Imaging α-GalCer–Activated iNKT Cells in a Hepatic Metastatic Environment

Colorectal cancer is the most prevalent cancer diagnosed worldwide (1), with high mortality because the cancer continues to have high metastatic rates (2) and very poor response to checkpoint inhibitors (3, 4). Metastasis is a complex process in which cancer cells acquire the ability to leave the primary tumor and seed distant organs (5). The sequential multistep process is characterized by intravasation into blood vessels from the primary site and arrest at a distant organ site where they colonize the microenvironment and grow to form secondary tumors. Each step in this cascade is rate limiting and depends on the intrinsic properties of cancer cells and the interactions with host cells within the new environment. The liver is frequently a preferred organ of metastasis occurrence (2) and predominantly determines the outcome of patients with colorectal cancer (6, 7). At the time of diagnosis, a substantial percentage of patients with colorectal cancer have already developed liver metastases (8), and 15% to 25%, which are free of metastases at the time of diagnosis, will develop liver metastasis (9). Surgery is the mainstay treatment, and the survival rate significantly improves after surgical excision (6, 9, 10). However, there is also increasing evidence from experimental studies and a few clinical studies that surgery itself increases the likelihood of metastasis development in the liver (11–15). Factors that facilitate the incidence of metastases are (i) the operative release of tumor cells into circulation during the surgical procedure, (ii) residual tumor cells that remain at the surgical site and are spilled into circulation at a later time, and (iii) postoperative local and systemic immunosuppression of cell-mediated immunity, followed by the generation of a tumor cell–permissive environment that facilitates tumor cell survival and growth in the liver (13, 14, 16–20). As a result, there is a need to find treatments that could be administered prophylactically to prevent subsequent metastasis.

The conceptual framework demonstrating the relationship between the immune system and tumor formation was originally postulated by Paul Ehrlich and set the basis for the use of immunotherapy to harness the immune system to fight cancer (21). The liver microenvironment harbors multiple innate immune cells, many of which are strategically located in the liver vasculature and could interact with metastasizing tumor cells that arrive in the hepatic microcirculation. Indeed, invariant natural killer T (iNKT) cells, a unique subset of lymphocytes that recognize glycolipid antigens loaded on CD1d molecules, and Kupffer cells are two of the more prevalent intravascular immune cell types in the liver. Although Kupffer cells are sessile and therefore cannot move toward tumor cells, the iNKT cells constantly patrol the liver, crawling along the sinusoids (22–24), and if appropriately engaged, they could serve as potent effector cells against circulating tumor cells that attempt to seed the hepatic sinusoids. Indeed, two studies with somewhat opposing strategies suggested that eradicating the microbiome or stimulating with lipopolysaccharide could activate iNKT cells to provide antitumor activity in a model of colorectal cancer–associated liver metastases (25, 26). To more selectively stimulate iNKT cells, one could use the therapeutic glycosphingolipid, α-galactosylceramide (α-GalCer), which is an immunomodulatory drug that very potently and specifically stimulates iNKT cells (27). Although initial studies showed great promise in activating iNKT cells to induce antitumoral effects (27–29), numerous clinical trials with various patients with solid tumor have only shown partial and inconsistent responses (30–34). Explanations for these disparate results included the development of iNKT-cell anergy with multiple α-GalCer stimulations, the potential patient pool which included patients with established solid tumors, and induction of resistance within the tumor to activated iNKT cells. Nevertheless, there remains interest in the use of α-GalCer as a potential antimetastatic agent (25, 26, 35).

Using intravital imaging to track and test in an experimental colon carcinoma liver metastasis model, we demonstrated that established and growing metastatic tumors were unaffected by α-GalCer. As such, we focused on a prophylactic approach that could be used at time of surgical removal of primary tumors to target potential metastasis. We found that multiple doses, not a single dose, in the prophylactic setting, but not delayed treatment, significantly reduced early tumor burden. α-GalCer reduced tumor growth but did not kill tumor cells that seeded the liver. Depletion of iNKT cells, but not NK or cytotoxic CD8⁺ T cells, revealed the former in the antitumor effects of α-GalCer, perhaps via increased numbers of iNKT cells contacting the tumor. Although initial robust cytokine production did decrease with time, suggesting iNKT-cell anergy, delivery of α-GalCer restored cytokines back to baseline and continued to provide benefit. In the tumor microenvironment (TME), we observed upregulated PD-1 on iNKT cells and PD-L1 on surrounding endothelium, but PD-1 blockade did not affect iNKT-cell efficacy. We found no improved changes in tumor resistance when tumors were serially transplanted, ruling out the development of inherent genetic or epigenetic resistance to iNKT cells. Finally, we found that very few tumor cells expressed CD1d in vivo, so we delivered monomers of CD1d–α-GalCer and found improved efficacy. We provide preclinical evidence that α-GalCer therapy could be used prophylactically and sequentially after primary tumor surgery to reduce metastatic burden.

BALB/cJ wild-type and Cd1d-deficient mice were purchased from The Jackson Laboratory. Cxcr6^gfp/+ knock-in mice on the BALB/cJ background were gifts from D.R. Littman (New York School of Medicine and HHMI Skirball Institute of Biomolecular Medicine). All mice were used at the age of 6 to 12 weeks and were maintained in a specific pathogen–free double-barrier unit at the University of Calgary. Procedures for all conducted experiments were approved by the Animal Care Committee of the University of Calgary and were in compliance with the Canadian Council for Animal Care.

