The human colon cancer cell line HCT-116, human cervical cancer cell line HeLa, human breast cancer cell lines MCF-7 and MDA-MB-231, and human chronic myelogenous leukemia cell line K562 were obtained from the American Type Culture Collection (ATCC, Manassas, Virginia, USA). All cell lines were maintained in Dulbecco's Modified Eagle Medium (DMEM) or Roswell Park Memorial Institute-1640 (RPMI-1640) media (Thermo Fisher Scientific, Waltham, Massachusetts, USA) supplemented with 5% heat-inactivated fetal bovine serum (FBS; Thermo Fisher Scientific), 1% penicillin and streptomycin (Thermo Fisher Scientific) and antimycoplasma reagent (InvivoGen, San Diego, California, USA). Cells were passaged every 2–3 days to maintain appropriate densities.

The buffy coats of healthy volunteers were obtained from local blood centers. After mononuclear cells were isolated using Ficoll-Paque (Sigma-Aldrich, St. Louis, Missouri, USA), CD56⁺CD3^- cells were purified using the NK Cell Isolation Kit (Miltenyi Biotec, Bergisch Gladbach, Germany). Expanding procedures of NK cells were adapted from a previous study.21 Briefly, NK cells were suspended at 1×10⁶ cells/mL in X-VIVO 10 media (Lonza, Basel, Switzerland) supplemented with 5% heat-inactivated FBS, 200 U/mL interleukin-2 (IL-2; R&D Systems, Minneapolis, Minnesota, USA), 10 ng/mL IL-15 (R&D Systems) and mitomycin-treated K562 cells at 1×10⁶ cells/mL. During the 15 days of expansion, NK cells were passaged every 2–3 days with replenishment of cytokines and freshly prepared K562 cells.

CCR5 was cloned from a cDNA library of human blood cells using PCR, after which the coding sequence was cloned into green fluorescent protein (GFP)-expressing pCDH lentivirus vector (System Biosciences, Mountain View, California, USA). Lentivirus production and concentration were performed as previously described.22 Lentiviruses were serially diluted and the titers determined by analyzing the GFP-positive rates of infected 293 T cells. A spinfection protocol was employed to enhance the transduction efficacies of NK cells. Briefly, NK cells were activated for 4 days with cytokines and feeder cells as described above, then mixed with lentivirus at a multiplicity of infection (MOI) of 10 and centrifuged at 1000×g for 2 hours at 32°C. Five days after virus infection, integration of target genes in NK cells was determined by fluorescence-activated cell sorting (FACS).

Cells were counted, collected and then washed with ice-cold phosphate buffered saline (PBS) containing 2% FBS, after which the cells were incubated with fluorochrome-conjugated antibodies (Biolegend, San Diego, California, USA) or apoptosis detection chemicals (Thermo Fisher Scientific) according to manufacturer’s instructions. Following that, cells were washed three times with ice-cold PBS containing 2% FBS and analyzed using a Canto II or Accuri C6 flow cytometer (Becton Dickinson, Franklin Lakes, New Jersey, USA). For in vivo assays of NK cells, tissues were dispersed and single cell suspensions prepared. Next, samples were stained with fixable viability dye (Invitrogen, Carlsbad, California, USA) and fluorochrome-conjugated anti-human CD45 antibody (Biolegend). The stained samples were then subjected to FACS analysis.

Cytotoxic activity of NK cells was determined using a Lactate Dehydrogenase (LDH) Detection Kit (Thermo Fisher Scientific). Tumor cells were incubated with NK cells at various effector to target (E:T) ratios for 6 hours, after which the supernatants were transferred to a new plate, mixed with reaction buffer for 0.5 hour, and subsequently incubated with stop solution. LDH activities were then monitored by measuring optical densities at 490 and 680 nm. Cytotoxic lysis was calculated according to the formula % cytotoxicity = (sample LDH activity-spontaneous LDH activity)/(maximal LDH activity-spontaneous LDH activity)×100.

NK cells were incubated with tumor cells at E:T=1 for 24 hours, after which supernatants were collected and cytokine (interferon-γ (IFNγ) and tumor necrosis factor α (TNFα)) secretions determined using enzyme-linked immunosorbent assay (ELISA) kits (R&D Systems) according to manufacturer instructions.

Transwell assays were carried out as previously described.23 Supernatants from tumor cells or media containing CCL5 protein (R&D Systems) were added to the bottom wells. NK cells were added to the 5 µm Transwell chambers (Corning, Corning, New Jersey, USA) and incubated for 1 hour. For the blocking assays, CCL5-neutralizing antibody (R&D Systems) was supplemented to the bottom wells before adding NK cells to the upper chamber. After incubation, cells in the bottom wells were counted to determine the difference in NK cell movement.

