T Lymphocytes Interact With Hepatocytes ThroughFenestrations in Murine Liver Sinusoidal Endothelial Cells

Alessandra Warren,1 David G. Le Couteur,1 Robin Fraser,2 David G. Bowen,3

Geoffrey W. McCaughan,4 and Patrick Bertolino4

The liver has an established ability to induce tolerance. Recent evidence indicates that this uniqueproperty might be related to its distinctive architecture allowing T cells to be activated in situ inde-pendently of lymphoid tissues. Unlike lymph node–activated T cells, liver-activated T cells are short-lived, a mechanism that might contribute to the “liver tolerance effect.” Although the potential role ofhepatocytes as tolerogenic antigen-presenting cells has been demonstrated, the question as to whetherthese cells are able to interact with CD8� T cells in physiological settings remains controversial.Contradicting the immunological dogma stating that naıve T lymphocytes are prevented from inter-acting with parenchymal cells within non-lymphoid organs by an impenetrable endothelial barrier,we show here that the unique morphology of the liver sinusoidal endothelial cell (LSEC) permitsinteractions between lymphocytes and hepatocytes. Using electron microscopy, we demonstrate thatliver resident lymphocytes as well as circulating naıve CD8� T cells make direct contact with hepato-cytes through cytoplasmic extensions penetrating the endothelial fenestrations that perforate theLSECs. Furthermore, the expression of molecules required for primary T cell activation, MHC class Iand ICAM-1, is polarized on hepatocytes to the perisinusoidal cell membrane, thus maximizing theopportunity for interactions with circulating lymphocytes. In conclusion, this study has identified, atthe ultrastructural level, a unique type of interaction between naıve T lymphocytes and liver paren-chymal cells in vivo. These results hold implications for the pathogenesis of viral hepatitis in whichhepatocytes may represent the main antigen-presenting cell, and for the development of immunetolerance as lymphocytes pass through the liver. (HEPATOLOGY 2006;44:1182-1190.)

See Editorial on Page 1083

According to current immunological tenets, vascu-lar endothelial cells form an efficient physical bar-rier that prevents access by naıve T lymphocytes to

the surrounding tissue.1-3 Naıve lymphocytes require ac-

tivation in peripheral lymphoid tissues by professionalantigen-presenting cells (APCs) before they are able tomigrate across the endothelium for subsequent interac-tion with parenchymal cells.

Recent studies provide evidence that the liver maybe an exception to this model.4-9 In contrast to othersolid organs, naıve CD8� T lymphocytes can be acti-vated in the liver independently of lymphoid tissues.Although T lymphocyte activation in the liver andlymph nodes (LN) can occur concomitantly, the siteof activation appears to drive T lymphocytes towarddiffering fates. T lymphocyte activation by professionalAPCs in LN generates an effective immune response.Conversely, T lymphocyte activation in the liver is aninefficient process, leading to diminished cytotoxicactivity and reduced T cell survival.8 This model pro-vides a plausible mechanism for the so-called “livertolerance effect”10 and suggests that the site of primaryactivation has a profound influence on intrahepaticimmunity.9

Within the liver, there is increasing evidence that hepa-tocytes may function as APCs, a finding highly relevant tosome hepatotropic viral infections such as those by thehepatitis C virus. Hepatocytes express high levels of majorhistocompatibility complex (MHC) Class I molecules,11

as well as intercellular adhesion molecule-1 (ICAM-1)

Abbreviations: APC, antigen presenting cell; LN, lymph nodes; LSEC, Liver sinusoi-dal endothelial cell; IHL, Intrahepatic lymphocytes; SEM, scanning electron microscopy;TEM, transmission electron microscopy; CFSE, 5-carboxyfluorescein diacetate succin-imidyl ester; TEHLI, trans-endothelial hepatocyte–lymphocyte interactions.

From the 1Centre for Education and Research on Ageing (CERA) and the ANZACResearch Institute, Concord RG Hospital and University of Sydney, Sydney, Australia;the 2Department of Pathology, Christchurch School of Medicine and Health Sciences,University of Otago, Christchurch, New Zealand; 3Center for Vaccines and Immunity,Columbus Children’s Research Institute, Columbus, Ohio; 4AW Morrow Gastroenter-ology and Liver Centre, Centenary Institute of Cancer Medicine and Cell Biology, RoyalPrince Alfred Hospital and University of Sydney, Sydney, Australia.

Received February 19, 2006; accepted August 8, 2006.Supported by a program grant of the National Health and Medical Research

Council of Australia (NHMRC) and the Ageing and Alzheimers Research Founda-tion. A.W. and D.G.B. were supported by an NHMRC Postgraduate ResearchScholarship and an NHMRC CJ Martin Fellowship respectively; A.W. was alsosupported by the Ageing and Alzheimer’s Research Foundation.

Address reprint requests to: Dr. Patrick Bertolino, Centenary Institute of CancerMedicine and Cell Biology, Locked Bag No 6, Newtown 2042, NSW Australia. E-mail:p.bertolino@centenary.usyd.edu.au; fax: (612) 95 65 6101.

Copyright © 2006 by the American Association for the Study of Liver Diseases.Published online in Wiley InterScience (www.interscience.wiley.com).DOI 10.1002/hep.21378Potential conflict of interest: Nothing to report.

