Secretory TRAIL-Armed Natural Killer Cell–Based Therapy: In Vitro and In Vivo Colorectal Peritoneal Carcinomatosis Xenograft

Since preclinical studies in mice and primates have shown that administration of TNF-related apoptosis-inducing ligand (TRAIL) can induce apoptosis in human tumors without cytotoxicity to normal organs or tissue (1), Genentech and Amgen jointly prepared TRAIL and performed clinical trials on low-grade non-Hodgkin's lymphoma (2). Although the preclinical/clinical studies demonstrated the safety and activity of TRAIL, its short elimination half-life (approximately 30 minutes) is a main obstacle (3, 4). To overcome this obstacle, researchers have devised several strategies, including using humanized anti-TRAIL receptor antibodies (Ab; ref. 5–8) and a chimeric human TRAIL/human IgG-Fc fusion protein (9). In this study, we attempted to develop secretory TRAIL-natural killer (NK) cell–based therapy. Because NK cells possess tumor-infiltrating capabilities and accumulate selectively at tumor sites (10–13), we used NK cells as carriers to deliver secretory TRAIL.

NK cells are cytotoxic lymphocytes of the innate immune system that kill a variety of tumors and infected cells in the absence of previous stimulation in vivo and in vitro (14–16). NK cells possess two types of surface receptors: activating receptors (NCR, NKG2D, CD16, LY49) and inhibitory receptors [KIR (killer-cell immunoglobulin-like receptors), ILT/LIR, LY49, PD-1; refs. 17–22]. NK-cell activation results from the balance of signals produced by these two receptors. NK cells are activated by several cytokines, such as IL2, IL12, IL15, IL18, IL21, and IFNα/β (23, 24). Upon activation, NK cells proliferate and upregulate effector molecules, such as perforin, granzymes, Fas ligand (FasL), TNFα, and TRAIL (24–27). In contrast, Velthuis and colleagues reported that IL2-activated NK cells induce apoptosis by secretion of granzyme B and perforin, but not via the FasL, TNFα, or TRAIL pathways (28). NK cells can play an important role in immunosurveillance of tumors by directly inducing the apoptotic death of tumor cells (29). These observations support that the mechanism of NK cytotoxicity mainly relies on secretory granules, granzyme B, and requires cell adhesion (22, 30). NK cells also have an immunoregulatory role as they secrete several cytokines, such as IFNγ, following their ligand interaction with cell-surface receptors (31).

Moreover, NK cells demonstrate the ability to infiltrate tumors (10, 11). Because NK cells can recognize tumor cells and infiltrate solid tumors, one of the main goals of this study was to develop secretory TRAIL-armed IL2-activated NK (A-NK) cells and assess their tumoricidal efficacy in in vitro and in vivo systems.

In this study, we constructed pLenti-FETZ vector, which contains three functional domains: a secretion signal domain (the extracellular domain of a ligand for Flt3 tyrosine kinase receptor), a leucine zipper domain for trimerization, and the extracellular domain of TRAIL (a.a. 95-281). NK-92MI-FETZ cells were generated via lentiviral transduction; they can secrete high levels of glycosylated TRAIL fusion protein and induce cell death and apoptosis in colorectal cancer cell lines. Notably, NK-92MI-FETZ cells can infiltrate mouse peritoneal tumors and inhibit peritoneal tumor growth in vivo, which sheds light on their therapeutic potential for the treatment of colorectal peritoneal carcinomatosis.

Human colorectal carcinoma CX-1 cells were obtained from Dr. J.M. Jessup (National Institutes of Health, Bethesda, MD) in 1999, and no authentification was performed by the authors. They were cultured in RPMI-1640 medium (Gibco BRL) containing 10% FBS (HyClone). Human colorectal cancer HCT116 and LS174T cells were purchased from the American Tissue Type Culture Collection (ATCC) in 2014, and no authentification was performed by the authors. They were cultured in McCoy's 5A medium or DMEM (Gibco-BRL) containing 10% FBS. NK-92MI cells were purchased from the ATCC in 2014, and no authentification was done by the authors. They were cultured in minimum essential medium-α (Life Technologies) supplemented with 2 mmol/L l-glutamine, 0.2 mmol/L inositol, 0.1 mmol/L 2-mercaptoethanol, 0.02 mmol/L folic acid, 12.5% horse serum, and 12.5% FBS. HEK293T cells were purchased from the ATCC in 2014, and no authentification was done by the authors. They were cultured in DMEM supplemented with 10% heat-inactivated FBS.