Oncolytic vaccinia virus WR (Western Reserve) strain (ATCC) was engineered to express the CCL5 gene; a plasmid containing the CCL5 gene regulated by the viral promoter P7.5 early/late and firefly luciferase genes regulated by the strong synthetic viral promoter SE/L was inserted into the viral thymidine kinase (TK) gene locus by homologous recombination, thus inactivating the TK gene.20 The coding sequence of CCL5 was cloned from the human cDNA library. OV expansion in tumor cells was monitored using bioluminescence or plaque formation assays. Viabilities of tumor cells infected with OV were determined by an MTS cell viability assay for assessing cell survival relative to uninfected controls.

NK cells or HCT-116 cells infected with vaccinia virus were collected and washed with PBS. DNA was extracted using the QIAamp DNA Mini Kit (Qiagen, Hilden, Germany), after which real-time PCR was performed to detect the genomic DNA of vaccinia virus. Briefly, 100 ng DNA from NK cells or 5 ng DNA from HCT-116 cells were mixed with primers (0.4 µm for each primer) and TB Green Fast qPCR Mix (TaKaRa, Kusatsu, Japan). Next, 40 cycles of expansion were performed according to the manufacturer’s instruction. Primers were synthesized as previously described24 and the sequences were as follows, forward, 5ʹ-CGGCTAAGAGTTGCACATCCA-3ʹ and reverse, 5ʹ-CTCTGCTCCATTTAGTACCGATTCT-3ʹ. Ct values were used to determine the expansion of viral DNA.

Female nude mice (NU/J strain) aged 8–10 weeks were obtained from the Jackson Laboratory (Bar Harbor, Maine, USA). HCT-116 cells (1×10⁶) were implanted subcutaneously per mouse, after which OVs (10⁵ or 10⁶ plaque-forming unit (PFU) as required) were intratumorally injected 10 days later. Virus proliferation was monitored for three successive days via bioluminescence assays. Imaging was on an in vivo imaging system (IVIS200; PerkinElmer, Waltham, Maine, USA) after intraperitoneal injection of luciferin substrate. On the fourth day after virus treatment, 5×10⁶ NK cells labeled with Cy5.5 NHS Ester were injected through the tail veins. Two days later, the tumorous accumulation of NK cells was examined using a fluorescence molecular tomography imaging system (FMT2500; Perkin Elmer). In some experiments, tumor homogenates were collected 4 days post virus injection for CCL5 detection. Tumor volumes were calculated twice a week until they reached 1500 mm³.

Data are presented as the mean±SD and are representative of at least three independent experiments. Student’s t-test was used to compare two groups of data, while one-way analysis of variance analysis was performed to determine the difference among three or more groups. Kaplan-Meier curves were used to estimate survival and the log-rank test was used to determine the difference in survival. Bonferroni correction was performed when multiple survival curves were compared. P values <0.05 were considered statistically significant.

We first analyzed CCR5 expression on NK cells purified from patients with cancer and healthy donors. The purities of expanded NK cells ranged from 85% to 90% (online supplementary figure S1A). As shown in figure 1A, CCR5 was expressed at low levels on both freshly isolated and expanded NK cells (Less than 30% of NK cells were CCR5⁺). Due to the limited expression of CCR5, NK cells would not efficiently respond to endogenous or virus-delivered CCL5, and thus remain in the peripheral blood. Therefore, we attempted to overexpress CCR5 in NK cells with lentiviruses; NK cells were successfully engineered and CCR5 expression was significantly increased after lentivirus infection (online supplementary figure S1B and figure 1B). The transduction efficacies with NK cells from different donors ranged from 30% to 70%. Next, we examined the proliferation, cytotoxicity and directional movement of these engineered NK cells. During expansion, untransduced (UTD), GFP- (NK-GFP) and CCR5-transduced NK cells (NK-CCR5) showed similar proliferating curves (online supplementary figure S1C), indicating that transgenic modifications have limited impact on NK cell expansion. When coincubated with tumor cells at various E:T ratios, different NK cells demonstrated similar cytotoxicity (figure 1C). To further dissect the effects of CCR5 modification on NK cell cytotoxicity, the apoptotic death of tumor cells and the secretion of effector cytokines were tested at an E:T ratio of 1:1. As shown in figure 1D, tumor cell apoptosis was efficiently induced by NK cells 6 hours post incubation. The percentages of dying tumor cells induced by different NK cells were not significantly different (p>0.05; figure 1D and online supplementary figure S2). Consistent with this, the IFNγ and TNFα concentrations in supernatants after 24 hours of incubation exhibited minor differences among the three groups (p>0.05; figure 1E). These findings indicate that lentivirus-mediated CCR5 overexpression does not affect cytotoxic cytokine production and the direct tumor-killing activity of NK cells.

