Constitutive Activation of Natural Killer Cells in Primary Biliary Cholangitis

Natural killer (NK) cells are innate immune cells that interface with the adaptive immune system to generate a pro-inflammatory immune environment. Primary Biliary Cholangitis (PBC) is a hepatic autoimmune disorder with extrahepatic associations including systemic sclerosis, Sjogren's syndrome and thyroiditis. Immunogenetic studies have identified polymorphisms of the IL-12/STAT4 pathway as being associated with PBC. As this pathway is important for NK cell function we investigated NK cells in PBC. Circulating NK cells from individuals with PBC were constitutively activated, with higher levels of CD49a and the liver-homing marker, CXCR6, compared to participants with non-autoimmune chronic liver disease and healthy controls. Stimulation with minimal amounts of IL-12 (0.005 ng/ml) led to significant upregulation of CXCR6 (p < 0.005), and enhanced IFNγ production (p < 0.02) on NK cells from PBC patients compared to individuals with non-autoimmune chronic liver disease, indicating dysregulation of the IL-12/STAT4 axis. In RNAseq studies, resting NK cells from PBC patients had a constitutively activated transcriptional profile and upregulation of genes associated with IL-12/STAT4 signaling and metabolic reprogramming. Consistent with these findings, resting NK cells from PBC patients expressed higher levels of pSTAT4 compared to control groups (p < 0.001 vs. healthy controls and p < 0.05 vs. liver disease controls). In conclusion NK cells in PBC are sensitive to minute quantities of IL-12 and have a “primed” phenotype. We therefore propose that peripheral priming of NK cells to express tissue-homing markers may contribute to the pathophysiology of PBC.


INTRODUCTION
Primary Biliary Cholangitis (PBC) affects 1 in 1,000 women over 40 and is common indication for liver transplantation (1). PBC predominantly affects the liver, but is associated with a number of other conditions including Sjogren's syndrome, scleroderma, thyroid disease, and autoimmune diabetes (2). It can thus be considered a syswtemic disease. Genome wide association studies (GWAS) have consistently shown an association of polymorphisms within the Interleukin (IL)-12/Signal transducer and activator of transcription (STAT) 4 pathway with the development of PBC (3)(4)(5)(6)(7)(8). This pathway has been implicated in other autoimmune and inflammatory diseases including rheumatoid arthritis, chronic active hepatitis, psoriatic arthritis and Crohn's disease (9)(10)(11).
PBC is characterized by an elevation of Th1 cytokines, including interferon gamma (IFNγ), and also Th17 cells as mediators of inflammation (12,13). Natural killer (NK) cells produce IFNγ and are being increasingly recognized for their role in adaptive immunity and having trained immunological memory (14,15). Furthermore, they have been demonstrated to modulate tumor microenvironments toward inducing cytotoxic T-cell responses (16). Thus, NK cells have the potential to play an important role in autoimmune disease. NK cells have been previously implicated in the pathogenesis of PBC (17,18). In PBC NK cells demonstrate enhanced cytotoxic functions (19), are found in increased numbers around the small bile ducts and can be induced to lyse biliary epithelial cell (BECs) following toll-like receptor stimulation (20). Consistent with a role in interfacing with the adaptive immune system, NK cells may potentiate CD4 T-cell responses in PBC (21).
Recent work has shown that NK cells may also have "adaptive" features, with one characteristic being enhanced IFNγ production (22,23). Additionally CXCR6 marks liver-resident NK cells (24) and CD49a is present on tissue-resident NK cells in humans, including some liver-resident NK cells (23,(25)(26)(27). Both are associated with adaptive NK cells in murine studies (28,29). Consistent with this observation, human liver-resident CD49a+ NK cells have a strong Th1 cytokine response and display adaptive features (23,25). We recently showed that IL-12 can upregulate both CXCR6 and CD49a expression on NK cells in the peripheral blood, with comparable phenotypic features to liver-resident CD49a+ NK cells, linking NK cell peripheral activation with liver-homing (25). Based on these observations we aimed to investigate NK cell dysfunction in PBC.

Patients
Patients were recruited from University Hospital Southampton NHS Foundation Trust with informed consent ( Table 1) and full ethical approval (Research Ethics Committee numbers 13/WA/0329 and 06/Q1701/120).

