Human liver infiltrating γδ T cells are composed of clonally expanded circulating and tissue-resident populations

Graphical abstract

Background & Aims: cd T cells comprise a substantial proportion of tissue-associated lymphocytes. However, our current understanding of human cd T cells is primarily based on peripheral blood subsets, while the immunobiology of tissueassociated subsets remains largely unclear. Therefore, we aimed to elucidate the T cell receptor (TCR) diversity, immunophenotype and function of cd T cells in the human liver.
Methods: We characterised the TCR repertoire, immunophenotype and function of human liver infiltrating cd T cells, by TCR sequencing analysis, flow cytometry, in situ hybridisation and immunohistochemistry. We focussed on the predominant tissue-associated Vd2 À cd subset, which is implicated in liver immunopathology.
Results: Intrahepatic Vd2 À cd T cells were highly clonally focussed, with single expanded clonotypes featuring complex, private TCR rearrangements frequently dominating the compartment. Such T cells were predominantly CD27 lo/À effector lymphocytes, whereas naïve CD27 hi , TCR-diverse populations present in matched blood were generally absent in the liver.
Furthermore, while a CD45RA hi Vd2 À cd effector subset present in both liver and peripheral blood contained overlapping TCR clonotypes, the liver Vd2 À cd T cell pool also included a phenotypically distinct CD45RA lo effector compartment that was enriched for expression of the tissue tropism marker CD69, the hepatic homing chemokine receptors CXCR3 and CXCR6, and liver-restricted TCR clonotypes, suggestive of intrahepatic tissue residency. Liver infiltrating Vd2 À cd cells were capable of polyfunctional cytokine secretion, and unlike peripheral blood subsets, were responsive to both TCR and innate stimuli.
Conclusion: These findings suggest that the ability of Vd2 À cd T cells to undergo clonotypic expansion and differentiation is crucial in permitting access to solid tissues, such as the liver, which results in functionally distinct peripheral and liver-resident memory cd T cell subsets. They also highlight the inherent functional plasticity within the Vd2 À cd T cell compartment and provide information that could be used for the design of cellular therapies that suppress liver inflammation or combat liver cancer.
Lay summary: cd T cells are frequently enriched in many solid tissues, however the immunobiology of such tissue-associated subsets in humans has remained unclear. We show that intrahepatic cd T cells are enriched for clonally expanded effector T cells, whereas naïve cd T cells are largely excluded. Moreover, whereas a distinct proportion of circulating T cell clonotypes was present in both the liver tissue and peripheral blood, a functionally and clonotypically distinct population of liver-resident cd T cells was also evident. Our findings suggest that factors triggering cd T cell clonal selection and differentiation, such as infection, can drive enrichment of cd T cells into liver tissue, allowing Introduction cd T cells are unconventional lymphocytes enriched in solid tissues, where they are thought to play critical roles in immunosurveillance. 1 Studies of mouse tissue-associated cd subsets suggest cd T cell function can be predominantly innate-like, involving semi-invariant T cell subsets that enable fast response kinetics without a requirement for clonal selection and differentiation. [2][3][4][5] This role may allow for rapid 'lymphoid stress surveillance', limiting damage to host tissues in the face of microbial or non-microbial challenges, prior to full activation In contrast, the paradigms underlying human cd T cell immunobiology are far from clear. In humans, the peripheral blood is dominated by the Vd2 + /Vc9 + T cell subset, polyclonally activated by bacterial 9 and endogenous phospho-antigens, 10 arguably conforming to an innate-like paradigm. 11 In contrast, human solid tissues are enriched for Vd2 À cd T cells, of which the Vd1 + subset is the most prevalent. It is far less clear if this dominant human tissue-associated subset also adopts an innate-like biology. Indeed, Vd2 À T cells have been linked to recognition of a diverse range of ligands including to date Endothelial Protein C Receptor, 12 CD1 molecules, 13 Annexin-A2, 14 and even phycoerythrin. 15 Moreover, recent data have provided strong evidence that Vd1 + cells display an unconventional adaptive biology, undergoing clonal selection and differentiation from a naïve T cell receptor (TCR)-diverse precursor pool, 16 with viral infection one trigger driving expansion. 17 However, such studies have focussed on the subset of Vd2 À cd T cells that are retained in peripheral blood. To date, the immunobiology of human tissue-associated cd T cells remains relatively unstudied, despite the Vd2 À T cell subset representing a considerable proportion of the total T cell infiltration in many human solid tissues, including gut, 2 lung 18 and liver. 19 To shed light on the function of tissue-associated cd T cells and how this relates to peripheral subsets, we characterised human intrahepatic Vd2 À T cells. The liver is a site of considerable blood flow, receiving 75% of the total blood in the body every 2 h, with a third of this originating directly from the antigen-rich gut via the portal vein. In addition to providing a generally immunosuppressive microenvironment to facilitate tolerization of T cells toward non-pathogenic antigens present in the portal blood flow, the liver is also home to a large population of innate lymphoid cells, including natural killer (NK) cells, invariant natural killer T (iNKT) cells, mucosal associated invariant T (MAIT) cells 20 and cd T cells, 19 in addition to CD8 + cytotoxic T cells. 21 This enrichment is believed to balance the need for tolerization with a requirement for rapid identification and elimination of potentially harmful pathogenic entities, for example via pathogen associated molecular pattern receptors and semi-invariant T cell populations. 22 To shed light on the immunobiology of cd T cells in this context we exploited next generation sequencing (NGS) approaches, allowing us to probe the TCR repertoire, in parallel with immunophenotype, and function.
