Lactate-Mediated Acidification of Tumor Microenvironment Induces Apoptosis of Liver-Resident NK Cells in Colorectal Liver Metastasis

Colorectal cancer represents a major health problem worldwide, with more than 1.3 million new cases per year (1). Advancing age of populations globally and rising rate of obesity are driving increasing rates of colorectal cancer (2). Metastasis is a major cause of morbidity and mortality in colorectal cancer patients. In the absence of distant metastases (stages I–III), 5-year survival ranges between 70% and 90%; however, in patients with metastatic disease (stage IV), this is only 14% (3). The liver is the most common site of distant metastasis, with up to 50% of patients developing colorectal liver metastasis (CRLM; refs. 4, 5). However, it remains unknown how metastatic colorectal cells colonize an immunocompetent organ and avoid local immunosurveillance.

The adult human liver contains an extensive repertoire of antitumor immune cells. Innate lymphoid cells, including natural killer (NK) cells, natural killer T (NKT) cells, mucosal associated invariant T (MAIT) cells, and γδ T cells, predominate in the liver. Conventional CD4⁺ and CD8⁺T cells are also present, with an increased proportion of cytotoxic CD8⁺ T cells (6–11). NK cells are group 1 innate lymphoid cells (ILC), important in antiviral and tumor immunity, and in particular the control of metastasis (12), with NK-cell therapies currently being trialed for the treatment of CRLM (13). The human liver contains resident NK cells possessing enhanced antitumor activity characterized by increased degranulation (14–16). These cells are characterized as CD56^bright NK cells, express the chemokine receptor CXCR6, and have a unique transcription factor profile (Eomes^hiTbet^lo).

The evidence that immune infiltration of colorectal cancer is a determinant of disease outcome has led to the development of the Immunoscore as a routine diagnostic test (17–19). The density of T- and NK-cell infiltration of CRLM tumors has been shown to positively predict survival (20). However, infiltration of T and NK cells was highly variable, and some tumors were almost devoid of NK cells. NK cells appear to play a central role in the outcome of CRLM, demonstrated by their depletion and compromised function in primary colorectal cancer tumors and blood (21, 22). It remains unknown whether liver-resident NK cells are dysfunctional in metastases to the liver. We believe this loss of immunosurveillance is in part driven by the unique tumor microenvironment. The liver microenvironment is dramatically altered in the presence of CRLM (23–25), and the metabolic changes associated with tumor growth can have profound effects on infiltrating immune cells, especially T and NK cells, which require metabolic reprogramming upon activation (26–29).

In this study, we have characterized the distribution of liver-resident NK cells in different regions of tumor-bearing liver, tumor, tumor adjacent, and distal tissue from the resection margin. Using in vitro culture techniques, generating conditioned media from tumor and healthy tissue, we have identified lactate, a metabolite of glycolysis, which induces the apoptosis of liver-resident NK cells. This pathway provides the rationale for potential therapeutic approaches for the reinvigoration of liver-resident NK cells to limit CRLM growth.

Patients attending the National Liver Unit at St. Vincent's University Hospital for hepatic metastasectomy secondary to tumors of colorectal origin were eligible for inclusion in this study. Eighteen patients were recruited preoperatively (Table 1). Donor liver biopsies and liver perfusate (LP) samples were collected during orthotopic liver transplantation (n = 17; Table 1). All patients provided written informed consent. This study was carried out in accordance with the Declaration of Helsinki and was reviewed and approved by the ethics committee of St. Vincent's University Hospital.

Hepatic mononuclear cells (HMNCs) were isolated from donor LP, as previously described (14, 30). Briefly, exsanguinated donor livers were flushed and immersed in University of Wisconsin solution for transport (transport medium). In theater, the donor liver was flushed with saline to remove remaining transport medium (flush). The transport medium and flush are combined to make up the liver perfusate. HMNCs were then isolated by density centrifugation after filtration of perfusate. HMNCs were isolated from biopsies from tumor, tumor adjacent, and distal tissue from hepatic resection for CRLM, by mechanical and enzymatic digestion (collagenase type V, Sigma-Aldrich), as previously described (31).