RNA Sequencing
CD49a+ and CD49a-peripheral CD3-CD56+ NK cells from PBC patients, and CD3-CD56+ NK cells from HC were sorted using a BD FACS Aria directly into TRIzol (ThermoFisher, MA). RNA was isolated using miRNeasy micro kit (Qiagen, Hilden, Germany) loaded on an automated platform (Qiacube, Qiagen). Samples were quantified as described previously (30,31) and quality of RNA assessed by Fragment Analyzer (Advance Analytical). All samples had an RNA integrity number > 7.5. Purified total RNA (≈5 ng) was amplified following the Smart-seq2 protocol (32,33). Briefly, mRNA was captured using poly-dT oligos and reverse-transcribed into full-length cDNA using the described template-switching oligo (32,33). cDNA was amplified by PCR, purified using AMPure XP magnetic beads (Beckman Coulter). One nanogram of cDNA was used to prepare a standard NextEra XT sequencing library (NextEra XT DNA library prep kit and index kits; Illumina). Barcoded Illumina sequencing libraries (Nextera; Illumina) were generated utilizing an automated platform (Biomek FXP, Beckman Coulter). Both whole-transcriptome amplification and sequencing library preparations were performed in a 96-well format to reduce assay-to-assay variability. Quality control steps were included to determine total RNA quality and quantity, the optimal number of PCR preamplification cycles, and fragment library size. Samples were pooled at equimolar concentration, loaded and sequenced on the Illumina Sequencing platform, HiSeq2500 (Illumina) to obtain more than 7 million 50bp single-end reads (HiSeq Rapid Run Cluster and SBS Kit V2; Illumina) mapping uniquely to mRNA reference. Reads were mapped to ENSEMBL (34) release 95 using kallisto (35) with bias correction, and 50 bootstrap samples. Differentially expressed genes (DEG) were found using EdgeR (36) aggregating transcripts to gene level. All models included a term to model individual variation. Main differences of PBC vs. HC were detected using a model with group effect. CD49a+ vs. CD49a-NK cells were compared using a paired design. Genes with a false discovery rate (FDR)-corrected P-value < 0.05 were identified as differentially expressed, resulting from a likelihood ratio test using a negative binomial generalized linear model fit. Normalization offsets were calculated using the TMM. Pre-ranked gene set enrichment analysis was performed using Fgsea R package (37) with 10,000 permutations and the Molecular Signatures Database v 6.2(MsigDB) (38). Heat maps were generated using gplots v3.01 (39) clustering the samples by spearman correlation. Pathways were created using ReactomeDB (40) and visualized using Cytoscape v3.7 using the fold changes obtained from EdgeR. The project number at Gene Expression Omnibus is PRJNA542532, https://www.ncbi. nlm.nih.gov/bioproject/542532. For IFNγ secretion, PBMCs were stimulated for 12 h with IL-12, GolgiStop TM (BD Biosciences) was added (4 µl/6 ml culture medium) for the last 4 h, prior to cell surface staining as described above. Cells were then fixed and permeabilized (BD Cytofix/Cytoperm TM Plus Kit, BD Biosciences) prior to incubation with IFNγ (B27, APC, Biolegend R ). pSTAT4 activation was assessed by flow cytometry of PBMCs at rest and following stimulation with IL-12 for 1 h. Cells were surface stained as above, then incubated at 37 • C for 10-12 min in Cytofix solution (BD Biosciences), prior to resuspension in 1 ml Perm III Buffer (BD Biosciences) and staining with pSTAT4 (PY693, AF647, BD Biosciences). For the CD107a degranulation assay, PBMCs were incubated overnight with 1 ng/ml rhIL-15 (R and D Systems) and then added to a 96-well round bottomed plate with either no target or 721.221 cells at a 5:1 or 10:1 effector:target ratio. Cells were coincubated for 4 h at 37 • C in the presence of anti-CD107a-AF647 (eBioscience Ltd, Hatfield, UK). GolgiStop (BD Biosciences) was added after the first hour of incubation and cells were then stained and analyzed by flow cytometry.

Statistical Analysis
Statistical analysis was performed using Graph Pad Prism 7. Surface marker expression and intracellular staining for IFNγ and pSTAT4 are presented using the median and interquartile range. The Wilcoxon matched pairs signed rank test was used to compare data within the same patient group and the Mann Whitney U-test was used to compare data from different patient groups. The Kruskal-Wallis' test with Dunn's multiple comparison test was used to compare three unpaired sample groups. Regression analysis and the coefficient of determination (r 2 ) was used to determine correlation between expression levels of CXCR6 and CD49a with age. Chi-squared test was used to compare categorical demographic data.

