The gut digests, absorbs and metabolises nutrients whilst also acting as a barrier to pathogens entering from the intestinal lumen. Not all intestinal microbes are harmful and the full development of the gut immune system is shaped by interactions with commensal bacteria in a complex interplay which allows responses to harmless bacteria and nutrients to be suppressed whilst developing protective immune response against pathogens.1 2 Most of these responses occur in the intestinal mucosa and underlying lamina propria; however, transport of nutrients or translocation of pathogens into the portal circulation means that a second level of protection is required in the liver to respond to antigens that evade the gut immune system.3 In addition, pathogens can enter the liver directly via the systemic circulation or from the gut lumen via the biliary epithelium. Immune responses in the gut and to a lesser extent liver are well-described but the mechanisms that control immunological cross-talk between these two linked sites are only beginning to be elucidated.4 They provide insights into the pathogenesis of diseases that affect both sites including infections and inflammatory bowel disease (IBD).

ORGANISATION OF THE GUT IMMUNE SYSTEM

Antigen recognition and acquired intestinal immune responses

The mucosal barrier provides the first line of protection against intestinal pathogens.5 It consists of a mucus layer rich in anti-bacterial substances,6 immunoglobulin A (IgA)7 and commensal bacteria. The epithelial tight junctions prevent bacterial translocation and pathogen-recognition receptors, including toll-like receptors (TLRs) and nucleotide-binding oligomerisation domain (NOD) molecules, sense potentially harmful antigens and trigger innate immune responses.8 9 The critical function of these receptors is demonstrated by the association of Crohn’s disease with mutations that disrupt the functioning of NOD2.10

Antigens that penetrate the mucosal barrier are sampled and processed by specialised antigen presenting cells (APCs) called dendritic cells (DCs) and carried via lymphatics to draining mesenteric lymph nodes (MLNs) where they activate naive lymphocytes to generate antigen-specific effector lymphocytes. DCs arise from haematopoietic stem cells11 as myeloid DCs (mDCs) or plasmacytoid DCs (pDCs). mDCs are found in peripheral tissues, which they enter from blood,11 whereas pDCs enter lymph nodes. Under normal conditions there is constant trafficking of gut mDCs to draining lymph nodes. These DCs take up apoptotic enterocytes but because of the lack of inflammation they are not fully activated and unable to stimulate effector responses thereby maintaining tolerance to self-proteins.12 The inflammatory response to injury or infection provides signals that drive mDC maturation13 and increased migration from tissue via lymphatics into draining lymph nodes. These DCs are fully activated and thus able to stimulate immune responses, the nature of which will be shaped, in part, by the type of activating stimulus which reflects the nature of the injury or infection.11 Replenishment of tissue mDCs occurs by recruitment of precursors from blood. In contrast, pDCs do not traffic between liver or gut tissues and draining lymph nodes and instead operate in inflamed tissues by secreting type 1 interferons (IFNs)14 in response to viral infection. In vitro pDCs activate naive T cells to become immunosuppressive regulatory T cells (T_reg) that suppress immune responses in the gut.15^–17 Similar populations of cells have been found in the liver18 19 (see below). Luminal antigens are sampled directly by specialised populations of DCs in Peyer’s patches which express the fractalkine receptor, CX3CR120 and extend dendrite-like processes through epithelial tight junctions21 into the lumen and by unique epithelial M cells that transport luminal antigens to underlying DCs.1 2 5

The intraepithelial compartment

Intraepithelial lymphocytes include conventional α/β T cells and unusual subsets including αα CD8⁺ cells and γ/δ cells. CD4⁺ and CD8⁺ T cells are localised to the small bowel epithelium by the chemokine CCR9 and interactions between the integrin α_Eβ₇ (CD103) and epithelial E-cadherin.22 23 Some of these cells display cytotoxic activity for epithelial cells24 whereas others secrete TGF-β, suppress inflammation and maintain tissue integrity. Recent studies suggest an important role for γ/δ cells in early responses at mucosal sites where they regulate T and B cell activation by secreting cytokines such as IFN-γ.25