We next analyzed the migratory ability of modified NK cells. Transwell assays showed that NK cells rely on the CCR5-CCL5 axis for directional migration and that CCL5 deprivation impaired NK cell movement (figure 1F). Compared with the UTD and GFP-transgenic counterparts, NK-CCR5 cells exhibited increased migration ability in response to CCL5 protein (p<0.005) or tumor cell culture supernatant (p<0.005; figure 1F). When CCL5 was neutralized, the migration of NK-CCR5 cells was completely attenuated (p<0.005; figure 1F), indicating that the CCR5-CCL5 axis enhances NK cell migration. Furthermore, NK cells and their modified counterparts showed similar proliferation rates after coincubation with target cells (figure 1G), indicating that CCR5 modification has limited impact on the proliferation of activated NK cells.

GFP-expressing and CCR5-expressing NK cells were labelled with fluorescent dyes and injected through the tail veins of mice. After 48 hours, we analyzed the tumorous accumulation of NK cells via FMT imaging. As shown in figure 2A and B, tumor infiltration of NK cells was enhanced by the CCR5 transgene (p<0.05). In agreement, NK-CCR5 cells markedly delayed tumor progression and improved mouse survival (p<0.005), although NK-GFP cell infusion led to tumor suppression and prolonged survival as well (p<0.01;figure 2C,D). Together with our in vitro data, the results indicate that CCR5 modification enhances tumor infiltration and subsequently the cytotoxic effects of NK cells.

CCR5-expressing NK cells delayed tumor growth. However, none of the treated mice reached complete regression (figure 2C and D). One possible explanation is the insufficient expression of the CCR5 ligand CCL5 in tumors, which largely decides the infiltrating intensities of CCR5-expressing immune cells. We, thus, explored whether the intratumorous production of CCL5 by the OV increases NK cell trafficking and tumor cell elimination. First, we constructed a human CCL5-producing oncolytic vector (OV-ffLuc-CCL5) based on the TK deficient vaccinia virus (OV-ffLuc); TK deficiency limits viral replication to tumor cells. Next, human cancer cell lines were infected with OV-ffLuc or OV-ffLuc-CCL5 at an MOI=0.1 or 1. Bioluminescent imaging (BLI) demonstrated that both the prototype and CCL5-producing vaccinia virus replicated efficiently in malignant cells (figure 3A). Plaque formation assays also supported that CCL5 insertion had limited effect on OV replication (p>0.05; figure 3B). Furthermore, we analyzed the viabilities of infected cells and found that the proliferation rates of all four cancer cell lines were similarly inhibited by OV-ffLuc and the derivate OV-ffLuc-CCL5 (p>0.05; figure 3C). These data indicate that the CCL5-modified vaccinia virus has similar infectivity and replicating potency to the prototype OV.

As the CCL5 transgene had limited impact on OV infectivity, we asked whether the transgenic CCL5 can be efficiently expressed in host cells. To test this, tumor cells were infected with different viruses and the supernatants collected. ELISA showed that CCL5 levels were increased in supernatants derived from OV-ffLuc-CCL5-infected tumor cells (p<0.005; figure 4A). CCL5 concentrations from tumor cells treated with OV-ffLuc were mildly increased compared with those of PBS-treated samples (figure 4A), indicating that enhanced CCL5 levels result from the translation of the CCL5 transgene in the virus and are not induced by virus infection. We next performed Transwell assays and interestingly found that supernatants from tumor cells infected with CCL5-expressing OV improved the directional movement of NK-CCR5 cells (p<0.005) but not NK-GFP cells (p>0.05; figure 4B). To assess whether the enhanced movement of NK cells was mediated by CCL5 and not other chemokines induced after virus infection, CCL5-specific neutralizing antibody was added to the supernatants. We, thus, found that the directional movement of both NK-GFP and NK-CCR5 cells was inhibited (figure 4C), indicating that the production of CCL5 by OV specifically promotes NK-CCR5 cell migration. Of note, CCL5 was present in supernatants from chemokine-expressing OV-infected cells in markedly high levels (figure 4A). We, thus, asked whether NK cells expressing different levels of CCR5 exhibit inconsistent responses to CCL5. To test this hypothesis, media containing various concentrations of CCL5 were prepared. As shown in figure 4D, CCL5 enhanced the movement of CCR5-overexpressed NK cells in a concentration-dependent manner; however, NK-GFP cell migration showed limited improvement when CCL5 levels were over 1000 pg/mL (figure 4D). These findings indicate that CCR5-expressing NK cells respond better to CCL5 at higher concentration ranges. Therefore, we further analyzed the possibility of combining NK-CCR5 cells with OVs in the following experiments.

