated from immune cells may represent promising immunotherapeutic agents. Exosomes from dendritic cells have been exploited in vaccine approaches, and shown therapeutic potential in patients suffering from metastatic melanoma and nonsmall cell lung cancer [21,22]. Macrophage-derived EVs loaded with cytostatics have been demonstrated to target pulmonary metastases in mice models [23].
NK-cell derived EVs are demonstrated to induce apoptosis of cancer cell lines. These studies have exploited either primary NK cells or the FDA-approved NK-92 cell line as source for generating NK-EVs. The cells have typically been expanded with IL-2 and/or IL-15, and in some instances with feeder cells [24][25][26][27][28][29][30]. NK-EVs variably kill cancer cell lines of both hematological and solid tumor origin [17,24]. Moreover, two independent studies showed tumor regression of neuroblastoma in mice treated with NK-EVs [28,31]. A series of studies suggest that NK-cell derived EVs induce tumor-dell death via their incorporation of cytotoxic proteins [17,25,26,28]. Thus, NK-EVs are equipped to induce apoptosis in target cells, but the specific mechanisms for how NK-EVs interact with and kill the cancer cell remains unclear.
As IL-12/15/18-activated NK cells have shown potent anti-tumor activities against a wide range of tumors, this motivated us to compare the effect of EVs derived from resting (cultured in IL-15 alone) versus IL-12/15/18-activated NK cells. Moreover, there has been little standardization of NK-EV products, making it difficult to compare different studies on NK-EVs. In this study, we provide a comparative evaluation of the characteristics and functional activity of EVs separated from primary NK cells and the NK cell lines NK-92 and KHYG-1 stimulated with either IL-15 or IL-12/15/18 in order to identify the most potent EV product.

Nanoparticle tracking analysis (NTA)
EVs diluted in PBS (1:100) were analyzed using an LM10 nanoparticle tracking analyzer with a 532-laser (Malvern Panalytics). Samples were analyzed under constant flow conditions (flow rate = 20) at 25 °C, and 3-10 × 60 s videos were captured. Data were analyzed using NTA software with a detection threshold of 5 and bin size 2.

Transmission electron microscopy
EVs were imaged by negative stain electron microscopy. A 400 mesh copper grid with carbon-coated formvar film were placed on 20 µl EVs and incubated for 10 min. The grid was washed twice in water, excess liquid removed, and transferred to a drop of 2% uranyl acetate for 30 s, dried and subjected to microscopy with a Tecnai G 2 Spirit TEM (FEI, The Netherlands) equipped with a Morada digital camera and RADIUS imagining software.

Histology
EVs (20 µg) were stained with 10 µM CMTMR (Ther-moFisher Scientific), and washed using exosome spin columns (MW3000, ThermoFisher Scientific). HCT116 spheroids were incubated with CMTMR-stained EVs and 12 µM of CellEvent Caspase-3/7 Green Detection Reagent for 3 days. Single tumor spheroids were fixed with 4% formaldehyde for 1 h, aligned on the bottom of a plastic mold and snap-frozen in OCT compound (Tissue-Tek) using liquid nitrogen. Cryosections of 10 µm, with 20 µm intervals, were obtained using a Leica cm3050s cryostat. Sections were stained with DAPI and embedded in Prolong Gold antifade mounting medium (ThermoFisher) and analyzed using Olympus fluorescence microscope.

Proteomic analysis
Triplicate samples of EVs in PBS were subjected to LC-MS/ MS analysis. EVs were lysed in SDS and proteins precipitated using MagReSyn Amine beads (20 mg/ml, ReSyn BioScience) in 70% acetonitrile. Beads were washed in acetonitrile and 70% ethanol, and proteins reduced in 10 mM DTT in 50 mM ammonium bicarbonate for 30 min at 37 °C, and alkylated by iodoacetamine (20 mM). Proteins were digested with 1 µg trypsin at 37 °C overnight, and peptides captured, concentrated, and desalted using Evotip C18. Samples were analyzed by an EvosepOneLC coupled to a quadrupole-Orbitrap (QExactive HF) mass spectrometer (ThermoElectron) using a 15 cm C18 column with setting 30 samples/day. The mass spectrometer was operated in data-dependent mode to automatically switch between MS and MS/MS acquisition. MS raw files were submitted to MaxQuant software version 1.6.17.0 for protein identification [33]. Minimal unique peptides were set to 1, and FDR allowed was 0.01 for identification. The Swissprot human database was used with a requirement of minimum 2 valid values in at least one group. Data visualization was done using FunRich version 3.1.4 [34] and GraphPad. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository (identifier PXD027284).