Our study is the first to define the interconnected clonotypic, phenotypic and functional features of human tissue-associated cd T cells. The findings suggest that the liver selectively retains Vd2 À T cells that are clonally expanded and adopt an effector phenotype, and which include a subset containing liverrestricted clonotypes that is phenotypically and functionally distinct from those present in peripheral blood.

Material and methods
Ethical approval and samples Explanted diseased liver tissue and matched blood were obtained from patients who underwent liver transplantation for end-stage liver diseases including primary sclerosing cholangitis (PSC), primary biliary cholangitis (PBC), alcoholic liver disease (ALD), non-alcoholic steatohepatitis (NASH), hepatitis C virus (HCV) and hepatitis B virus (HBV) (Local Research Ethics Committee reference No. 98/CA5192) or normal liver samples from donor liver tissue surplus to clinical requirements (Local Research Ethics Committee reference No. 06/Q2708/11). Unless otherwise stated (see Fig. 1), all diseased liver tissue analysed was from HCV/HBV-negative donors, and were noncancerous. Normal liver tissue donors had no known prior history of liver disease or HCV/HBV infection. All diseased livers were Child C decompensated. Adult peripheral blood was obtained from consenting healthy donors (protocol approved by the NRES Committee West Midlands ethical board; REC reference 14/WM/1254).
T cell isolation, culture and activation Human liver infiltrating lymphocytes were isolated from fresh liver tissue as described previously. 20 A whole slice of liver was processed, thereby reducing any effects of heterogeneous disease localisation. Briefly, explanted liver tissue was diced into 5 mm 3 cubes, washed with Phosphate Buffered Saline (PBS), and then homogenised in a Seward stomacher 400 circulator (260 rpm, 5 min). The homogenate was filtered through fine (63 lm) mesh (John Staniar and Co, Manchester, UK) and the lymphocytes were isolated by density gradient separation using Lympholyte (VH Bio, Gateshead, UK) at 800Âg for 20 min. The lymphocyte layer was collected and washed with PBS. Cell viability was assessed by trypan blue exclusion. Peripheral blood mononuclear cells (PBMCs) were isolated from heparinised venous blood by lymphoprepÓ (Stem Cell Technologies) density gradient centrifugation as per the manufacturer's instructions. The cell culture medium used throughout this study was RPMI-1640 medium (Invitrogen) supplemented with 2 mM L-glutamine, 1% sodium pyruvate, 50 lg/ml penicillin/streptomycin (Invitrogen) and 10% foetal calf serum (Sigma).
In summary, sections were de-paraffinized, endogenous peroxidase activity was quenched using 0.3% hydrogen peroxide (Sigma Aldrich) in methanol for 20 min, and antigen retrieval carried out, involving boiling sections in 1% EDTA solution for 15 min. After washing and blocking steps, sections were incubated for 1 h in primary antibody (goat polyclonal -anti- antibody (Vector Labs Laboratories) for 30 min at room temperature. Following washing, sections were developed using ImmPACTTM DAB reagent (Vector Laboratories). Excess DAB was then removed by rinsing and sections were counterstained with Mayer's haematoxylin solution (Leica Biosystems). Once dry, slides were mounted using DPX (Cellpath, Newtown Powys, UK) and imaged on a Zeiss Axioskop 40 Microscope. Regions of parenchymal and portal tract tissue were identified and numbers of CD3+ or cd-TCR+ cells were counted per region identified, with five high power fields, selected at random, scored for each section. For in situ hybridisation, TCR chain-specific localisation of gamma delta TCR+ cells was performed using two protocols, either the ViewRNA TM ISH Tissue 2-Plex Assay developed by Affymetrix and performed manually, or the RNAscope Ò 2.5 LS Duplex Assay (ACD. For both protocols, liver slices were cut and immediately fixed in formalin for 24-48 h prior to being embedded in paraffin and mounted. Immediately after which the assay slides were baked at 60°C for 1 h to immobilise the sections. TCR repertoire analysis RNA was purified from sorted cells (intrahepatic Vd2 À T cells: 8,000-50,000 cells) protected in RNAlater (Sigma Aldrich) using an RNAmicro plus kit (Qiagen) according to the manufacturer's instructions. For high throughput deep sequencing of cd TCRs, we used amplicon rescued multiplex (ARM)-PCR and a MiSeq (illumina) next generation sequencer to analyse all sorted Vd2 À T cell populations. Following initial first-round RT-PCR using high concentrations of gene-specific primers, universal primers were used for the exponential phase of amplification (Patent: WO2009137255A2), allowing deep, quantitative and non-biased amplification of TCRc and TCRd sequences. All cDNA synthesis, amplification, NGS library preparation and sequencing were performed by iRepertoire, Inc. (Huntsville, USA).