Donor liver biopsies and CRLM tumors were used to generate tissue-conditioned media (CM) as previously described (24, 32). Tissue was cut into sections approximately 5 mm³ and incubated in 500 μL of RPMI-1640, with 10% fetal calf serum and 1% penicillin/streptomycin (Gibco) for 72 hours at 37°C. The supernatant was centrifuged to remove cell debris and stored at −80°C until use.

HMNCs (1 × 10⁵) were stained in FACS buffer for 30 minutes in the dark with fluorescently labeled monoclonal antibodies to determine the phenotypic differences between hepatic NK cells. A full list of antibodies used can be found in Supplementary Table S1. Dead cell exclusion was performed using fixable viability stain 780 (BD Biosciences). Mitochondrial mass was assessed using MitoTracker Deep Red (Thermo Fisher); mitochondrial ROS production was measured using MitoSox (Thermo Fisher). Flow cytometry analysis was performed using an LSRFortessa, using fluorescence-minus-one controls (FMO; BD Biosciences), and data were analyzed using FlowJo (Version 7.6.5, Tree Star).

NK cells were isolated from fresh LP by FACS. Briefly, donor LP was stained with monoclonal antibodies. NK cells were sorted into two populations CD56^bright CD16^+/− and CD56^dim CD16⁺⁺ using a FACSAria cell sorter (BD Biosciences).

Total RNA was extracted from FACS-sorted NK cells from LP and tissue from tumor and healthy tissue, using TRIzol reagent (Thermo Fisher). Reverse transcription was performed using the SuperScript VILO Master Mix (Thermo Fisher). Quantitative PCR was performed using primers for SLC16A1, SLC16A3, and SLC2A1 (see Supplementary Table S1 for primer sequences) using the PowerUp SYBR Green Master Mix (Thermo Fisher) on a StepOnePlus Real-Time PCR System (Thermo Fisher). RPS15 and B2M were used as reference genes to generate normalized relative expression values for target genes.

The K562 and SW620 cell lines were purchased from ATCC between 2008 and 2010. Experiments were performed on cells less than 15 passages from purchase. Cells were routinely tested for contamination using the LookOut Mycoplasma PCR Detection Kit (Sigma-Aldrich). Cell identity of 168 SW620 was performed by IdentiCell (Denmark). Cell identity of K562 cells has not been performed.

Target cells (K562 or SW620) were labeled with CellTrace Violet (CTV). FACS-sorted liver NK cells were incubated with target cells at a ratio of 2:1 for 4 hours. Anti-CD107a was added to assess degranulation. Cytotoxicity was assessed by propidium iodide staining (PI). CTV⁺ cells were gated, and PI⁺ cells were measured as a percentage of target cells.

Chemotaxis of liver NK cells was assessed using transwell inserts. Briefly, the lower chamber of the well was filled with serum-free RPMI-1640 media supplemented with 10% v/v CM. Liver NK cells (1 × 10⁵) were applied to the upper chamber. Plates were incubated for 2 hours at 37°C. Cell numbers were assessed by flow cytometry using 123Count eBeads (eBiosciences) to enumerate absolute cell counts.

Apoptosis was assessed by annexin V/7-AAD staining. After treatment, NK cells were placed in annexin staining buffer and incubated for 20 minutes at RT with 2 μL annexin V. After washing, 1 μL 7-AAD was added to each tube before acquisition on flow cytometer. Apoptotic cells were defined as annexin V⁺ 7AAD⁺. Caspase 8 activity was blocked in experiments using Z-IETD-FMK (BD Biosciences). Cells were treated with Z-IETD-FMK (50 μmol/L) for 2 hours before the addition of TCM for 24 hours. α-Cyano-4-hydroxycinnamic acid (αCHC, Sigma-Aldrich) was used to block lactate transport. αCHC (500 μmol/L) was administered for 2 hours before the addition of TCM for 24 hours. MitoTempo (Sigma-Aldrich), a mitochondrial ROS scavenger, was used in further experiments. Cells were treated with MitoTempo (50 μmol/L) for 2 hours before the addition of TCM for either 2 or 24 hours.

