Background
Natural killer cells (NK) were identified more than 30 years ago as a population of lymphokine activated killer cells that showed the ability to kill tumor cells in vitro in the absence of prior immune sensitization of the host [
1‐
4]. Over the ensuing years, much has been learned about regulation of their biologic activity and, in particular, their potential use as an immunotherapeutic modality in cancer [
5].
It has become clear that the biologic activity of NK cells is controlled by a complex repertoire of surface receptors which, upon engagement by ligands on a target cell, signal either an inhibitory or activating response [
6]. The major inhibitory and activating receptors are products of germ line genes encoding killer cell immunoglobulin-like receptors (KIRs) and in an autologous environment, inhibition of NK cell cytotoxic activity is dominant and governed by epitopes on self HLA class I alleles. In general, cytotoxic activity of NK cells is triggered when the target cell lacks expression of some or all HLA class I molecules; the basis for the "missing self" hypothesis [
7].
Recognizing the possibility that NK cells have the ability to kill tumors that lack expression of the inhibitory HLA class I alleles, investigators have reported significant antitumor responses in clinical settings of allogeneic stem cell transplantation. Importantly, clinical effects are demonstrated when inhibitory effects are bypassed by utilizing haplo-identical NK cells and best results are achieved when, in addition, KIR-ligand mismatched NK cells are selected [
8,
9]. In turn, this approach requires extensive donor screening and careful depletion of allogeneic T cells from the NK cell product before administration to the host in order to avoid the risk of graft-versus-host disease (GvHD) [
10].
The possibility that infusion of autologous NK cells could serve as an effective treatment modality for solid tumors has long been considered [
11]. However, implementation is hampered by (i) the small number of NK cells in peripheral blood that could be isolated relative to the number of cells that would be required to be effective and the difficulties associated with large-scale production of cytolytic NK cells in compliance with Good Manufacturing Practices (GMP), (ii) the need to activate the NK cells in order to induce NK cell mediated killing of a resident tumor and (iii) the constraints imposed by autologous inhibitory receptor-ligand interactions.
The first issue has been addressed in a number of reports that demonstrate that large numbers of NK cells could be expanded from CD56
+ cells isolated from peripheral blood mononuclear cells (PBMC) obtained from healthy individuals and patients with hematological malignancies and solid tumors. Expansion was achieved by short term culture with cytokines alone, by cytokines and co-culture with irradiated feeder cells consisting of EBV transformed lymphoblastoid cell lines or cytokines and co-culture with K562 cells that had been transfected with and expresses cell membrane-bound IL-15 and 4-1BBL [
12‐
16]. In most instances, these expanded cells were generated from NK cells (CD56
+CD3
-) isolated from peripheral blood using magnetic beads. The expanded NK cells were highly cytotoxic when tested against variety of target cells that consisted primarily of allogeneic cancer cell lines established from hematologic malignancies [
12,
17].
In addition, a GMP compliant and closed system has successfully been established for the enrichment of monocytes from PBMC using counter current elutriation [
18]. Besides a highly enriched population of monocytes, lymphocyte-enriched fractions are also obtained. Currently, clinical studies are ongoing utilizing elutriation derived monocytes for large-scale generation of dendritic cells in order to treat a variety of metastatic cancers.
The objectives of this study were to evaluate if the aforementioned strategies could be combined in order to expand large numbers of NK cells from PBMC from normal individuals and patients with various solid tumors. Furthermore, the possibility to expand NK cells from lymphocyte-enriched cell fractions derived from PBMC by elutriation rather than utilizing isolated CD56+ cells as the starting cell population was determined. In addition, cytolytic allogeneic and autologous activity by direct cytotoxicity as well as antibody-mediated cellular cytotoxicity and NK specific receptor-ligand interactions that mediate target cell killing were defined.
We confirmed that large quantities of cytotoxic NK cells can be expanded from PBMC in the presence of K562 cells expressing membrane-bound IL-15 and 4-1BBLigand from normal individuals and patients with various solid tumors. Ex-vivo expansion tended to alter the balance of NK cell receptor expression towards those that activate and mediate cytotoxicity. This activity resulted in cytotoxicity against various allogeneic tumor targets and more importantly, against autologous-derived gastric tumor targets. Blocking studies identified multiple activating receptor-ligand interactions that would be predicted to mediate NK cell cytotoxicity. Moreover, these activating receptor-ligand interactions were operative in antibody-dependent cellular cytotoxicity (ADCC) in an allogeneic and autologous setting. Importantly, as a mean for future clinical translation, GMP compliant cytolytic NK cells could efficiently be expanded from lymphocyte-enriched cell fractions obtained from PBMC by counter current elutriation.