Murine CT26 cells (CT26.WT colon carcinoma cells originally purchased from the ATCC) were transfected to express the fluorescent protein, iRFP (near-infrared fluorescent protein; refs. 36, 37), using the PiggyBac Transposon Vector System (SBI System Biosciences). First, iRFP [piRFP702-N1 (Addgene)] was cloned into the PiggyBac transposon vector (PB510B-1) for puromycin selection. The PiggyBac transposon element including iRFP was cotransfected with the Super PiggyBac transposase (PB210PA-1) for integration into the DNA of CT26 cells. FuGENE6 Transfection Reagent (Roche) was utilized. CT26 cells that survived puromycin selection stably integrated iRFP. iRFP near-infrared fluorescence was determined by flow cytometry (excitation with 633 nm laser and emission capture with 685 long pass filter). Nontransfected CT26 cells were used as negative controls. We found two very distinguishable populations based on fluorescent intensity. Sixty percent of our cell clones expressed high iRFP, and 40% expressed low iRFP; therefore, we sorted the iRFP^hi-expressing cells from the iRFP^low-expressing cells via FACS. Only cells that highly expressed iRFP were used for in vivo experiments and named CT26-iRFP. Cell sorting was performed using the BD FACSAria II (BD Biosciences).

CT26-iRFP cells were cultured in RPMI 1640 medium (Invitrogen) with 10% heat-inactivated FBS (Invitrogen), supplemented with l-glutamine (1%; Gibco), penicillin (100 units/mL; Gibco), and streptomycin (100 mg/mL). Cell culture medium was changed every third day. Puromycin (Life Technologies, 3 μg/mL) was added to the cell culture flask every time the medium was changed. Passaging was done at 80% confluency, and at 70% confluency, tumor cells were used for in vivo inoculation. For single-cell suspensions, tumor cells were detached with sterile Puck's ethylenediaminetetraacetic acid (EDTA) solution [potassium chloride (0.4 g/L), sodium chloride (8 g/L), sodium bicarbonate (0.35 g/L), dextrose (1 g/L), HEPES (2.38 g/L), and EDTA (2.292 g/L); all from Sigma] and washed in cell culture medium. After two additional washing steps with Hank's Balanced Salt Solution (HBSS; Invitrogen), cells were resuspended in 1 mL HBSS and counted with trypan blue (0.4% in PBS, Invitrogen) for the exclusion of dead cells. Cells that took up trypan blue were considered as nonviable. A hemocytometer (Neubauer) was loaded with a 100 μL dilution (10 μL tumor cell suspension and 90 μL trypan blue), and the number of viable tumor cells was calculated. Tumor cells were only used for further procedures when the viability was ≥90%. Mycoplasma testing (ATCC Mycoplasma Universal Detection Kit) was done each time after a new vial was thawed, and cells were cultured for 5 days before in vivo injection.

Mice were anesthetized with isoflurane (1%–2%). The left upper abdomen was shaved with a clipper and swabbed with alcohol before a small incision (5 mm) was made into the skin and the peritoneum of mice (BALB/cJ wild-type, Cd1d-deficient mice, or Cxcr6^gfp/+ knock-in mice). The spleen was identified and exteriorized through the incision, and a knot was made with sterile nonabsorbable silk sutures between the spleen and connective tissue. CT26-iRFP cells (2 × 10⁵) were intrasplenically inoculated. Control mice received HBSS (Invitrogen). After 1 minute, the knot was tied off, and the spleen was removed to avoid tumor growth in the spleen. Subsequently, the peritoneum was sutured, and the incision was closed using staples. The incision site was then swabbed with LORIS-10% PVP-I solution. Mice received s.c. buprenorphine (100 μL, 0.03 mg/mL, Ceva) before they were removed from the isoflurane and 8 to 12 hours after surgery. Mice were monitored for signs of illness once a day for 7 days. Tumor cells metastasizing to the liver were monitored, as described under spinning disk confocal microscopy, on an ongoing basis (38). Experimental endpoints are indicated in the timelines of each figure.

α-GalCer (Funakoshi) was dissolved in 5.6% sucrose (Sigma-Aldrich), 0.75% L-histidine (Sigma-Aldrich), and 0.5% Tween20 (Sigma-Aldrich) to a final concentration of 1 mg/mL. For in vivo treatment of mice, 2 μg α-GalCer was injected i.v. (tail vein) with HBSS (Invitrogen) in a volume of 100 μL per mouse once every 48 hours. The treatment was given 2 hours prior to CT26-iRFP injection, unless otherwise specified, and was repeated every 48 hours or every 5 days. The duration of the treatment is indicated in each figure.

α-GalCer–loaded CD1d monomers (PBS-57–loaded mouse CD1d monomer) and unloaded CD1d monomers (unloaded mouse CD1d monomer) were kindly provided by the NIH Tetramer facility. For in vivo treatment of mice, 2 μg per mouse was injected i.v. in a volume of 100 μL of saline. The first treatment was given 2 hours prior to CT26-iRFP injection and repeated every other day for a total of 5 times.

BALB/cJ wild-type mice were initially injected with α-GalCer (2 μg/mouse) 2 hours prior to intrasplenic inoculation with 2 × 10⁵ CT26-iRFP cells. Subsequently, α-GalCer was given every 48 hours for varying durations as indicated in the respective figure. Control mice received HBSS injections every 48 hours. Mice were sacrificed when they showed severe signs of illness due to tumor growth, such as weight loss (over 20% of body weight), maximum tumor size of 2 cm², enlarged abdomen, lack of grooming, lack of movement/eating/drinking, hunched, and isolated.