Single-cell TCR sequencing PBMCs were labelled as described above and Vd1 + T cells were single-cell sorted directly into individual wells in a 96 well plate containing 2 ll of Superscript VILO cDNA synthesis kit reaction mix (ThermoFisher) containing 0.1% Triton X-100, and incubated according to manufacturer's instructions. TCRc and TCRd cDNAs were amplified by two rounds of nested PCR using GoTaq mastermix (Promega) and primers for or Vd1, CAAGCCCAGTCATCAG-TATCC (external) and CAACTTCCCAGCAAAGAGATG (internal); for Cd GCAGGATCAAACTCTGTTATCTTC (external) and TCCTTC ACCAGACAAGCGAC (internal); for Vd3, GGCACGCTGTGTGACAAA (external) and CTGCTCTGCACTTACGACACTG (internal); for Vc1-8 CTGGTACCTACACCAGGAGGGGAAGG (external) and TGT GTTGGAATCAGGAVTCAG (internal); for Vc9 AGAGAGACCTGGT GAAGTCATACA (external) and GGTGGATAGGATACCTGAAACG (internal) and for Cc CTGACGATACATCTGTGTTCTTTG (external) and AATCGTGTTGCTCTTCTTTTCTT (internal). PCR products were separated on 1.2% agarose gels, and products of successful reactions were incubated with ExoSAP-IT PCR cleanup enzyme (Affymetrix) before sequencing with BigDye Terminator v3.1 (Applied Biosystems) following manufacturer's instructions and running on an ABI 3730 capillary sequencer (Functional Genomics Facility, University of Birmingham).
TCR repertoire data analysis Sequences data was error corrected and V, D and J gene usage and complementarity-determining region 3 (CDR3) sequences were identified and assigned, and tree maps generated using iRweb tools (iRepertoire, Inc, Huntsville, AL, USA). Tree maps show each unique CDR3 as a coloured rectangle, the size of each rectangle corresponds to each CDR3s abundance within the repertoire and the positioning is determined by the V region usage. For more detailed analysis and error correction of the TCR repertoire, datasets were processed using the MiXCR software package to further correct for PCR and sequencing errors. Diversity metrics, clonotype overlap and gene usage were plotted in R, by VDJTools.

TCR sequence analyses
The CDR3 length was defined as the number of amino acids between the second cysteine of the V region and the phenylalanine of the J region, according to IMGT. N and P nucleotides were identified using the IMGT Junction Analysis tool.

Statistical analysis
Tabulated data were analysed in Graphpad PRISM 7 (Graphpad Software Inc). Each data set was assessed for normality using Shapiro-Wilk normality test. Differences between columns were analysed by two-tailed Student's t tests for normally distributed data and Mann-Whitney for non-parametric data. Differences between groups were analysed using one-way ANOVA with Tukey's post-tests for normally distributed data or with Kruskal-Wallis with Tukey's post-tests for non-parametric data and RM two-way ANOVA with Tukey's post-tests was used when comparing groups with independent variables. ⁄ p <0.05, ⁄⁄ p <0.01, ⁄⁄⁄ p <0.001 and ⁄⁄⁄⁄ p <0.0001.

Data availability
The sequence data that support the findings of this study have been deposited in the NIH NCBI sequence read archive database with the primary accession code SRP113556 and SRP096009, for cd TCR repertoires. For more detailed metadata relating to individual samples please contact the authors.
For further details regarding the materials used, please refer to the CTAT table and Supplementary information.

Results
Human Vd2 À cd T cell populations are reportedly tissue tropic in nature, with enrichment of this compartment previously highlighted in diseased human gut 23 and liver. 19 We used immunohistochemistry (IHC) analysis to assess the infiltration and localisation of liver cd T cells. Firstly, cd T cells were a significantly enriched proportion of infiltrating CD3 + T cells in normal livers compared with livers explanted from patients with chronic liver disease (Fig. 1A). Furthermore, we noted the majority of the infiltrating CD3 + T cells were localised to portal areas; however, analysis of sequentially stained sections from normal tissue revealed a high proportion of parenchymaassociated CD3 + T cells were cd TCR + (Fig. 1B). Importantly, while a significant increase in infiltrating CD3 + T cells was observed in diseased tissue, cd T cell numbers did not significantly change, suggesting that disease drives an increased infiltration of total CD3 + T cells but not cd TCR + cell infiltration from the periphery (Fig. 1B, C, Fig. S1A). Further analysis of sequentially stained sections from explanted livers confirmed that cd TCR + cells were also preferentially associated with the liver parenchyma (Fig. 1D, Fig. S1B). We then examined the TCRd chain expression of liver infiltrating cd T cell populations by flow cytometry, in homogenised single-cell suspensions of liver tissue from human explanted livers (Fig. S1C). Consistent with our IHC data, a significantly higher proportion of the CD3 + T cell compartment was comprised of cd T cells in healthy liver tissue compared with disease tissue (Fig. 1E-F), of which the majority were Vd2 À (Fig. 1G), a direct inversion of the predominance of Vd2 + T cells in the peripheral blood. 24,16 Moreover, the majority of the Vd2 À compartment was made up of Vd1 + cd T cells (Fig. 1H, Fig. S1D), with the remainder comprised of other undefined Vd chains. Disease aetiology had no observed impact on this observation (Fig. S1E). Consistent with pan-cd T cell IHC, infiltration of Vd1 + cd T cells into liver parenchyma was demonstrated using IHC and in situ hybridisation; again, IHC staining of sequential sections suggested a high proportion of parenchymaassociated CD3 + T cells were Vd1 + (Fig. S1F). Of note, Vd1 + cd T cells were significantly enriched as a proportion of intrahepatic T cells in diseased cytomegalovirus (CMV) + liver donors compared with diseased CMV À donors, while Vd2 + T cells were not (Fig. 1I).
