The discovery of CD49a+ liver-resident natural killer (NK) cells in mice alters our view of NK cells and provides another opportunity to study NK cells. Although evidence has suggested roles for NK cells in liver diseases, whether and how CD49a+ NK cells contribute to liver diseases remain unclear. In this study, we observed that accumulation of CD49a+ tissue-resident NK cells in human hepatocellular carcinoma (HCC) was higher than in peritumoral tissues. We studied the exhausted and regulatory phenotypes of CD49a+ tissue-resident NK cells by analysis of protein and mRNA. The proportion of CD49a+ NK cells was positively correlated to the proportion of NK cells expressing inhibitory receptors. In addition, CD49a+ NK cells expressed more of checkpoint molecules PD-1, CD96, and TIGIT. Transcriptomic analysis implicated CD49a+ tissue-resident NK cells in the negative regulation of immune responses. Comparison of murine and human CD49a+ NK cells revealed their distinct characteristics and functions. Finally, accumulation of tissue-resident CD49a+ NK cells in liver tumor was correlated to deteriorating disease condition and poor prognosis. Our findings show that CD49a+ NK cells accumulate in liver tumor and suggest a role for CD49a+ NK cells in the negative regulation of immune responses and the development of HCC.
CD49a, also known as integrin α1, binds integrin β1 (CD29) to form the CD49a/CD29 (VLA-1) heterodimer, which is responsible for retaining lymphocytes in tissues (1, 2). CD49a was initially defined as an extracellular protein receptor on activated T lymphocytes (3). Functional studies in the 1990s demonstrated its role as an adhesion molecule on T cells (3, 4). In 2003, CD49a was identified as a selection marker in the development and differentiation of human bone marrow mesenchymal stem cells (MSC; ref. 5). Ten years later, a new subset of natural killer (NK) cells, termed tissue-resident NK (trNK) cells, was introduced, and CD49a was identified as the typical marker of trNK cells (6). In the study, Peng and her colleagues defined two subsets of murine NK cells: CD49a+DX5− and CD49a−DX5+ (6). CD49a−DX5+ NK cells are conventional NK (cNK) cells that circulate in the blood, whereas CD49a+DX5− NK cells are trNK cells that reside in the liver sinusoidal blood and possess memory potential (6, 7). The development of this NK-cell subset depends on transcription factor T-bet, but not Eomes (6), similar to type 1 innate lymphoid cells (ILC1) in mucosal tissues (8–10). Liver trNK cells and mucosal ILC1s are believed to arise from a common progenitor (the CHILP; ref. 11) and exhibit similar phenotypes such as high expression of CD49a, CD160, and CD69 and low expression of Ly49 receptors (8). Liver-resident NK cells and mucosal ILC1s are efficient producers of IFNγ and TNFα, and both degranulate less efficiently than cNK cells. Therefore, it has been suggested that this cell population in mice should be considered as ILC1s (8). Whether human CD49a+ trNK cells are also ILC1s remains unclear.
Findings on liver trNK cells provide insight into the discovery of trNK cells in other tissues (12). High frequency of CD49a+DX5− trNK cells has been found in the uterus, skin, and adipose tissue (2, 13); however, uterus trNK cells are T-bet independent and express high amounts of Eomes (13, 14). trNK cells in salivary gland are positive for both CD49a and DX5 (15) and present in normal numbers in T-bet–, Eomes-, and NFIL3-deficient mice (15, 16). Inspired by these findings, several groups have attempted to define trNK cells in humans. The T-bet+Eomes−CD49a+ NK-cell subset has been identified in human liver, but not in afferent or efferent blood of the liver (17). These cells express more inflammatory cytokines such as IFNγ, TNF, and GM-CSF and degranulate poorly upon stimulation (17). Cheuk and colleagues have pointed out CD49a as a marker to differentiate CD8+ TRM cells in human skin epithelia. Similar to human liver trNK cells, CD8+CD49a+ TRM cells also produce IFNγ. However, unlike human liver trNK cells that degranulate poorly upon stimulation, CD8+CD49a+ TRM cells from healthy skin may induce expression of perforin and granzyme B upon stimulation with IL15, providing a cytotoxic response (18). These studies suggest that CD49a functions differently in trNK cells than in TRM cells (19).
