Group‐2 Innate Lymphoid Cells Promote HCC Progression Through CXCL2‐Neutrophil‐Induced Immunosuppression

Abstract

Background and Aims

Due to their inherent characteristics, the function of group‐2 innate lymphoid cells (ILC2s) varies in a context‐dependent manner. ILC2s are involved in certain liver diseases; however, their involvement in HCC is unknown. In the present study, we assessed the role of an HCC‐derived ILC2 population in tumor progression.

Approach and Results

Through FACS and single‐cell RNA sequencing, we discovered that ILC2s were highly enriched in human HCC and correlated significantly with tumor recurrence and worse progression‐free survival as well as overall survival in patients. Mass cytometry identified a subset of HCC‐derived ILC2s that had lost the expression of killer cell lectin‐like receptor subfamily G, member 1 (KLRG1). Distinct from their circulating counterparts, these hepatic ILC2s highly expressed CD69 and an array of tissue resident–related genes. Furthermore, reduction of E‐cadherin in tumor cells caused the loss of KLRG1 expression in ILC2s, leading to their increased proliferation and subsequent accumulation in HCC sites. The KLRG1⁻ ILC2 subset showed elevated production of chemotaxis factors, including C‐X‐C motif chemokine (C‐X‐C motif) ligand (CXCL)‐2 and CXCL8, which in turn recruited neutrophils to form an immunosuppressive microenvironment, leading to tumor progression. Accordingly, restoring KLRG1 in ILC2s, inhibiting CXCL2 in ILC2s, or depleting neutrophils inhibited tumor progression in a murine HCC model.

Conclusions

We identified HCC‐associated ILC2s as an immune regulatory cell type that promotes tumor development, suggesting that targeting these ILC2s might lead to new treatments for HCC.

Abbreviations

Arg1

arginase 1

BrdU

bromodeoxyuridine

Cdh1

cadherin‐1

CFSE

carboxyfluorescein succinimidyl ester

CM

conditioned medium

CXCL

chemokine (C‐X‐C motif) ligand

GATA3

GATA binding protein 3

GSEA

gene‐set enrichment analysis

ICOS

inducible T‐cell costimulatory

ILC2

group‐2 innate lymphoid cell

KLRG1

killer cell lectin‐like receptor subfamily G, member 1

LIN

lineage

n.s.

not significant

PBMC

peripheral blood mononuclear cell

PFS

progression‐free survival

PTGDR2

prostaglandin D2 receptor 2

PVMC

portal vein blood cell

RNA‐seq

RNA sequencing

RPMI‐1640

Roswell Park Memorial Institute 1640 medium

scRNA‐seq

single‐cell RNA‐seq

TCR

T‐cell receptor

TME

tumor microenvironment

TSLP

thymic stromal lymphopoietin

t‐SNE

t‐distributed stochastic neighbor embedding

In the liver, considered as the largest immune organ, resident immune cell populations increase the pathological complexity of HCC, which is a leading cause of cancer‐related deaths worldwide.^{( 1 )} Although tumor‐infiltrating lymphoid cells are abundant in HCC tissues,^{( 2 )} the presence of different subtypes and their divergent interactions with other tumor microenvironment compartments, rendering them incapable of being primed,^{( 3 )} culminate in tumor progression and even treatment failure. Therefore, identifying the various infiltrating immune cells and determining their interactions within the tumor microenvironment (TME) will lead to improved HCC therapies.

Recently, innate lymphoid cells (ILCs) have been identified in the TME of diverse tumor types.^{( 4 )} ILCs are a heterogeneous family of innate immune cells lacking rearranged antigen‐specific receptors, which are classified into ILC1, ILC2, and ILC3 subsets based on transcription factors, phenotypic markers, and cytokine expression.^{( 5 )} ILCs are mostly found in the lung, skin, and gastrointestinal mucosa, playing roles in clearing helminth infection, tissue homeostasis, and lymphoid tissue formation.^{( 6 )} Collectively, IL‐33, IL‐25, and thymic stromal lymphopoietin (TSLP) are essential for ILC2 activation, because ILC2s lack T‐cell receptors (TCRs) or B‐cell receptors; however, the important of this is tissue‐dependent.^{( 7 )} Different ILC2 subsets are involved in various inflammatory diseases^{( 8} , ^{9 )} and in liver disorders, such as chronic hepatitis^{( 10 )} and hepatic fibrosis.^{( 11 )} Although ILC2s have been investigated in several tumor types, their involvement in HCC is unknown.

Tumor‐resident neutrophils can restrict or augment tumor growth, depending on the TME.^{( 12 )} At the interface between innate and adaptive immunity, neutrophils are recruited by monocytes, macrophages, or epithelial cells through chemokine (C‐X‐C motif) ligand (CXCL) 5, CXCL2, CXCL1, or monocyte chemoattractant protein‐1 to regulate T cells or regulatory T cells.^{( 13 )} ILC2‐expressed Th2 cytokines, such as IL‐5 and IL‐13, might affect neutrophils functionally in an inflammatory context^{( 10} , ^{14 )}; however, how ILC2s regulate neutrophils in HCC is unclear.

In the present study, we identified a KLRG1⁻ hepatic ILC2 subpopulation in HCC tissue, which is independent of the ILC2 activators IL‐33, IL‐25, and TSLP that have been detected in other inflammatory diseases.^{( 9} , ^{15 )} Mediated by low E‐cadherin levels, relief of killer cell lectin‐like receptor subfamily G, member 1 (KLRG1) signaling induced ILC2 proliferation and CXCL2 and IL‐13 production. The latter attracted neutrophils to the tumor site and up‐regulated arginase 1 (Arg1), which suppressed T‐cell activation. Taken together, the hepatic ILC2‐CXCL2‐neutrophil axis induced an immunosuppressive TME that enabled tumor progression.

Materials and Methods

Human Clinical Specimens

The project was approved by the ethics committee of the First Affiliated Hospital and the Second Affiliated Hospital, Zhejiang University School of Medicine. All samples were anonymously coded according to local ethical guidelines. Informed patient consent was obtained.