Lamina propria

The sub-epithelial lamina propria contains large numbers of antibody-secreting IgA plasma cells, macrophages and DCs as well as T cells. Macrophages are critical for maintaining homoeostasis. They phagocytose pathogens and, as a consequence of constant exposure to endotoxin, show reduced inflammatory responses to activation via TLR-4 thereby helping to prevent constant uncontrolled inflammation in the gut.1 The lamina propria also contains myofibroblasts26 that maintain the integrity of the gut wall and respond to pathogens by secreting cytokines that, depending on the nature of the stimulus, either drive effector responses or in the case of transforming growth factor-β (TGF-β) and interleukin 22 (IL22) maintain epithelial integrity. A population of subepithelial myofibroblasts from human colon express MHC class II and co-stimulatory molecules and can process and present antigens suggesting they act as competent APCs.26

The outcome of immune activation in the LP is critically dependent on the balance between activation of CD4 T cells that secrete IFN-γ and drive inflammatory responses and regulatory cells that suppress and control inflammation.1 CD4⁺CD25⁺T_regs are generated in the thymus or periphery as a consequence of activation of naive T cells by immature DCs in the presence of specific cytokines such as IL10 and TGF-β.27 They provide a mechanism to maintain tolerance to antigens not present in the thymus including food antigens or self-antigens first exposed to the immune system as a consequence of tissue damage. T_regs in the LP express CTLA-4, glucocorticoid-induced tumour necrosis factor (TNF) receptor, and Foxp3 and suppress proliferation of responder CD4⁺T cells in vitro. The importance of T_regs is demonstrated by experiments showing that their depletion leads to fulminant colitis whereas transfer of LP T_regs prevents the development of colitis.28 29 A population of CD103(+) mesenteric lymph node DCs induce the development of Foxp3⁺T_regs but only in the presence of TGF-β and retinoic acid.30 31

Gut-specific lymphocyte trafficking

A system of tissue-specific lymphocyte trafficking has evolved to target lymphocytes to areas of infection or injury (fig 1). This is controlled by combinations of chemokines and adhesion molecules (addressins) expressed in target tissues that act as a molecular postal code to attract subsets of lymphocytes expressing appropriate counter-receptors.32 The gut postal code consists of MAdCAM-1 on mucosal endothelium and the chemokine CCL25 expressed by small-bowel epithelium and their lymphocyte ligands α₄β₇ and CCR9.33 MAdCAM-1 is widely expressed in mucosal vessels and the intestinal lamina propria34 whereas CCL25 is only found in the thymus and small bowel.35 36 CCR9 and α₄β₇ are induced on mucosal lymphocytes during activation by gut DCs (see below) and ligation of CCR9 by CCL25 activates α₄β₇ binding to MAdCAM-1 thus promoting recruitment to the small bowel.35^–38 α₄β₇ is also involved in recruitment of activated T cells to the colon although the chemokine receptors involved are less apparent as CCL25 is absent from colonic epithelium.33 During exacerbations of IBD, MAdCAM-1 expression is upregulated and promotes the sustained recruitment of circulating α₄β₇⁺ lymphocytes and the establishment of chronic bowel inflammation. Antibody inhibition of α₄β₇ and CCR9 reduces inflammation in animal models and both approaches are currently being assessed in patients with IBD.39^–41

Activation of naive lymphocytes in Peyer’s patches and MLN not only primes them for antigen recognition but also imprints them to express α₄β₇ and CCR9 and thus to preferentially traffic as effector cells to the gut.42^–45 This imprinting is dependent on the ability of gut-derived DCs to convert retinol to retinoic acid which activates intracellular retinoid receptors in lymphocytes and the transcription of genes encoding CCR9 and α₄β₇.46 This ability is restricted to gut-derived DCs and not shared by DCs from liver or portal lymph nodes.42 43 In contrast, dermal DCs metabolise sunlight-induced vitamin D₃ and program T cell homing to the skin.47 48 Thus, specialised gut-derived DCs determine the tissue tropism of lymphocytes activated in GALT and also generate intestinal T_regs thereby shaping the outcome of gut inflammation.49