Lactate concentrations were analyzed using a colorimetric enzymatic assay (Sigma-Aldrich). The lactate determination kit comes with premade lactate assay buffer, lactate probe, enzyme mix, and standards. Samples were prepared and analyzed as per the manufacturer's protocol. Briefly, CM samples were diluted 1:50 in lactate assay buffer. Sample (50 μL) was incubated with 50 μL reaction mixture for 30 minutes at RT. Absorbance was measured at 570 nm. Similarly, glucose concentrations were measured using an enzymatic assay as per the manufacturer's instructions (Abcam). Conditioned media were diluted 1:50 in glucose assay buffer and incubated with an equal volume of reaction buffer, at 37°C for 30 minutes. Absorbance was measured at 570 nm.

Intracellular pH was measured using the pH sensitive dye pHrodo (Thermo Fisher). This compound is retained within the cytosol and becomes increasingly fluorescent as pH decreases. Fluorescence was measured using flow cytometry. Controls were established by treating NK cells with buffers of pH 7.5, 6.5, and 5.5.

Intracellular ATP was measured by luciferase assay (Invitrogen). After treatment, NK cells were lysed using DTT buffer. A luciferase construct, requiring ATP for activity, was used to determine intracellular ATP concentrations. Luminescence was measured using Fluoroskan Ascent FL (Labsystems).

IL15 concentrations in CM were measured by ELISA according to the manufacturer's instructions (PeproTech).

Statistical analysis was carried out using Prism GraphPad Version 5.0. For comparison of two unmatched groups, Mann–Whitney U test was used. Comparison of more than two unmatched groups was performed using Kruskal–Wallis test, with Dunn multiple comparison test. For paired comparisons, Friedman test with Dunn multiple comparison tests was used for more than two groups, and Wilcoxon signed rank test was used for comparison of two matched groups. Within an experiment, *, ** and *** represent P < 0.05, P < 0.01, and P < 0.001, respectively.

We have investigated the presence of liver-resident CD56^bright NK cells in CRLM tumors and surrounding tissue. NK cells are significantly reduced in CRLM tumors (tumor 4.5% ± 0.8%, adjacent 13.6% ± 1.9%, distal 15.6% ± 2.2%, P < 0.0001; Fig. 1A–C). Patients with recurrent hepatic disease had significantly lower intratumoral NK-cell numbers (recurrence: 41.2 ± 24.1/mg, no recurrence: 460.7 ± 193.0/mg, P = 0.047; Fig. 1D). T cells appear unaffected in CRLM, with no significant difference in the population number throughout the tumor-bearing organ (Fig. 1E and F). No correlation between T-cell infiltration of the tumor and disease recurrence was observed (Fig. 1G). The largest reduction in NK cells was seen from the liver-resident CD56^bright population (tumor 38.1% ± 3.4%, adjacent 50.8% ± 3.2%, distal 56.0% ± 3.2%, P = 0.0009; Fig. 1H–J). This depletion of liver-resident CD56^bright NK cells was significantly worse in patients who went on to develop recurrent CRLM (recurrence: 10.4 ± 1.3/mg, no recurrence: 204.7 ± 124.2/mg, P = 0.0238; Fig. 1K).

Liver-resident NK cells display an activated phenotype compared with conventional CD56^dim cells isolated from liver perfusate. They display increased expression of activation markers, including NKG2D, NKp46, and NKp44 (Supplementary Fig. S1A–F). They express cytotoxic molecules such as perforin, granzyme B, granzyme A, and granzyme K (Supplementary Fig. S1G–K). Despite lower expression of granzyme B and perforin compared with CD56^dim NK cells from liver perfusate, liver-resident CD56^bright NK cells degranulate significantly more in response to tumor cells and are capable of direct cytotoxicity against the metastatic colorectal cell line SW620, equal to CD56^dim NK cells (Supplementary Fig. S1L–O).