1182

and CD1d,12-14 a ligand recognized by NKT lympho-cytes.15 These cells are therefore fully competent to inter-act and activate lymphocytes via MHC class I receptors.Experimental studies have demonstrated that these cellsare indeed efficient APCs in vitro.16,17 Support for a roleof hepatocytes as APCs in vivo comes from studies per-formed using transgenic mouse models. When Alb-Kb

and Met-Kb transgenic mice, expressing the allo-MHCH-2Kb molecule on hepatocytes,18,19 were adoptivelytransferred with TCR transgenic T cells specific forH-2Kb, T cell primary activation occurred in the liverindependently of lymphoid tissues, thus supporting thethesis that hepatocytes can interact with circulating T cellsin vivo.6,8 Such results suggest that hepatocytes canpresent antigen to both CD8� T lymphocytes and possi-bly also NKT lymphocytes, and thus play a major role inantigen-specific recruitment of these cells to the liver.However, the question arises as to how circulating naıve Tlymphocytes could interact with hepatocytes, given thepresence of an endothelial barrier. This question remainscontroversial, as contact between hepatocytes and circu-lating lymphocytes do not normally occur.20

The unique morphology of the hepatic sinusoid andliver sinusoidal endothelial cells (LSECs) provides a po-tential mechanism for this unique interaction between Tlymphocytes and the liver. Hepatic sinusoids are porous,gossamer-like, cylindrical structures that are slightly nar-rower than blood cells. They connect afferent portal triadsto exiting central hepatic venules, a distance of approxi-mately 1 mm. Approximately one billion sinusoids formthe rich capillary network of the liver, which permits thevast hepatic blood flow to course slowly and intimatelybetween the cords of hepatocytes. Because the sinusoidallumen is narrower than lymphocytes, lymphocytes havebeen suggested to “massage” the LSECs during their tran-sit, thus facilitating lymph circulation in the space ofDisse.21 LSECs occupy a strategic position in the hepaticsinusoid. They are highly specialized endothelial cells thatline the wall of the hepatic sinusoid and separate the si-nusoidal blood, derived primarily from the portal vein,from hepatocytes. LSECs are perforated by fenestrations,which are pores approximately 100 nm in diametergrouped together in clusters known as liver sieve platesand occupying approximately 5% to 10% of the endothe-lial surface.22 Fenestrations are true discontinuities in theendothelium, lacking either a diaphragm or underlyingbasal lamina.

T lymphocytes could interact with hepatocytes by sev-eral plausible mechanisms (Fig. 1). First, fenestrationscould provide a portal for interaction between hepato-cytes and lymphocytes. As a first possibility, extensions oflymphocytes might pass through fenestrations and con-

tact the hepatocellular membrane in the extracellularspace of Disse (Fig. 1A). Alternatively, we have previouslyshown in scanning electron micrographs that cytoplasmicextensions of the hepatocytes (microvilli) extend into thesinusoidal lumen through fenestrations in LSECs.23,24

Possibly these hepatic microvilli contact circulating lym-phocytes within the vasculature (Fig. 1B). Finally, con-tact between lymphocytes and hepatocytes could occuracross gaps between LSECs (Fig. 1C), although thepresence of such gaps is controversial because it mightbe artifactual.

Here we used electron microscopy to investigate theintimate interactions between lymphocytes, LSECs, andhepatocytes. Initially we focused on interactions betweenintrahepatic lymphocytes (IHL) and hepatocytes in thenormal liver. However, we also investigated whether naıveT lymphocytes interact with hepatocytes using theMet-Kb and Alb-Kb transgenic mouse models in whichtheir cognate antigen is expressed on hepatocytes. Immu-nogold labeling was performed to identify donor naıve Tlymphocytes and to differentiate them from recipient in-trahepatic T lymphocytes.

Materials and Methods

Mice. All mice were maintained on a B10.BR (H-2k)background in the Centenary Institute animal facility un-der SPF conditions. B10.BR and C57BL/6 mice werepurchased from the Animal Resources Centre (Perth,Australia). Met-Kb and Alb–Kb transgenic mice express-ing H-2Kb on hepatocytes under the control of the sheepmetallothionein19 or the mouse albumin promoters,18 re-spectively, were kindly provided by Drs Grant Morahanand Jacques Miller (WEHI, Australia) and Dr Bernd Ar-nold (DKZ, Germany). Des-TCR, which express anH-2Kb–specific TCR,25 were a gift of Dr. Bernd Arnold.Des-TCR RAG-1�/� mice were generated by backcross-ing Des-TCR mice with H–2k RAG-1�/� as previouslydescribed.8 All experimental procedures were approved bythe University of Sydney Animal Ethics Committee.

Fig. 1. Different possibilities for lymphocyte–hepatocyte interactions:(A) Through LSEC fenetrations via lymphocyte cytoplasmic extensionspenetrating the space of Disse and contacting hepatocyte microvilli in thespace of Disse. (B) Through fenestrations via hepatocyte microvilli pro-truding into the sinusoidal lumen. (C) Through gaps between two LSECs.

HEPATOLOGY, Vol. 44, No. 5, 2006 WARREN ET AL. 1183

Adoptive Transfer. Pooled LN cells from Des-TCRRAG-1�/� mice were labeled with CFSE (5-carboxyfluo-rescein diacetate succinimidyl ester), and 5 or 20 millioncells were adoptively transferred into recipient transgenicmice as previously described.6

Electron Microscopy. Five and thirty minutes afterthe lymphocytes were injected into the portal vein, micewere culled by CO2 narcosis, and livers were immediatelyperfusion-fixed through the portal vein with a 23-gaugeneedle at a rate of 3 to 4 mL/min with phosphate-bufferedsaline and fixative [1% glutaraldehyde, 4% paraformalde-hyde, 2 mmol/L CaCl2, 2%(w/v) sucrose 0.1 mol/L ca-codylate buffer pH 7.4]. The livers were removed and cutinto small pieces before being fixed in the same fixative for1 hour at 4°C. Fixed tissue was embedded in Spurr’s andLR White resin. Liver blocks (at least four per mouse)were analyzed by electron microscopy. For scanning elec-tron microscopy (SEM), fixed tissue was osmicated, de-hydrated, and incubated in hexamethyl-disilazane. Tissuewas then mounted on stubs, splutter-coated with gold.and examined using a Cambridge S360 Scanning Micro-scope.