CT26-iRFP cells were inoculated into HBSS- or α-GalCer–treated BALB/cJ wild-type mice as described under the liver metastasis model. After 10 days, livers were collected from perfused mice, cut into small pieces, and incubated for 30 minutes at 37°C in DNaseI (0.25 mg/mL; Sigma) and collagenase type IV (6 mg/mL; Worthington) on a shaker. Cell suspensions were then passed through 70 μm and then 40 μm filter, followed by red blood cell lysis with LCK lysis buffer (VWR). After multiple washing steps with sorting solution [PBS (Invitrogen), 2% FBS (VWR), and 2 mmol/L EDTA (Sigma)], iRFP-positive cells were sorted from liver cells using a BD FACSAria II (BD Biosciences). Each procedure step was performed under sterile conditions. Sorted CT26-iRFP cells (2 × 10⁵) were then injected into HBSS- or α-GalCer–treated mice as indicated in the respective figure. For the collection of CT26-iRFP cells from α-GalCer–treated mice, CT26-iRFP cells from two mice were pooled due to the low amount of cancer cells in α-GalCer–treated mice.

C57BL/6J mice received a single dose of 600 mg/kg acetaminophen (Sigma) in saline via oral gavage. Control treatment was saline.

Anti–PD-1 (clone RMP1-14, BioXCell) or isotype control (clone 2A3, BioXCell) treatment was initiated 1 day prior to α-GalCer (see “α-GalCer treatment”) and CT26-iRFP injection (see “Liver metastasis model”) in BALB/cJ wild-type mice. Two hundred micrograms per mouse were administered i.p. and repeated every 3 days for a total of 4 injections over 10 days, as shown in the respective figure.

Anti-IFNγ (XMG1.2, BioXCell) and anti-TNFα (clone XT3.11, BioXCell) or isotype control IgG1_K (BioXCell) injections were given 1 day prior to α-Galcer (see “α-GalCer treatment”) and CT26-iRFP injection (see “Liver metastasis model”) in BALB/cJ wild-type. Two hundred microgram per mouse in PBS (Invitrogen) were administered i.p. and repeated every 2 days for a total of 3 times, as highlighted in the respective figure.

Spinning disk confocal microscopy was performed on all animals from the liver metastasis model and splenectomized control mice using the customized Olympus IX81–inverted microscope for visualizing CT26-iRFP tumors and CXCR6-gfp⁺ cells in the liver. The microscope was equipped with an Olympus focus drive that had a motorized xyz stage (Applied Scientific Instrumentation, MS-2000 with piezo-z insert) and a motorized objective turret holding three objective lenses: 4×/0.16 UPLANSAPO, 10×/0.40 UPLANSAPO, and 20×/0.70 UPLANSAPO. The objective lenses were connected to a light path (WaveFx; Quorum Technologies) that was linked to a Yokogawa CSU-10 head (Yokogawa Electric Corporation). The acquired images were recorded with an EM-CCD camera. The image acquisition software used was Volocity (Quorum Technologies). A detailed description of the procedure and preparation of the animals has been published in ref. 38. For in vivo imaging of PD-1, PD-L1, and Kupffer cells in Cxcr6^gfp/+ knock-in mice, the following antibodies were used: anti–PD-1-APC (clone 29F.1A12, Biolegend), anti–PD-1-isotype control-APC (clone RTK2758, Biolegend), anti–PD-L1-PE (clone 10F.9G2, Biolegend), anti–PD-L1-isotype control-APC (clone RTK4530, Biolegend), and anti-F4/80-AF750 (clone BM8, AbLab).

Single-cell suspensions from livers were generated by passing them in PBS (Invitrogen) through a 40 μm nylon mesh. After two washing steps in PBS, liver immune cells were resuspended in 44% Percoll (VWR). Subsequently, 70% Percoll was layered underneath to establish a Percoll gradient. After centrifugation, the interface layer, containing lymphocytes, was collected and resuspended in FACS buffer (PBS containing 2% FBS and 2 mmol/L EDTA). After repeated washing steps, red blood cells were lysed using ACK Lysing buffer (Gibco). 10⁶ cells were resuspended in FACS buffer and incubated with Fc-block, anti-mouse CD16/CD32 (clone 2.4G2, BioXCell), before cells were stained with the following antibodies: CD45-PE-Cyanine7 (clone 30-F11, eBioscience), isotype control rat IgG1b,κ PE-Cyanine7 (clone eB149/10H5, eBioscience), CD3-eFluor 660 (clone 17A2, eBioscience), isotype control rat IgG2b,κ-eFluor 660 (clone eB149/10H5, eBioscience), anti-mouse CD335 (NKp46), eFluor 450 (clone 29A1.4, eBioscience), isotype control rat IgG2a,κ eFluor 450 (clone eBR2a, eBioscience), CD8α-BV786 (clone 53-6.7, BD Biosciences), isotype control rat LOU/M IgG2a,κ BV786 (cloneR35-95), PBS57 (α-GalCer)-loaded CD1d Tetramer-PE and unloaded CD1d-Tetramer-PE (kindly provided by the NIH Tetramer facility), and the fixable viability stain 510 (FVS510, BD Horizon). For PD-1 cell surface expression on iNKT cells, NK cells, and CD8⁺ T cells, anti–PD-1-eFluor450 (clone RMP1-30, eBioscience) and isotype control-eFluor450 (eB149/10H5, eBioscience) were used. 123count eBeads (Affymetrix, eBioscience) were used to determine absolute cell counts of the different cell populations. Data acquisition was done on a BD Canto2 in the Nicole Perkins Microbial Communities Core Labs in the Snyder Institute for Chronic Diseases at the University of Calgary.

For expression analysis of CD1d by CT26-iRFP cells, livers were perfused with saline through the vena cava. After cutting into small pieces, liver pieces were incubated for 30 minutes at 37°C in DNaseI (0.25 mg/mL; Sigma) and collagenase type IV (6 mg/mL; Worthington) on a shaker. Cell suspensions were then passed through 70 μm and then 40 μm filter, followed by red blood cell lysis with LCK lysis buffer (VWR). After washing, 10 × 10⁶ cells were incubated with Fc-block as stated before being stained with anti–CD1d-PE (clone 1B1, eBioscience).