Liver is an immunotolerant organ that often encounters chronic infections and tumorigenesis. The underlying mechanisms often involve cross-talk between liver-resident antigen-presenting cells and circulating lymphocytes. A number of studies have demonstrated the role of CD49a in T cell–mediated inflammatory diseases (1, 20, 21). However, little is known about CD49a+ NK cells in liver tolerance and diseases (22). A study using genetically engineered mouse models with global transcriptomic and flow cytometry analyses has revealed cytokine–TGFβ signaling–dependent conversion of cNK cells (CD49a−CD49b+Eomes+) into intermediate ILC1 (intILC1; CD49a+CD49b+Eomes+) and ILC1 (CD49a+CD49b−Eomesint) cells in the tumor microenvironment. IntILC1 and ILC1 were unable to control local tumor growth and metastasis, and tumor escape from the innate immune system was partially mediated by TNF-producing ILC1s (23), suggesting a role for CD49a+ liver-resident NK cells in the process of tumorigenesis in mice.
Human hepatocellular carcinoma (HCC) accounts for most liver malignancies and is often caused by common clinical risk factors, such as chronic infection with hepatitis B and C viruses, heavy alcohol intake, steatohepatitis, and diabetes (24). By using HCC as a model system, we evaluated the proportion and absolute count of CD49a+ NK cells in HCC patients. We explored the association of CD49a+ NK cells with clinical pathologic variables and patient outcomes. In addition, comparison between primary CD49a+ and CD49a− NK-cell subsets revealed an exhausted phenotype of CD49a+ NK cells, suggesting their possible role in the repression of immune responses and development of HCC.
Materials and Methods
Fresh liver tumor tissues were prospectively collected from 28 HCC patients during surgery at The First Affiliated Hospital of University of Science and Technology of China, Hefei, China (cohort 1). Among these samples, 25 consisted of an HCC tumor and its corresponding adjacent peritumoral tissue (collected 2 cm distal to the tumor site). None of the patients had received chemotherapy or radiotherapy prior to surgery. Normal fresh liver tissues collected distal to liver echinococcosis were obtained from The First Affiliated Hospital of Xinjiang Medical University, Urumqi, China. The clinical characteristics of all tissue samples from HCC patients are summarized in Supplementary Table S1. The details of all patients are provided in Supplementary Table S2. All samples were anonymously coded in accordance with the Declaration of Helsinki. Written informed consent was obtained from each patient included in the study, and the protocol of all study cohorts was approved by the Ethical Board of the Institutional Review Board of the University of Science and Technology of China.
Lymphocyte isolation and flow cytometry
Liver-infiltrating lymphocytes were obtained as previously described (25). To digest the specimens, liver tissue samples were cut into small pieces and digested in RPMI-1640 (HyClone Laboratories) supplemented with collagenase IV (10 mg/mL, Sigma-Aldrich), DNase I (33.3 mg/mL, Sigma-Aldrich), and 20% FBS (HyClone Laboratories) at 37°C for 1 to 2 hours. The peripheral lymphocytes and liver-infiltrating lymphocytes were then stained with fluorochrome-conjugated antibodies and analyzed through flow cytometry. Antibodies against the following proteins were used for staining: CD3 (SK7), CD56 (B159), CD16 (3G8), CD49a (SR84), CD160 (BY55), NKG2D (1D11), PD-1 (MIH4), NKp30 (p30-15), NKp44 (p44-8), NKp46 (9E2; BD PharMingen); CD200R (OX108), CD96 (NK92.39), CD244 (eBioDM244; eBioscience); NKG2A (131411), NKG2C (134591; R&D Systems); LAG-3 (17B4), BTLA (4D6; Abcam); and TIGIT (A15153G; BioLegend). The stained cells were analyzed using a FACSCalibur flow cytometer (Becton Dickinson), and the data were analyzed using FlowJo analysis software 7.6.1 (Treestar).