Anti-caspase 8, anti-caspase 9, anti-caspase 3, anti-PARP, and anti-human CD45 Ab were purchased from Cell Signaling Technology. Anti-TRAIL Ab was obtained from Santa Cruz Biotechnology. For production of TRAIL, a human TRAIL cDNA fragment (amino acids 114–281) obtained via RT-PCR was cloned into a pET-23d (Novagen) plasmid, and His-tagged TRAIL protein was purified using the Qiagen express protein purification system (Qiagen).

[3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium (MTS) studies were carried out using the Promega CellTiter 96 AQueous One Solution Cell Proliferation Assay (Promega). Cells (1 × 10⁵) were grown in RPMI-1640 medium containing 10% FBS in tissue culture–coated 96-well plates and treated with drugs for 24 hours. Cells were then treated with 20 μL MTS/phenazine methosulfate solution for 1 to 4 h at 37°C. Absorbance at 490 nm was determined using an ELISA plate reader.

pFETZ was obtained from Dr. Y He (Immunotherapy Center, Medical College of Georgia, GA). It contains the extracellular domain of Flt3L (a.a. 1–81), an isoleucine zipper domain, and the extracellular domain of TRAIL (a.a. 95–281). pFETZ was subcloned into pENTER-D-TOPO (Invitrogen) as an entry clone. In vitro recombination between an entry clone (containing a gene of interest flanked by attL sites) and a destination vector was performed to construct pLenti-FETZ/GFP expression vector. Clones with the right sequence were chosen. Lentivirus carrying a secretable trimeric TRAIL gene is called Lenti-FETZ, and Lenti-GFP virus served as a control.

Lentiviral particles are generated by transfection of the following plasmids [the control plasmid pLenti-GFP or the expression plasmid (i.e., pLenti-FETZ), plus pLenti-3A, pLenti-3B, and pLenti-3C] into 293-T cells using Lipofectamine 2000 (Life technologies). Culture media were harvested 48 hours after transfection, filtered through 0.45 μm filters, underwent ultracentrifugation at 100,000 × g for 2 hours at 4°C, and were stored at −80°C in single-use aliquots.

NK-92MI cells were transduced with the lentivirus (Lenti-GFP and Lenti-FETZ). Multiplicity of infection (MOI) was between 20 and 100. Upon infection, NK-92MI cells were selected with 2 μg/mL puromycin for 3 weeks.

Glycosylation of secreted TRAIL was examined by treatment with three different types of glycosidases. It is well known that O-Glycosidase can remove desialylated core 1 and core 3 O-linked disaccharides attached to Ser/Thr residues. Endo H is a recombinant glycosidase and can remove only high-mannose and some hybrid types of N-linked carbohydrates. Unlike Endo H, PNGase F can remove all types of N-linked (Asn linked) glycosylation regardless of their types (high-mannose, hybrid, bi-, tri-, and tetra-antennary). Supernatant of NK-92MI-FETZ was treated with three different types of glycosidases, and then glycosylated and deglycosylated TRAIL were determined by immunoblotting assay.

Protein was measured with BCA Protein Assay Reagent (Pierce) and separated with SDS-PAGE gel and transferred to nitrocellulose membrane. The membrane was then blocked with 5% nonfat dry milk in TBS-Tween-20 for 0.5 hours and incubated with primary Ab at 4°C overnight. The membrane was incubated with horseradish peroxidase–conjugated anti-rabbit or anti-mouse IgG at room temperature for 1 hour and then visualized using the chemiluminescence protocol.

The supernatant of each NK cell culture was collected and examined using ELISA to measure the concentrations of soluble TRAIL. The supernatants of the NK cell culture and cell protein extract were centrifuged for 10 minutes at 6,000 x g and analyzed with an ELISA kit (R&D systems) to determine the concentrations of TRAIL.