The antibody NKT14 (39) was kindly gifted from Dr. Robert Schaub (NKT Therapeutics Inc.). BALB/cJ mice were injected i.p. with 100 μg NKT14 antibody or isotype control (IgG2a) 24 hours before treatment with α-GalCer and CT26-iRFP cell inoculation. To confirm iNKT-cell depletion, flow cytometry was performed as described under “Flow cytometric analysis of liver cells.” Briefly, 1 × 10⁶ single cells isolated from livers were resuspended in FACS buffer, blocked with Fc block (indicated above), and stained using the following antibodies: CD45-FITC (clone 30-F11, eBioscience), isotype control rat IgG2b,κ-FITC (clone eB149/10H5, eBioscience), CD3-eFluor 660 (clone 17A2, eBioscience), isotype control rat IgG2b,κ-eFluor 660 (clone eB149/10H5, eBioscience), TCR-β-APC-eFluor-780 (H57-597, eBioscience), isotype control Armenian hamster IgG APC-eFluor 780 (clone eBio299Arm), and PBS57 (α-GalCer)-loaded CD1d Tetramer-PE and unloaded CD1d-Tetramer-PE.

CT26-iRFP metastases were quantified at different time points in the liver as indicated in figures. Single spinning disk confocal images (10× lens) were recorded with an electronic computer-controlled stage and subsequently stitched together (10% overlap for each image). We acquired a total of 60 images along the x/y axes and aligned them together to obtain an overview image of the liver, called a “stitched image.” The stitched image corresponded to a size of 11.5 mm² in the liver. Tumor cell aggregates consisting of at least three or more tumor cells were quantified as a metastasis. The size of the metastases was acquired and measured using the image acquisition and analysis software Volocity (Quorum Technologies). The size of metastases shown in figures is the average size of all quantified metastases per liver area for each mouse. The number of metastases was manually counted. The metastatic tumor burden was determined by dividing the sum of the areas of metastases through the imaged surface area of the liver. FITC-albumin (Sigma-Aldrich; 10 μL per mouse, stock 1 mg/mL in PBS) was i.v. injected to detect perfusion of hepatic sinusoids.

The crawling velocity and length of contact duration of CXCR6-gfp⁺ cells with CT26-iRFP metastases were recorded in Cxcr6^gfp/+ knock-in mice using spinning disk confocal microscopy and quantified over 1 hour. Image sequences were recorded at 4 frames/min (15 seconds between each frame) using the Volocity software. All CXCR6-gfp⁺ cells that established contact with metastatic cells per field of view (FOV) were manually tracked. For each mouse liver, 3 FOVs were simultaneously recorded, and the mean number of cells for each mouse interacting with metastatic cells was calculated and then plotted against the time points assessed. The percentage of all CT26-iRFP–interacting cells that moved at a velocity of ≤10 μm/min and 10 to 20 μm/min was calculated and graphed.

Mouse blood plasma was used to measure cytokines in circulation. Five hundred microliter of blood was collected by cardiac puncture and added to an Eppendorf tube containing 10 μL heparin (1,000 U/mL; Sandoz). After centrifugation for 10 minutes at 400 × g, the plasma was collected and stored at −80°C. The assessment of cytokines and chemokines (mouse 31-plex array) was done by Eve Technologies. The IFNγ/IL10 ratio was calculated by dividing the individual values measured for IFNγ by the individual values measured for IL10 at the indicated time points.

Mice were anesthetized with isoflurane (1%–2%). Blood samples (500 μL) were collected after cardiac puncture and added to an Eppendorf tube containing 10 μL of heparin (1,000 U/mL). After centrifugation for 10 minutes at 400 × g, blood plasma was filtered (0.22 μm) before alanine transaminase (ALT) was measured as per the manufacturer's protocol [Biotron Diagnostics Inc., ALT (SGPT) color kit]. Time points of blood collection and controls are indicated in the respective figure. ALT levels after acetaminophen treatment were determined in blood serum 12 hours after treatment with the ALT activity assay kit from Sigma (MAK052-1KT).

For all experiments, the data were the mean of a minimum of three independent experiments using different mice. Data are represented as mean ± SEM. Statistical tests are included in each figure legend and were completed using GraphPad Prism 9.1.0 for Windows.

We investigated the immunomodulatory role of α-GalCer in an experimental murine colorectal cancer liver metastasis model. To test α-GalCer for use as potential adjuvant therapy, we treated mice i.v. with a single dose of α-GalCer prior to tumor cell inoculation (Fig. 1A). To examine how α-GalCer affected seeding of tumor cells, we used intravital spinning disk confocal microscopy due to the composition of the tumors (often single cells). High-resolution imaging demonstrated that α-GalCer initially reduced the development of colorectal cancer liver metastases (Fig. 1B) over 3 days, but by day 5 (Fig. 1C), we found similar numbers and sizes of tumors in α-GalCer–treated mice (Fig. 1D) compared with untreated mice, indicating early tumor escape.

