Potential conflict of interest: Nothing to report.
Defects in natural killer (NK) cell functions are necessary for tumor immune escape, but their underlying regulatory mechanisms in human cancers remain largely unknown. Here we show, in detailed studies of NK cells in 294 untreated patients with hepatocellular carcinoma (HCC), that accumulation of functional NK cells in HCC tissues could predict improved survival of patients. However, in patients with advanced-stage HCC, NK cells were significantly decreased in number with impaired tumor necrosis factor alpha (TNF-α) and interferon-gamma (IFN-γ) production. High infiltration of peritumoral stroma monocytes/macrophages was positively correlated with impaired functional activities of NK cells in intratumoral areas. Further kinetic experiments revealed that soon after exposure to tumor-derived monocytes, NK cells underwent a rapid, transient activation, but then they became exhausted, and eventually died. The monocytes from HCC tissues, but not from nontumoral liver, strongly express CD48 proteins; and such monocyte-induced NK cell dysfunction was markedly attenuated by blocking CD48 receptor 2B4 on NK cells, but not by blockade of NKG2D and NKp30. Conclusion: These data reveal that human NK cells are regulated by a fine-tuned collaborative action between different types of immune cells, which may reflect a novel immune-escape mechanism by which tumors dynamically regulate their functions at distinct tumor microenvironments. (HEPATOLOGY 2013)
Natural killer (NK) cells constitute a major component of host defense against tumors by secretion of granules containing lytic enzymes or by triggering apoptosis.1, 2 Clinical and experimental studies have demonstrated that dysfunction of NK cells often leads to advanced disease progression in several types of human solid tumors.3, 4 Compared to other organs, NK cells constitute a predominant lymphocyte population in the liver,5 and studies in mice indicate that liver-infiltrating NK cells play a critical role in clearing viral infection.6, 7 However, very little is known about the nature, regulation, and functions of NK cells in human hepatocellular carcinoma (HCC).
NK cell activity is regulated by both activating and inhibitory receptors. Activation of NK cells is mediated through the interaction of NK cell surface activation receptors with their ligands on target cells. Alternatively, interaction of an inhibitory receptor with its ligand negatively regulates NK cell activity.8, 9 In addition to being expressed on target cells, the regulatory ligands for NK cell activation are also found on activated antigen-presenting cells (APCs).8 However, there is substantial evidence that the inflammatory response associated with activated APCs can be rerouted in a tumor-promoting direction.10, 11 Macrophages (Mψ) markedly outnumber other APCs in solid tumors,12-14 and we recently found that tumor environments can alter the normal development of Mψ that is intended to trigger transient early activation of monocytes in peritumoral stroma, which in turn induces formation of suppressive Mψ in the intratumoral region.15 Thus, functional data on NK cells in the presence of monocytes/Mψ in different niches of a tumor are essential for understanding their roles and potential mechanisms in tumor immunopathogenesis.
The CD48 molecule is a costimulatory ligand of the CD2 family receptors, and it has been found on various hematopoietic cells, particularly on APCs.16 CD48 binds CD2 and other molecules, although its high-affinity receptor in both mouse and human systems is 2B4. The interactions between 2B4 and CD48 are closely associated with inflammatory disorders.16, 17 The present study showed that high intratumoral NK cell density was associated with increased survival of HCC patients, but NK cells exhibited lowered infiltration and weakened functional activities in patients with advanced-stage HCC. Such attenuated infiltration and dysfunction of NK cells in the intratumoral region was positively associated with the increased level of activated monocyte/Mψ in peritumoral stroma of HCC tissues, and accordingly, activated monocytes isolated from HCC tissues caused transient activation, but subsequent exhaustion, and ultimate apoptosis of NK cells. This process was mediated by cell-cell interactions by way of 2B4-CD48, but not NKG2D and NKp30.
Ab, antibody; APCs, antigen-presenting cells; HCC, hepatocellular carcinoma; IL, interleukin; Mψ, macrophage(s); NK, natural killer; TAM, tumor-associated Mψ.
Materials and Methods
Patients and Specimens.
Detailed information about the patients and specimens is described in the Supporting Materials and Methods and Supporting Table 1.