Single-cell suspensions were stained with FITC- or allophycocyanin (APC)-conjugated CD45 Abs. To distinguish NK-92 cells from tumor cells, cell surface marker human CD45 was used. The conjugated Ab-specific to human CD45 was obtained from BioLegend. HCT116 cells have no expression of CD45, whereas NK-92MI cells are strongly positive (Supplementary Fig. S1B). An Annexin-V–FITC Apoptosis Detection kit (BD Pharmingen) was used to measure apoptosis. HCT116, NK-92MI, and NK-92MI-FETZ cells were stained with propidium iodide (PI) and FITC-conjugated Annexin V and analyzed with flow cytometry (Supplementary Fig. S1C). Data were collected on a FACSCalibur (BD Biosciences) and analyzed using CellQuest Pro software (BD Biosciences).

Cytolytic activity was assessed using a standard ⁵¹Cr release assay. Target cells (HCT116) were labeled with 50 μCi ⁵¹Cr, and 1 × 10³ cells/well were incubated in a U-bottom 96-well plate with effector cells (NK-92MI), at indicated conditions, at the indicated effector/target (E/T) ratios. After 4 hours, 50 μL aliquots of cell-free supernatant were harvested and assayed for ⁵¹Cr release in TopCounter reader (PerkinElmer) and then analyzed using the following formula: (test ⁵¹Cr release – spontaneous ⁵¹Cr release)/(maximal ⁵¹Cr release − spontaneous ⁵¹Cr release) × 100. Spontaneous ⁵¹Cr release represents the amount of ⁵¹Cr released by target cells in the absence of effector cells. The maximal ⁵¹Cr release was determined by measuring the ⁵¹Cr radioactivity from complete lysis of target cells with 1% Triton X-100.

An in vitro cell migration assay was performed to determine the tropism of NK-92MI cells for cancer cell–conditioned medium. Cancer cell culture supernatant (600 μL) was added in the bottom well of a Transwell plate (5-μm pore membrane; Corning) before 100,000 NK-92MI-GFP or NK-92MI-FETZ cells in 100 μL were added to the top well. NK-92MI cells were allowed to migrate across the membrane for 4 hours at 37°C. The cells in the lower chamber were counted.

LS174T cells (1 × 10⁶) expressing firefly luciferase were delivered i.p. NK-92MI cells (1 × 10⁶) were labeled using Cy5.5 NHS ester (Lumiprobe Corporation) for 30 minutes before injection, according to the manufacturer's instructions, and delivered either by i.v. or i.p. injections 7 days after tumor cell injection. Bioluminescence and fluorescence were determined using an in vivo imaging system (IVIS200 system, Perkin Elmer). At study termination, mice were sacrificed and samples of tumors, serums, peritoneal washes, spleens, and livers were collected for further analysis.

Fixed specimens were embedded in paraffin and cut into 4-μm sections for hematoxylin and eosin staining. GFP Ab (rabbit polyclonal; Invitrogen) and TRAIL Ab (rabbit monoclonal; Santa Cruz Biotechnology) were used as primary Abs and detected with biotinylated secondary Abs and diaminobenzidine (Vector Laboratories).

Fluorescent microscopy was used to detect Cy5.5-positive cells with 4′,6-diamidino-2-phenylindole (DAPI) counterstain. Terminal deoxynucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) staining was performed according to the manufacturer's instructions (Roche Diagnostics). Microscopy was conducted using a light (Olympus BX40), fluorescent (Carl Zeiss; Axioskop 2), or confocal (Bio-Rad MRC 1024) microscope.