Cells

Murine ILC2 cells were isolated from the liver of recombinant IL‐33‐treated C57BL/6 or CD45.1 mice and were cultured with 10 ng/mL recombinant murine IL‐2 (Pepro Tech, Cranbury, NJ) and IL‐33 (Pepro Tech) in Roswell Park Memorial Institute 1640 (RPMI‐1640) medium. Culture methods for other cell lines are described in the Supporting Information.

Mice

Six‐week‐old to 8‐week‐old male C57BL/6JNju were purchased from the Model Animal Research Center of Nanjing University (Nanjing, China), and congenic B6 mice with a CD45.1 allelic marker were obtained from the Jackson Laboratory (Bar Harbor, ME). The mice were bred and housed under pathogen‐free conditions at an animal facility. All animal experiments were conducted according to the protocols approved by the Ethics Committee of the First Affiliated Hospital, Zhejiang University School of Medicine.

Establishment of the In Situ Tumor Model

The in situ HCC model was established as described previously.^{( 16 )} Detailed procedures are given in the Supporting Information.

Isolation of Immune Cells From Different Tissues

A digestion solution containing 2% fetal bovine serum RPMI‐1640, type IV collagenase (1 mg/mL; Worthington, Lakewood, NJ) and DNase (15 µg/mL; Sigma‐Aldrich, St. Louis, MO) was used for chemical digestion. Methods of isolation and digestion for different tissues are described in the Supporting information.

Flow Cytometry and FACS

Single‐cell suspensions were stained with different antibody panels according to specific experimental designs, as described in the manufacturer’s instructions. Please refer to the Supporting information for the detailed procedures and antibody list (Supporting Table S1).

Mass Cytometry Staining and Analysis

Mass cytometry was performed by PLTTech Inc. (Hangzhou, China). The protocol has been described previously.^{( 17 )} The antibodies used in the assay are listed in Supporting Table S2. The detailed staining procedures are given in the Supporting Information.

Preparation of the Conditioned Medium

ILC2‐PCDH‐GFP, ILC2‐PCDH‐KLRG1, ILC2‐CRISPR‐GFP, and ILC2‐CRISPR‐KLRG1 cells (1 × 10⁵) were seeded in 24‐well plates and cultured in RPMI‐1640 with 10 ng/mL IL‐2 and IL‐33. The supernatant was collected and filtered after 3 days of culture.

Carboxyfluorescein Succinimidyl Ester Labeling

According to the manufacturer’s instructions, sorted ILC2s or T cells were incubated with 0.5 µM carboxyfluorescein succinimidyl ester (CFSE) labeling (BioLegend, San Diego, CA) in 1 mL of PBS for 15 minutes at 37°C, and quenched using a 5× volume of precooled complete medium. The labeled cells were then used for co‐culture experiments.

Western Blotting

Total protein from frozen human liver tissues or cell lines was extracted using radio immunoprecipitation assay buffer (Thermo Fisher Scientific, Waltham, MA). Please refer to the Supporting Information for the detailed procedures. The primary antibodies used are listed in Supporting Table S1.

RNA Extraction and Quantitative Real‐Time Reverse‐Transcription PCR

RNA was extracted using Trizol (Thermo Fisher Scientific) and reverse‐transcribed to complementary DNA (cDNA), which was subjected to quantitative PCR, and was performed on an Applied Biosystems 7500 Fast Real‐Time PCR System (Foster City, CA) according to the manufacturer’s instructions. Please see the Supporting Information for the detailed procedures and sequence information (Supporting Table S3).

RNA Sequencing and Gene‐Expression Analysis

RNA sequencing (RNA‐seq) was performed using the Illumina Hiseq XTEN platform (San Diego, CA) at Novogene Co. Ltd. (Beijing, China). Gene‐set enrichment analysis (GSEA) was performed using the online tool of the LC Bio Cloud Platform (https://www.omicstudio.cn/tool). Morpheus was used for heatmap visualization (https://software.broadinstitute.org/morpheus/). Please refer to the Supporting Information for the detailed procedures. Raw data were deposited in the GEO database (accession number GSE151895).

Single‐Cell RNA‐Seq and Analysis

Single‐cell RNA‐seq (scRNA‐seq) was performed using the Illumina HiSeq PE150 platform at Novogene following standard procedures, as described in the Supporting Information. The t‐distributed stochastic neighbor embedding (t‐SNE) analysis and cell clustering were performed using the Loupe Cell Browser (version 3.1.0; 10X Genomics, Pleasanton, CA).

In Vivo Bromodeoxyuridine Assay

At 4‐5 weeks after establishment of the in situ HCC model by hydrodynamic injection of oncogene plasmids (when the tumor nodules were macroscopically visible), the mice were injected intraperitoneally with 1 mg of bromodeoxyuridine (BrdU; Becton, Dickinson and Co., Franklin Lakes, NJ). After 24 hours, the hematopoietic cells from their livers were extracted and analyzed for BrdU incorporation using flow cytometry, according to the manufacturer’s instructions.

Plasmid Construction, Lentiviral Production, and Infection

The mouse cadherin‐1 (Cdh1) and Klrg1 cDNAs were cloned into the pCDH‐CMV‐MCS‐EF1a‐copGFP lentiviral vector (System Biosciences, Palo Alto, CA) and single‐guide RNAs of Klrg1, Cdh1, and Cxcl2 were ligated into the lentiCRISPRV2(GFP) vector (Hebio, Shanghai, China), based on previously published protocols.^{( 18 )} Please refer to Supporting Information Table S3 for the detailed procedures and sequence information.

Hepatic ILC2 Isolation From Murine Liver

Wild‐type mice were injected peritoneally with 400 ng of recombinant murine IL‐33 (Thermo Fisher Scientific) every day for 7 days. The mice were sacrificed, and hematopoietic cells from their livers were stained for ILC2s and isolated using FACS. Primary ILC2s were cultured in vitro in the presence of 10 ng/mL IL‐2 and IL‐33.