CXCR6+ and CD49a+ NK Cells Are Found at Higher Frequencies in the Peripheral Blood in Patients With PBC Compared to Controls
Thirty-six individuals with PBC were enrolled ( Table 1). Peripheral blood was also collected from 31 patients with HFe as non-autoimmune liver disease controls, and nine healthy volunteers. Individuals with PBC were more likely to be female and slightly older than the HFe population, however the liver disease groups were matched for disease stage. Median serum liver enzyme levels were as follows for individuals with PBC: alkaline phosphatase (ALP) 175 IU/l (range 64-1487), alanine transaminase (ALT) 27.5 IU/l (range 12-226) and Bilirubin 9 µmol/l (range 3-65).
The frequencies of ex-vivo unstimulated NK cells in the peripheral blood were not significantly different between participant groups: PBC 10.8%, HFe 11.4%, and HC 11.5% (Supplementary Figure 1A). Frequencies of CD56 bright NK cells were also comparable, with a non-significant trend toward a higher frequency of CD56 bright NK cells in the PBC group (PBC 8.9% vs. HFe 6.4% and HC 5.3%; Supplementary Figure 1B).
The frequency of NK cells expressing CXCR6 was significantly higher in PBC patients compared to HFe (3.4 vs. 2.4%, p < 0.05) and HC (3.4 vs. 2.0%, p < 0.01; Figures 1A,B). There was also increased expression of CD49a on NK cells from PBC patients compared to HFe (2.2 vs. 1.3%, p < 0.01) and HC (2.2 vs. 0.9%, p < 0.01; Figures 1A,B). These increases were not associated with Ursodeoxycholic acid (UDCA) therapy ( Supplementary Figures 2A,B) and were specific to NK cells, as there was no increase in the frequencies CD3+CD56-T-cells expressing CXCR6 or CD49a in PBC patients ( Figure 1C). Examination of all participants revealed there was no association between either gender or age with frequencies of CXCR6+ and CD49a+ NK cells, suggesting that the higher frequency of tissue-resident NK cells observed in the blood in PBC patients, compared to the chronic liver disease    Figure 1D). Thus, overall there is increased expression of markers associated with tissue-residency on resting peripheral blood NK cells in PBC. Liver-resident CXCR6+ and CD49a+ NK cells are known to be phenotypically and functionally distinct (25). To investigate the relationship between CXCR6+ and CD49a+ NK cells in the peripheral blood we examined the expression of CXCR6 on circulating CD49a+ and CD49a-NK cells ex vivo. We found that the CD49a+ NK cell population contained higher frequencies of CXCR6+ cells than the CD49a-population, for individuals with HFe, PBC, and HC ( Figure 1E). These frequencies were higher within the CD49a+ NK cell subset for patients with PBC vs. HFe (12.0 vs. 7.5%, p < 0.02) and vs. HC (12.0 vs. 5.6%, p < 0.05) (Figure 1E).

Circulating CD49a+ NK Cells in PBC Are Constitutively Activated
As PBC is an inflammatory autoimmune disorder and our previous work had shown that intrahepatic NK cells expressing CD49a are associated with higher levels of IFNγ than CXCR6 (25), we further characterized the CD49a+ NK cell population by performing RNA sequencing on paired sorted peripheral blood CD49a+ and CD49a-NK cells from three patients with PBC, and peripheral blood derived NK cells from three HC. Of the three individuals with PBC all were female, aged 48, 56, and 56 and one had cirrhosis. Principal Component Analysis revealed that both subsets of NK cells from PBC patients had a distinct transcriptional profile compared to NK cells from HC, and that CD49a+ NK cells from PBC patients had a divergent gene expression profile from CD49a-NK cells (Figure 2A).
In comparison to HC, CD49a+ NK cells from PBC patients represent an activated phenotype characterized by enrichment of the inflammatory response (normalized enrichment score . c-Myc is essential for IL-2/IL-12 induced metabolic reprogramming in NK cells and is associated with upregulation of SLC7A5 (41). We therefore analyzed cytokine signaling pathways. A total of 12 genes were upregulated (p < 0.05) in CD49a+ NK cells from PBC patients compared to NK cells from HC including key genes associated with cytokine signaling (IRF8, STAT1, SOCS1, IL7R, IL12A, and NFKB1). Consistent with the STAT4 pathway being dysregulated in PBC, components of the IL-12/STAT4 pathway were upregulated in CD49a+ and CD49a-NK cells compared to NK cells from HC. STAT4, JAK2, CD3D, MAPK8, CXCR3, and IL12A were significantly enriched in PBC derived CD49a+ NK cells compared to HC ( Figure 2E). Furthermore, the transcription factor ZBTB32, which is associated with adaptive NK cell responses following murine cytomegalovirus infection, was also upregulated in CD49a+ compared to CD49a-NK cells from PBC patients (p < 0.001, FDR < 0.01, Supplementary Table 1) (42,43).

