CXCR6 marks a novel subset of T-bet(lo)Eomes(hi) natural killer cells residing in human liver

Abstract

Natural killer cells (NK) are highly enriched in the human liver, where they can regulate immunity and immunopathology. We probed them for a liver-resident subset, distinct from conventional bone-marrow-derived NK. CXCR6+ NK were strikingly enriched in healthy and diseased liver compared to blood (p < 0.0001). Human hepatic CXCR6+ NK had an immature phenotype (predominantly CD56(bright)CD16-CD57-), and expressed the tissue-residency marker CD69. CXCR6+ NK produced fewer cytotoxic mediators and pro-inflammatory cytokines than the non-liver-specific CXCR6- fraction. Instead CXCR6+ NK could upregulate TRAIL, a key death ligand in hepatitis pathogenesis. CXCR6 demarcated liver NK into two transcriptionally distinct populations: T-bet(hi)Eomes(lo)(CXCR6-) and T-bet(lo)Eomes(hi)(CXCR6+); the latter was virtually absent in the periphery. The small circulating CXCR6+ subset was predominantly T-bet(hi)Eomes(lo), suggesting its lineage was closer to CXCR6- peripheral than CXCR6+ liver NK. These data reveal a large subset of human liver-resident T-bet(lo)Eomes(hi) NK, distinguished by their surface expression of CXCR6, adapted for hepatic tolerance and inducible anti-viral immunity.

Figure 1. CXCR6+ NK cells are enriched in the liver in health and disease.

(a) Gating strategy for identification of NK cells (singlets, live cells, CD14−, CD19−, CD3−, CD56+ cells) using multicolor flow cytometry and representative histograms of CXCR6 staining of NK cells from periphery and liver. (b) Summary frequencies of CXCR6+ NK cells from PBMC (n = 20), and liver (n = 15), comprising biopsies obtained from healthy donor livers (n = 10) or healthy margins of resections for colorectal metastases (n = 5). (c) Direct comparison of CXCR6 frequencies of NK cells isolated from perfusion liquid and liver tissue from the same healthy controls (n = 9). (d) Summary of CXCR6 frequencies of NK cells from paired PBMC and liver biopsies from patients with CHB (n = 30). (e) Summary of CXCR6 expression on NK cells from healthy liver (n = 15), compared to liver tissue from patients with CHB (n = 30). CHB patients were additionally divided by ALT (low ALT < 35 IU/l, n = 15; high ALT > 35 IU/l, n = 15). Diagrams show individual data for healthy controls and CHB. P-values indicated in each diagram. Mann-Whitney U-test (b), Kruskal-Wallis with Dunn’s multiple comparison test (e) and paired Wilcoxon test (c,d) were applied. ****p-value < 0.0001.

Figure 2. Intrahepatic CXCR6+ NK cells have an immature phenotype and express high levels of CD69.

(a) Representative contour plots for identification of CD56^bright and CD56^dim NK cells based on CD16/CD56 expression within intrahepatic CXCR6− and CXCR6+NK cells. Summary of frequency of CD56^bright NK cells in intrahepatic CXCR6− and CXCR6+ NK cells of liver tissue of healthy controls (n = 15) and patients with CHB (n = 30). (b) Representative histogram overlay of CD57 expression on intrahepatic CXCR6− and CXCR6+ NK cells. Summary of frequency of CD57 expressing intrahepatic CXCR6− and CXCR6+NK cells of healthy controls (perfusion liquid n = 6; liver tissue n = 1) and patients with CHB (n = 18). (c) Representative histogram overlaying CD69 expression on intrahepatic CXCR6− and CXCR6+ NK cells. Summary of frequency of CD69 expressing intrahepatic CXCR6− and CXCR6+ NK cells of healthy controls (fresh perfusion liquid n = 2; frozen perfusion liquid n = 6; liver tissue n = 4) and of patients with CHB (n = 7). (d) Representative histogram overlaying HLA-DR expression on intrahepatic CXCR6− and CXCR6+ NK cells. Summary of frequency of HLA-DR expressing intrahepatic CXCR6− and CXCR6+ NK cells of healthy controls (fresh perfusion liquid n = 2; frozen perfusion liquid n = 6; liver tissue n = 4) and patients with CHB (n = 7). Paired Wilcoxon test was applied for all figures. *p-value < 0.05, ***p-value < 0.001, ****p-value < 0.0001.

Figure 3. Functional capacity of CXCR6− and CXCR6+ intrahepatic NK cells.