Thus, we established a treatment protocol of repeated α-GalCer administration every 48 hours (Fig. 2A), but at a concentration that caused only a transient increase in liver enzymes and no detectable liver damage (Supplementary Fig. S1). Overt liver damage with acetaminophen induced a 10-fold greater concentration of ALT (Supplementary Fig. S1). The evaluation of hepatic metastases at days 5 and 10 revealed that liver metastases in untreated mice grew rapidly over time into large metastases, whereby at day 10, the individual number of tumors decreased, but the size increased, consistent with increased burden size (Fig. 2B and C). In contrast, in α-GalCer–treated mice, metastases remained as single tumor cells or were sufficiently small that all hepatic vasculature was perfused by FITC-albumin. We also noted that these very small metastases replicated by day 10 but failed to form large tumors (Fig. 2C). By using intravital microscopy, we were able to quantify the seeding of single intravascular tumor cells in the hepatic sinusoids. The amount of single tumor cells 3 days following injection was nearly identical in untreated and α-GalCer–treated mice (Fig. 2D), suggesting that seeding was not affected by α-GalCer. Untreated mice succumbed to disease as early as 14 days, and all died by 19 days, whereas mice that received consecutive α-GalCer treatments over 10 days showed enhanced survival (Fig. 2E). Because at day 10, iNKT-cell cytokines, like IFNγ, returned to basal concentrations (discussed in later section), we asked whether extending the length of treatment indefinitely, either by continuing to give α-GalCer every second day or once every week, would further prolong survival. These two regimens further doubled the lifespan of mice (Fig. 2E), suggesting beneficial effects beyond cytokines with repetitive α-GalCer treatment after day 10, and repeated α-GalCer from day 10 onward continued to suppress hepatic metastasis.

A significant number of patients with colorectal cancer have established metastatic disease in the liver at the time of diagnosis (9). To determine the therapeutic limitations of our findings, we asked whether α-GalCer was effective at reducing established hepatic metastases. The treatment with α-GalCer was started 8 days after tumor cell inoculation (Fig. 3A). Three treatments with α-GalCer did not reduce metastatic lesions (Fig. 3B). When treatment was started earlier at day 5 (Fig. 3C), although metastases appeared smaller, a significant reduction in tumor size was not achieved (Fig. 3D). Collectively, these results demonstrate that α-GalCer works most effectively as a pretreatment and does not dampen the growth of preexisting established liver metastases once they reach a certain size.

It is now well established that α-GalCer binds to and is presented on CD1d molecules. iNKT cells express CD1d-restricted T-cell receptors with lipid antigen specificity and are reactive to the glycosphingolipid α-GalCer. To test the mechanistic basis of hepatic tumor control, we repeatedly injected Cd1d^−/− mice, which are deficient of NKT cells, with α-GalCer after tumor cell inoculation (Supplementary Fig. S2A). Hepatic tumors in Cd1d^−/− mice after repeated treatment with α-GalCer were comparable with untreated mice (Supplementary Fig. S2B), confirming previous studies that showed the antitumor activity of α-GalCer depended on CD1d (28, 29, 40) and required NKT cells. Cd1d^−/− mice did not have greater tumor burden than wild-type mice, indicating that the lack of NKT cells did not change tumor growth. To validate that α-GalCer only affected type I NKT cells (iNKT) and not type II NKT cells, we depleted liver iNKT cells, using an mAb (Supplementary Fig. S2C–S2E; ref. 39). Treating mice with α-GalCer in the absence of iNKT cells resulted in hepatic metastases that grew similarly to metastases in untreated mice, despite having ample type II NKT cells, but in the presence of iNKT cells, metastases were significantly reduced in size (Supplementary Fig. S2D). These data support the notion that α-GalCer specifically stimulates iNKT cells to induce antitumor immunity.

To assess in vivo the functional role of hepatic iNKT cells during sequential α-GalCer stimulation, we used Cxcr6^gfp/+ reporter mice to visualize iNKT cells in real time by spinning disk confocal microscopy, as we and others have done previously (23, 24). CXCR6 expression is not restricted to NKT cells (41, 42); therefore, we refer to the fluorescent cells as Cxcr6^gfp/+ cells. Supplementary Video S1 shows the green-fluorescent Cxcr6^gfp/+ cells migrating within the sinusoids of a control liver. The white macrophages are sessile and illustrate how relatively dynamic the Cxcr6^gfp/+ cells, including the iNKT cells, are. Intravital imaging data at 1, 3, and 5 days revealed that without α-GalCer treatment, a transient increase in Cxcr6^gfp/+ cells occurred in tumor-bearing mice, but only at day 3, with numbers returning to baseline by day 5 (Fig. 4A and B). By contrast, administration of α-GalCer caused a progressive increase in Cxcr6^gfp/+ cells at days 3 and 5 in the absence of tumors, which was further enhanced in tumor-bearing mice, whereby at day 5, a 10-fold increase in the number of Cxcr6^gfp/+ cells was seen (Fig. 4A–C; Supplementary Video S2).

To confirm that the increased hepatic Cxcr6^gfp/+ cells were iNKT cells, we isolated immune cells from the liver and determined by flow cytometry quantities of different cell types within the Cxcr6^gfp/+ population (Fig. 4D and E; Supplementary Fig. S3A–S3C). We found significantly increased percentages of Cxcr6^gfp/+ iNKT cells and amplified iNKT-cell numbers after repeated α-GalCer treatment (Fig. 4D and E). However, we also observed higher quantities of Cxcr6^gfp/+ NK cells (Supplementary Fig. S3B), and there was nearly an order of magnitude fewer Cxcr6^gfp/+ NK cells than Cxcr6^gfp/+ iNKT cells with α-GalCer treatment (Fig. 4D and E). Because α-GalCer is very specific for iNKT cells, the increased NK-cell number is likely due to iNKT-cell activation. There was no change in the level of cytotoxic CD8⁺ T cells (Supplementary Fig. S3C).