ORAL TOLERANCE

Despite these mechanisms some pathogens and food antigens evade the mucosal immune system and enter the liver via the portal circulation.50 Orally administered antigens can be picked up, processed and presented on liver endothelial cells within 2 h of ingestion51 52 and the liver is critical in the regulation of immune responses to pathogens entering via the gut. The liver is ideally positioned for such a role. It receives 75% of its blood supply from the portal vein which drains the gut and oral tolerance is lost in the presence of a porto-systemic shunt which allows portal blood to pass directly from the gut to the systemic circulation.53

IMMUNE ACTIVATION AND REGULATION IN THE LIVER

Although the liver is characterised by immune tolerance in several settings it is also capable of generating vigorous immune responses to infections such as hepatitis A and hepatitis E viruses, both of which enter via the gut.52 In simple terms a vigorous intrahepatic immune response depends on activation of T cells by fully activated DCs within secondary lymphoid tissues whereas direct activation within the liver by resident APCs including endothelial cells and hepatocytes usually results in tolerance.54 55 This is logical as it allows the liver to tolerate soluble food antigens captured by liver endothelial cells and self-antigens on hepatocytes that fail to cause damage whilst responding appropriately to infections that cause injury, inflammation and full activation of DCs.

The biliary epithelium is an extension of the gut epithelial barrier

Similar barrier functions are shared by the biliary and gut epithelium including IgA secretion, the expression of pathogen pattern receptors and resident epithelial associated lymphocytes and DCs.56 57 Cholangiocytes participate in leukocyte recruitment by secreting chemokines and expressing adhesion molecules in response to proinflammatory cytokines or TLR ligands.48 58^–60 Recent evidence implicates two particular chemokines, fractalkine and CXCL16.61 62 Both exist in transmembrane forms and are detected on inflamed epithelium.63 The receptor for CXCL16, CXCR6, is found on effector T cells, NK cells, and NKT cells64 with high levels reported on liver-infiltrating and gut-infiltrating lymphocytes.61 65 66 The lack of CXCR6 protects animals from colitis and CD8⁺ T-cell infiltration into the liver in GVHD.67

Intrahepatic lymphocytes

The normal liver contains lymphocytes scattered throughout the parenchyma and within portal tracts.68 69 These include CD4⁺ and CD8⁺ T cells, B cells, NK and NKT cells.70 NK cells provide protection against viral infections and tumours by recognising cells that have downregulated MHC class I71 and NKT cells secrete cytokines in response to bacterial sphingolipids presented by CD1d on DCs and epithelial cells.72 NKT cells can also suppress inflammation by destroying APCs and secreting cytokines that polarise immune response towards tolerance.73 The role of intrahepatic B cells is less clear although recent work suggests they may be involved in fibrogenesis.74 Like the gut the liver also contains γ/δ T cells which have been implicated in regulatory networks (see later).

Antigen presentation by dendritic cells

The liver contains distinct populations of APCs including DCs19 75 and other cells capable of antigen presentation such as sinusoidal endothelial cells,76 Kupffer cells,77 78 stellate cells79 and hepatocytes.80 Both the site and nature of the APCs determine the outcome of intrahepatic immune activation.3 54 Like interstitial gut DCs, mDCs in the liver are immature and although efficient stimulators of naive T cells they secrete high levels of IL10 and tend to drive regulatory responses.81 82 In humans the CD16⁺ subset, which makes up 15% of human blood monocytes, gives rise to tissue DCs that tend to suppress immune responses83 and our unpublished data show that most mDCs in human liver are CD16⁺. Whereas mDCs are found in normal liver pDCs are recruited in response to inflammation.18 84 Intrahepatic BDCA-2⁺ pDCs are a discrete population of MHC class II and CD123^high cells that can be detected in inflamed liver but which are absent from normal liver. Defects in pDC function, particularly a lack of IFN secretion, are associated with a dysregulated immune response in viral hepatitis.18 85