Isolation of Mononuclear Cells from Peripheral Blood and Tissues.
Peripheral leukocytes were isolated by Ficoll density gradient centrifugation.15, 18 Tumor- and nontumor-infiltrating leukocytes were obtained from paired fresh tissue samples as described.19 The mononuclear cells were washed and resuspended in medium supplemented with 1% heat-inactivated fetal calf serum (FCS) for fluorescent-activated cell sorter (FACS) analysis.
Leukocytes were stained with surface markers, fixed, permeabilized with IntraPre Reagent (Beckman Coulter, Fullerton, CA), and further stained with antibodies against intracellular markers. Data were acquired on Gallios (Beckman Coulter, Brea, CA). For the measurement of intracellular cytokine production, cells were stimulated at 37°C for 5 hours with Leukocyte Activation Cocktail (BD Bioscience) before staining as described.20 The fluorochrome-conjugated monoclonal antibodies (mAbs) are listed in Supporting Table 2.
Immunohistochemistry and Immunofluorescence.
Paraffin-embedded and formalin-fixed samples were cut into 5-μm sections, which were then processed for immunohistochemistry as described.21 After incubation with an antibody against human NK-1 (Thermo Fisher Scientific, Fremont CA) or CD68 (Dako, Denmark), the adjacent sections were stained with diaminobenzidine or 3-amino-9-ethylcarbazole in an Envision System (Dako). For immunofluorescence analysis, tissues were stained with monoclonal mouse antihuman NK-1 and rabbit antihuman CD68 or with mouse antihuman NK-1 and goat antihuman CD69. Secondary antibodies included Alexa Fluor 488-conjugated goat antimouse IgG with Alexa Fluor 568-conjugated goat antirabbit IgG and Alexa Fluor 488-conjugated donkey antigoat IgG with Alexa Fluor 568-conjugated donkey antimouse IgG (Molecular Probes, Eugene, OR). Positive cells were quantified using ImagePro Plus software (Media Cybernetics) and expressed as the mean of the percentage of positive cells ± standard error of the mean (SEM) in 10 high-powered fields detected by confocal microscopy.
Evaluation of Immunohistochemical Variables.
The evaluation of immunohistochemical variables is detailed in the Supporting Materials and Methods.
In Vitro Coculture of Monocytes and NK Cells.
In vitro coculture of monocytes and NK cells is detailed in the Supporting Materials and Methods.
Preparation of Tumor Supernatant (TSN) and Coculture of TSN-Treated Monocytes and NK Cells.
Preparation of TSN and coculture of TSN-treated monocytes and NK cells are detailed in the Supporting Materials and Methods.
The statistical analysis is detailed in the Supporting Materials and Methods.
Accumulation of Functional NK Cells in HCC Tissues Suppresses Disease Progression and Predicts Improved Survival.
To evaluate the potential role of NK cells in HCC immunopathology, we first investigated their infiltration in human normal liver (distal normal tissues of liver hemangioma), chronic hepatitis liver (tissues from liver transplantation), nontumoral liver, and paired intratumoral tissues (Supporting Table 1). The presence of NK cells was visualized by immunohistochemical staining of NK-1 (CD57) in paraffin-embedded tissues. As shown in Fig. 1A and Supporting Fig. 1, NK cells were predominant in normal liver, chronic hepatitis liver, and nontumoral liver rather than in the intratumoral region (all P < 0.05). Such decreased infiltration of NK cells in the intratumoral region mainly happened in patients with advanced-stage HCC (stages I-II [n = 178] versus stages III-IV [n = 78]; P < 0.01; Fig. 1A). Similar infiltration levels of NK cells were observed in normal, chronic hepatitis, and nontumoral liver (Fig. 1A).