Our studies demonstrate that human NK cells acquire cytolytic activity against autologous gastric tumor cells after ex-vivo expansion and suggest a therapeutic potential for autologous expanded NK cells, both directly and in combination with monoclonal antibodies in future cell-based immunotherapy.
Methods
Cells and Cell Fractions
Human blood samples were purchased (BRT Laboratories, Baltimore, MD) and whole peripheral blood mononuclear cells (PBMC) were isolated using density-gradient centrifugation. Using leukapheresis products purchased from the same source, the constitutive cell populations were fractionated by continuous-counterflow elutriation following protocols established by the manufacturer of cell separator (Elutra, Gambro BCT). This instrument uses continuous counter-flow elutriation technology to separate cells fractions based primarily by size and secondarily by specific gravity. In brief, the leukapheresis product was loaded via an inlet pump into a constantly rotating (2400 rpm) elutriation chamber. Based on centrifuge speed and cell density, five elutriated cell fractions were collected. PBMC and various elutriated cell fractions were viably frozen in RPMI-1640 (Invitrogen Corp., Grand Island, NY) supplemented with 20% human AB serum (Gemini Bio-Products, Woodland, CA) and 10% Dimethylsulfoxide (Sigma, St. Louis, MO) using an automated cell freezer (Gordinier Electronics, Roseville, MI) and stored in the vapor phase of liquid nitrogen until used.
The myeloid cell line K562, prostate cancer cell lines LNCaP, PC-3 and DU-145 and breast cancer cell line MCF-7 were available in our laboratory. The lung cancer cell line H358 was kindly provided by Dr. S. Ostrand-Rosenberg (Department of Biological Sciences, University of Maryland Baltimore County, Catonsville, MD) and the Head and Neck cancer cell line TU-167 was kindly provided by Dr. S. Strome (Department of Otorhinolaryngology-Head and Neck Surgery, University of Maryland, Baltimore, MD).
With patient consent and under approval of the Institutional Review Board, peripheral blood mononuclear cells were obtained from 2 patients with gastric cancer undergoing treatment at the Tokyo Clinic and Research Institute. Cell lines (tumor 1 and tumor 2) were established from biopsies of metastatic gastric tumor lesions from the respective patients. All tumor cell lines were cultured in RPMI 1640 supplemented with 10% Fetal Bovine Serum, 1% P/S and 1% Glutamax-1 (cRPMI).
Ex-vivo NK cell expansion
NK cells were expanded from PBMC as previously described with some minor modifications [
12]. In brief, PBMC (1.5 × 10
6) were incubated with irradiated (14,000 rad) K562-mbIL15-41BBL cells (10
6) in a 24-well tissue culture plate in the presence of 200 IU/ml human IL-2 (R&D Systems Inc) in cRPMI. Half of the culture medium was replaced every 2-3 days with fresh culture medium for the first 6 days. After 6 days of expansion, cells were harvested, washed, counted and re-cultured at a starting cell density of 1 × 10
5-3 × 10
5/ml in T-25 or T-75 culture flasks in cRPMI supplemented with IL-2. Cells were expanded for and additional 8 days. Additional cRPMI was added to the flasks if necessary based on cell density.
Flow Cytometry
Cell surface expression was determined before and after 14 days of cell expansion by staining with directly conjugated mouse anti-human mAb's against CD3, CD56, αβTCR, γδTCR, HLA class I, HLA-DR, Fas, Fas-ligand, KLRD1, NKG2a, KIR3DL1, ILT2, CD62L, KIR3DL2/3, NKG2d, DNAM-1, NKp46, NKp44 and NKp30 (BD Biosciences). Gates were set around NK cells which were defined as CD3-CD56+ cells. Surface expression of NK cell ligands was determined on both autologous gastric tumor cell lines and included directly conjugated mouse anti-human nectin-2, PVR, MIC A/B, Fas, HLA class I, HLA class II, HLA-G and purified mouse anti-human HLA-E, ULPB-1, ULBP-2 and ULBP-3. For EGFR-mediated ADCC, gastric tumors were stained with mouse anti-human EGFR mAb. Mouse IgGs were used as isotype controls and purified mAbs were secondarily stained with FITC labelled goat anti-mouse mAb. A minimum of 10000 events were acquired using a BD™ LSR II flow cytometer. Data was analyzed with BD™ FACS DIVA Software.