Pathways of DC migration in the liver

mDC precursors migrate from bone marrow via the blood to the liver where they reside until signals from pathogens stimulate them to take up and process antigen. The signals that recruit DC precursors into the liver are poorly understood but probably involve a combination of chemokine and adhesive signals displayed on sinusoidal endothelium and Kupffer cells.86 87 In response to infection intrahepatic DCs migrate from the parenchyma via the space of Disse to the portal area where they interact with T cells. In some circumstances this leads to the development of portal tract-associated lymphoid tissue87 88 whereas in others the DCs migrate to draining lymph nodes where T cell activation takes place.54 89

Regulation of DC function in the liver

Some stimuli in the liver result in T cell priming and the generation of effective immune responses whereas others result in tolerance.3 54 Liver-derived DCs are inherently tolerogenic when compared with skin DCs; they produce IL1081 and express low levels of the co-stimulatory molecules required for full naive T cell activation.81 The differentiation and function of DCs within the liver is shaped by the local microenvironment including danger signals that reflect tissue damage responses; pathogen-associated molecular patterns that allow DCs to detect pathogens and tissue-specific signals that shape the liver-specific nature of DC.46 81 90 The latter signals are the least understood. Local secretion of IL10 and TGF-β by Kupffer cells and hepatocytes skews DC function towards the generation of regulatory as opposed to effector pathways78 91 and DCs cultured in the presence of liver tissue differentiate towards a regulatory phenotype.81 92 The importance of local IL10 is emphasised in murine models of liver injury93 and inhibition of IL10 restores the depressed function of human intrahepatic DCs.18 94

Interactions with stromal cells regulate the differentiation and activation of interstitial DCs. Thus splenic stromal cells drive the differentiation of regulatory DCs whereas dermal fibroblasts promote DCs that drive effector T cell responses.95 96 Myofibroblasts are present in both the liver and gut and have characteristics that allow them to shape DC differentiation as well as acting themselves as APCs.79 The gut and liver cooperate to regulate intrahepatic DC maturation and function. A recent study reports that the threshold for activation of intrahepatic DCs is increased by IL6-dependent STAT3 signalling driven by intestinal bacteria carried to the liver via the portal vein.97 Other studies report that Kupffer cells and sinusoidal endothelial cells respond to endotoxin by secreting immunosuppressive cytokines.98 99 Thus the liver avoids being in a state of perpetual inflammation as a consequence of continual exposure to gut-derived antigens and bacterial products97 100 and manipulating gut bacteria may be an effective strategy to alter intra-hepatic immune responses. Thus, like intestinal DCs, the default setting of intrahepatic DCs under homeostatic conditions is towards tolerance but in response to local inflammatory signals they can induce full effector responses.

Antigen presentation by resident liver cells

A distinctive feature of the liver is the ability of macrophages, endothelial cells, epithelial cells and stellate cells to act as APCs under specific conditions.3 52 54 79 The liver’s unique architecture permits interactions between circulating T cells via fenestrations in the sinusoidal endothelium with underlying hepatocytes and stellate cells.101