Based on the above observations, we predicted that the presence of NK cells in HCC tissues would have a favorable effect on patient survival. To test this assumption, HCC patients who had received curative resection with follow-up data were divided into two groups according to the median value of NK-1+ cell density in the intratumoral region. There was a striking positive association between NK-1+ cell density in the intratumoral region and both overall survival (OS) and disease-free survival (DFS) (n =256; P < 0.001 for both; Fig. 1B). By contrast, NK-1+ cells in the nontumoral liver (n = 95) were unrelated to the prognosis of either OS or DFS (Fig. 1B). The NK-1+ cell density in intratumoral region was also associated with Vascular invasion (P = 0.014) and TNM stage (P = 0.003) (Table 1). Multivariate analysis revealed that the number of NK-1+ cells in the intratumoral region was an independent prognostic factor for both OS and DFS (Table 2).
Table 1. Association of Tumoral CD57 Cells with Clinicopathological Characteristics
Tumoral CD57 cells
The underlined terms represent statistical significance. AFP, a-fetoprotein; ALT, alanine aminotransferase; HbsAg, hepatitis B surface antigen; TNM, tumor node metastasis. The cutoffs for intratumoral CD57, peritumoral stroma and intratumoral CD68 cells were selected according to their median values.
Tumor size (cm)
Table 2. Univariate and Multivariate Analyses of Factors Associated with Survival and Recurrence
Cox proportional hazards regression model. Variables used in multivariate analysis were adopted by univariate analysis. The underlined terms represent statistical significance. AFP, a-fetoprotein; ALT, alanine aminotransferase; CI, confidence interval; HbsAg, hepatitis B surface antigen; HR, hazard ratio; NA, not adopted; TNM, tumor node metastasis.
Age, years (>48 vs. ≤48)
Gender (male vs. female)
HbsAg (positive vs. negative)
Cirrhosis (present vs. absent)
ALT,U/L (>42 vs. ≤42)
AFP, ng/mL (>25 vs. ≤25)
Tumor size, cm (>5 vs. ≤5)
Tumor differentiation (III+IV vs.I+II)
Vascular invasion (present vs. absent)
Tumor multiplicity (multiple vs. solitary)
Capsulation (present vs. absent)
TNM stage (III+IV vs.I+II)
Tumoral CD57 (high vs. low)
Peritumoral CD68 (high vs. low)
Tumoral CD68 (high vs. low)
We subsequently used flow cytometry to examine the phenotypic features of NK cells in paired blood, nontumoral liver, and intratumoral tissues from HCC patients. Consistent with the above results, the ratios of total (CD3−CD56+) NK cells were significantly lower in intratumoral tissues from advanced-stage HCC patients than in paired nontumoral liver (Fig. 1C,D; Supporting Fig. 2). Further analysis revealed that the reduction of NK cells in intratumoral tissues could be ascribed to the marked decline of CD16−CD56bright NK cells, but not the CD16+CD56dim NK cells. Nontumoral liver contained a higher proportion of CD16−CD56bright NK cells than their blood counterparts. Supporting the general view that CD16−CD56bright NK cells have a greater capacity for cytokine production compared to that of CD16+CD56dim NK cells,22 the NK cells isolated from nontumoral liver produced significantly higher levels of interferon-gamma (IFN-γ) and tumor necrosis factor alpha (TNF-α). As expected, the NK cells from intratumoral tissues of advanced-stage HCC exhibited attenuated capacities for those two cytokine productions. These findings, together with the positive prognostic value of NK cells in the intratumoral region, suggest that infiltration of functional NK cells in HCC tissues may represent the host reaction to malignancy, and, however, that tumor environments administrate NK cell function during disease progression.
High Infiltration of Peritumoral Stroma Monocytes/Mψ Correlates with Impaired Functional Activities of NK Cells in HCC Tissues.