We next assessed the TCR repertoire of enriched populations of Vd2 À cd T cells from both healthy and diseased liver tissue by amplicon rescued multiplex (ARM)-PCR and deep sequencing (Fig. S2A). Tree plot and clonotype analysis of Vd2 À TCR repertoires indicated that both healthy and diseased liver tissue was generally dominated by a small number of highly prevalent clonotypes ( Fig. 2A-C), with the 10 most prevalent CDR3 sequences accounting for >40% of TCRc and TCRd sequences in 9 and 8 out of 10 samples, respectively, and one dominant clone representing >50% in 2 of the 10 TCRc and TCRd samples ( Fig. 2B-C). Comparison with D75 values obtained from adult and cord blood Vd1 + TCR repertoires placed liver Vd2 À TCR repertoires in a comparable range with other highly focussed cd TCR repertoires (Fig. 2D). Furthermore, when measuring the number of unique clonotypes detected in the first 10 5 CDR3 sequences obtained in each sample, an alternative measure of TCR diversity, liver samples displayed a significantly less diverse repertoire than blood cd TCR repertoires (Fig. S2B).
Comparison of Chao1 diversity metrics revealed no difference in the diversity of clonotypes between healthy and diseased liver TCR repertoires (Fig. S2C). Consistent with a broadly similar TCR repertoire in healthy and diseased tissue, comparison of normalised CDR3 lengths from healthy and diseased samples yielded no discernible difference (Fig. S2D). Previous studies have highlighted that peripheral blood Vd2 À TCRc repertoires contain few shared sequences. 16,17 We found that liver Vd2 À TCRc repertoires were in general more private than blood Vd2 À TCRc repertoires and had very limited shared sequences between unrelated donors (Fig. S2E). Consistent with flow cytometry analyses ( Fig. 1G-H), Vd chain usage was dominated by Vd1 (73.96% ±SEM 8.7) and Vd3 (24.05% ±SEM 9.3) chain usage, with little Vd4, Vd5 and Vd8 usage observed (Fig. 2E). Despite dominant clonotypes, Vc chain usage was highly heterogeneous, with all coding Vc chains utilised across our samples (Fig. 2F). Moreover, no significant difference was observed in Vd or Vc chain usage between healthy and diseased samples (Fig. 2E-F), consistent with the similar diversity metrics observed in diseased and healthy liver samples. These TCR sequencing data indicate the overwhelming prevalence of Vd1 + TCR sequences in liver tissue, while confirming previous findings demonstrating a relative enrichment of Vd3 + cd T cells in human liver compared to peripheral blood. 19 Next, we assessed individual Vd1 + and Vd3 + TCR repertoires for evidence of clonal expansion, initially using accumulated frequency curves to measure the 10 most prevalent clonotypes across all samples (Fig. S2F). These analyses provided evidence for clonal dominance in both liver Vd1 + and Vd3 + TCR repertoires, similar to clonotypically focussed peripheral blood Vd1 + TCR repertoires but different from unfocussed cord blood Vd1 + TCR repertoires (Fig. S2F).
This distinctive clonal dominance was unequivocally confirmed by sorting single intrahepatic Vd1 + and Vd3 + T cells and performing single-cell TCR sequencing. This approach highlighted that intrahepatic Vd1 + and Vd3 + (Fig. 3) T cell populations were composed of a small number of dominant clonotypes, using a variety of functional Vc and Jc gene segments. We also confirmed that concurrent clonal focussing can occur in both Vd1 + and Vd3 + TCR repertoires in the same donors (Fig. S3A). Moreover, analysis of CDR3d sequences revealed substantial complexity. As in peripheral blood, CDR3d1 were long, frequently using two diversity (D) gene segments and containing extensive non-templated nucleotide (nt) additions (Table. S1). CDR3d3 sequences were generally shorter than CDR3d1 sequences and contained fewer non-templated nt ( Table. S2; Fig. S3B), though there was no evidence of CDR3d3 length restriction, in contrast to CDR3c9 sequences in Vc9 + /Vd2 + T cells. 16 These data highlight the private nature of expanded clonotypes in intrahepatic Vd2 À TCR repertoires and the broad range of Vc chains that they collectively utilise.