Immunogold and Immunohistochemistry. Ultra-thin (70-90 nm) sections were cut and collected on form-var-coated nickel slot grids (ProSciTech, Australia). Goldpost-embedding immunocytochemistry was then per-formed to detect LFA-1 or CFSE-labeled cells. Grids wereincubated overnight at 4°C with rabbit anti fluorescein/Oregon green primary antibody (Invitrogen, Australia)and for 2 hours at 20°C with gold-conjugated (15 nm or10 nm, ProSciTech, Australia) secondary antibody. Sec-tions were analyzed for gold-positive lymphocytes, andfields of interest were photographed using a HitachiH7100FA Transmission Microscope. Part of the fixedtissue was paraffin-embedded, and immunohistochemis-try was performed using the same primary antibody. Us-ing a light microscope (magnification 200�), positive

cells were counted on 40 mm2 section area for each sam-ple. Immunohistochemistry detecting ICAM-1 andH-2Kb expression was performed on frozen liver sectionspost-fixation in cold acetone (15 minutes) of recipientmice. Primary antibodies, mouse anti-H-2Kb biotinylated(BD Pharmingen, San Diego, CA) and rat anti-mouseICAM-1 (Chemicon, Australia) were applied to sectionsfor 2 hours. ICAM-1 sections were incubated with anti-rat fluorescein isothiocyanate (FITC) antibody (CaltagLabs, Carlsbad, CA) or biotinylated anti-ICAM-1 anti-body (Sigma Aldrich, Australia) for 1 hour. A solution ofstreptavidin-horseradish peroxidase (HRPO) conjugate(Dako, Australia) was then applied, and immunolabelingwas revealed using 3,3�-diaminobenzidine (Sigma Al-drich).

Results

IHL Interact With Hepatocytes Through Fenestra-tions in LSECs. Previous data using Met-Kb and Alb-Kb

transgenic mouse models in which antigen expression wasrestricted to liver parenchymal cells suggest that T lym-phocytes can interact with hepatocytes within minutesafter adoptive transfer.6,8 The early onset of these eventsindicates that lymphocytes can access MHC/peptidecomplexes expressed by liver parenchymal cells. To inves-tigate how hepatocytes and lymphocytes interact, we an-alyzed the liver sinusoids using both scanning (SEM) andtransmission electron microscopy (TEM).

Fenestrations are unevenly distributed and clustered toform structures known as sieve plates (Fig. 2A-B). Fenes-trations are complete perforations in the LSEC and there-fore expose the underlying hepatocytes and hepaticstellate cells to the sinusoidal lumen (Fig. 2B-C). Further-more, neither basal lamina nor connective tissue underliethe LSECs.

Fig. 2. Structure of mouse liver sinusoidal endothelial cells. (A) Under scanning electron microscope (SEM), the liver sinusoidal endothelial cells(LSEC) present fenestrations organized in sieve plates (Original magnification 15,000�). (B) Higher magnification SEM picture of a liver sieve in LSECillustrating the lack of basal lamina between hepatocytes and LSEC. Hepatic microvilli can be noticed under the LSEC layer (Original magnification20,000�). (C) In ultra-thin sections, the sinusoid is delimited by endothelial cell thin cytoplasmic processes [e] which contain fenestrations (smallarrows). They appear as open connections between the sinusoidal lumen [S] and the space of Disse [sD]. In the space of Disse between the LSECand the hepatocytes, hepatic stellate cells [HSC, marked with an asterix] processes may be observed. Original magnification 7,000�.

1184 WARREN ET AL. HEPATOLOGY, November 2006

Analysis of scanning electron micrographs demon-strated that a large proportion of the IHL (47%) were indirect contact with the luminal surface of LSEC. Some ofthese lymphocytes were squeezed by the narrow wall ofthe sinusoids, favoring intimate contact between the twocells (Fig. 3A-B). All lymphocytes demonstrated numer-ous cytoplasmic extensions with diameters similar to fen-estrations (106.87 � 11.25nm; Fig. 3B-C). Supportingthe interaction model displayed in Fig. 1A, such lympho-cyte cytoplasmic extensions were seen within the lumen ofthe fenestrations, extending into the space of Disse and incontact with hepatocyte microvilli (Fig. 3D). In total,11% of all IHL found in the lumen (total number oflymphocytes examined, n � 60), were found to interactwith hepatocyte microvilli by this mechanism. Such lym-phocytes were shown to have up to five extensions travers-ing adjacent fenestrations on single sections examined byTEM (Fig. 3D), although the total number per lympho-cyte will obviously be much greater.

Hepatocyte Microvilli Extend Into the SinusoidalLumen Through Fenestrations But Do Not ContactLymphocytes. As suggested in earlier studies,23,26 hepato-cyte microvilli were seen protruding through the fenestra-tions and into the sinusoidal lumen (Fig. 4A). However,on no occasion (out of the 60 IHL analyzed) were these

microvilli seen to be in contact with lymphocytes in thesinusoid. This indicates that this type of interaction (Fig.1B) is rare, or alternatively that such interactions are un-stable and dislodged during the fixation process.