APCs are critical for initiating and maintaining NK cell responses,23 and Mψ markedly outnumber other APCs in solid tumors.14 We have recently observed that HCC environments can alter the normal developmental process of Mψ that is intended to dynamically regulate monocyte activation in peritumoral stroma, and in that way create conditions that are conducive to tumor progression.15 Therefore, we next set out to elucidate the possible association between monocytes/Mψ and NK cells in HCC patients, paying particular attention to the tissue microlocalization of those cells (Supporting Fig. 3). Both monocytes/Mψ (CD68+ cells) and NK cells were present throughout the HCC tissues, and often accumulated in peritumoral stroma (42 ± 7 cells/field and 39 ± 5 cells/field, respectively; n = 10), but not in intratumoral areas (12 ± 3 cells/field and 9 ± 2 cells/field, respectively; n = 10) (Fig. 2A). However, inconsistent with our hypothesis, the level of NK cells, neither in peritumoral stroma nor in the intratumoral region, was associated with the density of monocytes/Mψ in the same area of HCC tissues (Fig. 2B). Unexpectedly, we observed an inverse correlation between the densities of peritumoral stroma monocytes/Mψ and intratumoral NK cells, but a positive association between the densities of peritumoral stroma NK cells and intratumoral NK cells (Fig. 2B), which suggests that recruiting NK cells are educated by monocytes/Mψ in peritumoral stroma, which in turn lead to NK cell dysfunction in the intratumoral region of HCC tissues.
To address that hypothesis, we then analyzed the activation status of NK cells in HCC samples dual-stained for CD57 and CD69 (marker for lymphocyte activation). Most NK cells in peritumoral stroma showed marked expression of CD69 (68.5% ± 6.8%; n = 10), which implies that they acquired an activated phenotype (Fig. 2C). In contrast, the majority NK cells in the intratumoral region were negative for CD69 (6.6% ± 2.1%; n = 10), whereas the NK cells in nontumoral liver exhibited moderate expression of CD69 (23.7% ± 2.6%; n = 10) (Fig. 2C). The activation of NK cells in peritumoral stroma was further confirmed with flow cytometry analysis showing that most NK cells isolated from the invading tumor edge of HCC tissues exhibited increased capacities for production of IFN-γ and TNF-α as well as up-regulated expression of the activation marker CD69 and the proliferation marker Ki67 (Fig. 2D). Remarkably, most of these activated NK cells belonged to the CD16−CD56bright NK cell subsets (Fig. 2E). These data, together with activation of monocytes in peritumoral stroma11, 15 and dysfunction of NK cells in intratumoral tissues (Fig. 1), indicate that NK cells might be preactivated in peritumoral stroma and thereafter become dysfunctional in the intratumoral region, and this process can be possibly regulated by activated monocytes. In support of this, NK cells isolated from intratumoral tissues exhibited significantly higher expression of surface degranulation marker CD107a but reduced expression of perforin, TNF-associated apoptosis-inducing ligand (TRAIL), and Granzyme B, revealing a dysfunctional form of cells (Fig. 2D,F). Also, high infiltration of peritumoral stroma CD68+ cells was positively associated with impaired production of IFN-γ in intratumoral NK cells (Fig. 2F).
Tumor Monocytes/Mψ Caused Early Activation and Subsequent Dysfunction of NK Cells.
To further elucidate the effect of tumor monocytes/Mψ on NK cell dysfunction, we purified monocytes (CD14high cells) from nontumoral liver and paired tumor tissues, and then cultured those cells with allogeneic circulating NK cells. The results showed that the expression of Ki67, CD69, TRAIL, and Granzyme B was significantly up-regulated in/on NK cells after exposure to monocytes from tumor tissues (>70% of them were HLA-DRhigh) for 2 days, but was reduced remarkably on day 8 (Fig. 3A,B). Similar patterns of cytokine productions were obtained in tumor monocyte-treated NK cells, including the marked expression IFN-γ and TNF-α on day 2 and a subsequent exhaustion on day 10 (Fig. 3C,D). Furthermore, analysis of the survival of NK cells after 10-day exposure to tumor monocytes revealed that over 55% of the NK cells were positive for annexin V, implying they were undergoing apoptosis (Fig. 3E). Of note, the monocytes isolated from nontumoral liver (<15% of them were HLA-DRhigh) did not trigger such sequential activation, exhaustion, and apoptosis of NK cells (Fig. 3). Furthermore, we also incubated monocytes with culture supernatant from hepatoma cells (TSN) to generate tumor-educated monocytes,15 and then cultured those cells with purified autologous NK cells. Similar sequential activation and exhaustion were observed in NK cells after exposure to TSN-treated monocytes (Supporting Fig. 4A,B). Collectively, these findings show that activated monocyte-mediated early NK cell activation in peritumoral stroma leads to NK cell exhaustion/reduction in the intratumoral region.