Sinusoidal endothelial cells

Low flow rates through the narrow hepatic sinusoids promote interactions between circulating lymphocytes and liver sinusoidal endothelial cells (LSECs). LSECs express C-type lectins and scavenger receptors which allow them to take up, process and present antigen to naive lymphocytes.51 52 102 The mannose receptor, which is expressed at high levels on LSECs, feeds antigen into an early endosomal compartment committed to presentation on MHC class I which may explain how LSECs cross-present antigens to CD8⁺ T cells.103 In addition to presenting soluble antigens from portal blood LSECs also take up apoptotic cells and cross-present antigens from these cells to naive T cells.104 The outcome of antigen presentation by LSEC is usually tolerance with apoptosis of antigen-specific CD8⁺ T cells and IL10 and IL4 secretion by responding CD4⁺ T cells.51 105 In addition, some activated T cells are trapped by ICAM-1-dependent mechanisms within the sinusoids before undergoing Fas-mediated apoptosis, a homeostatic mechanism that contributes to the control of systemic CD8⁺ T cell responses.106^–108 PD-L1, the ligand for the immunoinhibitory receptor PD-1, is expressed constitutively on sinusoidal endothelium and Kupffer cells and provides another mechanism to inhibit activation of effector T cells.109

Hepatocytes

Although hepatocytes can function as APCs to activate naive T cells in vitro it was believed that they are inaccessible to naive T cells in vivo and thus unlikely to be biologically important. However, Bertolino and colleagues54,101 have shown that naive T cells in the sinusoids can access hepatocytes via fenestrations within sinusoidal endothelium. Lymphocyte pseudopods penetrate the fenestrations and come into contact with microvilli on the hepatocyte surface. This allows naive T cells to be activated in vivo by antigen restricted to hepatocytes. In most circumstances activation by hepatocytes leads to antigen-specific tolerance; however, a recent study of murine liver transplantation reported complete differentiation of CD8⁺ T cells after activation within the transplanted liver leading the authors to suggest that activation of T_reg cells rather than incomplete activation by hepatocytes is the dominant mechanism of intrahepatic tolerance.110

The role of inflammation in determining the outcome of immune activation

Thus, local presentation of antigens in the liver whether they be soluble antigens taken up by LSECs or intracellular antigens presented by hepatocytes usually results in tolerance.54 This default setting may be related to the relative insensitivity of these tissues to lipopolysaccharide, a vital property that prevents the liver being in a constant state of immune activation in response to gut-derived bacterial products in the portal circulation. Recent evidence suggests that the hepatic endotoxin response is also regulated by the gut. In Crohn’s disease gut inflammation results in the recruitment and activation of Th1 and Th17 T effector T cells which secrete IL22111 which is detectable at high levels in the serum.112 IL22 has effects on non-immune cells including epithelial cells where it promotes healing and bacterial defence and on hepatocytes it has a hepatoprotective effect and promotes liver regeneration.113 114 It also induces hepatocytes to secrete LPS-binding protein (LBP) which, in turn, binds portal endotoxin and suppresses intrahepatic inflammation.112 Thus IL22 secreted in the inflamed gut is able to modulate inflammatory responses within the liver, preventing the inappropriate systemic spread of inflammation. However, other forms of inflammation can overcome local LPS resistance and provide an environment in which T cells can be effectively activated. Thus the recruitment of small numbers of activated CD8⁺ T cell blasts can prime hepatocytes to induce full T cell activation in some models115 and induction of local type 1 interferons by murine cytomegalovirus results in a vigorous intrahepatic immune response.116 Furthermore, increased endotoxin levels in portal blood as a consequence of increased gut permeability have been associated with the development of steatohepatitis demonstrating that endotoxin resistance can be overcome.117