Gene-expression profiling analysis
Purified NK cells from normal liver tissues were first enriched by MACS using NK Cell Isolation Kit (Miltenyi Biotec). CD3−CD56+CD49a+ and CD3−CD56+CD49a− hepatic NK cells were isolated by FACSAria cell sorter (BD Biosciences) using CD3 (SK7), CD56 (B159), and CD49a (SR84; BD PharMingen) to attain a purity greater than 95%. For analysis of the molecular signatures of human CD49a+/− NK cells, purified CD49a+/− NK cells (samples from three healthy donors were pooled for each cell type) were submitted for microarray analysis using the Whole Human Genome Microarray Kit (G4112F; Agilent Technologies). Transcription profile chip service was provided by Shanghai Biotechnology Cooperation. Microarray image analysis was performed using Agilent's Feature Extraction V9.1.3 software (Agilent Technologies). Expression values were log2 transformed, and subsequent analyses were conducted using SAS statistical software online (http://www.ebioservice.com/). Differential genes with fold change greater than 2 were selected and analyzed according to the Gene Ontology (GO) terms. Gene Set Enrichment Analysis (GSEA) is a computational method that determines whether an a priori set of genes shows statistically significant, concordant differences between two biological states. GSEA was interpreted as previously described (26, 27). The microarray data were deposited into the National Center for Biotechnology Information GEO repository under accession number GSE109198.
The Cancer Genome Atlas database
A database (https://cancergenome.nih.gov/), The Cancer Genome Atlas (TCGA), has generated comprehensive, multidimensional maps of the key genomic changes in 33 types of cancers. In our study, 442 HCC samples, 413 bladder urothelial carcinoma samples, and 479 skin cutaneous melanoma samples with detailed ITGA1 (CD49a gene) expression data were selected from the updated TCGA database (raw data at the NCI; source mutation data from GDAC Firehose). Patients with fully characterized tumors, intact disease-free survival (DFS) and overall survival (OS) data, complete RNA-seq information, and no pretreatment were included. We used this database to explore the prognostic value of ITGA1 in patients with various cancers as previously described (28, 29).
Statistical analysis and figure creation were performed using GraphPad Prism 6.0 and SPSS Statistics 21. Comparisons between two groups were performed using either the Wilcoxon matched-pairs signed rank test or the Mann–Whitney test. Results are expressed as means ± SEM. Simple correlations were summarized using the Pearson correlation coefficients (r). Kaplan–Meier analysis and the log-rank test were used to analyze DFS and OS of patients with cancer from TCGA database. P values of less than or equal to 0.05 were identified as significant in all analyses (*, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001).
CD49a+ NK cells accumulate in the intratumoral tissues of HCC
Although CD49a+ liver-resident NK cells account for around 30% to 50% of all NK cells in the murine livers, human CD49a+ NK cells account for only 0.11% to 12.7% (average 2.3%) of the total NK cells in healthy human liver (30). However, the percentage of CD49a+ NK cells in the transformed liver tissues remains obscure. Here, we isolated primary lymphocytes from both intratumoral tissue (IT) and peritumoral tissue (PT) of HCC patients and analyzed their phenotypes through antibody labeling and flow cytometry (Supplementary Fig. S1A). CD49a expression on NK cells was upregulated in the tumor compared with PT counterpart (Fig. 1A). The percentage of CD49a+ NK cells was significantly upregulated in IT compared with PT, which was also apparent in comparison between paired PT and IT individually (Fig. 1B). In addition, mean fluorescence intensity (MFI) and absolute count of CD49a+ NK cells were also significantly higher in IT than in PT (Fig. 1C).
Human NK cells are often classified into CD56bright and CD56dim NK cells (31). CD49a expression was upregulated on both CD56bright and CD56dim NK cells in liver tumor tissues (Fig. 2A). The percentages of both CD49a+CD56bright NK cells and CD49a+CD56dim NK cells in IT were significantly higher than those in PT, which was also apparent when comparing PT and IT of each patient individually (Fig. 2B and C). Although the absolute count of CD49a+CD56bright NK cells was not significantly higher in IT (Fig. 2D) due to a significant reduced number of total NK cells in tumor (Supplementary Fig. S1B), the percentage of CD49a+CD56bright NK cells remained increased in IT. On the other hand, the absolute count of CD49a+CD56dim NK cells was significantly higher in IT (Fig. 2D), and the number of total NK cells in tumor was lower (Supplementary Fig. S1B). The percentage of CD49a+CD56dim NK cells also increased in IT. The changes in the IT:PT MFI ratio for CD49a+CD56bright, CD49a+CD56dim, and CD49a+ NK subsets were similar (Fig. 2E), suggesting an overexpression of CD49a on intratumoral CD56bright, CD56dim, and total NK cells of HCC tissues.