Low Dose IL-12 Preferentially Stimulates NK Cells From Individuals With PBC
To investigate whether the altered phenotype of NK cells in PBC may be related to changes in the cytokine signaling pathway, we studied the expression of CXCR6 and CD49a following cytokine activation in short term culture. We cultured purified NK cells from non-cirrhotic individuals with PBC and HFe in increasing concentrations of IL-12 (0.005, 0.5, 5.0, and 10 ng/ml) and IL-15 (1 and 25 ng/ml) for 12 h. Stimulation with 0.005 ng/ml IL-12 led to significant upregulation of CXCR6 on NK cells from individuals with PBC, from 2.9 to 4.9% (p < 0.001). No upregulation was observed in the HFe group (2.2 to 2.7%, p > 0.05). At this concentration of IL-12 NK cells from PBC patients showed significantly greater CXCR6 expression than NK cells from HFe patients (p < 0.005) (Figure 3A). IL-15 did not significantly upregulate CXCR6 in either patient group (Figure 3A). Conversely, we observed similar responses in CD49a expression to both IL-12 and IL-15 in the PBC and HFe groups ( Figure 3B). However, stimulation with IL-12 0.005 ng/ml led to an increase in CD49a+CXCR6+ doublepositive NK cells in individuals with PBC compared to those with HFe (0.24 vs. 0.17%, p < 0.05), and compared to culture of PBC NK cells in media alone (0.24 vs. 0.17%, p < 0.005) (Figure 3C). Responses to IL-15 were similar between PBC and HFe NK cells (Figures 3B,C). Thus, in PBC peripheral NK cells are sensitive to IL-12 and this preferentially upregulates the liver homing marker CXCR6.
To determine if the IL-12 sensitivity of NK cells in PBC is associated with a functional consequence we measured IFNγ secretion. NK cells from individuals with PBC, but not from HFe patients or HC, responded to 0.005 ng/ml IL-12 with an increase in IFNγ expression from 1.1% at rest to 2.9% (p < 0.001) and produced greater quantities of IFNγ compared to HFe patients (2.9 vs. 1.2%, p < 0.02) and HC (2.9 vs. 0.4%, p < 0.001; Figures 4A,B). At higher concentrations of IL-12 responses were similar between the groups, suggesting that this is a hypersensitivity to IL-12 and not an increase in the potential for NK cells to secrete IFNγ (Figure 4B).
Activated and "memory" NK cells have been shown to express CD49a, which is associated with enhanced IFNγ secretion (23,25). We therefore investigated which subsets of NK cells were associated with IFNγ release in response to low dose IL-12. CD49a+ NK cells from individuals with PBC were more readily stimulated to release IFNγ than CD49a-NK cells (median 8.6 vs. 2.5% p < 0.001) consistent with their activated phenotype by RNAseq. Whilst CD49a+ NK cells from HFe and HC groups also secreted IFNγ in response to low dose IL-12, this was significantly less than from PBC patients [PBC 8.6% (range 0-43.0%) vs. HFe 2.6% (range 0-8.2%), p < 0.02 and HC 1.4% (range 0-5.1%), p < 0.005] (Figure 4C). IFNγ production by the CD49a-NK cell populations was comparable between the two liver disease groups (p > 0.05). No significant difference in IFNγ secretion was noted between CXCR6+ and CXCR6-NK cells.
On stimulation with low doses of IL-12, the frequencies of pSTAT4+ NK cells did not increase further in either patient group until concentrations reached 5 ng/ml (Figures 5A,C). Differences between the groups were lost at this point. This indicates that the increased levels of IFNγ production observed following IL-12 0.005 ng/ml are most likely related to increased baseline levels in STAT4 activation rather than de novo STAT4 phosphorylation during culture, and that NK cells in PBC are constitutively activated.