Regulatory T cells and liver tolerance

Peripheral T_regs are generated if naive T cells are activated by immature DCs or in the presence of IL10 and TGF-β27 suggesting that the liver will be a fertile environment to generate T_regs.118 We reported a subset of intrahepatic T_regs in human liver119 which secrete IL10 and suppress immune responses. They use the chemokine receptor CXCR3 to respond to IFN-γ-dependent chemokines produced within the inflamed liver120 and CCR10 to localise to CCL28 secreted by biliary epithelial cells resulting in accumulation around bile ducts. CCL28 is also expressed by intestinal epithelium and thus similar signals may localise T_regs in the gut and liver.119 CD4⁺CD25⁺ T_regs can suppress activation of CD4⁺ T cells by LSEC, Kupffer cells or hepatocytes in vitro.78 94 However, this can be overcome by activation of TLR-4 suggesting that interactions between T_regs, pathogens and other liver cells will determine the outcome of immune activation in the liver and the transition from tolerance to inflammation.78 T_regs are required to maintain the stability of chronic inflammation as demonstrated by studies in murine models of colitis in which an absence of T_regs results in fulminant colitis.28 Other cells can mediate suppression, including NKT cells, which are found in the gut and at high frequencies in the liver.72 A prominent role for NKT cell-derived IL4 in suppressing T-cell-mediated hepatitis has been demonstrated in mice73 121 and oral antigens can activate intrahepatic NKT cells that have the ability to suppress experimental colitis, suggesting that these cells may be involved in regulating tolerance across both the gut and liver.122 γ/δ T cells are also found in the gut and liver where they play important regulatory roles, particularly in suppressing macrophage activation in infections such as Listeria monocytogenes123 and in transplant tolerance.124 Thus the outcome of any inflammatory response in the liver will be determined by complex local interactions between several cell types shaped by the local cytokine environment.52

DISEASE AFFECTING THE GUT AND LIVER

In light of the close integration of the mucosal and hepatic immune systems and shared exposure to antigens it is not surprising that the liver can be affected in immune-mediated diseases primarily affecting the gut.

Coeliac disease

An increased appreciation of the link between coeliac disease and liver disease has developed as more cases are diagnosed using serological tests to detect antibodies against tissue transglutaminase.125^–127 The mechanisms that link coeliac disease to liver dysfunction are unknown but abnormal liver enzymes are detected in up to 60% of patients with coeliac disease and 10% of patients with unexplained elevated transaminases have endomysial antibodies.127 Liver biopsy shows a non-specific lymphocytic infiltrate which usually resolves on a gluten free-diet. Coeliac disease has been reported in association with autoimmune hepatitis and primary sclerosing cholangitis (PSC) but perhaps the strongest association is with primary biliary cirrhosis (PBC) and this may be part of the autoimmune diathesis associated with coeliac disease.126 Around 3% of patients with coeliac disease have PBC and, conversely, 6–7% of PBC patients have coeliac disease.126 128 129 Unfortunately, treatment with a gluten-free diet does not improve the outcome of autoimmune liver disease although it may alleviate symptoms of fatigue. Coeliac disease has also been associated with fatty liver disease where increased gut permeability and consequent increased portal endotoxinaemia has been implicated as a mechanism.117

Inflammatory bowel disease and primary sclerosing cholangitis

Liver disease, in the form of autoimmune hepatitis or PSC develops in 2.4–7.5% of patients with IBD, and 70–85% patients with PSC will suffer from IBD at some point in their lives.130 The strongest association is with ulcerative colitis (90%), with Crohn’s colitis predominating in the remaining 10% (fig 2). The estimated incidence of PSC is one to six cases per 100 000 in the UK and two-thirds affect men.131 132 PSC is commonly associated with right-sided colitis characterised by involvement of the terminal ileum, known as “back-wash ileitis”.133 In Crohn’s disease it is almost always associated with large bowel involvement.134 135

Several studies have reported involvement of the liver in animals with IBD. Deficiency of the signalling protein Sin leads to excessive T cell responses in young mice associated with the spontaneous development of small intestinal inflammation. Seventy per cent of affected animals also develop a granulomatous hepatitis although whether this is driven by lymphocytes activated in the gut or as part of systemic activation is unclear.136 The SAMP-1/yit strain of senescence-prone mice also develops a spontaneous ileitis associated with hepatitis characterised by focal clusters of lymphocytes throughout the liver.137 However, both of these models are characterised by small-bowel inflammation whereas the link with clinical IBD is with colitis.