CD49a+ NK cells are associated with exhausted and regulatory characteristics
Immune cells in the tumor microenvironment are often marked by exhausted phenotypes and impaired functions. Exhaustion of these immune cells is one of the reasons why tumor cells escape immunosurveillance. The percentage of CD49a+ NK cells was positively correlated with the percentage of NKG2A+ NK cells (r = 0.5804, P < 0.0001), LAG3+ NK cells (r = 0.6428, P = 0.0022), CD200R+ NK cells (r = 0.5804, P = 0.0184), CD96+ NK cells (r = 0.9241, P < 0.0001), BTLA+ NK cells (r = 0.5183, P = 0.0135), or PD-1+ NK cells (r = 0.5362, P = 0.0723; Fig. 3A–F). Among these inhibitory receptors, LAG-3, CD200R, CD96, and PD-1 are often upregulated on exhausted immune cells, suggesting that the CD49a+ NK-cell population is likely to be an exhausted cell population, especially in the tumor. On the other hand, the percentage of CD49a+ NK cells was negatively correlated to the percentage of CD160+ NK cells (r = −0.4995, P = 0.0008), CD244+ NK cells (r = −0.5797, P = 0.0482), or NKG2D+ NK cells (r = −0.4281, P = 0.1890; Fig. 3G–I). One exception was the positive association between the percentages of CD49a+ and NKp44+ NK cells (r = 0.8155, P = 0.0012; Fig. 3J). No significant association was evident with NKp30+ or NKp46+ NK cells (Fig. 3K and L). We further analyzed the expression of typical NK-cell receptors and exhaustion-related phenotypic molecules on intratumoral CD49a+ and CD49a− NK cells through flow cytometry. Exhaustion-related checkpoint molecules such as PD-1, CD96, and TIGIT were upregulated on intratumoral CD49a+ NK cells compared with CD49a− counterparts (Fig. 3M), suggesting an exhausted phenotype of CD49a+ NK cells in the tumor microenvironment.
Transcriptomic differences between human and murine CD49a+ and CD49a− NK-cell subsets
Although CD49a+ NK cells are known in human liver, their characteristics and functions have yet to be revealed. To compare and contrast primary CD49a+ and CD49a− NK-cell populations, we isolated primary hepatic lymphocytes from human livers and purified CD3−CD56+CD49a+ and CD3−CD56+CD49a− NK cells through negative selection and flow cytometry sorting. By using high-resolution microarrays, we found differences between CD49a+ and CD49a− NK-cell subsets. A greater than 2-fold change was found in 10,517 expressed genes (Fig. 4A and B), which fell into 11 functional categories (Supplementary Fig. S2A). Quantitative analysis of differential gene expression reveals that genes encoding inhibitory receptors, such as CD200R1, BTLA, and LAG3, were overexpressed in CD49a+ NK cells, whereas genes encoding activating receptors, such as CD226, CD69, and CD27, were underexpressed in CD49a+ NK cells compared with CD49a− NK cells (Fig. 4C). Genes encoding cytokines TGFB2, IFNG, TNF, and PRF1 were upregulated in CD49a+ NK cells, whereas GZMB, IL15, and IL18 were downregulated compared with CD49a− NK cells (Fig. 4C). A pathway enrichment network was created by Cytoscape to show the upregulated (red) and downregulated (green) genes involved in the regulation of immune response and cellular processes: Downregulated genes were involved in the inflammatory response, immune response, and regulation of cytokine production (Supplementary Fig. S2B). Immune response–related genes exhibited the most and largest green nodes, indicating enormous downregulation of these genes in CD49a+ NK cells, suggesting the contribution of CD49a in the negative regulation of NK cells (Supplementary Fig. S2B). On the other hand, GSEA has revealed that most highly enriched genes in CD49a+ NK cells overlapped with published gene signatures in the negative regulation of immune response and innate immune response (Fig. 4D); in addition, the most enriched gene sets in CD49a+ NK cells were related to the immune response, regulation of immune system process, immune system process, negative regulation of innate immune response, and negative regulation of immune response (Fig. 4E), suggesting a regulatory characteristic of human hepatic CD49a+ NK cells and a role in the negative regulation of immune responses and liver tolerance.