No Gaps Between LSECs. Electron micrographs alsoshowed that although LSECs do not form tight junctions,they do form a relatively continuous endothelial barrier.No gaps were seen between adjacent LSEC on any occa-

Fig. 3. Interactions between intrahepaticlymphocytes (IHL) and liver cells within thesinusoids. (A) Transmission electron micros-copy (TEM) picture of an IHL circulating in thelumen of a liver sinusoid. The narrow diameterof liver sinusoids [S] forces the lymphocyte tosqueeze through the sinusoids and establishintimate contact with LSEC [e] (Original mag-nification 12,000�). [H]: hepatocyte (B) and(C) SEM pictures of IHL, which present villi ontheir surface. (B) IHL in the sinusoidal lumendisplaying numerous cytoplasmic extensionswith diameters similar to fenestrations (Originalmagnification 10,000�). In (C) the IHL appearto contact the surface of liver sinusoidal wallsusing its villi (Original magnification 18,000�)(D) TEM cross-section picture of a lymphocytein a sinusoid with five cytoplasmic extensions[indicated by arrows] contacting the basal sur-face of the underlying hepatocyte (originalmagnification 10,000�)

Fig. 4. Liver sinusoidal endothelial cell (LSECs) are not separated bygaps but allow hepatocytes microvilli to protrude through their fenestra-tions. (A) Hepatocyte microvilli are occasionally observed to protrude intothe sinusoidal lumen [S]. Original magnification 40,000. (B) High mag-nification scanning electron micrograph (SEM) indicates the absence ofgaps between two LSECs (3) (Original magnification 35,000�).

HEPATOLOGY, Vol. 44, No. 5, 2006 WARREN ET AL. 1185

sion, nor were any interactions visualized between lym-phocytes and hepatocytes through gaps between adjacentlymphocytes (Fig. 4B). Therefore, contacts between Tcells and hepatocytes are unlikely to occur betweenLSECs (Fig. 1C).

Naıve T Lymphocyte Interact With HepatocytesThrough Fenestrations in LSECs. As primary activa-tion of antigen-specific T cells by hepatocytes has beenevidenced in vivo6,8; we next investigated whether naıve Tcells also establish interactions with hepatocytes throughLSEC fenestrations. For this purpose, LN T cells fromDes-TCR RAG-1�/� mice, containing a pure and mono-clonal population of naıve CD8� T cells specific forH-2Kb, were labeled with CFSE and injected into controlB10.BR mice or Met-Kb and Alb-Kb transgenic mice ex-pressing H-2Kb on hepatocytes.6,8 Analysis was per-formed at very early stages of T lymphocyte activation (5and 30 minutes after injection of lymphocytes).

By counting CFSE-labeled T cells by immunohisto-chemistry, twice as many labeled donor T lymphocyteswere detected in the livers of Met-Kb compared with thecontrol B10.BR mice (Fig. 5A), a result consistent withflow cytometric and radiolabeling studies6,27; V. Benselerand P. Bertolino unpublished result). Most of the donorT lymphocytes (93%) were found in the sinusoids ratherthan in portal tracts (Fig. 5B).

Naıve T lymphocytes displayed villi similarly to IHL(Fig. 5C) and were recognized by the presence of CFSEdetected using immunogold staining. Naıve T lympho-cytes (2/17 cells in B10BR mice and 8/52 cells in Met-Kb

mice) were observed that had cytoplasmic extensions tra-versing the fenestrations and in direct contact with hepa-tocytes (Fig. 5D-E). However, these were less frequentlyseen compared with the IHL, with only one such interac-tion seen per TEM image of each naıve lymphocyte (com-pared with five for IHL). These experiments suggest thatnaıve T cells are also able to interact with hepatocytesthrough LSEC fenestrations.

Expression of MHC Class I Molecules and ICAM-1on Hepatocytes Is Polarized Toward the SinusoidalLumen. The initiation of primary T cell activation re-quires TCR and LFA-1 cross-linking. TEM immunogoldstaining of LN cells demonstrates that LFA-1 is expressedon the lymphocyte villae and potentially able to interactwith its ligand ICAM-1 (Fig. 6). Both MHC/peptidecomplex and ICAM-1 have been shown to be expressed atrelatively high levels on purified hepatocytes11,28 (Berto-lino P., unpublished results). To investigate whether thesemolecules were physiologically accessible to circulating Tcells, we investigated the in situ localization of bothH-2Kb and ICAM-1 in the liver using light and fluores-cence microscopy. Interestingly, expression of ICAM-1 in

Fig. 5. Naıve T lymphocytes can contact hepatocytes through fenestra-tions. (A) Most naıve T cells injected into Met-Kb and B10.BR animals werelocalized along the sinusoids rather than around the portal tracts. 5-Car-boxyfluorescein diacetate succinimidyl ester (CFSE)-labeled T cells detectedby immunohistochemistry using an anti-fluorescein antibody coupled toHRPO. The histograms show the average number of CFSE� cells in a tissuearea of 1 mm2 by light microscopy in B10.BR and Met-Kb mice at 5 minutesand 30 minutes after T cell transfer. An area of 40mm2 was analyzed for eachliver. Most positive cells (93%) were located in the parenchyma rather thanaround portal tracts (B). (C) SEM micrograph of purified naıve T cells. Asshown for IHL, naıve T cells isolated from the LN of Des-TCR RAG-1�/� micealso display cytoplasmic extensions (Original magnification 8,000�). (D)LR-White embedded ultra-thin TEM section of an immunogold labeled naıveT lymphocyte establishing one TEHLI (indicated by the large arrow). Liver wasperfusion-fixed 5 minutes after adoptive transfer of CFSE� cells (Originalmagnification 30,000�). (E) Another example of an immunogold labeled cellcontacting both LSEC [*] and hepatocyte (Original magnification 12,000�). TheTELHI is indicated by the large arrow. In D and E, 10 nm gold beads, indicatedby small arrows and used to identify the donor T cells, are shown in the enlargeddetail of the section. No gold labeling was seen in liver cells or in control mice notinjected with CFSE-labeled T cells (data not shown).