What makes the relationship between the gut and the liver particularly intriguing is the fact that patients can develop PSC for the first time many years after total colectomy for colitis, and colonic inflammation can occur for the first time after patients have undergone liver transplantation for PSC.138 This contrasts with most extra-intestinal manifestations of IBD such as skin and eye disease and acute non-erosive arthropathy where relapses are linked to colonic inflammation and improve on resolution of colitis (table 1).139 140 Consequently, any hypothesis linking liver disease with gut inflammation must take this into account. The final common pathway in IBD and its hepatic complications is a destructive inflammatory infiltrate and evidence implicates mucosal lymphocytes in extra-intestinal disease.141 142 Recent insights into the molecular basis of lymphocyte homing have suggested novel mechanisms to explain how extra-intestinal complications can occur many years after inflammation in the gut has resolved.4 138

A ROLE FOR MUCOSAL T CELLS IN LIVER DISEASE

The existence of a population of long-lived memory T cells capable of homing both to the liver and the gut could explain the link between IBD and liver disease (table 2). There is a teleological rationale for shared lymphocyte recirculation between the gut and liver. The liver develops from the ventral floor of the foregut as the liver diverticulum from the undifferentiated gut endoderm. Structure is provided by the mesenchymal-derived blood sinusoids143 with subsequent growth driven by growth factors from the cardiac mesoderm or foregut endoderm. Subsequently, the gut is populated by lymphocyte precursors derived from the developing liver.144 Thus the two organs share common embryological origins. Because the liver receives most of its blood supply from the portal vein it makes sense for mucosal lymphocytes to provide protection across both sites against pathogens that have crossed the mucosal barrier and entered the liver. However, dysregulation of this system could allow uncontrolled inflammation to spread from the gut to the liver. Evidence suggests that a proportion of the lymphocytic infiltrate in PSC consists of cells generated in the gut during episodes of inflammation which enter the liver in response to aberrantly expressed gut-homing molecules.4

The first evidence linking mucosal T cells with extra-intestinal disease came from observations that activated mucosal lymphoblasts bind to synovial vessels141 using generic adhesion receptors such as ICAM-1 and VCAM-1 but not gut-specific receptors.145 This model is consistent with systemic activation of vascular beds by TNF-α and could explain why pyoderma gangrenosum and acute seronegative arthritis respond so well to treatment with anti-TNF-α antibodies.146 However, this is not the case for hepatic disease which can occur in the absence of a diseased colon excluding the involvement of recently activated effector cells from the gut, though the existence of a population of long-lived mucosal memory cells that migrate between the gut and liver could explain the link. Evidence in support of such an entero-hepatic lymphocyte pathway comes from observations that the gut addressin MAdCAM-1 and the gut-associated chemokine CCL25 are detected on liver endothelium in PSC whereas under normal conditions they are restricted to the gut.4 34 147 CCL25 can trigger adhesion of α₄β₇⁺ lymphocytes to MAdCAM-1 on hepatic vessels148 providing a mechanism for the recruitment of α₄β₇⁺CCR9⁺ mucosal lymphocytes to the liver.37 Furthermore, VAP-1 which supports lymphocyte adhesion to normal human liver vessels149 is markedly upregulated on mucosal vessels in IBD.150

The best evidence that PSC is driven by mucosal T cells comes from the finding that 20% of the intrahepatic T lymphocytes in PSC are CCR9⁺/α₄β₇⁺,37 a unique adhesion molecule combination that is imprinted during activation of lymphocytes by gut DCs.33 42 An alternative explanation is that antigen presentation in the liver or draining nodes in PSC is altered to promote the generation of CCR9⁺α₄β₇⁺ lymphocytes in the liver. However, we found that this ability is restricted to gut-derived DCs and DCs isolated from other tissues including the liver or draining lymph nodes of patients with PSC, were unable to induce the gut homing phenotype on responding T cells.42 Thus we conclude that liver-infiltrating T cells in PSC must have been activated in the gut.37 The liver-infiltrating CCR9⁺α₄β₇ cells are primed to secrete IFN-γ on stimulation suggesting that they can be rapidly expanded into effector cells in the liver thereby mediating inflammatory damage although the triggering antigen is still unknown.37 Several animal models are associated with inflammation in the gut and liver. The Samp-1 mouse develops an ileitis which has a close resemblance to human Crohn’s disease and also biliary inflammation although, at least in early stages, the liver inflammation does not appear to be associated with increased hepatic CCL25 expression.41 The IL2Rα (CD25) deficient mouse develops colitis and has recently been shown to develop biliary disease although this has features more closely resembling primary biliary cirrhosis rather than PSC. It is presumed that the underlying defect in this mouse is of regulatory T cells emphasising their importance in maintaining immune control in both the liver and the gut.151 152