One concern of combining OVs with NK cells is whether the immune cells would be infected and consequently destroyed. To address this, we performed the following assays; NK-CCR5 were incubated with OV-ffLuc or OV-ffLuc-CCL5 at an MOI=1. Real-time PCR showed that small amounts of OV infected NK cells but the virus did not expand in those cytotoxic cells (online supplementary figure S3A). Consistently, BLI demonstrated that OV genome transcription was hampered in NK cells (figure 5A). These findings suggest that both the infection and replication of OV were inefficient in NK cells. Furthermore, MTS assays demonstrated that NK cell viability was not compromized after inoculation with OV, even when the MOI was increased to 5 (p>0.05; figure 5B). Next, we examined the cytotoxic activity of NK cells; following coincubation with OV for 24 hours, NK cells were washed with fresh media and then mixed with tumor cells for 6 hours (to detect cytolysis of tumor cells) or 24 hours (to detect cytokine secretion of NK cells). Compared with the counterparts pretreated with PBS, NK cells preincubated with OV-ffLuc or OV-ffLuc-CCL5 induced similar levels of tumor cell death (figure 5C and online supplementary figure S3B). Moreover, ELISA confirmed that inflammatory cytokine secretion was not affected in NK cells previously incubated with OV (online supplementary figure S3C). Meanwhile, the trafficking ability of NK cells responding to CCL5 was not altered by OV pretreatment (figure 5D). Our results indicate that NK cells are resistant to OV infection and retain their potent cytolytic activity after encountering OVs.

As depicted in figure 6A, we explored whether CCL5-producing OV can improve NK cell infiltration and antitumor activity. Immunodeficient mice engrafted with HCT-116 were inoculated with different doses of OV (10⁵ or 10⁶ PFU) intratumorally. Next, the expanding potentials of OV were monitored by BLI. OV-ffLuc and OV-ffLuc-CCL5 showed similar replicating kinetics in vivo (figure 6B and C); virus replication peaked 48 hours after infection and then decreased (figure 6C and online supplementary figure S4). On day 10, the virus was hardly detected in vivo (online supplementary figure S4). Further analysis showed that the replication and persistence of OV in vivo were not affected by the infusion of NK cells (online supplementary figure S4). Compared with OV-ffLuc, CCL5-transgenic OV increased the specific chemokine concentrations in a dose-dependent manner within tumors (figure 6D). Four days after virus injection, when virus titers had decreased but CCL5 levels were significantly increased, NK-CCR5 cells were administered (figure 6A). Two days after NK cell injection, the tumorous accumulation of engineered immune cells was markedly improved by chemokine-expressing OV, especially at a high dose of OV (figure 6E,F). Consistent with this, FACS analysis showed that the directional movement of NK-CCR5 cells was enhanced by CCL5-expressing OV, especially by a high dose of specific OV on day 6 (p<0.0005; online supplementary figure S5B). In agreement with these findings, enhanced accumulation of NK cells was noted within tumor lesions from OV-ffLuc-CCL5-treated mice on day 9 and 14 (p<0.05; online supplementary figure S5B). On day 24, there were still more NK cells within CCL5-expressing OV-treated tumors than in those treated with NK alone or in combination with OV-ffLuc; however, the difference was not statistically significant (online supplementary figure S5B). Similarly, NK cell accumulation was enhanced by OV-ffLuc-CCL5, especially by a high dose of CCL5-producing OV, in HeLa cell-xenografted models (online supplementary figure S6). In addition, transient retention of NK-CCR5 cells was observed within normal tissues, including the liver, lung and spleen (online supplementary figure S5B); these accumulations were comparable between mice treated with NK alone or in combination with OVs (online supplementary figure S5B). These findings suggest that CCL5-expressing OV can enhance the tumor-targeted movement of NK cells in vivo.

After chemokine-transgenic OV was found to improve NK cell filtration into tumors, we monitored the long-term antitumor efficacy of combined therapies. As expected, NK cells or OV alone delayed tumor growth rather than completely eradicating the malignant cells (figure 7A,B). In mice receiving both NK cells and OV-ffLuc treatment, tumor growth was slower than that of single treatment due to sequential injections of cytolytic virus and cells (figure 7A,B); nevertheless, no complete responses were noted after such therapy. In contrast, the combination of OV-ffLuc-CCL5 with CCR5-NK cells led to better prognoses, especially in the group injected with a high dose of CCL5-producing OV, with more than half of mice showing a complete response (figure 7). These in vivo results indicate that CCL5-induction within tumors can improve NK cell accumulation and tumor regression.