In Vivo Application of ILC2s

On day 14 after the in situ HCC models were established, 1 × 10⁵ of ILC2s, ILC2‐PCDH‐KLRG1, ILC2‐CRISPR‐KLRG1, and ILC2‐CRISPR‐CXCL2 cells, and related control cells were resuspended in 200 μL of PBS and injected through the caudal vein, every 3 days, following the experimental design. The mice were sacrificed after 2‐4 weeks of injection.

Other Techniques

Please consult the Supporting Information for the detailed procedures used for the following techniques: magnetic‐activated cell sorting, immunohistochemistry, cell proliferation assay of CD8⁺ T cells, chemoattractant assay, recombinant human E‐cadherin binding experiments, in vitro co‐culture of ILC2 and Hepa1‐6 or AML‐12 cells, ELISA, immunofluorescence, adoptive infusion of CD45.1‐ILC2s, in vivo neutrophil depletion assay, and hematoxylin and eosin staining.

Statistical Analysis

Statistical analysis was conducted using GraphPad Prism 6 (GraphPad Inc., San Diego, CA) or SPSS (IBM Corp., Armonk, NY). Please see the Supporting Information for the detailed procedures.

Results

Abundance of ILC2s in Tumor Correlates With HCC Prognosis

As previously defined in human tissues, we isolated and defined the CD45⁺lineage (LIN)⁻CD127⁺CRTH2⁺ cells expressing high levels of GATA binding protein 3 (GATA3) from the paired human tumor and nontumoral liver tissues (n = 150) as ILC2s (Supporting Fig. S1A,B). Taking normal liver tissues without inflammation and fibrosis from excised liver hemangiomas as controls, we found that the percentage of ILC2s (gated as CD45⁺LIN⁻CD127⁺CRTH2⁺) in the LIN⁻ subset was substantially elevated in tumors (Fig. 1A). Consistent results were obtained regarding the percentage of ILC2s in the LIN⁻ subset (Fig. 1B) when the ILC2s were defined as CD45⁺LIN⁻CD127⁺CRTH2⁺GATA3⁺RORγt^lo (Supporting Fig. S1C) in another cohort of paired HCC and nontumoral tissues (n = 28). The absolute number of ILC2s per gram of tumor tissue of these two patient cohorts were elevated (Supporting Fig. S1D‐F). In addition, the percentages of ILC2s in the CD45⁺ subset and total lymphocytes in the liver, especially in tumors, were higher compared with those in peripheral blood mononuclear cells (PBMCs) or portal vein blood cells (PVMCs) (Fig. 1C, left two panels). The proportions of total ILCs relative to the number of CD45⁺ cells were also elevated in tumor tissues (Fig. 1C, right panel), which suggested that the ILC2s in the HCC microenvironment were proliferating and accumulating. Similar results were obtained from the second cohort (Fig. 1D). Given the differential frequency of ILC2s between HCC and nontumoral livers, we next investigated whether there was a correlation between ILC2s and patient survival. To minimize the effect caused by patient heterogeneity, we normalized the proportion of ILC2s in the LIN⁻ population of HCC tissues to that of their respective nontumoral liver controls. The median value was set as the cutoff point to divide the 150 cases of HCC into ILC2 high‐abundance or low‐abundance groups. Patients with a high tumor/nontumor ILC2 ratio survived worse, with both of their progression‐free survival (PFS) and overall survival affected (Fig. 1E). Among various clinical pathological factors (Supporting Table S4), HBV infection, microvascular invasion, and recurrence correlated positively with hepatic ILC2 abundance in patients with HCC (Fig. 1F). Univariate and multivariate analyses showed that tumor‐node‐metastasis stage and a high tumor/nontumor ILC2 ratio were independent factors for the incidence of recurrence and patient survival (Supporting Tables S5‐S7). These data indicate that hepatic ILC2 abundance correlates strongly with HCC progression.

FIG. 1.

HCC‐derived hepatic ILC2s correlate with HCC survival. (A) Percentages of ILC2s (gated as CD45⁺LIN⁻CD127⁺CRTH2⁺) in the CD45⁺LIN⁻ subsets of 150 paired tumor and nontumor tissues and 20 normal liver tissues. (B) Percentages of CRTH2⁺GATA3⁺RORγt^lo ILC2s in tumor, paired nontumor tissues (n = 28), and normal liver tissues (n = 11) relative to the LIN⁻ subset. (C) Percentages of ILC2s in tumor, paired nontumor tissues (n = 150), and autologous peripheral blood (n = 38) and portal vein blood (n = 12) relative to the CD45⁺ cells, lymphocytes, and ILC subsets in CD45⁺ cells. (D) Percentages of GATA3⁺ ILC2s in tumor, paired nontumor tissues (n = 28), and normal liver tissues (n = 11) relative to the CD45⁺ cells, lymphocytes, and ILC subsets in CD45⁺ cells. (E) The PFS and overall survival of low and high tumor/no‐tumor ILC2 ratios. The cohort was separated based on the median value of the tumor/nontumor ILC2 ratio. (F) Potential correlations of tumor/nontumor ILC2 ratios with HBV infection, microvascular invasion, and recurrence. Statistical analysis was performed using the Mann‐Whitney U test (A‐D,F) or Kaplan‐Meier curves using the log‐rank test (E). *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. Abbreviations: n.s., not significant; OS, overall survival.