We next assessed the relationship between peripheral blood and intrahepatic Vd1 + TCRs in the same individuals. Flow cytometry analysis of these matched samples indicated the enrichment of cd T cells in the liver (Fig. 4A), which occurred alongside the previously noted enrichment of CD8 + ab T cells (Fig. 4B). [25][26][27] Moreover, while Vd1 + T cells were specifically enriched there was an overall reduction in the proportion of infiltrating Vd2 + T cells in the liver compared to the blood (Fig. 4C). Peripheral blood Vd1 + T cells comprise both clonotypically focussed effector and separate TCR-unfocussed naïve subcompartments, which can be delineated based on distinct CD27 lo/À CD45RA + and CD27 hi CD45RA +/À expression patterns, respectively. 16 We assessed liver and blood Vd1 + T cells for the expression of CD27 and CD45RA surface markers ( Fig. 4D-E); we noted a loss of CD27 hi Vd1 + T cells (Fig. 4D) in intrahepatic cd T cells, consistent with the lower diversity we observed in liver TCR repertoires than that of peripheral blood. While CD27 lo/À CD45RA hi cells were present in both liver and blood, we noted the presence of an intrahepatic CD27 lo/À CD45RA lo/À Vd1 + T cell population that was present in all livers to varying degrees, but that was found at only very low levels in peripheral blood (Fig. 4E). The extent of this enrichment in liver was unaffected by liver disease aetiology (Fig. 4E) or CMV infection (Fig. S4).
We then explored the clonality of intrahepatic CD27 lo/À CD45RA hi and CD27 lo/À CD45RA lo/À populations by single-cell TCR sequencing. In a representative liver sample, sorted intrahepatic CD27 lo/À CD45RA lo and CD27 lo/À CD45RA hi Vd1 + T cell populations each comprised single prominent, distinct clonotypes using single-cell sort identities (i.e. CD45RA hi or lo ), allowing the direct alignment of clonotype to phenotype at the single-cell level (Fig. 5A). Notably, within intrahepatic cd T cells, both the CD45RA hi and CD45RA lo populations were predominantly clonally expanded ( Fig. 5A; B, left panel). Consistent with previous findings, 16 in blood the CD27 hi compartment (reduced in frequency in liver) was polyclonal, whereas the CD27 lo/À CD45RA hi compartment was dominated by clonal expansions (Fig. 5B, right panel); notably the CD27 lo/À CD45RA lo compartment was essentially absent in blood. We then systematically examined the relationship between clonotypic and phenotypic identity from matched pairs of blood and liver Vd1 + cd T cells (Fig. 5C). Overall in our paired samples, we identified clonotypes present in both the blood and liver, however importantly we also identified clonotypes unique to either liver or blood (Fig. 5C). The phenotype of clonotypes found only in the blood or shared between blood and liver generally mapped to the CD27 lo/À CD45RA hi compartment found both in blood and liver. In contrast, the clonotypes present exclusively in the liver mapped between CD27 lo/À CD45RA lo and CD27 lo/À CD45RA hi compartments, with a trend towards a CD27 lo/À CD45RA lo phenotype (Fig. 5C). As examples, the highly expanded Vd1 CALGGGGFPQKPGGAGPPTAQLFF and CALGEHPHFFLHLIGTIKLIF clonotypes present in the livers of Donor 0886 and Donor 1421 (both ALD) respectively were CD27 lo/À CD45RA hi in phenotype and also present in the respective matched peripheral blood samples, whereas in each case liver-restricted expanded clonotypes were also observed, but predominantly CD27 lo/À CD45RA lo (Fig. S5A). Taken together, while considerable clonotypic overlap between liver and blood subsets is observed, we identified a distinct population of intrahepatic CD27 lo/À CD45RA lo Vd1 + T cells largely absent from the blood, and which frequently contains TCRs restricted to the liver. This paradigm is likely to extend to intrahepatic Vd3 + cd T cells, which also exhibited a significant proportion of CD45RA lo cells (Fig. S5B).
We sought to further characterise intrahepatic CD27 lo/À CD45RA lo and CD27 lo/À CD45RA hi Vd1 + T cells for markers associated with tissue retention. Firstly, while the surrogate marker of tissue-resident memory T cells (T RM ), CD69, was expressed    widely by Vd1 + T cells, it was markedly higher on CD27 lo/À CD45RA lo Vd1 + T cells and comparable to CD45RA lo CD8 + ab T cells (Fig. 6A). In keeping with functional tissue retention, Vd1 + T cells expressed CXCR3 and CXCR6, with expression predominantly associated with the CD27 lo/À CD45RA lo population (Fig. 6B). In contrast, the endothelial homing receptor CX 3 CR1 (highly expressed by peripheral blood CD27 lo/À CD45RA hi Vd1 + T cells 16 ) was retained on intrahepatic CD27 lo/À CD45RA hi cells but was markedly reduced on CD27 lo/À CD45RA lo Vd1 + T cells (Fig. 6B). Interestingly, intrahepatic CD45RA lo Vd1 + T cells did not express significantly more CD103 than CD45RA hi Vd1 + T cells, which contrasts with CD8 + CD45RA lo T cells isolated from the same livers (Fig. 6B). We next assessed the functionality of intrahepatic Vd1 + T cell populations by ex vivo stimulation with recombinant cytokines or by TCR activation. Following TCR stimulation, intrahepatic Vd1 + T cell populations in general strongly upregulated the T cell activation marker CD25, with equivalent responses in CD8 + ab T cells from the same samples, although Vd1 + T cells from some liver samples responded more robustly than others. Importantly, intrahepatic CD27 lo/À CD45RA lo Vd1 + T cells displayed a greater sensitivity to innate associated cytokines IL-12 and IL-18, than CD27 lo/À CD45RA hi Vd1 + T cells (Fig. 6C). Notably, peripheral blood CD27 lo/À CD45RA hi Vd1 + T cells are unresponsive to IL12/IL-18 stimulation. 16 In keeping with a clonally expanded intrahepatic Vd1 + T cell population, significant responses were observed with IL-15 but not IL-7 cytokines (Fig. 6C). We next assessed effector potential, by analysing intracellular expression of cytolytic granzyme B and perforin. Intrahepatic CD27 lo/À CD45RA hi Vd1 + T cells expressed marked levels of both effector molecules while CD27 lo/À CD45RA lo Vd1 + T cells had much lower expression (Fig. 6D). Conversely, stimulation of the CD27 lo/À CD45RA lo population with PMA and ionomycin produced significantly more of the pro-inflammatory cytokines IFN-c and TNFa than the CD27 lo/À CD45RA hi population (Fig. 6E). These data suggest that intrahepatic CD27 lo/À CD45RA lo Vd1 + T cells have a more prominent tissue-associated phenotype than that of the CD27 lo/À CD45RA hi Vd1 + T cell population, which are more similar to peripheral blood CD27 lo/À CD45RA hi Vd1 + T cells. Moreover, these two populations possess either enhanced cytolytic (CD45RA hi ) or pro-inflammatory cytokine (CD45RA lo ) responses, suggesting distinct roles in intrahepatic immunity.