Statistical analysis
All data are expressed as mean ± SEM of at least three independent experiments using GraphPad Prism. Statistical analysis were performed using the unpaired Mann-Whitney U-test.

Resting and cytokine-activated NK cells yield similar output of EVs
The level of EV production, EV size, and expression of EV markers was compared in EVs prepared from primary NK cells, NK-92, or KHYG-1 cells cultured in either IL-15 alone or in combination with IL-12 and IL-18 for 48 h. For primary NK cells, stimulation through the activating receptor CD16 was included as an alternative activation pathway. This receptor is not expressed by the NK cell lines. EVs were precipitated from cell culture supernatants, and resuspended in a volume reflecting the cell numbers of the cultures.
Electron microscopy images of EVs from NK cells confirmed successful isolation of vesicles of cup-shaped morphology in the size range 60-125 nm (Fig. 1a). Particle concentration measurements by NTA showed EV concentration from the two NK-cell lines compared to NK cells (Fig. 1b), but a similar EV-concentration from IL-15 or IL-12/15/18 stimulated cells for all three cell types (Fig. 1b). CD16-stimulated NK cells generated lower amounts of EVs compared to cytokine-treated cells (Fig. 1d). EV size was comparable between the different stimuli conditions and cell types (Fig. 1c). Protein concentration measurements largely mirrored the NTA measurements (Fig. 1e). Finally, the different EV isolates expressed the EV markers CD63 and CD81 (Fig. 1f). Notably, KHYG-1 EVs had strong expression of CD81 relative to CD63, whereas CD81 and CD63 signal ratios were comparable for NK-and NK-92 EVs. Expression of both markers were low in EVs from CD16stimulated NK cells. In summary, although the IL-12/15/18 cytokines yield NK cells with enhanced effector functions, this is not reflected in EV output.

EVs from IL-12/15/18-stimulated cells have enhanced ability to kill tumor spheroids.
We next compared the killing potential of the different EV isolates against the colon cancer cell line HCT116. EVs from IL-15-or IL-12/15/18-stimulated NK cells or NK-92 cells induced all cell death, and EVs from activated NK-92 cells induced more death than EVs from IL-15-stimulated cells (Fig. 2a). KHYG-1 EVs induced surprisingly little cell death. To better mimic a solid tumor, we generated HCT116 spheroids and assessed induction of apoptosis and spheroid growth over 3 days (Fig. 2b-d). Both NK-EVs and NK-92-EVs induced an increased caspase 3/7 activity in the spheroids, with slightly higher apoptosis from IL-12/15/18genereated EVs. Again, KHYG-1 EVs did not induce apoptosis over background.
For comparison against EVs, the donor NK cells and NK-92 cells were tested in killing assays (Fig. 3j-r). They showed highest killing capacity of HCT-15 and DU145 cells. PC3 and SK-BR-3 were relatively resistant to both cellular cytotoxicity and EV-induced apoptosis. Interestingly, although EVs induced high levels of apoptosis of HCT116, T-4D7 and WM9, this was not reflected in cellular cytotoxicity, which was particularly low toward WM9 (Fig. 3r). This suggest that EVs may target cancer cells not killed by the cellular counterpart.

EVs contain cytolytic proteins dictated by their cell of origin
The observed differences in tumor-targeting abilities prompted a comparative protein profiling. Triplicate samples were subjected to LC-MS/MS, and a total 534 proteins were identified across all samples and replicates after subtraction of contaminants. Further analyses were restricted to proteins detected in at least two of the three replicates (Suppl. Table 1). 127 proteins were identified in EVs from primary NK cells, 45 of which were shared by EVs derived from the three different stimulations (Fig. 4a). 222 and 279 proteins were identified in NK-92 EVs or KHYG-1 EVs, respectively (Fig. 4b-c). The similarity of the proteome recovered from EVs isolated from either IL-15-or IL-12/15/18-stimulated cells ranged from 61.6 to 71.2% (Fig. 4d). PCA analysis showed that EVs tended to cluster based on their cell of origin (Fig. 4e). Cellular component analysis showed high enrichment of exosomal proteins for all samples (p < 0.001) (Fig. 4f). A core of 34 proteins were identified in all seven EV isolates, and showed a similar pathway enrichment as individual EV isolates (Fig. 4f). Cytolytic proteins were differentially distributed. While FasL was detectable in EVs from NK cells, NK-92 and KHYG-1 cells, granzyme B and perforin were only detected in EVs from NK cells and NK-92 cells ( Fig. 4G and H). Granzyme B was similarly expressed in EVs from IL-15 vs IL-12/15/18-stimulated cells. Both granzyme B and perforin were present at low levels in EVs from CD16-stimulated cells. Taken together, the data show that the EV protein profile varies between NK-cell sources, and may also explain why KHYG-1 EVs appear less potent.