Gene-expression comparison between murine CD49a+, murine CD49a−, human CD49a+, and human CD49a− NK-cell subsets revealed differences between murine and human CD49a+ NK-cell subsets: Downregulated genes in human CD49a+ NK cells were upregulated in murine CD49a+ NK cells (Fig. 5A). These downregulated genes overlapped with published gene sets of known NK cell–related pathways, including cell-cycle phase, cell division, positive regulation of IFNγ production, immunoglobulin production, mediator of immune response, and heart development (Fig. 5B). Further analysis of the transcriptomic data revealed higher gene expression of GZMK, CXCR4, LAG3, PRDM1, and IL15 in human CD49a+ NK cells compared with murine CD49a+ NK cells (Fig. 5C), and lower gene expression of IFNG, GZMA, CXCR6, KLRK1, TNFSF10, KLRG1, IL10, CTLA4, and NOTCH1 in human CD49a+ NK cells compared with murine CD49a+ NK cells (Fig. 5C). Thus murine and human CD49a+ NK cells are different.
Accumulation of CD49a+ NK cells predicts tumor progression and poor clinical outcome
Tumor size is one of the primary indicators of tumor differentiation and clinical stage. To investigate the role of CD49a in the progression of HCC, we analyzed the percentage of CD49a+ NK cells in IT using flow cytometry and plotted against tumor size. There was no statistically significant correlation between the two (Fig. 6A). The percentage of CD49a+ NK cells was categorized into two groups based on the presence or absence of cancer thrombus and tumor capsule. The results indicated that HCC patients with tumor thrombus or without tumor capsule exhibited significantly more CD49a+ NK cells compared with those without tumor thrombus or with tumor capsule (Fig. 6B and C). Furthermore, HCC patients whose tumors recurred or who died also had significantly more intratumoral CD49a+ NK cells compared with patients without tumor recurrence or death (Fig. 6D and E). Patients were further divided into two groups based on the minimum P value cutoff values of their CD49a+ NK-cell proportions. Patients with a higher percentage of CD49a+ NK cells within tumor had shorter OS (P < 0.05) and DFS (P < 0.01; Fig. 6F). The TCGA database of 442 HCC patients showed that patients who had a family history of HCC occurrence exhibited significantly higher ITGA1 (encoding CD49a) expression compared with those without a family history (Fig. 6G). Moreover, significant differences were shown between groups of ITGA1high and ITGA1low patients with bladder urothelial carcinoma: Patients with higher ITGA1 mRNA expression exhibited a higher recurrence rate and lower survival rate (Fig. 7A). However, the correlation did not reach statistical significance in cases of skin cutaneous melanoma (Fig. 7B). Overall, these data suggest a pathogenic role of CD49a+ NK cells in the progression and prognosis of HCC.