The remaining CD56^bright NK cells maintained a tissue-resident phenotype (Eomes^hi Tbet^lo; Fig. 2A and B) throughout the tumor-bearing tissue. In addition, they maintained expression of other markers of tissue residency, including CXCR6, CD69, CD49a, and TRAIL (Fig. 2C–G). This indicated that CD56^bright NK cells were not infiltrating peripheral blood cells and that liver-resident NK cells throughout the liver may be activated in CRLM. A decrease in the absolute count of CD56^dim NK cells was also observed in the tumor; however, the scale of this change was smaller than liver-resident NK cells (Supplementary Fig. S2A–C). The remaining CD56^dim cells were Eomes^lo Tbet^hi and failed to highly express markers of tissue residency, CXCR6, CD69, and CD49a. However, they did express TRAIL, which may indicate an activated phenotype (Supplementary Fig. S2D–H). Tumor-infiltrating liver-resident NK cells displayed increased expression of NKp44 (a marker of NK-cell activation) but no associated increase in activating receptors such as NKG2D, NKp46, or NKG2C (Fig. 3A–E). We next analyzed the expression of cytotoxic molecules in liver-resident NK cells from tumor-bearing liver. Similar expression of perforin, granzyme A, granzyme B, or granzyme K was observed in each site, indicating that cytotoxic potential was maintained in tumor-infiltrating CD56^bright NK cells (Fig. 3F–J).

To address the depletion of NK cells from CRLM tumors, we used a CM culture system to identify the mechanism by which CD56^bright NK cells are lost from tumor tissue. Tumor and donor liver tissue were cultured for 72 hours, and supernatants were used to treat NK-cell subsets isolated from donor liver perfusate. Treatment with CRLM-CM had no effect on the migration ability of the liver-resident NK cells in vitro (Supplementary Fig. S3A–S3C). CD56^dim NK cells had a reduced capacity to migrate toward TCM compared with healthy LCM, possibly accounting for their decreased numbers in CRLM tumors (Supplementary Fig. S3D).

In the absence of a defect in migration, we hypothesized that the tumor microenvironment was inducing apoptosis of liver-resident NK cells. NK-cell subsets from donor liver perfusate were treated with liver (LCM) and CRLM (TCM) CM for 24 hours. There was a significant increase in the proportion of late apoptotic (Annexin V⁺/7-AAD⁺) liver-resident NK cells treated with TCM compared with LCM but no change in early apoptotic cells (LCM: 21.1% ± 2.8%, TCM: 39.2% ± 8.1%, P = 0.034; Fig. 4A–C), with no similar increase observed in CD56^dim NK cells (P = 0.92; Fig. 4D and E). The proportion of late apoptotic cells was significantly higher in CD56^bright NK cells treated with TCM, compared with CD56^dim cells (Fig. 4F). This apoptosis was not reversible when extrinsic cell signals (i.e., TRAIL, Fas signaling) were blocked through the Caspase 8 inhibitor Z-IETD-FMK (P = 0.88; Fig. 4G).

The tumor microenvironment can be inhospitable to immune cells due to the reduced availability of growth factors, essential amino acids, and metabolites. We therefore investigated whether liver-resident NK cells showed signs of mitochondrial stress in the CRLM tumor microenvironment. There is a decrease in the mitochondrial mass of CD56^bright NK cells in CRLM tumors compared with surrounding tissue, which is not seen in tumor-infiltrating CD56^dim NK cells (Fig. 4H–J). This is accompanied by the production of mitochondrial ROS, which is significantly increased in tumor-infiltrating CD56^bright NK cells compared with surrounding tissue (Fig. 4K–L). This is not the case with CD56^dim NK cells, which do not produce mitochondrial ROS (Fig. 4M). Mitochondrial mass is significantly lower in liver-resident NK cells compared with CD56^dim NK cells in liver perfusate (Supplementary Fig. S4A and B), suggesting an intrinsic difference in metabolic activity may underpin the sensitivity of liver-resident CD56^bright NK cells to apoptosis.