Discussion
Tissue-associated T cells are thought to play a critical role in tissue immunosurveillance and homeostasis. [28][29][30] In mice, cd T cells have been implicated in epithelial homeostasis, 31 cutaneous wound healing 32 and maintenance of gut mucosa, 33 and have been highlighted as innate-like, expressing canonical TCRs. 34 In humans, solid tissues are known to be enriched for cd T cells but the immunobiology of the T cells present has remained largely unclear. Recent studies on Vd1 + T cells, the canonical tissue-associated human cd T cell subset, have revealed an adaptive biology. 16,17 However, these results were based exclusively on peripheral blood Vd1 + cells, and the immunobiology of solid tissue-associated Vd1 + lymphocytes, often assumed to be innate-like, is of particular interest. We chose to probe these issues by characterising intrahepatic cd T cells as a human model system. We used NGS approaches to show the hepatic Vd2 À compartment is comprised of highly clonal, private expansions, based on complex TCR rearrangements. Importantly these were evident in both diseased and healthy livers, with no skewing of the TCR repertoire chain usage observed between the two scenarios.
Moreover, the proportion of Vd2 À cd T cells decreased upon liver inflammation compared with healthy livers, because of an influx of ab T cells. Therefore, the accumulation of cd T cells in human liver is not driven by the diseased hepatic microenvironment present in these patients, and may reflect a response to other immune challenges such as infection. Of relevance, CMV infection has recently been highlighted as one of a number of drivers of Vd2 À T cell clonal expansion (specifically of Vd1 + T cells) in peripheral blood. 16,17 Moreover, studies on murine CMV have highlighted the potential of expanded cd T cell subsets to populate a range of peripheral tissues, including the liver. 35,36 These observations raise the significant possibility that the expanded clonotypes that contribute so dominantly to human intrahepatic cd T cells both in normal and diseased settings have arisen due to previous infections. Consistent with this, Vd1 + cd T cells were significantly enriched in liver explants from CMV + vs. CMV À donors. Therefore, CMV represents one likely driver of Vd1 + infiltration in the liver. However, it is notable that similar clonotypic focussing and immunophenotypic profiles of intrahepatic Vd2 À T cells were observed in both CMV + and CMV À individuals, consistent with the idea that the Vd2 À subset can mount tissue-localised responses to multiple infections. This mirrors the situation with human Vd1 + T cells  in peripheral blood, where although CMV is linked with an increased proportion of Vd1 + T cells 16,37 and clearly drives clonal expansions of Vd1 + clonotypes, 17 such expansions are commonly observed in CMV À individuals, suggestive of other infectious drivers. 16 While the candidate drivers of intrahepatic Vd2 + T cell expansion would include HCV/HBV, notably we did not study HCV/HBV-related liver disease, and therefore other non-CMV/HCV/HBV drivers must exist. In principle, an alternative to infection representing a main driver of Vd2 À clonal expansion is that intrahepatic Vd2 À T cells are populated in the liver during development. However, both their Vd2 À chain usage and the highly complex nature of the intrahepatic Vd2 À TCR CDR3 regions would argue against this possibility, since foetal cd TCRs would be expected to utilise more simple CDR3 sequences and have also been highlighted as predominantly Vd2 + , 38 thereby highlighting post-natal stimuli such as infection as a more likely underlying driver. Given previous observations regarding peripheral blood Vd1 + T cells, 16 which like those in the liver were frequently highly clonal and also featured private expansions based on complex TCR rearrangements, a key question was the extent to which liver Vd2 À cd T cells mirrored those in the blood. Our study provides compelling evidence that despite the profound link between the liver and the peripheral circulatory system, there is a distinct profile of Vd2 À cd T cells in each compartment, indicative of compartmentalisation of certain Vd2 À subsets.