(a) Summary of percentage of NK cells isolated from perfusates expressing CXCR6 with or without stimulation with IL-12/IL-18 (5 ng/ml, 50 ng/ml) or PMA and Ionomycin at low (3 ng/ml, 100 ng/ml) or high (25 ng/ml, 1 μg/ml) dose (n = 7; 7; 7; 5). (b) Representative histograms overlaying IFNγ production of unstimulated, and high dose PMA/Ionomycin stimulated, for intrahepatic CXCR6− and CXCR6+ NK cells. Summary of IFNγ production by CXCR6− and CXCR6+ NK cells from healthy liver perfusates with stimuli as indicated (n = 7; 7; 5), after subtraction of the unstimulated control. (c) Representative histograms overlaying TNF, MIP-1β and GM-CSF production of unstimulated, and high dose PMA/Ionomycin stimulated, intrahepatic CXCR6− and CXCR6+ NK cells. Summary of TNF, MIP-1β and GM-CSF production by CXCR6− and CXCR6+ NK cells from healthy liver perfusates with stimuli as indicated (n = 7; 5), after subtraction of the unstimulated control. (d) Representative histograms overlaying CD107a expression of unstimulated, and high dose PMA/Ionomycin stimulated, intrahepatic CXCR6− and CXCR6+NK cells. Summary of CD107a expression by CXCR6− and CXCR6+NK cells from healthy liver perfusates with stimuli as indicated (n = 7; 5), after subtraction of the unstimulated control. (e) Representative histogram overlaying ex vivo granzyme B expression in intrahepatic CXCR6− and CXCR6+ NK cells. Summary of frequency of granzyme B expressing intrahepatic CXCR6− and CXCR6+ NK cells from healthy controls (perfusion liquid n = 9; liver tissue n = 1). (f) Representative histogram overlaying ex vivo perforin expression in intrahepatic CXCR6− and CXCR6+ NK cells. Summary of frequency of perforin expressing intrahepatic CXCR6− and CXCR6+ NK cells from healthy controls (perfusion liquid n = 7). (g) Representative histogram overlaying TRAIL expression on intrahepatic CXCR6− and CXCR6+ NK cells from a patient with CHB. Summary of frequency of TRAIL expressing intrahepatic CXCR6− and CXCR6+ NK cells from patients with CHB (n = 10). Kruskal-Wallis with Dunn’s multiple comparison test (A) and Paired Wilcoxon test (B–G) was applied. *p-value < 0.05, ***p-value < 0.001.

Figure 4. The liver contains a distinct T-bet^loEomes^hi NK cell subset identified by CXCR6, but not CD49a expression.

(a) Representative contour plots for identification of T-bet^hiEomes^lo and T-bet^loEomes^hi subsets of total NK cells from periphery and liver tissue. (b) Summary of frequency of T-bet^loEomes^hi cells of total NK cells in PBMC (n = 20) and liver tissue of healthy controls (n = 10), and paired PBMC and liver biopsies from patients with CHB (n = 19). (c) Representative contour plots for identification of T-bet^loEomes^hi and T-bet^hiEomes^lo subsets in intrahepatic CXCR6− and CXCR6+ NK cells. (d) Summary of frequency of T-bet^loEomes^hi cells among intrahepatic CXCR6− and CXCR6+ NK cells of liver tissue of healthy controls (n = 10) and patients with CHB (n = 19). (e,f) Correlation of frequency of T-bet^loEomes^hi and CXCR6 expression of NK cells from liver tissue of healthy controls ((e) n = 10) and liver biopsies from patients with CHB ((f) n = 19). (g) Representative contour plots for identification of T-bet^loEomes^hi and T-bet^hiEomes^lo subsets among intrahepatic CD49a− and CD49a+ NK cells. (h) Summary of frequency of T-bet^lo Eomes^hi cells among intrahepatic CD49a− and CD49a+ NK cells of healthy liver tissue (n = 8). (i) Summary of frequency of CD49a expression on intrahepatic NK cells from paired perfusion liquid and tissue from pre-implant donor livers (n = 7). (j) Representative contour plots for identification of T-bet^loEomes^hi and T-bet^hiEomes^lo subsets among peripheral CXCR6− and CXCR6+ NK cells. (k) Summary of frequency of T-bet^loEomes^hi cells among peripheral CXCR6− and CXCR6+ NK cells (n = 20) and intrahepatic CXCR6+NK cells of liver tissue of healthy controls (n = 10); and of peripheral CXCR6− and CXCR6+ NK cells and paired intrahepatic CXCR6+ NK cells of patients with CHB (n = 19). P-values and r indicated in each diagram. Mann-Whitney U-test ((b) left), Kruskal-Wallis with Dunn’s multiple comparison test ((k) left), Friedmans with Dunn’s multiple comparision test ((k) right), paired Wilcoxon test ((b) right, (d,h,i)) and Spearman’s correlation (e,f) were applied. *p-value < 0.05, **p-value < 0.01, ****p-value < 0.0001.