We next attempted to identify the factor responsible for inducing mitochondrial stress in tumor-infiltrating NK cells. NK cells are dependent on IL15 for survival, and withdrawal of this growth factor can cause mitochondrial dysfunction. However, in supernatant from CRLM, IL15 was detected at similar concentrations to donor tissue (TCM: 335.6 ± 74 pg/mL, LCM: 300 ± 15.4 pg/mL, P = 0.84; Fig. 4N). Availability of glucose in the tumor microenvironment can also be a rate-limiting step in immune activation and survival; however, in this in vitro system glucose did not differ between the LCM and TCM but may play a greater role in vivo (TCM: 17.1 ± 3.6 mmol/L, LCM: 17.2 ± 1.5 mmol/L, P = 0.49; Fig. 4O). Furthermore, expression of glucose transporter (GLUT1), amino acid transporter (CD98), and transferrin receptor (CD71) did not differ between donor NK-cell subsets and was unchanged in tumor-infiltrating CD56^bright NK cells, which indicates nutrient availability may not be drastically altered (Supplementary Fig. S5A–H). Despite comparable concentrations of glucose, CRLM tumors appeared more glycolytically active compared with nontumor tissue, producing significantly higher concentrations of lactate (TCM: 11.5 ± 2.8 mmol/L (range, 1.59–24.9 mmol/L), LCM: 1.7 ± 0.3 mmol/L, P = 0.002; Fig. 4P) and increased expression of GLUT1 mRNA in tumor samples compared with adjacent tissue, as well as increased expression of the lactate transporter SLC16A3 (Supplementary Fig. S6A–S6C). This increased production of lactate led to a significant decrease in the pH of CM (TCM: pH 6.6 ± 0.16, LCM: pH 7.4 ± 0.08, P = 0.0043; Fig. 4Q).

We used two forms of lactate to investigate its effect on donor liver–resident NK cells, sodium lactate (basic form), and lactic acid (acidic form) at 10 and 20 mmol/L. No significant change was observed in the proportion of early apoptotic NK cells under any condition tested (Supplementary Fig. S7A–S7G). Sodium lactate did not induce an increase in late apoptotic cells in either CD56^bright or CD56^dim NK cells (Fig. 5A and B). Lactic acid induced a significant increase in late apoptotic CD56^bright NK cells (media: 13.5% ± 2.8%, 10 mmol/L lactic acid: 20.7% ± 4.6%, 20 mmol/L lactic acid: 43.4% ± 9.1%, P = 0.01; Fig. 5C) but not in CD56^dim NK cells (Fig. 5D). In order to determine if this process was pH mediated, medium was altered to pH 7.4 and 6.4 to mimic the pH of normal tissue and tumor tissue with 20 mmol/L lactic acid. Liver-resident CD56^bright NK cells undergo significantly higher frequency of apoptosis in pH 6.4 (media: 11.5% ± 2.7%, pH 7.4: 23.9% ± 6.5%, pH 6.4: 32.9% ± 5.9%, P = 0.041; Fig. 5E). Apoptosis did increase in CD56^dim NK cells; however, this was not statistically significant (Fig. 5F). Although there appears to be a trend toward increased apoptosis in CD56^dim NK cells treated with lactic acid and low pH media, this was not significant and highlights the increased susceptibility of liver-resident CD56^bright NK cells to these environments.

In order to determine the mechanism by which CD56^bright NK cells undergo apoptosis, donor liver–resident NK cells were treated with lactic acid for 2 hours at 10 and 20 mmol/L. After 2 hours, the intracellular pH of NK cells decreases significantly (media: pH 7.4 ± 0.15, 10 mmol/L lactic acid: pH 6.9 ± 0.2, 20 mmol/L lactic acid: pH 6.5 ± 0.4, P = 0.0046; Fig. 5G and H). This was accompanied by an increase in mitochondrial ROS production, similar to that seen in NK from CRLM tumor samples (media: 10.4% ± 1.9%, 10 mmol/L lactic acid: 13.3% ± 2.6%, 20 mmol/L lactic acid: 20.0% ± 0.9%, P = 0.021; Fig. 5I). In addition, intracellular ATP dropped significantly in lactic acid–treated cells (media: 1070 ± 334 pmol/10⁵ cells, 10 mmol/L lactic acid: 688 ± 230 pmol/10⁵ cells, 20 mmol/L lactic acid: 589 ± 178 pmol/10⁵ cells, P = 0.008; Fig. 5J). Furthermore, when liver-resident NK cells were treated with TCM for 2 hours, a decrease in intracellular pH was observed (TCM: pH 6.1 ± 0.15, LCM: pH 7.38 ± 0.16, P = 0.018; Fig. 5K), accompanied by the production of mitochondrial ROS (TCM: 16.5% ± 3.5%, LCM: 5.3% ± 0.86%, P = 0.037; Fig. 5L). Together, these results indicate that lactic acid in the tumor microenvironment causes pH-dependent mitochondrial stress and metabolic dysfunction, which promotes apoptosis.