NOD.CB17-Prkdc^scid/J mice (six weeks old, Taconic) were challenged i.p. with 1 × 10⁶ LS174T cells expressing firefly luciferase, which was stably transduced by lentiviral transfection of the pGL4 Luciferase Reporter Vector (Promega) and selected with puromycin. The luciferase signal was monitored by injecting the luciferase substrate luciferin (150 mg/kg, i.p.; GoldBio) 5 minutes after anesthesia with 2% isoflurane prior to imaging on an IVIS200 system. Bioluminescence signal was quantified using the LivingImage software (Perkin Elmer). The mice were examined for bioluminescent signal, indicating basal tumor loading. Mice with the same level of bioluminescence were divided into four groups (seven per group) and given treatments of PBS, NK-92MI-GFP (5 × 10⁶, i.p.), or NK-92MI-FETZ (5 × 10⁶, i.p.) twice per week for 2 weeks. The tumor load, represented by bioluminescent signal determined using the IVIS bioluminescent Imaging System, was checked twice a week. Mice were fed ad libitum and maintained in environments with a controlled temperature of 22 to 24°C and 12-hour light and dark cycles. All animal experiments were carried out at the University of Pittsburgh in accordance with the Guide for the Care and Use of Laboratory Animals.

Statistical analysis was performed using GraphPad Prism 5 (GraphPad Software). In vivo experiments with multiple groups were analyzed using repeated measures ANOVA, and single group data were assessed using the Student t test. Differences were considered statistically significant when the P value was <0.05. All in vitro experiments were performed in triplicate unless specified.

To investigate whether TRAIL is cytotoxic to NK-92MI cells, we measured cell viability with MTS assay. NK-92MI is an IL2-independent NK-cell line derived from a male patient with rapidly progressive non-Hodgkin lymphoma (32). HCT116/LS174T/NK92MI cells were treated with various concentrations of TRAIL, and then viability was determined. As shown in Fig. 1A, HCT116 and LS174T cell viability decreased as TRAIL concentration increased. HCT116 cells were more sensitive than LS174T cells to TRAIL. However, unlike HCT116 and LS174T cells, NK-92MI cells were resistant to TRAIL. Similar results were obtained in terms of TRAIL-induced PARP cleavage, the hallmark of apoptosis (Fig. 1B). These results demonstrated that NK-29MI cells can be used as carriers to deliver secretory TRAIL.

We constructed a lentiviral vector (pLenti-FETZ) producing a secretable form of TRAIL or pLenti-GFP as a control lentiviral vector (Fig. 2A). We chose lentiviral vectors because they are able to efficiently transduce NK-92MI cells (Fig. 2B). We established NK-92MI-GFP and NK-92MI-FETZ stable cell lines by selecting the lentiviral vector–infected NK-92MI cells with puromycin treatment for 3 weeks. Interestingly, we observed an expression of 40 kD TRAIL fusion protein in NK-92MI-FETZ cell lysate as well as 50 kD TRAIL fusion protein in the culture medium (Fig. 2C). These results suggest that the chimeric gene FETZ was expressed, modified, and secreted into the medium. To examine whether secreted TRAIL fusion protein was modified, three different types of glycosidases were treated to inhibit glycosylation of secreted TRAIL in the supernatant of NK92MI-FETZ cell–cultured medium. The molecular weight decreased after PNGase F (N-Glycosidase F) treatment, which was able to remove the N-linked glycosylation, but not O-Glycosidase and Endoglycosidase H (Endo H; Fig. 2D). The results from Fig. 2D indicate that secretory TRAIL contains N-linked Endo H insensitive carbohydrates. To quantify the level of secreted TRAIL from NK-92MI-FETZ cells, ELISA was performed. We observed that the level of secreted TRAIL gradually increased as a function of time (Fig. 2E). Results from immunoblot assay and ELISA demonstrate that NK-92MI-FETZ cells are able to secrete substantial amounts of TRAIL fusion protein.