Comparison of matched liver and blood samples indicated the differentiation status of the Vd2 À T cell subset was distinct in each compartment. Strikingly, liver Vd2 À T cells were uniformly CD27 lo/À , a phenotype previously linked to a clonally expanded effector subset present in peripheral blood, and essentially entirely lacked the CD27 hi subset, even when such populations were relatively prevalent in matched blood. Previously we have shown that CD27 hi Vd1 + T cells in peripheral blood are TCR-diverse and naïve in phenotype. Consistent with selective exclusion of this clonally diverse CD27 hi naïve population, liver Vd2 À cells lacked CCR7, CD62L and CD27 present on such naïve populations, and diversity metrics indicated liver Vd2 À T cells displayed an even more focussed repertoire in liver than in peripheral blood. Furthermore, the phenotype of liver Vd2 À T cells closely matched that of peripheral blood CD27 lo/À Vd1 + T cells, and there was substantial clonotypic overlap between these two populations. While we cannot exclude the possibility that such hepatic CD27 lo/À originated in the liver, these results support the concept that at least some hepatic CD27 lo/À cells may derive from those present in peripheral blood. Such a scenario would fit an adaptive model whereby naïve peripheral blood Vd2 À CD27 hi cells, which express secondary lymphoid homing markers but are devoid of CX 3 CR1, recirculate between blood and lymph, whereas the peripheral blood CD27 lo/À population, which is clonally expanded and likely antigen-experienced, is capable of accessing solid tissues, potentially because of increased CX 3 CR1 expression, and may also upregulate tissue retention markers following liver localisation.
A second indication of compartmentalisation was that in addition to being devoid of CD27 hi naïve cells, the hepatic Vd2 À T cell compartment comprised both a CD45RA hi and also a distinct CD45RA lo subset. By contrast, the peripheral blood CD27 lo/À Vd1 + cells are almost entirely CD45RA hi . Importantly, CD45RA hi clonotypes overlapped substantially between blood and liver within individuals. Such cells in the peripheral blood express a high level of the endothelial homing receptor CX 3 CR1 as well as increased CD16, low CD27/28, low CD127, and   enhanced levels of adhesion molecules relative to naïve CD27 hi cells. 16 While this could suggest capability of homing from peripheral blood to tissues, alternatively it could imply a vascular association, as has been suggested for effector memory CD8 T cells, 39 which include virus-specific CD8 +40 and CD4 +41 T cell subsets. The predominantly sinusoidal localisation of these cells identified in this study is consistent with this possibility, and may suggest a role in immunosurveillance at this site, as suggested for NKTs. 42 In light of the recent report that Vd1 + clonotypes can expand in response to CMV, 17 a virus that infects the endothelial compartment in vivo, and our observation here that Vd1 + T cells are enriched in CMV + vs. CMV À liver explants, these findings suggest this subset may contribute to unconventional T cell protection of the vascular niche, including within solid tissues, against chronic viral infection. Moreover, the observation CMV serostatus correlates with an enhanced proportion of intrahepatic Vd1 + T cells but not with a disturbed CD45RA hi vs. CD45RA lo Vd1 + ratio might suggest the potential within both phenotypic sub-compartments to respond to CMV. In contrast to CD45RA hi clonotypes and consistent with a reduced frequency of CD45RA hi Vd2 À cells in liver compared to peripheral blood, the same analyses of matched blood/liver samples revealed CD45 lo clonotypes were enriched for those restricted to the liver. In addition, this liver CD45RA lo compartment frequently contained clonal expansions. These cells demonstrate striking phenotypic correlation with liverresident lymphocytes identified in previous studies, including enhanced expression of CD69, CXCR3 and CXCR6, which has been noted in liver-resident NK populations 43,44 and CD8 + ab populations. 25 CD27 lo/À CD45RA lo Vd2 À T cells may therefore represent a liver-resident subset, although conceivably they may be able to access other solid tissues. Of note, CD45RA lo Vd1 + T cells exhibited considerably lower expression of CD103 relative to their CD8 + counterparts, suggesting other mechanisms may underly their tissue retention. The origin of this subset is unclear. One possibility is that it originates from a subset of blood CD45RA + cells that alter phenotype once in tissues and are retained there, perhaps following activation in the hepatic microenvironment. This route of generation is supported by our detection of liver-restricted clonotypes in both the CD45RA lo and CD45RA hi compartments. In addition, it is possible they may be locally generated. highlight that a liver-resident phenotype can be induced in CD8 + ab T cells via IL-15 followed by TGF-b signalling, 25 and based on the parallels between Vd1 + and CD8 + ab T cells identified in this study, a similar mechanism may be at work here.