Phenotype of Hepatic ILC2s in the HCC Microenvironment

To identify the human liver tumor–derived ILC2s, we sorted LIN⁻ hematopoietic cells from one HCC and nontumoral liver tissues, and applied scRNA‐seq. As shown in the t‐SNE plots, the ILC2 cluster from HCC (18.3%) was enlarged compared with that from the nontumoral tissue (4.2%) when confined to LIN⁻CD127⁺ cells (Fig. 2A,B), which was consistent with the statistical results from flow cytometry (Fig. 1). The mRNA expression pattern of known ILC2 selective (e.g., GATA3, prostaglandin D2 receptor 2 [PTGDR2], inducible T‐cell costimulatory [ICOS], killer cell lectin like receptor B1 [KLRB1], KLRG1, IL‐2RA) and exclusive (e.g., T‐box transcription factor 21 [TBX21], eomesodermin [EOMES], RAR related orphan receptor C [RORC]) genes validated the identity of these ILC2s under HCC conditions (Fig. 2C). We further confirmed our findings using CNGB Nucleotide Sequence Archive scRNA‐seq data (CNP0000650) acquired from 18 patients with HCC (Supporting Fig. S2A,B and Fig. 2D). To verify our finding at the protein level, we used mass cytometry and confirmed a series of ILC2 phenotypic (ICOS⁺CD25⁺CD161⁺CD117^+/−KLRG1^+/−; Fig. 2E) and transcriptional (T‐bet⁻RORγt^lo; Supporting Fig. S2C) markers. Surprisingly, a substantial subset of HCC‐enriched ILC2s did not express KLRG1 (orange oval circle), a well‐documented marker for mature ILC2s in the blood, mucosal tissues, and nonmucosal tissues in the resting state,^{( 19 )} suggesting the existence of an HCC‐derived ILC2 subtype. Furthermore, levels of known ILC2 activators, including IL‐33, IL‐25, and TSLP, were diminished in HCC samples (Fig. 2F and Supporting Fig. S3), indicating that the HCC‐derived ILC2s might be functionally divergent from canonical ILC2s. Thus, a distinct tumor‐derived ILC2 population accumulates in the HCC microenvironment.

FIG. 2.

The phenotype of hepatic ILC2s in tumor and nontumoral tissue. (A) Two‐dimensional visualization of single‐cell cluster of ILC2s (gated on LIN⁻ IL7R⁺PTGDR2⁺GATA3⁺ ) in HCC and nontumoral tissue by t‐SNE, indicated by a red dotted circle. Each dot corresponds to one single cell. (B) The proportion of ILC2s across LIN⁻IL7R⁺ cell subsets in paired HCC and nontumoral tissue (blue, ILC2s; gray, other cells). (C) ILC2 cluster in t‐SNE plots (every single cell captured for sequencing was included; T = 5,585 cells, n = 5,116 cells) marked by the mRNA expression of surface markers (IL‐7R, PTGDR2, ICOS, KLRB1, KLRG1, IL‐2RA, and KIT) and transcription factors (GATA3, TBX21, EOMES, and RORC) (green, high; gray, low). (D) Magnified t‐SNE visualized plots showing the expression of selected identification and exclusion marker genes for ILC2s (Supporting Fig. S2A,B) from scRNA‐seq data from the CNGB Nucleotide Sequence Archive (CNP0000650) (red, high; blue, low). (E) Visualized t‐SNE map of cells from tumor and nontumor tissue pregated on CD45⁺LIN⁻CD127⁺ cells from mass cytometry analysis. Normalized expression intensities of related surface markers were calculated and overlaid on the t‐SNE plot (red, high; blue, low). Orange and blue oval dotted circles indicate KLRG1⁻ and KLRG1⁺ cells, respectively. Data shown are from 3 patients. (F) Concentration of IL‐33 (n = 18), IL‐25 (n = 18), and TSLP (n = 9) in the supernatants of liver tissues as determined using ELISA. Statistical analysis was performed using a two‐tailed paired Student t test. Data are shown as the mean ± SD. *P < 0.05, **P < 0.01. Abbreviations: EOMES, eomesodermin; KLRB1, killer cell lectin like receptor B1; RORC, RAR related orphan receptor C; TBX21, T‐box transcription factor 21.

KLRG1‐ ILC2s are Tissue‐Derived and More Proliferative in HCC

We found that the HCC‐induced ILC2s expressed higher CD69 levels than ILC2s from nontumoral tissue, whereas those isolated from human peripheral and portal vein blood lacked CD69 (Fig. 3A), prompting us to speculate that these HCC‐accumulated ILC2s were derived from local tissues, not the circulation. From the transcriptomic data of ILC2s sorted from the paired HCC and nontumoral tissues, we detected high expression levels of certain core tissue residency genes in the tumor‐derived ILC2s, whereas that of nontumoral tissues had a transcriptomic signature similar to circulating naïve CD3⁺ T cells (Fig. 3B, left; and Supporting Table S7), which were inhibited in the HCC‐derived ILC2s (Fig. 3B, right; Supporting Table S7). Immunofluorescence assays showed that hepatic ILC2s were distant from the circulating T (CD3⁺ CD5⁺ TCRα/β⁺ TCRγ/δ⁺) cell aggregation pool, which was thought to reside either in the central vein or the portal area in the liver, indicating a noncirculating characteristic of hepatic ILC2s in tumors (Fig. 3C).

FIG. 3.

HCC‐originated hepatic ILC2s are tissue‐derived and more proliferative. (A) Representative flow cytometry histogram showing CD69 expression in ILC2s from tumor and nontumor tissues and peripheral blood from 1 representative patient (left panel). Percentage of CD69⁺ ILC2s from liver tissues (tumor and nontumor) and peripheral blood of 9 patients and the portal vein blood of 3 patients (right panel). (B) Relative expression of 38 core tissue‐resident and 26 core circulating genes in ILC2s derived from HCC, nontumor tissue, and peripheral blood naïve CD3⁺ T cells (blue, low; red, high; see Supporting Table S7 for gene list). (C) Representative immunofluorescence images of human HCC tissue sections stained with antibodies against CD3, CD5, TCRα/β, TCRγ/δ (pink), CD127 (green), and GATA3 (red), and stained with 4′,6‐diamidino‐2‐phenylindole (blue). Magnification, ×904. Cells marked by white arrows are ILC2s. (D) Representative flow cytometry plots showing KLRG1 expression in ILC2s from tumor and paired nontumor liver tissue (left panel). Percentage of KLRG1 expression in ILC2s from tumor, paired nontumor liver tissues (n = 15), and peripheral blood (n = 8) pregated on CD45⁺LIN⁻CD127⁺CRTH2⁺ cells (right panel). (E) Correlation of the percentage of KLRG1⁻ hepatic ILC2s to that of the total hepatic ILC2s in lymphocytes from HCC tissues (n = 15). (F) Proliferation of the sorted KLRG1⁻ ILC2s and KLRG1⁺ ILC2s co‐cultured with OP9 cells. CFSE‐labeled cells were detected using flow cytometry at days 1, 3, 5, and 7 (left panel). The number of KLRG1⁻ and KLRG1⁺ ILC2s recorded on days 5, 10, 15, and 20 (right panel). Statistical analysis was performed using the Mann‐Whitney U test (A), the Pearson correlation test (E), two‐way ANOVA (F), or a two‐tailed unpaired (D) Student t test. *P < 0.05, **P < 0.01, ****P < 0.0001. Abbreviations: DAPI, 4′,6‐diamidino‐2‐phenylindole; N, nontumor ILC2s; P, PBMC‐naïve CD3⁺ T cells; T, tumor ILC2s.