Next, we investigated whether NK-92MI-FETZ cells are able to kill cancer cells. We used the Transwell assay as a noncontacting coculture system to evaluate their cytotoxic effect. As shown in Fig. 3A, HCT116 cells demonstrated clear morphologic features typical of apoptotic death, such as shrinkage and blebbing, when NK-92MI-FETZ cells, but not NK-92MI cells, were cocultured for 24 hours. We also confirmed that secretory TRAIL-armed NK-92MI-FETZ cells and conditioned medium from these cells remarkably induced PARP cleavage (the hallmark feature of apoptosis) and activation of caspases in HCT116 cells (lanes 4 and 6 in Fig. 3B). Unlike secretory TRAIL-armed NK-92MI-FETZ cells, unarmed NK-92MI cells and conditioned medium from these cells induced minimal apoptosis in HCT116 cells (lanes 3 and 5 in Fig. 3B). As shown in Fig. 3B and C, there is not much difference between NK-92MI cells and NK-92MI-GFP cells. These data clearly demonstrate that secretory TRAIL from NK-92MI-FETZ cells effectively induces cytotoxicity in HCT116 cells. Similar results were observed in LS174T cells, but we incubated NK-92MI-FETZ cells with LS174T cells for 48 hours to see the apoptotic cell death, indicating that LS174T cells are less sensitive to secretory TRAIL than HCT116 cells (Fig. 3D). Data from Annexin V/PI assay confirmed secretory TRAIL-induced apoptosis in HCT116 cells (Fig. 3E; apoptotic death cells in top right plot quadrants). Apoptosis occurred in a NK-92MI-FETZ cell number (secretory TRAIL dose)–dependent manner. Results from Western blot assay also confirmed that apoptosis occurred in a secretory TRAIL dose–dependent manner (Fig. 3F).

To examine the cytotoxic effect of NK-92MI-FETZ cells on colorectal cancer cells in a contacting culture system, CX-1 cells were cocultured with NK-92MI-FETZ cells in a cell culture dish. As shown in Fig. 4A, detachment of GFP stably transfected CX-1 cells was observed, and a high ratio of effector to target (E:T) cells increased the detachment of CX-1 cells (Fig. 4A). We observed that NK-92MI-FETZ cells were more effective than NK-92MI cells at inducing the detachment of CX-1 cells. Similar results were observed when HCT116 cells were cocultured with NK-92MI/NK-92MI-FETZ cells (Supplementary Fig. S1A). To quantitatively analyze this, HCT116 cells and NK-92MI/NK-92MI-FETZ were cocultured with various ratios, and cells that underwent apoptotic death were detected with flow cytometric analysis (Fig. 4B). Using CD45 staining, the tumor cells were able to be gated as non-CD45 cells when tumor cells were mixed with NK-92MI cells (left plots in Fig. 4B). After mixing NK-92MI or NK-92MI-FETZ with HCT116 cells at an increased E:T ratio, HCT116 cells underwent increased apoptotic cell death (early apoptotic death cells in lower right plot quadrants, Annexin V⁺ PI⁻ and late apoptotic death cells in upper right plot quadrants, Annexin V⁺ PI⁺) and necrosis (upper left plot quadrants: Annexin V⁻ PI⁺) in NK-92MI-FETZ–treated tumor cells compared with NK-92MI cells (middle and right plots in Fig. 4B). We also used luciferase quantitative assay to measure tumor-specific NK-cell cytotoxic activity. Reduction in tumor cell–associated luciferase activity closely reflected the killing activity of the NK cells against tumor targets. NK-92MI-GFP or NK-92MI-FETZ cells were incubated with luciferase cDNA stably transfected HCT116-luc+ cells at a ratio of 4:1 for 24 hours. The luciferase activity of NK-92MI-FETZ cells mixed with HCT116 cells was significantly decreased compared with that of NK-92MI-GFP cells (Fig. 4C). Similar results were observed with ⁵¹Cr release assay (Fig. 4D). Moreover, NK-92MI-FETZ–induced cytotoxicity was suppressed by treatment with neutralizing anti-TRAIL Ab. These results revealed that enhanced cytotoxic effects of NK-92MI-FETZ are due to secretion of TRAIL. In summary, results from microscopic examination, flow cytometric analysis, luciferase quantitative assay, and ⁵¹Cr release assay demonstrate that NK-92MI-FETZ cells have more cytotoxic efficacy than NK-92MI cells.