To further examine the effect of α-GalCer on the in vivo activity of iNKT cells and tumor suppression, we characterized iNKT-cell dynamics. In the presence of α-GalCer, hepatic iNKT cells elicited more iNKT cell–metastatic cell contacts than in control mice (Fig. 5A and B; Supplementary Videos S3 and S4). The length of time that iNKT cells interacted with metastatic tumors (dwell times) was between 5 and 10 minutes in both untreated and α-GalCer–treated animals (Fig. 5C; Supplementary Videos S3 and S4). When the numbers of metastatic cell–iNKT cell interactions were divided into less than and more than 10-minute dwell times, the percentages were not different in the presence and absence of α-GalCer (Fig. 5D). Indeed, 10% to 30% of CXCR6-gfp⁺ cells moved at a velocity comparable with the normal crawling velocity of iNKT cells in the liver vasculature (10–20 μmol/L/min; ref. 24), whereas the remainder crawled slower in both control and α-GalCer–treated mice (Fig. 5C). Unlike the firm synapses described for target cell killing by CD8⁺ T cells in vitro that lasts longer than 30 minutes (43), we observed no stopping of iNKT cells in the metastatic environment in vivo. It has been shown that CD8⁺ T-cell synapses are not necessary for killing of target cells in vivo and instead kinapses (moving junctions between effector cells and target cells) form, which are much more dynamic and characterized by continuous motility of the effector cells on tumors (43, 44). Hence, intravital microscopy revealed that increased abundance of iNKT cells was responsible for increased contact with liver metastases in α-GalCer–treated mice and not overt retention of cells in the TME. It has previously been shown that NK cells can kill target cells within 5 to 10 minutes in vivo (43, 45), consistent with the time that iNKT cells in this study interacted with tumors.

α-GalCer treatment triggered the upregulation of the immunosuppressive molecules PD-1 on iNKT cells (Supplementary Fig. S4A) and PD-L1 on the sinusoidal endothelium of the liver, where the iNKT cells constantly patrol (Supplementary Fig. S4B). This suggests that increased expression of both receptor and ligand for immunoinhibition of iNKT cells following α-GalCer potentially facilitates a tolerant, immunosuppressive liver environment that supports tumor escape from immune surveillance. However, consistent with the poor efficacy of immunotherapy with checkpoint inhibitors in patients, treating mice with PD-1–blocking antibody during the course of α-GalCer administration did not further reduce liver metastasis (Supplementary Fig. S5A and S5B).

An immediate innate cell–like feature of iNKT cells is the rapid release of copious amounts of immunoregulatory cytokines without the need for differentiation, unlike CD4⁺ and CD8⁺ T cells (46, 47). However, previous studies in mice show that repeated stimulation of iNKT cells with intravenous α-GalCer induces anergy and cessation of proinflammatory cytokine production (48), which is not consistent with the improved survival with multiple, but not single, administration of α-GalCer that we observed. We gave α-GalCer every 2 days and measured multiple cytokines 4 hours after each stimulation (Fig. 6; Supplementary Fig. S6A). Significant increases in plasma proinflammatory (IFNγ) and anti-inflammatory (IL10) cytokines were seen over the first 8 days, with the former peaking early and the latter peaking later (Fig. 6A). The ratio of IFNγ and IL10 further highlighted the significant rise of IFNγ very early after α-GalCer, which declined over time, and the increase of IL10 over time (Fig. 6B). Release of IL4 and TNFα also peaked early and decreased over time, although concentrations were orders of magnitude less than IFNγ and IL10 (Fig. 6A). α-GalCer also induced high IFNγ up to day 8 (Fig. 6A). Consistent with cytokine release was the increase in granularity and size in iNKT cells, as well as a significant increase in granularity of NK cells (Supplementary Fig. S6B). NK cells do not respond directly to α-GalCer, so this may be a result of the cytokine release by iNKT cells.

We hypothesized that the early and robust cytokine response elicited by α-GalCer–activated iNKT cells created an antitumor liver environment that effectively suppressed liver metastasis, as seen in Fig. 2B. A significant elevation of proinflammatory and regulatory cytokines can engage multiple arms of the antitumor response (49), and therefore, we examined the role of inflammatory cytokines in our liver metastasis model. Despite previous reports (29, 50), neutralization of IFNγ alone (Supplementary Fig. S7A and S7B) did not reverse the effects of α-GalCer stimulation. Inhibition of IFNγ and TNFα (Supplementary Fig. SS7C and S7D) did not further induce a significant antitumor effect with α-GalCer treatment. Clearly, the change from a tolerant TME into a more proinflammatory TME was not sufficient to reverse the antitumoral effects of iNKT cells.

We showed (Fig. 2C) that the metastatic tumors led to mortality (Fig. 2C). To determine whether tumor cells in α-GalCer–treated mice underwent functional genetic or epigenetic immunoediting leading to resistance (51), we fluorescently sorted CT26-iRFP cells from the livers of either untreated or α-GalCer–treated mice and subsequently intrasplenically injected into mice that were left untreated or repeatedly injected with α-GalCer (Fig. 7A). After 5 days, when we observe the most robust differences, we performed intravital imaging to assess the metastases in the liver. Tumor cells transferred from α-GalCer–treated mice seeded the liver to the same extent as tumor cells from untreated mice (Fig. 7B). α-GalCer delivery to serially transplanted naïve or α-GalCer–treated tumor cells revealed that both tumors were still responsive to treatment (Fig. 7B), demonstrating that CT26-iRFP cells do not acquire a genetic or epigenetic survival benefit in response to α-GalCer. It has been reported that cancer cells will upregulate coinhibitory molecules in response to specific cytokines to resist elimination by immune cells (52). Indeed, we did observe upregulation of PD-L1 on the cell surface of CT26-iRFP cells following α-GalCer treatment (Supplementary Fig. S8A); however, this was insufficient to induce further treatment resistance in metastatic tumor cells in the serial transplant model (Fig. 7B).