The discovery of CD49a+ liver-resident NK cells in mice not only changes the classic concept of NK cells but also opens a new window for the study of NK cells. Two distinct subsets of NK cells have been identified in murine liver: CD49a−DX5+cNK cells that constantly circulate in the blood and CD49a+DX5− trNK cells that remain resident in the tissues (6, 22). Liver-resident NK cells are often accompanied by higher expression of CXCR6, CXCR3, CD69, and TNF-related apoptosis-inducing ligand (TRAIL). Transcription factors T-bet, Hobit, and PLZF are necessary for their development, whereas Eomes is not required (32). Following the study on liver-resident NK cells, CD49a+DX5− NK cells have also been identified in other organs such as the uterus, skin, kidney, and adipose tissue (2, 13, 33, 34). Homologous to murine CD49a+ liver-resident NK cells, the T-bet+Eomes−CD49a+ NK-cell subset has been defined in the human liver (17). Marquardt and colleagues have identified CD3−CD56+CD49a+ lymphocytes in normal liver tissues; this subset made up 0.11% to 12.7% (average 2.3%) of the total CD3−CD56+ lymphocyte population (17). In our study, CD3−CD56+CD49a+ NK cells made up 1.3% to 26.9% (average 8.2%) of CD3−CD56+ NK cells in PT (the relatively normal tissues). The proportion of CD49a+ NK cells from 19 of 26 PT examined ranged from 0.11% to12.7%; therefore, most of our data in PT are consistent with findings in normal liver tissues. The differences raised between the two studies may be due to the following: (i) We collected PT 2 cm distal from tumor tissues. Although the samples are free of tumors, we cannot rule out the possibility that some may exhibit different degrees of liver injury (cirrhosis, fibrosis, inflammation, etc.) that might contribute to a higher proportion of CD49a+ NK cells observed in our study than the Marquardt study. (ii) Our patients are all Chinese and most of them are HBV+, which may also contribute to the difference between the two studies.
Recruitment and retention of NK cells in the liver are very important to the progression of liver diseases including viral infection, fibrosis, cirrhosis, and even liver tumor (35, 36). Indeed, CD49a+ NK cells are enriched in cirrhotic liver when compared with tumor-free liver section samples (37). In our study, we identified accumulation of CD49a+ NK cells in HCC tissues and found distinct genetic profiles between primary hepatic CD49a+ and CD49a− NK cells. In line with our findings, engagement of CD49a+ NK cells expressing high levels of NK1.1, CD49a, CD103, and the cytolytic molecule granzyme B has also been observed in oncogene-induced murine cancer models (38). Furthermore, NK cells from malignant pleural effusion of patients with primary and metastatic tumor have shown increased expression of CD49a and CD69 and decreased expression of CD57 (39). Boudjadi and colleagues showed the presentation of CD49a protein and transcript (ITGA1) in a large proportion of colorectal cancers (40).
Conversion of NK cells into ILC1s has been proposed. Cortez and colleagues showed that NK cells acquire an ILC1-like gene signature when SMAD4 is deficient (41). Gao and colleagues identified three subsets of ILC1s [NK cells, intermediate ILC1s (intILC1s), and ILC1s] from a subcutaneous MCA1956 tumor and pointed out the conversion of NK cells (CD49a−CD49b+Eomes+) into intILC1s (CD49a+CD49b+Eomes+) and ILC1s (CD49a+CD49b−Eomesint) in the tumor microenvironment (23). As the tumor grows, the proportion of tumor NK cells decreases, and the fraction of intILC1s increases (23). In accordance with these findings, our study also shows an increased proportion of CD49a+ NK cells and a decreased proportion of CD49a− NK cells in the liver tumor. The divergent outcomes from the study by Marquardt and colleagues on healthy liver and our study could be due to the difference between healthy and diseased liver. Gao and colleagues have also shown that TGFβ induces the differentiation of NK cells into intILC1s and ILC1s in the tumor microenvironment (23), which may explain our observations. Whether CD49a+ trNK cells in human livers can be further classified or a conversion exists, as in murine tumor models, remains to be discovered.