Flow cytometry confirmed an increased proportion of KLGR1⁻ ILC2s in HCC tissue than in nontumoral tissue and PBMCs (Fig. 3D). To study the origin of KLGR1⁻ ILC2s in HCC, we co‐cultured hepatic ILC2s (CD45⁺LIN⁻CD127⁺ST2⁺) obtained and identified from recombinant murine IL‐33‐treated mice (Supporting Fig. S4) directly with Hepa1‐6 cells, and observed a significant decrease in KLRG1 in the former, whereas co‐culture in Transwell inserts without direct cell‐to‐cell contact had no effect (Supporting Fig. S5). Interestingly, KLRG1 expression on ILC2s was also unaffected when co‐cultured with AML‐12 cells (Supporting Fig. S5). KLRG1 signaling is mediated by an interaction with cadherin, particularly E‐cadherin; therefore, we investigated their relationship in ILC2s. Corroboratively, E‐cadherin was absent from Hepa1‐6 and H22 mouse HCC cell lines (Supporting Fig. S6A) and was decreased in most human HCC cell lines (Supporting Fig. S6B). E‐cadherin expression, represented by fragments per kilobase million values from the GEO database, in HCC were lower compared with that of nontumoral tissues (Supporting Fig. S6C), and the protein level was lower in a higher proportion of HCC samples (Supporting Fig. S6D). Importantly, overexpression of Cdh1 (encoding E‐cadherin) in Hepa1‐6 cells maintained (Supporting Fig. S8A,B), whereas knockout of Cdh1 in AML‐12 cells decreased, KLRG1 expression in co‐cultured ILC2s (Supporting Fig. S7C,D), which helped to determine the interactions between E‐cadherin and KLRG1. Further verified in HCC samples shown in Fig. 3C, the positive correlation between E‐cadherin expression and the KLRG1⁺ILC2 ratio (Supporting Fig. S8) confirmed the relevance of these two markers.

KLRG1 could induce inhibitory signaling; therefore, we hypothesized that KLRG1 controls the proliferation of hepatic ILC2s. Indeed, we observed a positive correlation between the proportion of KLRG1⁻ hepatic ILC2s and the total ILC2s of hepatic lymphocytes (Fig. 3E). Human primary hepatic ILC2s isolated from tumor tissues displayed a typical lymphoid morphology, expressed GATA3 and IL‐13, and could proliferate dramatically in vitro in the presence of IL‐2 and IL‐33 (Supporting Fig. S9). KLRG1⁻ ILC2s isolated from HCC tissues showed significantly higher proliferation rates compared with KLRG1⁺ ILC2s when co‐cultured with OP9 bone marrow stromal cells in the presence of 10 ng/mL IL‐2, IL‐7, IL‐25, and IL‐33 (Fig. 3F, left), leading to a drastic increase in their total numbers (Fig. 3F, right). Taken together, the self‐renewal capacity of HCC‐derived ILC2s contributes to their accumulation at tumor sites.

HCC‐Derived Hepatic ILC2s Produce Neutrophil Attractants CXCL2, CXCL8, and Th2 Cytokine IL‐13

Transcriptome analysis of hepatic ILC2s from paired HCC and nontumor tissues showed a significant enrichment of cell chemotaxis function, which was reflected in the up‐regulation of the CXC chemokine family members, especially CXCL2 and CXCL8, in HCC‐associated hepatic ILC2s (Fig. 4A). GSEA further revealed that the difference between tumor and nontumoral ILC2s resembled that between lipopolysaccharide‐treated or control‐treated monocytes (normalized enrichment score [NES] = 1.647, p.adjust = 0.011; Fig. 4B). As a dominant member of Th2 cytokines, IL‐13, along with other secretory factors, drove a significant enrichment of cytokine and cytokine receptor interaction–related genes in HCC‐derived ILC2s (NES = 1.56, p.adjust = 0.016; Fig. 4C). After treatment with a leukocyte stimulation cocktail in vitro, using the CD127⁺CRTH2⁻GATA3⁻ subset as a control, we confirmed that HCC‐derived ILC2s secreted higher levels of IL‐13 and CXCL8 (IL‐8) than those from nontumoral tissues (Fig. 4D). Accordingly, the frequency of HCC‐induced ILC2s correlated positively with the number of neutrophils relative to the CD45⁺ population (Fig. 4E), because both CXCL2 and CXCL8 chemokines are powerful neutrophil attractants. After co‐culture with AML‐12 cells, ILC2 expression of Il‐13 and Cxcl2 was strongly inhibited, compared to those cultured with Hepa1‐6 cells (Fig. 4F). Taken together, HCC‐derived ILC2s are functionally enhanced.

FIG. 4.