1186 WARREN ET AL. HEPATOLOGY, November 2006

the liver was not evenly distributed but was strongly con-centrated toward the surface lining the sinusoids (Fig.7A-B). Polarized expression of H-2Kb toward the basolat-eral surface of the hepatocyte was also evidenced inC57BL/6 mice in which all liver cells express H-2Kb (Fig.7E) but also in Met-Kb and Alb-Kb mice where H-2Kb

expression is restricted to hepatocytes (Fig 7C-D).These results suggest that expression of two of the mol-

ecules required for primary T cell activation, ICAM-1 andMHC Class I, is polarized toward the basolateral surfaceof the hepatocyte, thus maximizing the primary activationsignal that hepatocytes might provide to T cells

To test whether ICAM-1 expressed by liver cells iscritical for the retention of antigen-specific T cells whenH-2Kb is expressed by hepatocytes, Met-Kb and B10. BRmice were injected with CFSE-labeled Des-TCR LN Tcells in the absence or in the presence of Fab fragmentsspecific for ICAM-1 and ICAM-2 that block the interac-tions of these molecules with their common ligandLFA-1. As shown in Fig. 5A, the total number of donorCD8� T cells isolated from the liver of Met-Kb mice at 1hour after T cell transfer was approximately twice thenumber isolated from B10.BR mice. Anti–ICAM-1 andICAM-2 treatment inhibited this antigen-specific reten-tion to control levels observed in B10.BR mice (Fig. 7F).These results suggest that T cell retention mediated byhepatocytes is ICAM-1/LFA-1 dependent.

DiscussionThe electron micrographs presented in this study sup-

port the model of interaction depicted in Fig. 1A. and

Fig. 6. Transmission electron micrograph of immunogold staining ofLFA-1 on the membrane of a representative lymph node cell. Lymph nodecells with a naive morphology isolated from a wild-type C57BL/6 mousewere stained using an anti-LFA-1 antibody and an anti-rat antibodycoupled to gold particles. Gold particles (indicated by arrow heads) weredistributed relatively evenly on the membrane, including on microvilli.Control sections (no primary anti-LFA-1 antibody) did not display anystaining (data not shown).

Fig. 7. ICAM-1 and H-2Kb expression are polarized and localizedpredominantly toward the lumen. (A) ICAM-1 expression in the liver ofB10.BR mice using immunofluorescence. (B) ICAM-1 expression inMet-Kb mice by immunohistochemistry on liver frozen sections. (C-E)H-2Kb expression by immunohistochemistry in Met-Kb mice (C), Alb-Kb

(D), and C57BL/6 mice (E). [S]: sinusoid (Original magnification 400�).(F) Retention of CFSE-labeled Des-TCR T cells in the liver of Met-Kb miceis inhibited by blocking ICAM-1 and ICAM-2. CFSE-labeled LN Des-TCRcells were injected into Met-Kb and B10.BR mice as previously de-scribed27 in the absence or in the presence of 300 �g of blocking Fabfragments specific for ICAM-1 (KAT-1) and ICAM-2 (3C4), provided byProf A. Hamann. Liver retention of donor CD8� T cells was measured at1h after T cell transfer by calculating the total number of CD8� CFSE� Tcells harvested from this organ using flow cytometry as previouslydescribed.8 Histograms represent the mean and standard error of 2 to 4mice per group.

HEPATOLOGY, Vol. 44, No. 5, 2006 WARREN ET AL. 1187

indicate that both IHL and naıve T lymphocytes insertcytoplasmic extensions through fenestrations in the LSECand that such extensions make direct contact with thehepatocellular membranes. On two-dimensional TEM ofultra-thin sections, we observed up to five such extensionson each IHL and one on naıve T lymphocytes. This indi-cates that these interactions are quite common in eachlymphocyte, and they were seen in at least 10% to 15% ofnaıve T lymphocytes in the liver. Our study probablyunderestimates the prevalence of these interactions be-cause TEM is limited in the number of cells that can bestudied and provides only a single section of each lympho-cyte. We propose to term these interactions trans-endo-thelial hepatocyte–lymphocyte interactions (TEHLI).

Cytoplasmic extensions, or pseudopods, on lympho-cytes are present in both naıve and activated T lympho-cytes, and their size is similar to LSEC fenestrations,which provides opportunity for the extensions to passthrough the fenestrations and interact with hepatic mi-crovilli. Recent reports indicate that only 10 peptide/MHC complexes are necessary to reach a maximal Ca2� Tcell response and form a stable synapse,29 suggesting thatthe sizes of the extensions and fenestrations are largeenough to initiate and accommodate nascent immuno-logical synapses. T lymphocyte activation through theTEHLI seems therefore possible. We also have shownthat naıve T lymphocytes establish TEHLI. Althoughthese results do not provide direct evidence that TEHLImediate T cell activation, speculating that naıve T lym-phocytes might be activated by hepatocytes via TEHLI istempting. If this is the case, this provides a pivotal obser-vation underpinning the potential role of hepatocytes asAPCs in vivo. Naıve T lymphocytes established relativelyfew TEHLI, perhaps suggesting that they are more ran-dom or less stable than those established by activated lym-phocytes, where adhesion molecules and integrinsrecognition is likely to facilitate the formation of immu-nological synapses and enhance the opportunity for, andstability of, TEHLI.

The number of TEHLI established by naıve T cells at 5and 30 minutes after T cell transfer in Met-Kb andB10.BR mice was similar, suggesting that naıve T cellsscan the hepatocyte basolateral surface independently ofantigen expression, similarly to dendritic cell/T cell inter-actions.30 Thus serial contacts with peptide/MHC com-plexes may lead to a cumulative activation signal in the Tcell, upregulation and affinity changes of LFA-1 resultingin T cell arrest. We propose that these changes allow ac-tivated T cells to establish more stable TEHLI. Thismodel explains why IHL, known to be mostly activatedcells, displayed more TEHLI than naıve T cells.