Because CD49a+ trNK cells accumulate in the tumor microenvironment, we propose that these cells may exhibit exhausted/regulatory characteristics. Our study shows that the percentage of CD49a+ NK cells was positively associated with the percentage of inhibitory receptor NKG2A+, LAG3+, CD200R+, CD96+, BTLA+, or PD-1+ NK cells. On the other hand, the percentage of CD49a+ NK cells was negatively associated with the percentage of activating receptor CD160+, CD244+, or NKG2D+ NK cells. Higher expression of exhaustion-related checkpoint molecules (PD-1, CD96, and TIGIT) and lower expression of activating receptors (NKp46, NKp30, CD244, and CD160) were observed on intratumoral CD49a+ NK cells compared with their CD49a− counterparts, as indicated by the heat map, reconfirming the exhausted phenotype of CD49a+ NK cells in the tumor microenvironment. A study by Zhou and colleagues reported an immunoregulatory phenotype of murine CD49a+ liver-resident NK cells expressing PD-L1, LAG-3, and CD200R (42). In humans, expression of genes encoding inhibitory checkpoint receptors, such as CTLA-4, CD96, and LAG-3, is higher in tumor CD49a+ NK cells than in tumor CD49a− NK cells (23). Tumor CD49a+ NK cells also show higher expression of genes encoding the inhibitory NK-cell receptors NKG2A and KLRG1 (23). Furthermore, intratumoral NK cells from HCC patients exhibit reduced CD160 expression and elevated CD96 expression (43, 44), and NK cells from malignant pleural effusion exhibit a decrease of activating receptor NKp30 and an enhancement of inhibitory receptor NKG2A (39).
To better define and distinguish CD49a+ and CD49a− NK-cell subsets, we compared healthy primary CD49a+ and CD49a− NK cells using transcriptomic analyses. The two NK-cell subsets differ in their transcriptomes, with 10,517 differentially expressed genes identified. Genes encoding inhibitory receptors CD200R (CD200R1), BTLA (BTLA), and LAG-3 (LAG3) were enriched in CD49a+ NK cells, whereas genes encoding activating receptors such as CD226 (CD226) and CD69 (CD69) were less transcribed in CD49a+ NK cells. Cytokine-induced murine hepatic CD49a+ NK cells produce more IFNγ and TNFα (45), and human hepatic CD49a+ NK cells express more inflammatory cytokines (IFNγ, TNF, and GM-CSF) and degranulate poorly upon stimulation (17). In line with these findings, genes IFNG and TNF were enriched in CD49a+ NK cells, whereas GZMB, IL15, and IL18 were lowered in CD49a+ NK cells. Furthermore, GSEA plot has revealed that the most enriched genes in CD49a+ NK cells overlapped with published gene signatures in the negative regulation of immune response. These data suggest an exhausted/regulatory characteristic of the human hepatic CD49a+ trNK-cell subset and its possible role in the negative regulation of immune responses.
Unlike murine CD49a+ trNK cells, which constitute 30% to 50% of all NK cells in the liver, human CD49a+ trNK cells constitute an average of only 2.3% of all NK cells in the liver (30). An Eomeshi subset of NK cells stays permanently in the human liver (46), accounts for more than 50% of human liver NK cells, and overlaps with CD56brightCXCR6+ NK cells. However, Eomeshi NK cells do not overlap with CD49a+ liver-resident NK cells (30). Therefore, two distinct NK-cell subsets have been defined in the human liver: CD49a+ NK-cell subset and Eomeshi (largely CD56brightCXCR6+) NK-cell subset. In our study, transcriptomic comparison between murine CD49a+, murine CD49a−, human CD49a+, and human CD49a− NK-cell subsets revealed differences between murine and human CD49a+ NK cells. Most downregulated genes in human CD49a+ NK cells were upregulated in murine CD49a+ NK cells. Furthermore, higher gene expression of GZMK (encoding granzyme K), CXCR4 (encoding CXCR4), LAG3 (encoding LAG-3), PRDM1 (encoding Blimp-1), and IL15 (encoding IL15) and lower gene expression of IFNG (encoding IFNγ), GZMA (encoding granzyme A), CXCR6 (encoding CXCR6), KLRK1 (encoding NKG2D), TNFSF10 (encoding TRAIL), KLRG1 (encoding KLRG1), IL10 (encoding IL10), CTLA4 (encoding CTLA-4), and NOTCH1 (encoding Notch 1) were observed in human CD49a+ NK cells compared with murine CD49a+ NK cells. These data suggest that liver-resident CD49a+ NK cells in humans are not equivalent to CD49a+ NK cells in mice, and may exhibit different characteristics and functions in association with their liver microenvironment.