HCC‐derived ILC2s express higher levels of CXCL2, CXCL8, and IL‐13. (A) Relative expression of chemotaxis signature genes in paired tumor and nontumor hepatic ILC2s isolated from 3 patients. Data were derived from RNA‐seq. (B,C) GSEA analysis showing the genes from HCC‐derived ILC2s enriched primarily in lipopolysaccharide‐stimulated monocytes (B) and interactions between cytokines and receptors (C). (D) Representative cytometry plots (left panel) and percentage analysis (n = 8) of IL‐13 and IL‐8 (CXCL8) expression from tumor‐originated ILC2s (red), nontumoral ILC2s (solid), and non‐ILC2s (CRTH2⁻GATA3⁻, dotted). (E) Correlation of the percentage of total hepatic ILC2s to that of CD45⁺CD11b⁺CD66b⁺ neutrophils (n = 30). (F) 4 × 10⁴ murine ILC2s were co‐cultured with Hepa1‐6 or AML‐12, respectively, for 3 days. Relative mRNA expression of Il‐13, Cxcl2, Il‐10, and Cxcl1 in murine ILC2s isolated from the co‐culture system, as examined using quantitative real‐time PCR. Statistical analysis was performed using a two‐tailed paired Student t test (D,F) and Pearson correlation test (E). *P < 0.05, ***P < 0.001, ****P < 0.0001. Abbreviations: CTRL, control; CCRL2,C‐C motif chemokine receptor like 2; CCL5,C‐C motif chemokine ligand 5; CCR9,C‐C motif chemokine receptor 9; CCL3L3,C‐C motif chemokine ligand 3 like 3; CRTH2,chemo‐attractant Receptor‐ homologous molecule expressed on Th2 cells; KEGG, Kyoto Encyclopedia of Genes and Genomes; LGALS3, lectin, galactoside‐binding, soluble, 3; LPS, lipopolysaccharide; LGALS3,lectin, galactoside‐binding, soluble, 3; Non‐T‐ILC2,nontumor‐ILC2; S1PR1,sphingosine‐1‐phosphate receptor 1; XCL1,X‐C motif chemokine ligand 1.

Loss of KLRG1 Causes Enhanced Proliferation of ILC2s, Attracts Neutrophils, and Induces an Immunosuppressive Phenotype

We then validated these findings in an established c‐Myc/NRas‐induced murine HCC model,^{( 16 )} in which we detected a higher frequency and larger number of CD45⁺LIN⁻CD127⁺GATA3⁺RORγt⁻ ILC2s in the tumor‐bearing livers (Supporting Figs. S10 and S11), similar to the observations in human specimens. Additionally, the c‐Myc/NRas‐induced HCC also harbored fewer KLRG1⁺ ILC2s (Fig. 5A). An in vivo BrdU assay showed that hepatic KLRG1⁻ ILC2s exhibited a significantly stronger proliferative capacity than the KLRG1⁺ ILC2s in HCC tissues (Fig. 5B). To determine whether KLRG1 is a key factor modulating ILC2 proliferation, we generated Klrg1 knockout (ILC2‐CRISPR‐KLRG1) and overexpressing (ILC2‐PCDH‐KLRG1) ILC2s (Supporting Fig. S12) and cultured them with 10 ng/mL IL‐2/IL‐33. Klrg1 overexpression suppressed, whereas deprivation of KLRG1 inhibitory signaling promoted, ILC2 proliferation (Fig. 5C).

FIG. 5.

KLRG1⁻ ILC2s promote HCC by self‐proliferation, and attracting and inducing an immunosuppressive phenotype in neutrophils. (A) KLRG1 expression in hepatic ILC2s from the c‐Myc/Nras‐induced HCC model and control mice. (B) Proliferation of KLRG1⁻ ILC2s and KLRG1⁺ ILC2s determined by BrdU incorporation in oncogene‐induced liver tumors and controls. (C) Proliferation curves of ILC2s‐CRISPR‐GFP, ILC2‐CRISPR‐KLRG1, ILC2‐PCDH‐GFP, and ILC2‐PCDH‐KLRG1 cells that were cultured in vitro with 10 ng/mL IL‐2 and IL‐33 for 8 days. (D) Relative mRNA levels of Il‐5, Il‐10, Il‐13, and Cxcl2 in ILC2‐CRISPR‐GFP, ILC2‐CRISPR‐KLRG1, ILC2‐PCDH‐GFP, and ILC2‐PCDH‐KLRG1 cells. (E) Relative expression of Arg1 in neutrophils cultured in the CM of ILC2‐CRISPR‐GFP, ILC2‐CRISPR‐KLRG1, ILC2‐PCDH‐GFP, and ILC2‐PCDH‐KLRG1 cells. (F) Percentage of proliferative CD8⁺ T cells labeled with CFSE and interferon‐γ‐producing CD8⁺ T cells detected using flow cytometry after 2 days of co‐culture, with neutrophils pretreated with the CM of ILC2‐PCDH‐KLRG, ILC2‐CRISPR‐KLRG1, and related control cells. Statistical analysis was performed using two‐way ANOVA (C) or two‐tailed unpaired Student t test (A‐B,D‐F). *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Abbreviation: GFP, green fluorescent protein; HDI, hydrodynamic injection.