We next attempted to block lactate transport using α-cyano-4-hydroxycinnamic acid (αCHC). However, pretreatment with αCHC did not reduce mitochondrial ROS production after exposure to lactic acid (Fig. 6A and B). Furthermore, NK-cell subsets do not differ in their expression of lactate transporters SLC16A1 or SLC16A3 (Fig. 6C and D). These results indicate that the effects of lactic acid were independent of transport into the cell and are more likely related to changes in the pH of the microenvironment.

In order to determine whether NK-cell apoptosis was ROS mediated, we utilized MitoTempo, a mitochondria-specific ROS scavenger. Treatment with MitoTempo significantly reduced mitochondrial ROS (lactic acid: 36.1% ± 9.3%, lactic acid + MitoTempo: 18.0% ± 8.5%, P = 0.002; Fig. 6E and F) and rescued liver-resident NK cells from apoptosis after 24-hour treatment with lactic acid, decreasing the proportion of late apoptotic cells (lactic acid: 30.7% ± 2.5%, lactic acid + MitoTempo: 18.1% ± 2.0%, P = 0.0013; Fig. 6G–I).

In this study, we have characterized the NK-cell repertoire of tumor-bearing liver in patients with CRLM and demonstrated a significant depletion of liver-resident NK cells from CRLM tumors. We have evidence that NK-cell depletion correlates with CRLM recurrence. This depletion is induced by an accumulation of lactate in the tumor microenvironment, causing a reduction in intracellular pH in hepatic NK cells, leading to mitochondrial dysfunction and apoptosis.

NK cells are essential components of the hepatic antitumor immune repertoire. Liver NK cells have been shown to be cytotoxic against tumor cells and should provide the liver with ample antitumor immunity (33–35). The liver-resident NK-cell subpopulation displays increased expression of cytotoxicity receptors, expresses cytotoxic molecules, and is capable of similar direct cytotoxicity against SW620 cells, compared with conventional NK cells. Furthermore, TRAIL expression was significantly increased throughout the tumor-bearing liver, indicating that NK-cell immunosurveillance is heightened in CRLM.

Previous reports have been contradictory, with one claiming that the presence of NK cells in CRLM is correlated with a reduced life expectancy (36) and another with improved disease outcome (20). However, specific NK-cell subsets were not investigated in either of these studies, perhaps accounting for the opposing conclusions. Here, we have identified a significant reduction in liver-resident NK cells from CRLM tumors, despite effective migration toward CRLM-CM. This depletion correlated with hepatic recurrence postresection. This finding parallels results from primary colorectal cancer tumors, which are relatively devoid of NK cells, with surrounding tissue containing normal frequencies of NK-cell populations (21).

We hypothesized that NK cells undergo apoptosis after migration into the tumor microenvironment. Ex vivo tumor-infiltrating CD56^bright NK cells displayed increased mitochondrial ROS production and decreased mitochondrial mass compared with cells from surrounding tissue and infiltrating CD56^dim NK cells. CRLM tumors appear highly glycolytic, with higher GLUT1 expression, lactate transporters, and extracellular lactate. Lactate is produced by tumor cells during Warburg metabolism and is then exported via monocarboxylate transporters (MCT1/4) along with protons, lowering the extracellular pH (26). In vitro, lactic acid induced significant apoptosis of liver-resident NK cells. The acidic environment provoked a change in intracellular pH of liver-resident NK cells, production of mitochondrial ROS, and reduced ATP generation, inducing intrinsic ROS-mediated apoptosis. Lactate is a potent immune modulator in the tumor microenvironment, known to polarize macrophages and neutrophils, suppress DC activity, and dysregulate and kill T and NK cells (29, 37–40), and our results implicate lactate as the intrinsic cell death signal for liver-resident NK cells. Mechanistically, our results are similar to that described in murine NK cells in a metastatic B16 melanoma model by Brand and colleagues (29), and our results extend on this work to illustrate how a ROS scavenger (MitoTempo) is able to mitigate the effect of pH-induced damage and prevent apoptosis in human tissue-resident NK cells. This result highlights the potential of targeting lactate and the microenvironment pH in order to treat metastatic malignancies in humans.