Graft-versus-host disease (GVHD) provides further evidence that mucosal T cells are recruited to the liver under some conditions. In allogeneic stem-cell transplantation donor T cells mediate GVHD in response to alloantigens expressed in the skin, intestine, liver and thymus. Peyer’s patches are required to activate host-specific cytotoxic T cells in GVHD and blocking α₄β₇-integrin prevents the disease developing in the liver.153 Animals given α₄β₇-depleted donor T cells develop significantly less GVHD of the intestines and liver, whereas cutaneous and thymic GVHD is unaffected in recipients of α₄β₇-depleted T cells.154

INFECTIONS OF THE GUT AND LIVER

Most T cells that infiltrate the liver are primed cells, including those with specificity for persistent viruses, suggesting that trafficking of memory T cells through the liver might contribute to immune surveillance.155 Indeed in murine influenza the liver can act as a reservoir of antigen-specific CD8⁺ effector T cells in both acute and recall immune responses despite the fact that virus is not detected outside the lung.156 The hypothesis that memory lymphocytes activated in the gut can be recruited to the liver is supported by the observation that the intestinal CD8⁺ T cell response to L monocytogenes is associated with a clonal expansion of memory CD8⁺ T cells in the liver but not the spleen.157 After activation by gut-restricted antigens CD8⁺ memory T cells can be detected in extra-intestinal sites including the liver.157 However, these studies do not confirm that mucosal memory T cells preferentially recirculate through the liver under normal conditions. Indeed, adoptive transfer studies in parabiosis models show that recruitment of memory T cells to the liver under non-inflammatory conditions does not require α₄β₇ integrins.158

The ability of mucosal memory lymphocytes to respond to antigens in the liver may be restricted by the tolerogenic milieu. Mechanisms that prevent intrahepatic activation of lymphocytes could result in the death of most enteric lymphocytes that enter the liver. IL10 is a key mediator of liver tolerance and it is implicated in the resolution of gut inflammation and the regulation of immune responses to gut parasites some of which also infect the liver.159 Oral infection with Trichinella spiralis results in a severe hepatitis in IL10 knockout animals as a consequence of a potent CD4⁺ T-cell-mediated response to parasites migrating via the portal vein into the liver.93 160 An ongoing intestinal immune response is necessary for the development of hepatitis which is driven by gut-derived CD4⁺ cells recruited to the liver in a MAdCAM-1/α₄β₇-dependent manner.93 In wild-type animals IL10 prevents the development of hepatitis in response to parasites in the liver allowing the worm to survive outside the gut but if IL10 is absent a vigorous immune response ensues.160 This infection provides a very clear example of how local IL10 can completely suppress immune responses to pathogens entering the liver from the gut.

CONCLUSIONS

Knowledge of the mechanisms that underpin co-ordinated immune responses between the liver and the gut not only allows us to explain why and how some intestinal diseases are associated with disease at distant extra-intestinal sites, but also suggests novel therapeutic targets. The important role played by homing of mucosal T cells expressing α₄β₇ in response to aberrantly expressed gut addressins suggests that blocking these pathways might prevent the recruitment of effector cells into the liver. Therapeutic inhibitors of both CCR9 and α₄β₇ are currently in development for the treatment of IBD and it will interesting to see if these reagents also show benefit in treating extra-intestinal inflammation in the liver.