Granzyme B and perforin are concentrated in purified vesicles
Isolation of EVs by precipitation may result in co-precipitation of protein aggregates. To verify that the observed tumor cell apoptosis was induced by EVs, the EV precipitate was further purified by SEC. Ten fractions were analyzed with the highest protein concentration contained in fractions 7-10, and the highest particle concentrations were measured in fractions 5-7 (Fig. 5a,b). Western blot analysis showed that both perforin and granzyme B co-localized in fraction 4 together with the small EV markers CD81 and CD63, indicating that small EVs indeed contain cytolytic proteins. Low levels of granzyme B and perforin were also detected in the later fractions as putative soluble proteins (Fig. 5c). Finally, we demonstrate that the contents of fraction 4 induced apoptosis of HCT116 tumor spheroids (Fig. 5d).

EVs enter the spheroid core, and apoptosis is partly dependent on NKG2D.
We next tested the ability of EVs to penetrate the spheroids. EVs derived from IL-12/15/18-stimulated NK cells were labelled with the red dye CMTMR to allow their identification. Spheroids were fixed and sectioned after 3 days culture with EVs. Figure 6a shows that EVs are detected within the tumor spheroids and co-localizing with apoptotic cells, indicating ability to enter the spheroid. We finally addressed the mechanism by which the EVs interact with the tumor cells, and hypothesized the involvement of activating NK cell receptors or death receptors. The tumor cell lines all variably express the DNAM-1 ligands CD112 and CD155 (Suppl Fig. 2), low levels B7-H6 (NKp30 ligand), while the NKG2D ligands MICA/B were expressed by HCT-15, HCT116, OVCAR-3, T-4D7, WM9, and DU145. HCT116 spheroids co-incubated with NK-92 EVs in presence of antibodies toward either MICA/B, NKG2D, or both, showed a decrease in apoptotic signal (Fig. 6b, c). Incubations with antibodies toward DNAM-1, NKp30, NKp46, or FasL had no effect (data not shown). These data suggest that specific receptor-ligand interactions are involved in NK-cell derived EV-mediated tumor cell apoptosis.