Several researchers have reported that NK cells possess tumor-infiltrating capability and accumulate selectively at tumor sites (10–13). First, we investigated whether NK-92MI or NK-92MI-FETZ cells can migrate to cancer cells in vitro. We harvested conditioned medium from CX-1/LS174T/HCT116 cells and added them into the lower chamber of the Transwell plate and seeded NK-92MI-GFP/NK-92MI-FETZ cells into the upper chamber. We observed that migration of NK-92MI cells significantly increased to the conditioned medium from tumor cells in comparison to medium alone (Fig. 5A). No differences were observed regarding the migration ability of NK-92MI-GFP and NK-92MI-FETZ cells (Fig. 5A). Second, we investigated whether NK-92MI cells can infiltrate peritoneal carcinomatosis in vivo. To establish xenograft tumor formation, we i.p. injected LS174T-luc+ cells into NOD.CB17-Prkdc^scid/J mice. These mice are characterized by an absence of functional T cells and B cells and have low NK-cell activity. We compared the efficiency of i.p. versus i.v. delivery of NK-92MI cells to the xenograft tumor. Figure 5B shows an accumulation of NK-92MI cells at tumor sites; i.p. injection was more effective than i.v. injection for delivery of NK-92MI cells. To confirm these observations, tumor tissues were removed, sectioned, and stained with DAPI. DAPI (tumor cell) and Cy5.5 (NK-92MI cell) fluorescent-stained cells were detected. Tumor-infiltrating NK-92MI cells were detected in i.p.-injected mice, but not many were detected in i.v.-injected mice (Fig. 5C). Time sequence examinations confirmed these results. Figure 5D shows that the Cy5.5 fluorescence signal radiance in the i.v. group was less than that of the i.p. group at every time point. To examine accumulation of NK-92MI cells in various tissues, tumor-bearing mice were i.p. or i.v. injected with NK-92MI-Cy5.5 cells, and the presence of NK-92MI-Cy5.5 cells was examined in the liver, spleen, lymph nodes, and tumor tissues. We observed that NK-92MI cells had mainly accumulated in the liver with i.v. injection, but not much in tumor tissues, whereas NK-92MI cells had accumulated in the liver, lymph nodes, and tumor tissues with i.p. injection (Fig. 5E). We also collected blood and peritoneal wash from i.p.- or i.v.-injected mice. The CD45 marker was used to distinguish human NK-92MI cells from the mouse-originated cells. We observed that human CD45-expressing cells (NK-92MI) were present in both the peritoneal wash and blood in both groups. There was 15-fold more NK-92MI cells in the peritoneal wash in the i.p. group than in the i.v. group (Fig. 5F and G). All these data confirmed that NK-92MI cells were able to infiltrate the LS174T peritoneal tumor, and that i.p. injection was more effective than i.v. injection at delivering NK-92MI cells to tumors.

To investigate the tumoricidal efficacy of TRAIL, we investigated the effect of NK-92MI-FETZ on the growth of LS174T intraperitoneal tumors in vivo. Figure 6B and C show that administration with NK-92MI-FETZ cells significantly delayed tumor growth compared with the sham group and the NK-92MI-GFP cell group (P < 0.05). Interestingly, no statistically significant delay of tumor growth was found with treatment of NK-92MI-GFP cells (Fig. 6A and B), even though NK-92MI-GFP cells selectively accumulated in tumor tissues (Fig. 6C). There was no difference between the tumor-infiltrating efficiency of NK-92MI-FETZ cells and that of NK-92MI-GFP cells (data not shown). Data from immunohistochemistry staining assay show the presence of both NK-92MI-GFP cells and NK-92MI-FETZ cells in tumor tissues (CD45 in Fig. 6D). However, NK-92MI-FETZ cells effectively induced apoptosis (TUNEL in Fig. 6D). Notably, ELISA data showed a 6-fold and 7-fold increase in TRAIL concentrations in the blood (Fig. 6E) and tumor tissues (Fig. 6F) of the NK-92MI-FETZ cell–treated group, respectively, compared with those of the sham group and the NK-92MI-GFP cell–treated group. Collectively, these results show that NK-92MI-FETZ cells can effectively accumulate at tumor sites, infiltrate into peritoneal tumors via i.p. injection, and induce apoptotic tumor cell death via the secretion of TRAIL in vivo.