Role of 2B4 and CD48 Interaction in Regulating NK Cell Function by Tumor Monocytes/Mψ.
APCs can regulate NK cell responses by way of membrane-bound molecules and secretion of soluble mediators.23, 24 Thus, we cultured purified tumor monocytes with allogeneic circulating NK cells in different chambers of a transwell plate. As shown in Fig. 4A, a transwell system almost completely blocked the ability of intratumoral monocytes to induce NK cell dysfunction after 8 days, indicating that membrane-bound molecules are required to trigger NK cell dysfunction. Among the numerous activating receptors expressed on NK cells, 2B4 that is constitutively expressed by NK cells has been indicated in the reciprocal interactions between monocytes/Mψ and NK cells.25 Supporting Fig. 5 shows that most of the NK cells in peritumoral stroma were in close contact with CD68+ monocytes/Mψ, and accordingly, monocytes isolated from tumor tissues exhibited significantly higher expression of the 2B4 ligand CD48 (n = 5; P < 0.01 compared with nontumoral liver-infiltrating monocytes/Mψ; Fig. 4B), which suggests that tumor monocytes may regulate NK cell function by way of CD48 signals. To address that possibility, we conducted experiments using the 2B4 mAb against these receptors, and then exposed the cells to monocytes isolated from tumor tissues. In support, the anti-2B4 Ab effectively attenuated NK cell activation during the early phase of coculture with tumor monocytes and markedly restored the ability of these NK cells to produce IFN-γ and TNF-α at later coculture periods (Fig. 4C,D), whereas the control Ab only had a marginal effect on cytokine production. Moreover, we also observed that pretreatment of NK cells with anti-2B4 mAb also inhibited their apoptosis after 10-day exposure to tumor monocytes (Fig. 4E). This finding was further confirmed in an autologous system showing that blockade of 2B4 effectively restored the production of IFN-γ and TNF-α in NK cells cultured for 8 days with TSN-treated monocytes (Supporting Fig. 4C,D).
Previous studies have shown that human dendritic cells can induce NK cell activation by way of interacting with surface receptor NKG2D and NKp30.23, 26 Inasmuch as we had detected high levels of these two receptors on NK cells isolated from both nontumoral liver and tumor tissues (Supporting Fig. 6), we performed new experiments using blocking Abs against NKG2D and NKp30, respectively. However, neither of these Abs had any effect on tumor monocyte-induced early NK cell activation (Fig. 4F). These findings indicate distinct mechanisms between dendritic cells and monocytes/Mψ in regulating NK cell activation.
Defects in NK cell functions have been recognized as important mechanisms for tumor immune escape.27 The present study showed that, although high infiltration of functional NK cells in intratumoral region of HCC tissues predicts improved survival, NK cells were significantly decreased with impaired functional activities in patients with advanced-stage HCC, and their levels were negatively correlated with the density of activated monocytes/Mψ in peritumoral stroma. Activated monocytes isolated from HCC tissues dynamically modulated NK cell functions by way of two opposing functional stages, i.e., transient early activation and subsequent exhaustion/apoptosis. This dynamic regulation of NK cell activity may represent a novel immune-editing mechanism by which tumors co-opt the crosstalk between activated monocytes and NK cells to counteract the potent antitumor responses from NK cells.