Discussion
In this study, we have performed a comparative characterization of EVs derived from either primary NK cells or the NK cell lines NK-92 and KHYG-1 cultured under either resting or activating conditions. The study is thus the first to directly compare EVs from primary NK cells versus NK cell lines.
Short-term cultures of NK cells with the combination of cytokines IL-12, IL-15, and IL-18 yields NK cells with enhanced ability to eradicate tumors in vivo [5][6][7]. We hypothesized that NK cells cultured under these conditions yield a higher EV output, but we show in this study that there is comparable release of EVs from resting versus activated NK cells. Previous data are in line with our data, in that comparable amounts of exosomes were found when comparing resting versus IL-2-activated primary human NK cells [25,35], although these studies were based on protein quantifications alone. We also subjected NK cells to activation through the Fc receptor CD16 that mediate antibodydependent cellular cytotoxicity. Stimulation through CD16 yields a strong activation signal in NK cells, but generated only low levels of EVs as judged by NTA and testing for CD63 and CD81 expression. This could indicate that EV release in NK cells is predominantly induced via cytokines.
Previous studies of NK-cell derived EVs were based on separation via ultracentrifugation or precipitation. We isolated EVs via precipitation, although the EV isolate will not be as pure as that obtained via ultracentrifugation. A major concern was the possible co-precipitation of protein aggregates containing perforin and/or granzymes that could be mediating the observed tumor cell apoptosis. Further purification of the EV precipitate by SEC demonstrated the co-localization of the small EV markers CD81 and CD63 with perforin and granzyme B, demonstrating that cytolytic proteins are enriched within the EV fraction and not merely present as soluble contaminants. Importantly, apoptosis was induced by the purified EV fraction, indicating that tumor cell death is indeed mediated by the EVs.
Interestingly, a recent study demonstrated that NK cells secrete cytolytic proteins in the form of particles surrounded by a coat of glycoproteins [36]. We cannot exclude that such particles may be present in our EV precipitates, but we detected low levels of perforin and granzymes in EV isolates from CD16-stimulated NK cells, which should induce release of cytolytic granule contents. Cytolytic granules in cytotoxic T cells are shown to contain intraluminal vesicles that may be secreted [37], but T cells also release small microvesicles upon contact with target cells [38]. We hypothesize that the EVs secreted from CD16-stimulated NK cells could represent microvesicles, in which case these would not incorporate cytolytic proteins.
KHYG-1 EVs were functionally different from EVs derived from primary NK cells or NK-92 cells in that they did not induce tumor cell apoptosis. We observed that KHYG-1 EVs contained high levels of CD81 relative to CD63, a skewing that was not apparent for EVs from NK cells or NK-92. Western blots indicated no skewing in expression of CD81 versus CD63 in KHYG-1 cell lysates. Interestingly, a recent study suggested that small EVs expressing mainly CD81 represent vesicles budding off from the plasma membrane [39] [40]. This could explain the absence of granzyme B within KHYG-1 EVs, which would be sorted to intraluminal vesicles en route to cytolytic granules.
We observed a tendency for higher levels of spheroid apoptosis of IL-12/15/18-generated EVs compared to EVs derived from cells cultured in IL-15 alone, which would argue for the use of EVs derived from cytokine-activated NK cells for therapeutic use. This difference was pronounced for WM9, OVCAR-3 and SK-RB-3 spheroids. The explanation for this functional difference is currently unknown, and the MS analysis did not reveal major differences between EVs from IL-15 vs IL-12/15/18-stimulated cells that could explain differential function, but we speculate there may be differences in surface markers that were not picked up by the MS analysis. Interestingly, the WM9 and SK-RB-3 cell lines represent cell lines not efficiently killed by NK cells, indicating that NK-cell derived EVs may target cells independently of their donor cell.
Mechanistically, we pinpoint NKG2D as a receptor mediating the interaction between NK-cell derived EVs with HCT116 spheroids. The HCT116 cells expressed high levels of the NKG2D ligands MICA/B, as did the cell lines that were targetable by NK-EVs. Interestingly, we show that the three least targetable tumor cell lines, U87, PC3, and SK-RB-3 expressed very little MICA/B. In support of our data, exosomes from NK cells were previously shown to contain relatively high levels of NKG2D [24,26]. The tumor cell lines expressed ligands for the DNAM-1 receptors, but we did not observe any reduction of apoptosis of HCT116 spheroids when attempting to block DNAM-1. This was also the case for FasL, although FasL was expressed by all EV isolates. In contrast, a recent study showed that blockade of DNAM-1 ligands led to a slight reduction in killing potential of EVs derived from primary NK cells against by the lymphoblastic cell line NALM-18 [25]. The same study suggested that the EVs were internalized, and that there was a correlation between amount of internalized EVs and the cytotoxic effect on the tumor cells [25]. Our own study indicates that EVs are able to infiltrate the spheroid structure, as we were able to detect EVs within cells in the HCT116 spheroid core.
In summary, we show that EV output is similar under resting or cytokine-activating conditions, and that EVs derived from either primary NK cells or the NK-92 cell lines show similar ability to induce apoptosis of tumor spheroids, with a slight enhanced function of EVs derived IL-12/15/18stimulated cells. We thus conclude that an optimal NK-cell derived EV therapeutic product could be derived from NK-92 cells stimulated with IL-12/15/18 cytokines. Importantly, future studies should address methodology to separate subsets of EVs with cytolytic function, in contrast to bulk EVs obtained by precipitation, SEC, and even ultracentrifugation. Bulk EVs are heterogeneous in terms of size, cargo, and function, and also derive from different sources within Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. Fig. 6 EVs derived from NK cells enter the spheroid core, and spheroid apoptosis is dependent on NKG2D-MICA/B interaction. a HCT116 spheroids were incubated with CMTMR-stained EVs (red) derived from NK cells cultured in IL-12/15/18 for 24 h in presence of caspase 3/7 green detection reagent. The spheroids were fixed, sectioned, stained with the nuclear dye DAPI (blue), and analyzed by fluorescence microscopy. The images are representative of five different spheroids. b-c HCT116 spheroids were cultured with EVs derived from NK-92 cells stimulated with IL-12/15/18 in the presence of blocking antibodies toward NKG2D, MICA/B, or a combination of both. The images (b) is representative of four independent experiments, and the graph (c) shows the median SEM of four independent experiments