In a final series of experiments, we measured the cell surface levels of CD1d on CT26-iRFP cells in vivo. Although up to 7% of cancer cells expressed CD1d in wild-type mice, none was detected in CD1d^−/− mice, suggesting that CD1d⁺ tumor cells in wild-type mice might be immune cells that phagocytosed tumor cells and therefore expressed iRFP (Supplementary Fig. S8B). This also suggests that α-GalCer presentation to iNKT cells could be limiting, unless other immune cells presented α-GalCer to the iNKT cells. Previous studies suggest that α-GalCer–loaded CD1d monomers, α-GalCer–pulsed DCs (53, 54), nanoparticle-formulated α-GalCer (55), or liposome-loaded α-GalCer (56) all increase efficacy of the glycolipid. Because of this, we used α-GalCer bound to CD1d monomers and subjected mice to repeated treatment, similar to free α-GalCer treatment (Fig. 7C). Surprisingly, in 60% of the mice, liver metastases were not present at all at day 10 after tumor inoculation, while in 40% of the mice, the tumors were very small (Fig. 7D and E). It is worth mentioning that the monomers are made in a lentivirus expression system in cell lines (and not in bacterial systems), so there is little chance of Toll-like receptor ligand contaminants. This is further validated by the complete lack of biological efficacy of the unloaded CD1d monomer (Fig. 7D and E).

Because the prognosis for many patients with cancer depends on the extent of primary tumor metastasis (5), surgery to eliminate cancer is the standard treatment and is intended to be curative. However, there is emerging evidence that surgical operation enhances the recurrence of cancer through postoperative immunosuppression and establishes a cancer cell–permissive environment (9, 13–15, 57). The incidence of metastasis could be mitigated by creating a less conducive environment in the liver microcirculation, where we noted tumor cells to adhere and iNKT cells that constantly patrol the sinusoids. Indeed, our data highlight that administration of multiple doses of α-GalCer, given just before the time of tumor cell seeding in the liver, created a tumor-hostile, proinflammatory environment and had therapeutic efficacy, which is concordant with previous reports (29, 48, 57–59). Delaying α-GalCer treatment by just 5 days resulted in significant metastasis. In line with these results, the success of α-GalCer in clinical trials of established solid tumors is very limited (35).This suggests that there is a short perioperative and postoperative period that allows modulation of the immune response to reduce and inhibit cancer occurrence in the liver after surgery. Therefore, we would advocate for α-GalCer in a situation like colorectal surgery, in which there is increased risk of metastasis to the liver but not for patients with already existing malignancies.

Previous intravital imaging reveals that although the macrophages of the liver (Kupffer cells) are able to engulf opsonized tumor cells, the tumor cells have to be within proximity of the macrophages, as they are sessile and shown not to patrol the sinusoids (22). As such, we attempted to activate motile iNKT cells that could access all tumor cells that resided in the sinusoids. Previous work has shown that iNKT cells have cytotoxic capacity toward tumor cells (60, 61). The metastatic colon carcinoma cells did not express CD1d, but the iNKT cell still limited tumor growth (62). Most in vitro experiments have reported that effector cells adhere avidly to target tumor cells before imparting cytotoxicity and killing. However, in vivo work has raised the possibility that firm synapses described for target cell killing by CD8⁺ T cells in vitro that lasted 30 minutes or longer (43) are not readily apparent in vivo. During a dynamic interaction of a motile effector cell with a target cell, kinapses (moving junctions) are formed and are a more likely scenario of cytotoxic killing (44). We observed no stopping of CXCR6-gfp⁺ cells, the majority of which were iNKT cells, in the metastatic environment in vivo, and the time of interaction also did not increase per individual CXCR6-gfp⁺ cell. Although no one has imaged iNKT-cell killing in vivo, NK cells have effective cytotoxic functions within a 5-minute contact interval (43), which is consistent with our 8- to 10-minute contact time between CXCR6-gfp⁺ cells and metastatic tumor cells. α-GalCer increased the number of iNKT cells and NK cells (10-fold), as well as their granularity. As such, we adopted the concept of “per capita killing” (44), which describes the killing properties of the entire population of CXCR6-gfp⁺ cells rather than individual cells. iNKT cells have been shown to induce killing via FAS ligand, TRAIL, and perforin, supporting our notion that iNKT cells could prevent growth of tumors via a cytotoxic contact mechanism (60). Intravital imaging also identified that single tumor cells or very small metastases still seeded the liver microenvironment, persisted, and were not killed. It is possible that kinapses that integrate signaling and maintain communication between effector and a single nonreplicating tumor cell could function to induce tolerance (63). Indeed, increased metabolism during replication may be necessary for iNKT cells to impart their cytotoxic activity, a concept that is well documented for senescent tumor cells (64). As a whole, α-GalCer increased survival of mice; however, ultimately, tumor cells escaped iNKT-cell antitumor responses, large tumors formed, and the mice succumbed to disease.