Despite recent advances in understanding the crosstalk between APCs and NK cells in different disease models,28-30 little is known about their underlying regulatory mechanisms in human tumors. The present study provides evidence that tumor-activated monocytes/Mψ play a dominant role in regulating both the function and life span of NK cells in HCC, as indicated by the results of four sets of experiments. First, we observed that the level of NK cells was remarkably lower in the intratumoral region of advanced-stage HCC than in paired nontumoral liver, and there were significant negative correlations between the densities of NK cells in the intratumoral region and CD68+ monocytes/Mψ in peritumoral stroma. Second, coculture with tumor-derived activated monocytes for 8∼10 days impaired NK cell functions, as rendering them exhibit phenotypic features similar to those isolated from HCC tumors. Third, kinetic experiments revealed an early activation, but subsequent exhaustion, and ultimate apoptosis process in NK cells cultured with tumor monocytes. Fourth, blockade of the interaction between 2B4 and CD48, but not NKG2D or NKp30, significantly attenuated the ability of tumor monocytes to cause the sequential activation and exhaustion/apoptosis of NK cells. These observations suggest that activation of monocytes/Mψ in peritumoral stroma may not represent host reaction to the malignancy but instead they are rerouted in a tumor-promoting direction by triggering NK cell dysfunction. This notion is supported by our recent findings that the density of monocytes/Mψ in peritumoral stroma correlated with advanced disease stages and could serve as an independent predictor of poor survival in HCC patients.11
Immune exhaustion occurs concomitantly with immune activation, which represents a common mechanism in the regression of acute inflammation.11, 15 We and others have recently found that soluble tumor-derived factors elicited sequential activation and exhaustion of newly recruited monocytes, resulting in the formation of immunosuppressive Mψ in the intratumoral region, and in that way avoid the potentially dangerous actions of Mψ.15 These findings suggest that tumor can mimic some of the signaling pathways of the immune system to propagate conditions that favor tumor immune tolerance and promote escape from tumor immunity. Apparently, such sequential preactivation and exhaustion of cells is a general phenomenon that may also apply to other stimuli or physiological processes.31-33 This concept is well complemented by our current study showing that NK cells were educated by activated monocytes to adopt a cytotoxic phenotype during their early migration stage and subsequently subjected to activation-induced cell death in tumors. Consistent with this, our observations in human HCC indicate that most NK cells showing high expression of CD69 molecules colocate with HLA-DRhigh monocytes/Mψ in peritumoral stroma,15 while they exhibit an impaired functional state in intratumoral region, which contains monocytes/Mψ that have become immunosuppressive.
Although tumor NK cells originate from circulating blood NK cells, they are strongly impaired with regard to various functions related to cytotoxicity.34, 35 The present study provided evidence that tumor monocyte-associated CD48 molecules were essential for early transient activation and subsequent dysfunction of NK cells by way of its receptor 2B4. First, most activated monocytes isolated from HCC tissues strongly expressed CD48 molecules, and they were in close contact with NK cells in peritumoral stroma. Second, these CD48-expressing monocytes effectively triggered early activation and subsequent exhaustion/apoptosis of NK cells, and that effect was attenuated by blocking 2B4 on NK cells. Third, blockade of NKG2D or NKp30 did not inhibit such tumor monocyte-mediated NK cell dysfunction in vitro. Consistent with our results, other investigators have found that 2B4-CD48 interactions were important for the activation of NK cells by lipopolysaccharide (LPS)-activated Mψ.25
NK cells can be divided into subsets based on their expression of CD56 and CD16.22, 36 Increased levels of CD56brightCD16− NK cells are found in normal human liver, albeit CD56dimCD16+ NK cells dominate in peripheral blood. In the current study, we observed that the ratios of CD56brightCD16− NK cells were significantly lower in tumor from patients with advanced-stage HCC than in paired nontumoral liver. Therefore, it is possible that, in the presence of tumor-activated monocytes, the CD56brightCD16− NK cells are first activated to produce proinflammatory IFN-γ and TNF-α and then become exhausted and subsequently die. Consistent with our results, other investigators reported that, upon encountering APCs, CD56brightCD16− NK cells have a greater capacity for cytokine production compared to that of CD56dimCD16+ NK cells.22, 37
Although cancer patients exhibit a generalized immunosuppressive status,38-40 there is substantial evidence that the inflammatory reaction can also promote tumor progression by fostering immune privilege.41 These results increase our understanding of the formation of NK cells phenotypes in tumors. Soluble factors derived from cancer cells can trigger transient activation of newly recruited monocytes in peritumoral stroma and thereby induce the monocytes to express high levels of CD48 molecules, which, in turn, leads to the transient early activation and subsequent exhaustion/death of NK cells. These findings provide important new insights into the mechanisms by which activated monocytes in tumors may perform a suppressive role by regulating NK cells functions. These data will be helpful for the rational design of novel immune-based anticancer therapies.