Confirming our previous observations, we have shownhere, using both SEM and TEM, the presence of hepaticmicrovilli extending through the fenestrations into thesinusoidal lumen. Although we have not found any clearevidence that these microvilli interact with circulatinglymphocytes (Fig. 1B), we cannot totally exclude thatsuch brief interactions occur and contribute to T cell ac-tivation.

Other factors might facilitate TEHLI. The low bloodflow rate within the liver sinusoids31-33 favors interactionsbetween blood lymphocytes and liver cells.10 This mightallow primary activation to occur in the absence of selec-tins.8,34 In addition, lymphocytes are larger than somesinusoids and therefore can massage the LSEC as theytraverse the liver21 (Fig. 3A). This increases the possibilityof cytoplasmic extensions on the lymphocytes passingthrough fenestrations. The absence of basal lamina andcollagen in the space of Disse also would facilitate accessto the hepatocellular membrane. Finally, we found thatthe expression of ICAM-1 and MHC Class I molecules,which are required for primary T cell activation, is limitedto the basolateral membrane of the hepatocyte, whichwould maximize opportunity for receptor recognitionand immunological synapse.

Many implications follow the discovery of TEHLI. Thisis the first ultrastructural demonstration of any interactionbetween naıve T lymphocytes and parenchymal cells in vivo.The general immunological paradigm is that naıve T lym-phocytes must be activated by professional APCs in LN be-fore they develop the capacity to traverse any endothelialbarrier and invade the surrounding tissue. However, here wehave clearly observed direct contact between naıve T lym-phocytes and hepatocytes in the absence of inflammation,suggesting that the liver is an exception to this paradigm.Contact between hepatocytes and naıve T cells is suggestedby our previous studies using transgenic mouse models. Na-ıve T lymphocytes expressing a transgenic TCR specific forthe allo-MHC class I molecule H-2Kb are activated withinminutes after adoptive transfer in the liver of Met-Kb andAlb-Kb transgenic mice expressing the H-2Kb molecule onhepatocytes. Although Met-Kb mice do display some ectopicexpression of H-2Kb in the LN on some professional APCs,the expression of the antigen in the liver is restricted to hepa-tocytes.8 Similarly, in Alb-Kb mice, H-2Kb expression is de-tected on hepatocytes but not on LSECs and Kupffer cells.8

As the transgenic TCR recognizes the intact H-2Kb class Imolecule rather than a peptide derived from H-2Kb, cross-presentation is excluded in this model. Antigen-specific re-tention and activation of naıve T lymphocytes in the liverwithin minutes of adoptive transfer implies that H-2Kb ex-pressed by hepatocytes is readily available and recognized bynaıve T lymphocytes. We confirmed that the lymphocytes

1188 WARREN ET AL. HEPATOLOGY, November 2006

that were seen on electron microscopy were naıve T lympho-cytes by using immunogold to detect the marker CFSE,which is present only in the naıve T lymphocytes derivedfrom the donor mice.

Clearly, the observation of TEHLI between naıve Tlymphocytes and hepatocytes consolidates the evidencefor the role of hepatocytes as APCs. In vitro experimentshave demonstrated that hepatocytes can act as very effi-cient APCs for high-affinity transgenic T lymphocytesand that this activation is ICAM-1 dependent.16 The ob-servation that antigen-specific T cell activation occurs inboth Met-Kb and Alb-Kb transgenic mice indicates thathepatocytes also can act as efficient APCs in situ. Thisproperty would be favored by the high local expression ofMHC Class I and ICAM-1 on the basolateral surface ofthe hepatocytes. Based on a large number of immunohis-tochemical studies in mice and humans, expression ofMHC molecules on hepatocytes has been generally re-garded as very low.35,36 However, a recent report11 hasshown that these cells express abundant and conforma-tionally stable MHC class I/peptide complexes with sur-face densities that are nearly as high as on splenocytes.Using quantitative flow cytometry techniques and cali-bration standards to adjust for the differences in cell sizeand autofluorescence between hepatocytes and spleno-cytes, it was demonstrated that, on a per cell basis, mousehepatocytes express 7- to 16-fold higher levels of MHCclass I molecules than splenocytes, whereas membranedensities were at least 30% to 75% as high as those esti-mated on splenic lymphocytes.11 Expression of bothICAM-1 and MHC class I is normally polarized. Conse-quently, it is likely that local densities of these moleculesare higher at the basolateral surface of hepatocytes.Freshly isolated hepatocytes would be expected to redis-tribute expression of these molecules on the whole plasmamembrane surface after dissociation, thus decreasingoverall membrane ligand density. Because peptide/MHCand ICAM-1 density influences T cell activation,37 wetherefore predict that in vitro experiments using hepato-cytes underestimate the APC capability of these cells un-der physiological conditions.