Cytotoxicity assays
Cytolytic NK cell activity was measured by 4 hour chromium 51 (
51Cr)-release assays as previously described [
19]. K562 cells were included as target cells in all cytotoxicity assays to assess overall cytotoxicity performance (data not shown). Expanded day 14 cells were purified into separate populations of NK cells (CD3
-CD56
+) and NKT/T (CD3
+CD56
+/CD3
+CD56
-) cells using MACS human CD3 microbeads and non-expanded NK cells were purified from PBMC using a MACS human NK cell isolation kit. (Miltenyi Biotec Inc). Cell purity was determined to be >92% and 95% respectively. To determine ADCC, 10 μg/ml human IgG1 (huIgG1, Sigma-Aldrich Corp, St. Louis, MO, USA) or Cetuximab (University of Maryland Marlene and Stewart Greenebaum Cancer Center Pharmacy) was added to defined wells. Purified mouse IgG1, mouse anti-DNAM-1, NKp46, NKp44, NKp30 or all four together (all at 10 μg/ml) were added to defined wells during 4 hours of cytotoxicity in order to assess specific activating NK cell receptor-tumor ligand interactions. Reduction in cytotoxicity was calculated based on percentage cytotoxicity in the presence of indicate blocking mAb(s) versus percentage cytotoxicity in the presence of mouse control mAb. The % reduction in ADCC was calculated with percentage cytotoxicity in the presence of human IgG1 set at 100%. To minimize changes that may occur when cell lines are established from primary tumors, the gastric cell lines used in these studies were cultured for less than 10 passages after isolation from the primary tumor tissue.
Statistics
Paired two-tailed Student's t tests were used to calculate p values. P < 0.05 was considered to be significant.
Discussion
The use of NK cells as a cancer treatment modality in the absence of allogeneic stem cell transplant requires that large quantities of NK cells are generated that kill the tumor cells directly or augment the cytotoxic effect of tumor directed monoclonal antibodies. The critical findings of the studies reported herein are that (i) the technology that was reported to successfully expand NK cells isolated from PBMC [
12,
17] was also effective in generating large numbers of NK cells from PBMC and various elutriated cell fractions without previous isolation of the CD56
+ cells, (ii) activation receptors and other cell surface structures that mediate target cell killing are increasingly expressed on the expanded NK cells, (iii) NK cells expanded from PBMC from patients with solid tumors kill allogeneic and autologous cancer cell lines by direct cytotoxicity and, importantly, mediate ADCC to autologous tumor cell targets in the presence of chimeric mAb and (iv) the interactions of activating receptors DNAM-1, NKp30 and NK-p46 with cognate ligands on the tumor target appear to mediate direct and ADCC mediated autologous cytotoxicity in gastric carcinoma.
For translation into the clinic it is important to observe that besides NK cells, relatively small numbers of NKT and T cells are expanded in this system. These cell populations may mediate GvHD when infused together with NK cells in adoptive allogeneic immunotherapy protocols. GvHD is a serious, potentially life-threatening, condition resulting from transplanted or infused allogeneic donor cell recognition of the recipients' tissues as non-self, and is predominantly mediated by CD3
+ T cells [
30]. These cells are often depleted to prevent GvHD, as could be accomplished with the cells expanded by the protocol presented here. Depletion of T cells from the NK cell product before administration to the host is likely to be less critical in the autologous setting.
An important observation in our studies was that the expanded NK cells did not kill autologous and allogeneic PBMC, an indication that despite the increase in surface expression of activating receptors on the NK cells, the inhibitory ligands expressed on normal PBMC were dominant and able to control cytolytic activity against non-malignant cells. This is further illustrated in that both gastric tumor cell lines were susceptible to autologous cytotoxicity despite the expression of high levels of inhibitory classical and non-classical HLA class I molecules. These data suggest that, under certain conditions, activating receptor-ligand recognition may override receptor-ligand interactions that inhibit NK activity. Emerging data indicates that important triggers in this interaction are surface structures (ligand) that are expressed on cells that have undergone malignant transformation. In addition, it is well recognized that HLA class I expression the major NK cell inhibitory structure, is often down regulated in many solid tumors. In the case of autologous NK cell cytotoxicity against PBMC, inhibitory signals still predominated over activating signals, since no cytotoxicity of NK cells against autologous or allogeneic PBMC was observed. Our results indicate that the NK cells expanded and activated by the methods described do not recognize and kill non-transformed cells.