Our results also highlight that hepatic cd T cells are functionally distinct from equivalent subsets in peripheral blood. While still responsive to TCR stimulation/co-stimulation, compared to blood Vd2 À T cells they displayed markedly increased responsiveness to IL-12/IL-18 in line with CD8 + T cells isolated from the same tissue. This responsiveness extended to the liverrestricted CD45RA lo subset, which appeared to display enhanced production of pro-inflammatory cytokines relative to CD45RA hi cells. These observations suggest CD45RA hi and CD45RA lo subsets may have different roles, the former more vascular focussed and cytotoxic, the latter an immunoregulatory tissue-associated subset more focussed on cytokine production and potential induction of a wider T cell response to stress challenges. It is unclear if these distinct features stem directly from the nature of the clonotypes present and their antigenic targets, or whether they reflect the influence of hepatic microenvironmental factors that may also influence intrahepatic retention. 45 Importantly, we note several limitations of our study. Firstly, all diseased samples were derived from end-stage liver disease. While the closely matched clonotypic focussing and immunophenotypic profiles present in normal tissue would predict similar profiles at earlier disease stages, we cannot exclude the possibility that disease stage influences the nature of the intrahepatic cd T cell population, and use of biopsy material from early disease stages with longitudinal follow-up could be an interesting avenue of future investigation. Secondly, while we examined several disease pathologies, these were predominantly restricted to fatty/alcoholic liver disease (ALD, NAFLD) or autoimmune liver disease (AIH, PBC, or PSC). While HCV/HBV+ liver samples showed similar frequencies of cd T cells, we did not study cd T cell immunophenotype or clonotypic focussing   Fig. 6. Intrahepatic Vd1 + T cells segregate into cytokine producing and cytotoxic subsets. (A) Representative histograms from one donor and summary data of CD69 surface expression by CD45RA lo (orange) and CD45RA hi (grey) intrahepatic Vd1 + and CD8 + T cells (n = 8). (B) As in (A), but displaying representative histograms and summary data for CXCR3, CXCR6 and CX 3 CR1 surface expression by intrahepatic Vd1 + and CD8 + T cells (n = 6). (C) Representative histograms and summary data from sorted intrahepatic CD3 + T cells were incubated with indicated medium, cytokines or anti-CD3/CD28 beads for 72 h. CD45RA lo (orange) and CD45RA hi (grey) Vd1 + T cells were then assessed for the upregulation of the T cell activation marker CD25 (n = 5-6). (D) Representative histograms and summary data for intracellular granzyme B and perforin expression by CD45RA lo (orange) and CD45RA hi (grey) intrahepatic Vd1 + and CD8 + T cells (n = 5-6). (E) Representative flow cytometry plot and summary data of intrahepatic CD3 + T cells stimulated with PMA/Ionomycin and assessed for the production of intracellular IFNc and TNFa in CD45RA lo (orange) and CD45RA hi (grey) Vd1 + and CD8 + cells (n = 6). Error bars indicate mean ± SEM; data analysed by Kruskal-Wallis ANOVA with Dunn's post-test comparisons, n.s. p >0.05, *p <0.05, **p <0.01 and ***p <0.001. (A-E) Disease aetiologies analysed included ALD, NASH, PSC, and normal liver no significant differences were observed between different individual disease groups in any of the comparisons highlighted. ALD, alcoholic liver disease; IFN, interferon; NASH, non-alcoholic steatohepatitis; PSC, primary sclerosing cholangitis; TNFa, tumour necrosis factor alpha. (This figure appears in colour on the web.) in such samples and cannot therefore exclude the possibility that HCV/HBV infection may drive development of distinct intrahepatic cd T cell profiles 47 or clonality, although we hypothesise they would follow broadly similar principles to those observed in this study; moreover, while we did not observe differences between the different disease types we did analyse, conceivably with larger samples sizes differences may have emerged, for example in the extent of cd TCR clonotypic focussing or cd T cell phenotypes. Finally, a comparison of the data presented here with cd T cell clonotype and immunophenotype profiles in other solid tissues, including during chronic inflammation, would shed light on tissuespecific cd T cell responses.
Our study establishes that in humans, clonally expanded cd T cell effector subsets can be selectively deployed to at least some solid tissues, including the liver, thereby providing ongoing immune surveillance against previously encountered infectious or non-infectious challenges, with CMV infection one likely driver of Vd1 + T cell intrahepatic infiltration. Importantly, both Vd1 + and Vd3 + intrahepatic T cell compartments displayed clonotypic expansion and a CD45RA lo subset, suggesting their immunobiology may be closely aligned. Moreover, the finding that intrahepatic cd T cell subsets can be phenotypically, clonotypically and functionally distinct from those in peripheral blood suggests distinct contributions to intrahepatic immune responses, and provides a basis for future investigation of human tissue-resident cd T cell populations. Notably, cd T cells are of increasing therapeutic interest, due partly to their potential to mount either anti-tumour, [47][48][49] or alternatively immunosuppressive 50 responses, but also their MHC-unrestricted recognition of target cells, which raises the prospect of broad applicability of cd T cell-based therapies in patient cohorts.
Our finding that there appears to be selective recruitment of cd T cell subsets of an effector phenotype into the hepatic pool may inform design of cd T cellular therapies that rely on administration/expansion of systemic cd T cells. Secondly, the finding that a number of distinct differentiation states exist within the Vd1 + compartment (including naïve, circulating effector, tissueresident effector) indicates a degree of plasticity that could be investigated further and potentially exploited therapeutically, either to increase immunosuppressive functionality during inflammatory liver disease, or for improved anti-tumour effector function in liver cancer. Finally, our finding that CMV infection represents one likely factor driving infiltration of potentially highly inflammatory Vd1 + T cells into the liver could have clinical relevance in chronic liver disease and CMVassociated hepatitis. Specifically, future studies correlating CMV titres with biomarkers of liver damage and with Vd1 + cd T cell frequency may shed light on whether the cd T cell response to CMV infection impacts the severity of chronic liver disease.