By dissecting the role of iNKT cells in α-GalCer–mediated tumor control, we confirmed that iNKT cells were the therapeutic target of α-GalCer (29). In contrast to previous reports where chronic treatment of mice with α-GalCer led to long-lasting anergy (48, 65) and to an almost complete loss of iNKT cells in the liver, we showed, using imaging, that iNKT cells robustly expanded and retained motility similar to inactivated iNKT cells (24). This difference may hint at the dose and treatment regimen that was used for α-GalCer. Regardless, our data suggest that some iNKT-cell effector functions dampened with time. For instance, we observed a very robust increase in cytokine production over a few days, which waned by 8 days. Nevertheless, continuing α-GalCer administration after 10 days still showed significant benefit, suggesting additional effector mechanisms. Although α-GalCer only stimulates iNKT cells, and no other cells, through its invariant T-cell receptor, iNKT-cell activation also led to increased NK-cell numbers and increased granularity. It is known that IFNγ can induce TNFα production of Kupffer cells (66), and indeed, we observed increased levels of both cytokines very early. It is likely that these and numerous other effector cells contributed to increased antitumor properties associated with α-GalCer, and inhibiting a single or even two effectors was not sufficient. Several previous reports have shown that the α-GalCer–mediated antitumor activity is abrogated in Ifng^−/− mice (29, 67). We attempted to inhibit IFNγ alone in an effort to reverse the α-GalCer antimetastatic properties, but this failed. Blocking both IFNγ and TNFα did cause some reversal, albeit only 30%, suggesting that the iNKT-cell effector functions are multifactorial including contact cytotoxicity, cytokine production, and perhaps controlling cancer progression through modulating and inhibiting tumor-promoting myeloid cells in the TME (61, 68, 69). Although it remains unclear why the mice treated with α-GalCer eventually succumbed to cancer, we can rule out a number of possibilities. There are increasing numbers of studies showing that tumor cells can adopt an immunoresistant phenotype through genetic or epigenetic modifications (51, 70, 71). Rather than trying to measure the many potential changes, we serially transferred tumors that had previously been exposed to α-GalCer–stimulated iNKT cells and observed no resistance to a second α-GalCer stimulation regimen. This finding suggests that perhaps effector cells, rather than tumor cells, eventually become resistant to therapy. In fact, high IL10, which we also measured, can decrease NK-cell cytotoxicity (72). We also observed some increase in PD-1 on the tumor cell surface with α-GalCer, but this did not induce any subsequent immune inhibition. The sinusoidal endothelium upregulated PD-L1, whereas the iNKT cells had increased expression of PD-1; however, inhibition of these molecules failed to improve antitumor immunity. This is consistent with data from patients with human colorectal cancer, whereby the response rate to PD-1/PD-L1 checkpoint blockade is very low (3). Although we cannot formally rule out anergy of iNKT cells after multiple rounds of α-GalCer therapy, as cytokines did decrease, there was an expansion of granular iNKT cells that maintained a patrolling phenotype, suggesting that not all aspects of the iNKT cells were turned off. It is likely that anergy will be more prevalent with increasing doses of α-GalCer in our model; however, human iNKT cells expand without the induction of anergy after repeated α-GalCer stimulation (73), further supporting the use α-GalCer prophylactically in patients with cancer undergoing surgery.

Experimental approaches to limit the induction of iNKT-cell anergy (48, 53) involve the use of a vehicle for α-GalCer administration, such as α-GalCer–pulsed DCs (53, 54), nanoparticle-formulated α-GalCer (55), liposome-loaded α-GalCer (56), and α-GalCer–loaded CD1d monomers, the latter having an antibody against a tumor-specific antigen. We took the latter approach, but without the tumor-seeking antibody, as the majority of iNKT cells in mice and humans are in the liver and should be activated by the circulating monomers. Indeed, we saw significant improvement of the antitumor response in vivo with α-GalCer bound to CD1d monomers. In fact, 60% of the mice liver metastases were absent at day 10 after tumor inoculation, whereas 40% of the mice had very small tumors. Additional work will be required to elucidate the mechanism behind this enhanced antitumor function.

There are limitations to our experimental liver metastasis model. iNKT cells and B cells interact through CD1d after lipid presentation (74) and mount type 2–dependent T-cell responses in germinal centers with noncognate help from iNKT cells. The spleen provides a reservoir of naïve B cells, and splenectomy reduces the amount of available circulating B cells. Nevertheless, it is worth noting that the liver has a very large endogenous population of B cells (75) and other CD1d-expressing cells, such as Kupffer cells. Therefore, the noncognate help by iNKT cells would still be present in the liver of splenectomized mice (24, 76). Another limitation of our study is the use of Cxcr6^gfp/+ reporter mice (24) for the imaging of iNKT cells due to the lack of iNKT-cell–specific reporter mice. In naïve mice, 60% to 70% of liver CXCR6-gfp⁺ are CD1d-restricted iNKT cells. This proportion further increases after α-GalCer activation so that the majority of responses observed could attribute to iNKT cells. Finally, mice have more liver iNKT cells than humans, which could also limit therapy in humans, but α-GalCer does increase numbers of iNKT cells in humans.

In summary, our experiments provide evidence that intermittent, i.v. α-GalCer administration activates a regulatory function in iNKT cells, and this could be used for therapeutic immunomodulation to suppress cancer growth. More specifically, the use of α-GalCer in the adjuvant setting could provide a potential survival benefit for patients with cancer undergoing primary tumor resection of solid tumors when administered before or during surgery to prevent the incidence of early relapse by modulating the postoperative immune response. Although for proof of principle, we used α-GalCer, and future work, especially in the clinic, could consider modified α-GalCer molecules that would more specifically release selective cytokines such as IFNγ and not IL4 and IL10. We speculate that this could be how a monomer of CD1d–α-GalCer functions and should be tested in the future.

No disclosures were reported.

L. Babes: Conceptualization, data curation, formal analysis, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing. R. Shim: Data curation. P. Kubes: Conceptualization, resources, supervision, funding acquisition, writing–review and editing.

The authors thank Dr. Karen K.H. Poon for conducting the cell sorting of CT26-iRFP cells, Dr. Xuequing Lun for performing some of the intravenous α-GalCer injections (Fig. 2E), as well as Bruna Araujo David for providing ALT levels after acetaminophen treatment (Supplementary Fig. S1).

This work was supported by the Canadian Cancer Society and the Alberta Innovates Health Solutions graduate studentship.

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.