DISCUSSION
We have identified that NK cells in PBC are activated, as determined by both phenotypic and functional analysis. Furthermore, they can be readily induced to express the liver homing marker CXCR6. The observation that CD49a can also be upregulated with low dose IL-12 in individuals with other chronic liver diseases, could be related to differences in transcriptional regulation between CXCR6 and CD49a, in that CD49a expression on NK cells appears more sensitive to short term culture with IL-12 than CXCR6. Consistent with this, previous work has shown that exogenous IL-12 has a greater effect on CD49a expression than CXCR6 expression, on NK cells in long term culture (25). However, there does appear to be a dysregulation of the IL-12 signaling pathway in PBC. We found that STAT4 was constitutively phosphorylated to greater levels in PBC than in the liver disease controls, and of functional relevance NK cells in PBC could be induced to secrete more IFNγ. Immunogenetic studies have correlated polymorphisms in the IL-12/STAT4 signaling pathway with the development of PBC (3)(4)(5)(6)(7)(8). In our RNAseq analysis, genes involved in this pathway were upregulated in NK cells from PBC patients, consistent with an underlying defect of NK cells in this autoimmune disease. Interestingly STAT4 polymorphisms have been associated with other major autoimmune diseases (9)(10)(11). Thus, this could represent a common mechanism for the potentiation of autoimmune disorders.
Liaskou et al. have previously shown that T reg cells from patients with PBC were highly sensitive to IL-12, but in that study IL-12 concentrations of 10 ng/ml were used (44). This is 2,000 times the concentration we used to demonstrate upregulation of CXCR6 and CD49a on NK cells from PBC patients. As the mean level of IL-12p70 found in the serum of PBC patients is around 10 pg/ml, and levels in the liver around 60 pg/ml, then the concentrations of IL-12 we tested are well within the physiological range (45). Thus in PBC, NK cells are several orders of magnitude more sensitive to IL-12 levels than T reg cells. We therefore propose that the single nucleotide polymorphisms in the IL-12/STAT4 pathway are more likely to have their effects on NK cells.
The transcriptomic analysis presented here demonstrates that CD49a+ NK cells in PBC have a "primed" phenotype. Activation of NK cells is associated with metabolic reprogramming and we found upregulation of the amino acid transporter SLC7A5 in the CD49a+ NK cell subset in PBC. This transporter is thought to be important for c-Myc mediated metabolic reprogramming of NK cells. Furthermore, there was downregulation of genes associated with oxidative phosphorylation. Thus, in PBC NK cells have features of metabolic reprogramming. Intriguingly, both metabolic reprogramming and expression of CD49a are associated with adaptive NK cells (23).
Additionally we observed upregulation of the transcription factor ZBTB32 in CD49a+ NK cells in PBC. This has previously been described in association with cytomegalovirus seropositivity, and is thought to be essential for the proliferative burst of NK cells during viral infections (46,47). This raises the possibility that infectious agents may prime NK cells in PBC, by stimulating secretion of IL-12, upregulating CXCR6 on NK cells, and so signaling NK cells to traffic to the liver. Such IL-12 release could be triggered by bacterial or viral infections, which could occur anywhere in the body, including urinary tract infections which have previously been associated with PBC (48). PBC has multiple extrahepatic associations and so the acquisition of a tissue-residency phenotype could be associated with recruitment of NK cells to tissues other than the liver.
Our work demonstrates that NK cells in PBC circulate with an activated phenotype and readily express markers of liver-homing and tissue-residency, suggesting that they could migrate to the liver and contribute toward a pro-inflammatory hepatic environment. Furthermore, circulating CD49a+ NK cells isolated from patients with PBC demonstrate enhanced degranulation and thus may be capable of lysing BECs. Recent work in both humans and mice has demonstrated that NK cells can be recruited from the circulation into the liver, and in humans this has been related to recruitment of CXCR6 + Eomes lo NK cells into the liver to subsequently become CXCR6 + Eomes hi (24,49). Consistent with this, CXC Chemokine Ligand 16 (CXCL16), the chemokine that attracts CXCR6+ lymphocytes is upregulated in PBC (50). However, further studies of intrahepatic NK cells are required to determine the extent to which NK cells contribute toward hepatic inflammation. These studies may reveal if NK cells are involved in initiating inflammation, and augmenting T-cell mediated inflammation in PBC, in addition to contributing toward the lysis of BECs (51).
In conclusion our data demonstrate that NK cells in PBC circulate with a primed phenotype, suggestive of a role in the underlying pathogenesis of the disease. Investigation of NK cells in other autoimmune disorders is warranted as they represent a novel target for therapeutic intervention.

DATA AVAILABILITY STATEMENT
The datasets generated for this study can be found in the Gene Expression Omnibus, PRJNA542532, https://www.ncbi.nlm.nih. gov/bioproject/542532.

ETHICS STATEMENT
The studies involving human participants were reviewed and approved by (1)

AUTHOR CONTRIBUTIONS
TH was responsible for patient recruitment, experimental design, acquisition of data, data analysis, and interpretation and drafting of the manuscript. MB was responsible for data acquisition, analysis and interpretation. JN was responsible for data acquisition and analysis. AV was responsible for data analysis and interpretation. GS and AW were responsible for data acquisition and analysis. PV was responsible for study and experimental design with respect to RNA sequencing and data interpretation. MP was responsible for data interpretation. SK supervised the project, was responsible for the study concept, experimental design, data interpretation and critical revision of the manuscript.