Hepatocytes may act as APCs only in situations inwhich they express the relevant antigen, for example, theMet-Kb transgenic mouse model of autoimmune hepatitisand the early phase of viral hepatitis. Intrahepatic T lym-phocyte activation during early hepatitis C infectioncould contribute to the impaired immune response ob-served in chronic hepattis C virus.9,38 Using the Met-Kb

and Alb-Kb models, we have demonstrated that T lym-phocyte activation in the liver is an inefficient process,leading to diminished cytotoxic activity and reduced cellsurvival,8 a type of response consistent with immune tol-

erance. Based on these results, we propose that TEHLIplay a critical role in mediating activation of naıve T lym-phocytes by hepatocytes and in the development of toler-ance. By transplanting mouse livers expressing theovalbumin antigen into C57BL/6 mice, a recent study39

has confirmed our initial finding that antigen-specific na-ıve transgenic T cells are activated intrahepatically. How-ever, in their model, T cells activated in the liver matureinto full effector cells rather than being deleted, thus con-tradicting the concept that primary activation in the liverleads to tolerance. Whether these discrepancies reflect dif-ferences between transgenic models or, as we have re-cently proposed, whether inflammation occurring duringtransplantation alters the fate of intrahepatically activatedT cells, is unclear.9

As well as providing insight into the normal immunesystem, our observations might have implications for liverconditions associated with altered LSEC morphology and inparticular those conditions associated with loss of fenestra-tions such as cirrhosis and old age. We have shown that oldage is associated with dramatic reductions in the fenestra-tions of LSECs23,40,41; therefore, the altered immune re-sponses of older people might in part be mechanisticallylinked to reduced opportunity for TEHLI in old age.

In conclusion, the results of this study indicate thatdirect contact occurs between lymphocytes and hepato-cytes via processes we term TEHLI. This report providesultrastructural evidence of interaction between naıve Tlymphocytes and parenchymal cells in vivo. We proposethat activation of naıve T lymphocytes by antigens pre-sented on the hepatocellular membrane occurs viaTEHLI. If they result in T cell activation, TEHLI providea new mechanism that impacts on the immune response;provides a plausible mechanism for altered immune re-sponses in conditions associated with loss of fenestrationssuch as aging; and generates a potential therapeutic targetfor modulating the immune response in viral hepatitisand autoimmune disease.

Acknowledgment: The authors thank Dr Alf Ha-mann for providing the anti-ICAM-1 and anti-ICAM-2Fab fragments, Drs Bard Smedsrod and Randi Olsen fortheir precious help with the immunogold staining ofLFA-1, Drs Bernd Arnold, Grant Morahan, and JacquesMiller for providing us with the Alb-Kb, Des-TCR, andMet-Kb mice, as well as Jenny Kingham and the staff ofthe Centenary Institute animal facility.

References1. Mackay CR, Marston WL, Dudler L. Naive and memory T cells show

distinct pathways of lymphocyte recirculation. J Exp Med 1990;171:801-817.

HEPATOLOGY, Vol. 44, No. 5, 2006 WARREN ET AL. 1189

2. Mackay CR, Marston W, Dudler L. Altered patterns of T cell migrationthrough lymph nodes and skin following antigen challenge. Eur J Immu-nol 1992;22:2205-2210.

3. Salmi M, Jalkanen S. How do lymphocytes know where to go: currentconcepts and enigmas of lymphocyte homing. Adv Immunol 1997;64:139-218.

4. Ando K, Guidotti LG, Cerny A, Ishikawa T, Chisari FV. CTL access totissue antigen is restricted in vivo. J Immunol 1994;153:482-488.

5. Ando K, Guidotti LG, Wirth S, Ishikawa T, Missale G, Moriyama T, et al.Class I-restricted cytotoxic T lymphocytes are directly cytopathic for theirtarget cells in vivo. J Immunol 1994;152:3245-3253.

6. Bertolino P, Bowen DG, McCaughan GW, Fazekas De St Groth B. An-tigen-specific primary activation of CD8� T cells within the liver. J Im-munol 2001;166:5430-5438.

7. Bowen DG, Warren A, Davis T, Hoffmann MW, McCaughan GW, De StGroth BF, et al. Cytokine-dependent bystander hepatitis due to intrahe-patic murine CD8� T-cell activation by bone marrow-derived cells. Gas-troenterology 2002;123:1252-1264.

8. Bowen DG, Zen M, Holz L, Davis T, McCaughan GW, Bertolino P. Thesite of primary T cell activation is a determinant of the balance betweenintrahepatic tolerance and immunity. J Clin. Invest. 2004;114:701-712.

9. Bowen DG, McCaughan G, Bertolino P. Intrahepatic immunity: a tale oftwo sites? Trends Immunol 2005;26:512-517.

10. Bertolino P, McCaughan GW, Bowen DG. Role of primary intrahepatic T-cellactivation in the ’liver tolerance effect’. Immunol Cell Biol 2002;80:84-92.

11. Chen M, Tabaczewski P, Truscott SM, Van Kaer L, Stroynowski I. Hepa-tocytes express abundant surface class I MHC and efficiently use trans-porter associated with antigen processing, tapasin, and low molecularweight polypeptide proteasome subunit components of antigen processingand presentation pathway. J Immunol 2005;175:1047-1055.

12. Bumgardner GL, Matas AJ, Chen S, Cahill D, Cunningham TR, PayneWD, et al. Comparison of in vivo and in vitro immune response to purifiedhepatocytes. Transplantation 1990;49:429-436.

13. Mandal M, Chen XR, Alegre ML, Chiu NM, Chen YH, Castano AR, et al.Tissue distribution, regulation and intracellular localization of murineCD1 molecules. Mol Immunol 1998;35:525-536.

14. Agrati C, Martini F, Nisii C, Oliva A, D’Offizi G, Narciso P, et al. CD1dexpression by hepatocytes is a main restriction element for intrahepaticT-cell recognition. J Biol Regul Homeost Agents 2005;19:41-48.

15. Yu KO, Porcelli SA. The diverse functions of CD1d-restricted NKT cellsand their potential for immunotherapy. Immunol Lett 2005;100:42-55.

16. Bertolino P, Trescol-Biemont MC, Rabourdin-Combe C. Hepatocytesinduce functional activation of naive CD8� T lymphocytes but fail topromote survival. Eur J Immunol 1998;28:221-236.