In addition, while significantly higher levels of the inhibitory CD94/NKG2A complex were expressed after ex-vivo cell expansion, it did not affect the potential of autologous gastric tumor cell recognition. The CD94/NKG2A complex is reported to directly inhibit NK cell cytotoxicity through recognition of HLA-E [
31]. Although both autologous gastric tumor cell lines did not express HLA-E (Table
3), we can not rule out that autologous tumor cell recognition will not be affected in HLA-E expressing tumors.
The ligands for natural cytotoxicity receptors NKp30, NKp44 and NKp46 are currently unknown. However; we postulate that at least NKp46 and NKp30 may be involved in autologous gastric tumor cell recognition since lytic activity was abrogated in the presence of blocking antibody against these receptors. Since no significant change was observed in NKG2D expression on expanded NK cells, we did not directly test the involvement of this activating receptor in autologous gastric tumor cell cytotoxicity. The fact that autologous cytotoxicity was not completely inhibited by a combination of anti-DNAM-1, NKp46, NKp44 and NKp30 may indicate that NKG2D or other unidentified receptors may also be involved. Importantly, the interaction between NK cell receptors and their ligands has recently been shown to abrogate NK cell mediated cytotoxicity of human and mouse melanoma cell lines [
32].
Of note, both tumor cell lines also expressed high levels of Fas which is recognized to establish cell death upon interaction with its ligand, Fas-ligand [
33]. In order to test the possibility of target cell-induced killing of the expanded NK cells, all NK cells were evaluated for Fas and Fas-ligand expression before and after ex-vivo expansion. Although expanded NK cells up-regulated high levels of Fas, they did not express Fas-ligand (data not shown).
It has been suggested that in order to overcome self tolerance, multiple activating receptor-ligand interactions should be engaged [
31]. Indeed, multiple activating interactions appear to be involved in autologous cytotoxicity of tumor cells derived from patient 1 when the inhibition of cytotoxicity, in the presence of all 4 antibodies, is compared with DNAM-1 or NKp30 alone (P = 0.0356 and P = 0.0165, respectively). In contrast, no significant additional decline in autologous cytotoxicity was observed for patient 2 when cytolytic activity of all four activating receptors was compared to NKp46 alone (P = 0.7359). We postulate that these data reflect variation in expression of receptor-ligand combination in humans that are known to be operative in the control of NK cell cytotoxic activity. These variations include HLA and KIR polymorphism as well as tumor type and tumor origin (e.g. primary versus metastatic tumor cells). This is illustrated in a recent report on studies in patients with multiple myeloma [
34] where the investigators demonstrated no specific association of autologous NK cell cytotoxicity with a single activating NK cell receptor. In fact, autologous cytotoxic effects were more likely mediated by several activating NK cell receptors which is also in agreement with a previous report [
35] demonstrating that natural cytotoxicity of resting NK cells requires co-activation by more than one receptor.
While currently elutriated cell fraction 5 is used for monocyte and dendritic cell-based immunotherapy therapy protocols [
36,
37], we demonstrated that cytolytic NK cells can be expanded from elutriated cell fractions 2, 3 and 4 regardless of the cellular content of these fractions. However, since NK cell expansion from fraction 4 failed in two out of four experiments, while expansion from PBMC and elutriated cell fractions 2 and 3 was highly successful, and considering the relative high amount of erythrocytes in fraction 2, it may be best to primarily utilize fraction 3 in NK cell expansion protocols. Of note, variability in expansion rates between donors is observed and requires further testing to determine the extent of this variation in the general population. Overall, these data provide a foundation for the large-scale generation of cytolytic NK cells from elutriated cell fractions, which could be employed alone or in combination with other cellular components such as dendritic cells for application in cellular therapy of cancer.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
CJV participated in the design of the experiments, conducted laboratory studies, prepared figures and tables and drafted the manuscript. RW established gastric tumor cell lines. SR conducted laboratory studies and assisted in the manuscript preparation. DC provided the transfected cell line and advice on NK cell expansion. KH cared for patients in the study and biopsied tissue. DLM oversaw the entirety of the project and assisted in the manuscript preparation. All authors read and approved the manuscript.