The accumulation of CD49a+ NK cells in the tumor may suggest a role for these cells in the progression and prognosis of HCC. Our study shows that patients with more CD49a+ NK cells within their tumor are more likely to exhibit cancer thrombus and no tumor capsule, suggesting that HCC patients with higher CD49a expression are accompanied by deteriorating disease conditions. Moreover, patients with a higher percentage of CD49a+ NK cells within tumors had shorter OS and DFS, indicating that the percentage of CD49a+ NK cells has value in predicting the clinical outcomes in HCC patients. Aside from NK cells, CD49a+ T cells have also been identified in tumors (47, 48). Blocking CD49a decreases intratumoral CD8+ T-cell infiltration and the efficacy of cancer vaccines on mucosal tumors (49, 50) and significantly impairs the control of subcutaneous B16-OVA tumors in mice (51). These findings suggest that CD49a plays a different role in tumor-infiltrating NK cells than it does in T cells, in which it has a prosurvival role. Therefore, we suggest that the protumor effect of ITGA1 in bladder urothelial carcinoma and skin cutaneous melanoma might be mainly caused by NK cells instead of T cells.
Few studies have touched on CD49a+ trNK cells in human tumor tissues. Here, we characterized and compared CD49a+ trNK cells in normal and tumor liver tissues and described their association with the progression of HCC. From our study, we propose that the accumulation of CD49a+ NK cells and their higher expression of inhibitory receptors could be a result of the tumor microenvironment. Our results raise many questions. First, although the phenotypes in human and mice are homologous, human hepatic CD49a+ NK cells are not equivalent to what has been found in the murine liver. Based on our global transcriptomic analysis, human and murine hepatic CD49a+ NK cells exhibit different characteristics and functions. Further studies are required to better define the characteristics and functions of human hepatic CD49a+ NK cells and to explore their roles in the tumor microenvironment. Second, induction and retention of the CD49a+ NK-cell subset in human liver remain unresolved. Murine hepatic CD49a+ NK cells can be induced by culturing with cytokines IL2, IL12, IL15, or IL18 or the cytokine cocktail (IL2/IL12/IL15/IL18; ref. 45). In addition, TGFβ may convert NK cells into intILC1s (CD49a+CD49b+Eomes+) and ILC1s (CD49a+CD49b−Eomesint) both in vivo and in vitro (23). However, molecules required for the induction and retention of human trNK cells remain unresolved. Third, whether CD49a+ NK cells play a protective or pathogenic role in the tumor microenvironment remains obscured. For example, CD49a+CD103+ NK cells may contribute to immunoserveillance in oncogene-induced murine cancer models, the absence of which results in accelerated tumor growth (38). Although other findings have suggested that conversion of CD49a− NK cells to CD49a+ NK cells in the tumor microenvironment accelerates tumor growth and metastasis (23, 41), tumor cell escape from the innate immune system is partially mediated by TNF-producing ILC1s (23).
Overall, our study has shown that more CD49a+ NK cells accumulate in human HCC tumor tissues than in PT. We studied exhausted phenotype and regulatory characteristic of this population. Comparison between murine and human CD49a+ NK cells showed differences in their characteristics and functions. The CD49a+ NK-cell subset in HCC was correlated to the progression and prognosis of HCC. Accumulation of CD49a+ NK cells in the human liver may result in deteriorating disease condition and poorer prognosis. Our findings not only reveal the accumulation of CD49a+ NK cells in liver tumor but also suggest possible immunologic mechanisms driven by this cell subset in the progression of HCC.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: W. Xiao, R. Sun, Z. Tian, C. Sun
Development of methodology: H. Sun, C. Sun
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): L. Liu, H. Liu, J. Wang, R. Lin, R. Sun, C. Sun
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Sun, K. Li
Writing, review, and/or revision of the manuscript: H. Sun, Z. Tian, C. Sun
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): L. Liu, Q. Huang, H. Liu, M. Huang, J. Wang, H. Wen, R. Lin, H. Wei
Study supervision: K. Qu, W. Xiao, R. Sun, Z. Tian, C. Sun
This work was supported by the National Natural Science Foundation of China (#81788101, #81761128013, #31670908, #81701631, #31390433, #81821001, and #91542000), the Chinese Academy of Sciences (XDB29030201), and the Fundamental Research Funds for the Central Universities (#WK2070000107).
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.