17. Bertolino P, Trescol-Biemont MC, Thomas J, Fazekas de St Groth B,Pihlgren M, Marvel J, et al. Death by neglect as a deletional mechanism ofperipheral tolerance. Int Immunol 1999;11:1225-1238.

18. Schonrich G, Momburg F, Malissen M, Schmitt-Verhulst AM, MalissenB, Hammerling GJ, et al. Distinct mechanisms of extrathymic T cell tol-erance due to differential expression of self antigen. Int Immunol 1992;4:581-590.

19. Morahan G, Brennan FE, Bhathal PS, Allison J, Cox KO, Miller JFAP.Expression in transgenic mice of class I histocompatibility antigens con-trolled by the metallothionein promoter. Proc Natl Acad Sci U S A 1989;86:3782-3786.

20. Knolle PA, Gerken G. Local control of the immune response in the liver.Immunol Rev 2000;174:21-34.

21. Wisse E, De Zanger RB, Charels K, Van Der Smissen P, McCuskey RS.The liver sieve: considerations concerning the structure and function of

endothelial fenestrae, the sinusoidal wall and the space of Disse. HEPATOL-OGY; 1985;5:683-692.

22. Fraser R, Bosanquet AG, Day WA. Filtration of chylomicrons by the livermay influence cholesterol metabolism and atherosclerosis. Atherosclerosis1978;29:113-123.

23. Le Couteur DG, Cogger VC, Markus AM, Harvey PJ, Yin ZL, AnsselinAD, et al. Pseudocapillarization and associated energy limitation in theaged rat liver. HEPATOLOGY 2001;33:537-543.

24. Cogger VC, Mross PE, Hosie MJ, Ansselin AD, McLean AJ, Le CouteurDG. The effect of acute oxidative stress on the ultrastructure of the per-fused rat liver. Pharmacol Toxicol 2001;89:306-311.

25. Schonrich G, Kalinke U, Momburg F, Malissen M, Scmitt-Verhulst AM,Malissen B, et al. Down-regulation of T cell receptors on self-reactive Tcells as a novel mechanism for extrathymic tolerance induction. Cell 1991;65:293-304.

26. Cogger VC, Mross PE, Hosie MJ, Ansselin AD, McLean AJ, Le CouteurDG. The effect of acute oxidative stress on the ultrastructure of the per-fused rat liver. Pharmacol Toxicol 2001;89:306-311.

27. Bertolino P, Schrage A, Bowen DG, Klugewitz K, Ghani S, Eulenburg K,et al. Early intrahepatic antigen-specific retention of naive CD8� T cells ispredominantly ICAM-1/LFA-1 dependent in mice. HEPATOLOGY 2005;42:1063-1071.

28. Bumgardner GL, Orosz CG. Unusual patterns of alloimmunity evoked byallogeneic liver parenchymal cells. Immunol Rev 2000;174:260-279.

29. Irvine DJ, Purbhoo MA, Krogsgaard M, Davis MM. Direct observation ofligand recognition by T cells. Nature 2002;419:845-849.

30. Delon J, Bercovici N, Raposo G, Liblau R, Trautmann A. Antigen-depen-dent and -independent Ca2� responses triggered in T cells by dendriticcells compared with B cells. J Exp Med 1998;188:1473-1484.

31. MacSween RNM, Scothorne RJ: Developmental anatomy and normalstructure. In: MacSween RNM, Anthony PP, Scheuer PJ, eds. Pathologyof the Liver. New York: Churchill Livingstone, 1979; 1-49.

32. Fraser R, Dobbs BR, Rogers GW. Lipoproteins and the liver sieve: the roleof the fenestrated sinusoidal endothelium in lipoprotein metabolism, ath-erosclerosis, and cirrhosis. HEPATOLOGY 1995;21:863-874.

33. Lalor PF, Shields P, Grant A, Adams DH. Recruitment of lymphocytes tothe human liver. Immunol Cell Biol 2002;80:52-64.

34. Wong J, Johnston B, Lee SS, Bullard DC, Smith CW, Beaudet AL, et al. Aminimal role for selectins in the recruitment of leukocytes into the in-flamed liver microvasculature. J Clin Invest 1997;99:2782-2790.

35. Barbatis C, Woods J, Morton JA, Fleming KA, McMichael A, McGee JO.Immunohistochemical analysis of HLA (A, B, C) antigens in liver diseaseusing a monoclonal antibody. Gut 1981;22:985-991.

36. Momburg F, Koch N, Moller P, Moldenhauer G, Hammerling GJ. In vivoinduction of H-2K/D antigens by recombinant interferon-gamma. EurJ Immunol 1986;16:551-557.

37. Gonzalez PA, Carreno LJ, Coombs D, Mora JE, Palmieri E, Goldstein B,et al. T cell receptor binding kinetics required for T cell activation dependon the density of cognate ligand on the antigen-presenting cell. Proc NatlAcad Sci U S A 2005;102:4824-4829.

38. Willberg C, Barnes E, Klenerman P. HCV immunology-Death and themaiden T cell. Cell Death Differ 2003;10:S39–47.

39. Klein I, Crispe IN. Complete differentiation of CD8� T cells activatedlocally within the transplanted liver. J Exp Med 2006;203:437-447.

40. Le Couteur DG, Fraser R, Cogger VC, McLean AJ. Hepatic pseudocapil-larisation and atherosclerosis in ageing. Lancet 2002;359:1612-1615.

41. Warren A, Bertolino P, Cogger VC, McLean AJ, Fraser R, Couteur DG.Hepatic pseudocapillarization in aged mice. Exp Gerontol 2005;40:807-812.

1190 WARREN ET AL. HEPATOLOGY, November 2006