Reasons for increased liver-resident NK-cell susceptibility to pH induced cellular stress compared with CD56^dim NK cells are unknown. Lymphocytes have been shown to be capable of maintaining intracellular pH in acidic conditions to a certain point (∼pH 6.8; ref. 41). HCO⁻ provides most of the buffering capacity in the cell and is produced from CO₂ generated from mitochondrial respiration (42). Cells with greater mitochondrial mass and an active TCA cycle produce more CO₂ and have greater buffering capacity. We have shown that liver-resident NK cells have significantly lower mitochondrial mass, and increased mitochondrial ROS production, compared with CD56^dim NK cells, reducing their capacity to generate CO₂, possibly making them more susceptible to extracellular pH changes.

The recurrence of CRLM after successful resection occurs in up to 70% of patients. Evidently, immune surveillance mechanisms in the liver remain compromised after removal of the tumor. We found that disease recurrence correlates with reduced NK-cell numbers in CRLM tumors, more so than T cells in the intestine, which predict colorectal cancer outcome (17, 18). Treatment targeting lactate production may benefit these patients, improving immune surveillance in the liver and limiting future tumor growth. We believe this approach will substantially reduce the recurrence rate after surgical resection.

Targeting metabolic pathways in oncology is not a new therapeutic concept, and the use of antimetabolites such as methotrexate, folic acid, or 5-fluorouracil is a common treatment for many cancers, despite their toxicity (43). A lactate-mediated mechanism, by which CRLM subverts local antitumor immunity, provides new opportunities for therapies that target specific metabolic pathways. Inhibition of glycolysis in tumor cells is of growing interest to oncologists; however, identifying components in this pathway that specifically inhibit tumor growth and spare immune cells is challenging. In preclinical testing, drugs that target GAPDH (Koningic acid) and LDH (FX11) are effective in limiting tumor growth (43–45). The use of systemic bicarbonate buffering, which neutralizes tumor acidity, has been reported to reduce tumor invasiveness and improve NK-cell responses (46, 47); however, patient adherence to treatment can prove difficult. Alternatively, lactate transport can be targeted specifically (AZD3965, NCT01791595) to prevent release of lactate from tumor cells, increasing it to toxic concentrations within tumor cells, reducing their growth, and inducing their apoptosis (48–50).

Our results highlight the immunosurveillance role of liver-resident NK cells, which are depleted in patients with recurrent CRLM. This depletion is driven by tumor-derived lactate, identifying a potential immunometabolic therapeutic target in patients with CRLM. Therapies capable of reducing lactate concentrations within the tumor microenvironment may restore liver-resident NK-cell populations, benefiting patients with both resectable CRLM and inoperable CRLM, by reducing tumor growth and restoring local tumor immunity. We believe this approach could substantially reduce the recurrence rate after surgical resection.

No potential conflicts of interest were disclosed.

Conception and design: C. Harmon, M.W. Robinson, C. O'Farrelly

Development of methodology: C. Harmon, M.W. Robinson, L. Lynch

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C. Harmon, F. Hand, K. Mentor, E. Hoti, J. Geoghegan, C. O'Farrelly

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): C. Harmon, M.W. Robinson, L. Lynch

Writing, review, and/or revision of the manuscript: C. Harmon, M.W. Robinson, F. Hand, D.D. Houlihan, C. O'Farrelly

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D. Almuaili, C. O'Farrelly

Study supervision: M.W. Robinson, C. O'Farrelly

We wish to thank donor families for participating in this research project. The support of the whole liver transplant team at St. Vincent's University Hospital is gratefully acknowledged. The authors would also like to thank all members of the Comparative Immunology Group for helpful discussion. This work was supported by grants from the Health Research Board of Ireland (RP 2008/189) and a Science Foundation Ireland Investigator Award (12/IA/1667).

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.