Dramatic clinical antitumor effects have been observed with activated NK cells in refractory leukemia, but not ovarian cancer (33). The delivery mode of NK cells may affect tumor cell killing and circumvent the lack of NK-cell expansion, especially for ovarian cancer. We established a mouse peritoneal carcinomatosis xenograft model via i.p. injection of LS174T cells and confirmed that i.p.-, but not i.v.-delivered, NK-92MI cells were able to effectively infiltrate LS174T peritoneal tumors (Fig. 5C and E). These results were consistent with previous observations: NK cells not only accumulate selectively at tumor sites independent of tumor antigen expression, but also penetrate deeply into tumor tissues (10–13). Upon activation, infiltrated NK cells proliferate and upregulate effector molecules, such as perforin, granzymes, FasL, TNFα, and TRAIL (24–27). However, we observed that the effect of NK-92MI-GFP cells was limited (Fig. 6A and B), although NK-92MI-GFP cells infiltrated into tumors (Fig. 6C and D), engrafted within the peritoneal cavity, and circulated (data not shown). One possible limitation of infiltrated NK cells is due to tumor microenvironments. Tumor cells may secrete tumor-derived exosomes containing IL10 cytokine, TGFβ growth factor, and miRNAs, which recruit immunosuppressive cells and impact the NK functions (34). The other possibility is the absence of danger signals (22). A failure of integration of signals for activation and inhibition may cause the impairment of NK-cell functions. Another possibility is an insufficient ability to kill cancer cells through Fas-, perforin-mediated apoptosis, or NK-cell exhaustion (35). Among these possibilities, at the present time, we can only speculate on the inactivation of NK cells, which results from the unbalance of signals produced by activating receptors and inhibitory receptors. It is possible that the PD-1/PD-L1 signaling axis modulates the tumoricidal efficacy of NK-92MI cells (21). This possibility needs to be investigated by blocking the PD-1/PD-L1 signaling axis using PD-1– and PD-L1–specific Abs.

A major hindrance in studying NK cells is the resistance of these cells to gene transfer, such as lipid-based transfection and electroporation, etc. Among various technologies available for gene transfer, lentiviral-mediated transduction has been successful in introducing genes into NK cells (36). We also successfully transduced Lenti-GFP and Lenti-FETZ vector into NK-92MI cells (Fig. 2). We confirmed the secreted TRAIL fusion protein (FETZ) with Western blot and ELISA assay (Fig. 2C and E). We also confirmed that the secreted TRAIL fusion protein is the N-linked glycoprotein (Fig. 2D). Glycosylation serves a variety of structural and functional roles in membrane and secreted proteins, including proper folding and enhanced protein stability, whereas unglycosylated protein degrades quickly (37). The glycosylated secretory TRAIL revealed cytotoxicity and tumoricidal activity (Figs. 3, 4, and 6)

Taken together, our findings show that NK-92MI-FETZ cells enhanced cell death and apoptosis in several colorectal cancer cell lines through secretion of glycosylated TRAIL fusion protein in vitro. NK-92MI-FETZ cells delivered i.p. can infiltrate colorectal peritoneal tumors, inhibit xenograft tumor growth, and increase apoptotic tumor cell death in vivo. Adoptive transfer of allogeneic NK-92 cells has been demonstrated to be a safe and potentially beneficial therapy with successful antitumor effects, receiving FDA approval for testing in cancer patients (38–40). We believe that the present findings will stimulate further clinical validation of i.p. delivery of genetically engineered NK-92MI-FETZ cell immunotherapy in colorectal peritoneal carcinomatosis patients.

No potential conflicts of interest were disclosed.

Conception and design: X. Song, W.T. Kwon, Y.J. Lee

Development of methodology: X. Song

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): X. Song, S.-H. Hong, W.T. Kwon, L.M. Bailey, P. Basse

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): X. Song, W.T. Kwon, Y.T. Kwon

Writing, review, and/or revision of the manuscript: X. Song, D.L. Bartlett, Y.T. Kwon, Y.J. Lee

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y.T. Kwon

Study supervision: Y.T. Kwon, Y.J. Lee

The authors thank Christine Heiner (Department of Surgery, University of Pittsburgh) for her critical reading of the article.

This work was supported by NCI grant R01CA140554 (to Y.J. Lee). This project used the University of Pittsburgh Cancer Institute Core Facility, and all authors except Y.T. Kwon were supported in part by award P30CA047904.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.