We next determined the effect of Klrg1 on the expression levels of Th2 cytokines and Cxcl2 in murine ILC2s, because the RayBio mouse cytokine antibody array (Mouse 23‐plex) showed that isolated murine ILC2s primarily produced Th2 cytokines (Supporting Fig. S13), and the transcriptome data indicated that human HCC‐induced ILC2s expressed equal levels of IL‐13, CXCL2, and CXCL8 (Supporting Fig. S14). The CXCL8 gene was not examined because it is absent in mice.^{( 20 )} Klrg1 overexpression down‐regulated Il‐13 and Cxcl2 mRNA levels, whereas its knockout enhanced the production ability of ILC2s (Fig. 5D). A chemotaxis assay showed that the conditioned medium (CM) of ILC2‐PCDH‐KLRG1 cells decreased neutrophil migration significantly, whereas ILC2‐CRISPR‐KLRG1 CM promoted chemotaxis (Supporting Fig. S15), in accordance with their chemokine production capacity. Furthermore, the application of anti‐CXCL2 antibodies inhibited chemotaxis significantly in both groups (Supporting Fig. S16). The results were also verified using hepatic ILC2s isolated from human HCC: Stimulation of KLRG1 signaling using E‐cadherin hindered cytokine and chemokine expression, and neutrophil attraction (Supporting Fig. S17). The ILC2 CM also induced an immunosuppressive cytokine profile in neutrophils (Supporting Fig. S18), likely by up‐regulating Arg1 through Th2 cytokines. Consistent with the regulation of Il‐13 by Klrg1, Arg1 levels were up‐regulated by the ILC2‐CRISPR‐KLRG1 CM and down‐regulated by the ILC2‐PCDH‐KLRG1 CM (Fig 5E). In addition, neutrophils pre‐incubated with the CMs of ILC2‐CRISPR‐KLRG1 or ILC2‐PCDH‐KLRG1 significantly decreased or enhanced the proliferation and cytotoxicity of the co‐cultured CD8⁺ T cells, respectively (Fig. 5F and Supporting Fig. S19). Taken together, these results showed that KLRG1⁻ hepatic ILC2s induce an immunosuppressive TME.

ILC2s Promote HCC Development Through CXCL2/Neutrophils In Vivo

We then confirmed our speculation about the role of the ILC2/CXCL2/neutrophil pathway in HCC development in vivo. From the normal liver, early stage, and late stage of HCC in the mouse model, the ratio and absolute number of ILC2s, Cxcl2 expression, and the neutrophil infiltration rate increased with increasing tumor burden. Meanwhile, in the tissue that lost E‐cadherin expression, the KLRG1 expression of ILC2s decreased significantly (Supporting Fig. S20). We adoptively transferred CD45.1⁺ ILC2s into tumor‐bearing and control mice of the CD45.2 strain, and found that at 3 days following injection, the transferred CD45.1⁺ ILC2s with diminished KLRG1 expression were detectable in the tumor sites (Supporting Fig. S21). Fourteen days after establishing the in situ HCC model, tumor‐bearing mice were injected with ILC2s or PBS every 3 days. Transferring ILC2s resulted in significantly increased neutrophils, arginase expression, and tumor burden (Fig. 6A and Supporting Fig. S22). To test the contribution of KLRG1, we transferred ILC2s with either Klrg1 overexpression or knockout, and observed elevated neutrophils, higher arginase expression, and a heavier tumor burden in the ILC2 Klrg1 knockout group (Fig. 6B and Supporting Fig. S23). Furthermore, after ILC2‐CRISPR‐CXCL2 cells were generated and purified (Supporting Fig. S24), the tumor‐promoting ability of ILC2s lacking CXCL2 expression was demonstrated to be markedly reduced, together with decreased neutrophil accumulation (Fig. 6C and Supporting Fig. S25A), although there were no differences in arginase expression, which might reflect the unaffected IL‐13 levels (Supporting Fig. S25B). Finally, antibody‐mediated depletion of neutrophils (Fig. 6D) in the tumor‐bearing mice injected with ILC2s alleviated the tumor burden (Fig. 6E and Supporting Fig S26) and partially restored the CD8⁺ T‐cell population (Fig. 6F). Therefore, ILC2s promote HCC development through the creation of a neutrophil‐mediated suppressive microenvironment.

FIG. 6.

ILC2s promote HCC development through CXCL2/neutrophils in vivo. (A) Murine in situ HCC models were established through hydrodynamic injection of oncogene plasmids. Murine ILC2s and PBS were injected into the caudal vein every 3 days. The percentage of neutrophils in CD45⁺ cells, MFI of arginase in neutrophils, and liver weight relative to body weight are shown. (B) HCC model mice were treated with injection of ILC2‐CRISPR‐KLRG1 or ILC2‐PCDH‐KLRG1 cells. The percentage of neutrophils in CD45⁺ cells, mean fluorescence intensity (MFI) of arginase in neutrophils, the liver/body weight ratio, and liver tumor nodules were calculated. (C) HCC model mice were injected with ILC2‐CRISPR‐GFP or ILC2‐CRISPR‐CXCL2 cells. The percentage of neutrophils in CD45⁺ cells, MFI of arginase in neutrophils, and liver/body weight ratio are shown. (D) ILC2‐treated HCC mice were injected peritoneally with 300 μg of anti‐Ly6G or anti‐IgG antibody every 2 days. Representative flow cytometry plots (gated on alive CD45⁺CD3⁻) showing the neutrophil (CD11b⁺Ly6G^hi) depletion efficiency and the percentage of neutrophils in CD45⁺ cells. (E) Tumor burden was measured as liver/body weight and the number of liver tumor nodules. (F) The percentage of interferon‐γ⁺ CD8 T cells are shown. Statistical analysis was performed using a two‐tailed unpaired Student t test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001. Abbreviation: IFN‐γ, interferon‐γ.

Discussion

HCC is a heterogeneous malignancy with a highly complex immune microenvironment, which induces both anti‐tumor and pro‐tumor responses under distinct circumstances. We observed a higher proportion of hepatic ILC2s in HCC relative to nontumoral and normal liver tissues. Different ILCs show substantial plasticity, and reversible trans‐differentiation between ILC1s and ILC3s is frequently observed.^{( 21 )} Previously, we showed that Notch signaling converted natural ILC2s into inflammatory ILC2s that displayed dual ILC2/ILC3 capabilities.^{( 22 )} The present study revealed a significant increase in the overall ILC proportion among HCC tissue hematopoietic cells relative to the paired nontumoral liver tissues, strongly indicating active proliferation and self‐renewal of hepatic ILC2s in the HCC microenvironment, rather than trans‐differentiation from other subsets. Studies have shown that ILCs are tissue‐sessile cells with limited migratory abilities.^{( 23 )} Consistent with this, we observed similar ILC2 proportions between peripheral and portal vein blood, indicating the lack of migration into HCC tissues. Comparisons between HCC‐induced hepatic ILC2s and peripheral naive CD3⁺ T cells confirmed that the increased ILC2s levels in HCC originated within the tumor.

The role of ILC2s in tumor initiation and progression is ambiguous, partly because of the complex TME. However, proliferative ILC2s have a pro‐tumorigenic role in gastric,^{( 24 )} breast,^{( 25 )} and bladder^{( 26 )} cancers. Here, we report an immunosuppressive role of hepatic ILC2s in HCC. The ILC2‐initiating cytokines IL‐25, IL‐33, and TSLP, verified by numerous previous studies, appear not to be potent activators of these hepatic ILC2s under HCC circumstance. Jeffery et al.^{( 27 )} reported that intrahepatic ILC2s expressed low levels of the IL‐33 receptor, ST2, as confirmed by mass cytometry, which also showed lower levels of IL‐17RB and TSLPR in hepatic ILC2s (data not shown). Consistently, the levels of IL‐25, IL‐33, and TSLP in HCC were significantly diminished compared with those in the paired nontumoral tissues. Another study also identified ILC2‐like cells that were independent of IL‐33,^{( 28 )} indicating the existence of tissue‐specific pathways that determine ILC2 phenotype and function.

The HCC‐induced ILC2s were polarized to a less‐mature KLRG1⁻ phenotype, showing enhanced self‐renewal and proliferation. KLRG1 is expressed primarily by mature natural killer (NK) cells^{( 29 )} and terminally differentiated T cells,^{( 30 )} and correlates with reduced proliferation, increased apoptosis, and impaired IFN‐γ secretion in NK cells.^{( 31 )} KLRG1 interaction with tissue‐specific E‐cadherin could be an essential mechanism to maintain tissue homeostasis by ameliorating ILC2‐Th2 cytokine‐driven inflammation. Therefore, we hypothesized that low hepatic tumor ILC2 KLRG1 expression might result from diminished HCC cell E‐cadherin expression. Without E‐cadherin stimulation, hepatic ILC2s escape from KLRG1‐mediated inhibitory signaling. Moreover, low ST2 expression on hepatic ILC2s and/or low IL‐33 release into the HCC microenvironment augmented KLRG1 loss, because IL‐33 is known to up‐regulate KLRG1 in ILC2s.^{( 32 )} These KLRG1⁻ ILC2s were highly proliferative compared with the KLRG1⁺ ILC2s in human and murine HCC, supporting our hypothesis that enriched hepatic ILC2s in HCC result from self‐renewal rather than trans‐differentiation across ILC subgroups or migration from other tissues.

In addition to Th2 cytokines, ILC2s also produce amphiregulin,^{( 33 )} granulocyte‐macrophage colony‐stimulating factor, CXCL1, and CXCL2.^{( 34 )} We found that the HCC‐derived ILC2s expressed higher levels of CXCL2 and CXCL8, which are potently chemotactic for neutrophils.^{( 35} , ^{36 )} The expansion of hepatic KLRG1⁻ ILC2s attracted neutrophils to the tumor site through enhanced CXCL2 expression, confirming that innate lymphoid cells are important for neutrophil attraction, in addition to known epithelial, endothelial, or myeloid cells. In mouse lungs following severe trauma, IL‐33‐mediated IL‐5 secretion from ILC2s induces neutrophil IL‐5 production,^{( 14 )} and in the liver of a mouse hepatitis model, ILC2s induced immunosuppressive neutrophils through IL‐33‐mediated IL‐13 expression.^{( 10 )} These interactions showed that the functional changes of neutrophils are mediated by IL‐33; however, neither of them involved neutrophil recruitment. ILC2s were also interact with other immune cells in the TME, either by secreting IL‐13 to induce suppressive monocytic‐myeloid derived cells,^{( 26 )} or through the programmed death ligand 1/programmed death‐1 (PD‐1) interaction that inhibits T cells.^{( 37 )} Contrastingly, recent research showed that in pancreatic ductal adenocarcinoma, IL‐33‐mediated ILC2s amplify PD‐1 blockade and activate CD8⁺ T cells by recruiting CD103⁺ dendritic cells to enhance antitumor immunity.^{( 38 )} Therefore, there are several other pathways through which hepatic ILC2s might shape the tumor milieu. In the present study, hepatic KLRG1⁻ ILC2s induced immunosuppressive phenotypes of neutrophils, and the depletion of CXCL2 and neutrophils could reduce the liver tumor burden in mice partially.

To summarize, we identified a tumor tissue–derived KLRG1⁻ hepatic ILC2 subset that was enriched in HCC and showed enhanced self‐renewal. The KLRG1⁻ ILC2/CXCL2/neutrophil axis in HCC induces an immunosuppressive microenvironment (Fig. 7). The abundance of hepatic ILC2s in HCC is associated with tumor recurrence, and correlate with poor survival of HCC. Our findings provide insights into the role of innate lymphoid cells in HCC progression, and suggest hepatic ILC2s as an attractive target for HCC immunotherapy.

FIG. 7.

Schematic model depicting the HCC‐derived ILC2s inducing a suppressive immune microenvironment in HCC. When mediated by HCC with diminished E‐cadherin expression, ILC2s were relieved from suppression of KLRG1 signaling by down‐regulated KLRG1. Through acquisition of enhanced capacity of self‐renewing as well as secreting IL‐13 and CXCL2/CXCL8, ILC2s were capable of recruiting and changing neutrophils into immunosuppressive ones, to finally form an immunosuppressive microenvironment.

Author Contributions

X.X., L.Y., Q.Z., and T.L. contributed to the study concept and design. X.X., L.Y., and Q.Z. contributed to the analysis and interpretation of data. X.X., Q.Z., H.S., and X.Z. contributed to the methodology. X.X. and L.Y. contributed to the statistical analysis. X.X., S.L., and M.Y. provided technical or material support. X.X., L.Y., Q.Z., and T.L. wrote the original draft. T.L. supervised the study. L.Y., Q.Z., and T.L. acquired the funding.

Supporting information

Supplementary Material