Abstract
Activation-induced cell death resulting in peripheral deletion of CD8+ T cells is associated with the accumulation of large numbers of apoptotic T cells in the liver. The hypothesis that this accumulation results from the intrahepatic trapping of T cells from the circulating pool predicts that the liver should retain T cells, which subsequently undergo apoptosis. Here we test this prediction. Perfusion of the liver with lymphocyte mixtures showed retention of activated, but neither resting nor apoptosing, T cells. This trapping was selective for CD8+ cells and was mediated primarily by ICAM-1 constitutively expressed on sinusoidal endothelial cells and Kupffer cells. T cells trapped in the liver became apoptotic. The normal liver is therefore a “sink” for activated T cells.
Systemic activation of T cells in a peptide-specific, MHC class I-restricted TCR transgenic mouse resulted in deletion of the activated cells from the lymph nodes (LNs)3 and spleen and an 8- to 10-fold increase in the number of T cells in the liver. The cells accumulating in the liver were TCRlow CD8+, and TCRlow CD8− CD4− (double negative, DN), in cell cycle, and a significant number were also undergoing apoptosis (1). This observation has been confirmed in two other MHC class I-restricted TCR transgenic systems (2, 3) and recently in the control and resolution phases of influenza virus infection of the lung (4). We have proposed that the T cells accumulating in the liver on T cell activation originate from conventional peripheral T cells that are undergoing activation-induced cell death (AICD) (1, 5), in which case the liver would be a major site for the removal of T cells undergoing AICD. An understanding of the phenomenon of hepatic T cell accumulation during T cell activation is important because the liver has a number of unique and enigmatic immunological properties. These include the ability of liver grafts to induce tolerance across MHC disparities (6), the ability of Ag introduced into the portal vein to induce tolerance (7), and the accumulation of T lymphocytes in the liver in a number of systemic autoimmune diseases, including systemic lupus erythematosus and rheumatoid arthritis in humans and diabetes in the BB rat (8, 9, 10).
The normal liver contains a complex population of intrahepatic lymphocytes (IHL). These include T cells with conventional levels of TCR, expressing either CD4 or CD8. However, the majority of IHL are TCRlow, lacking CD4 and CD8, and are in cell cycle as well as undergoing apoptosis (1, 11). In addition to these T lymphocytes, there are also NK cells, NK T (NK-T) cells, and a small population of c-Kit+ cells lacking lineage markers (12, 13, 14). The cells accumulating in the liver on peripheral T cell activation are indistinguishable from the CD8+ and DN T cells usually resident in the liver. The origin of the DN T cell population, in particular, is the subject of much debate. These cells have been proposed to originate by extrathymic maturation, based on the presence of DN cells in athymic (nude) mice and in the livers of adult thymectomized irradiated bone marrow chimeras (15, 16). An extension of this thesis proposes that the accumulation of cells in the liver on peripheral T cell activation is also due to extrathymic maturation. An alternative possibility is that the increase in hepatic T cells is due to intrahepatic priming and proliferation of thymically selected T cells. A number of cell populations in the liver, including Kupffer cells and sinusoidal endothelial cells, express class I MHC and the B7 costimulatory molecules, and there is in vitro evidence of their ability to prime T cell clones in an Ag-specific and MHC-restricted manner (17, 18, 19).
We have proposed that the hepatic accumulation that occurs during AICD is due to the selective retention by the liver of activated CD8+ cells that originate from peripheral lymphoid organs, including the LN. On activation, LN T cells down-regulate L-selectin and leave the LN via the efferent lymphatics, draining into the vascular compartment (20, 21). The hypothesis that accumulation results from the intrahepatic trapping of T cells from the circulating pool predicts that the liver should retain such activated T cells. Here we test whether the normal liver retains T cells flowing through it and determine the importance of T cell activation, the relative susceptibility of CD4+ and CD8+ T cells to trapping, and the fate of trapped T cells inside the liver.
Materials and Methods
Animals
C57BL/6J and C57BL/6J-Icam1tm1Bay mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and housed in a specific pathogen-free environment in conformance with institutional guidelines for animal care.
Lymphocyte activation and isolation
Axillary and inguinal LN cells for organ perfusion were obtained from 6- to 8-wk-old C57BL/6J mice 2 days after i.p. PBS injection (for unactivated cells) or after injection of 100 μg of anti-CD3 Ab (clone 145-2C11; to obtain activated cells). Inguinal and axillary LN were dissected under sterile conditions and mechanically disrupted to obtain a cell suspension.
Organ perfusion and lymphocyte labeling
Six- to eight-week-old C57BL/6J mice were anesthetized with 0.5 mg of pentobarbital, and the abdominal cavity was exposed by a midline incision. One hundred units of heparin was injected into the inferior vena cava, and the portal vein was cannulated with a 24-gauge catheter (Critikon, Tampa, FL). The inferior vena cava was cut below the renal veins, and the liver was perfused with CO2-buffered Click’s medium (Life Technologies, Gaithersburg, MD) supplemented with 0.6 mg/ml l-glutamine, 0.6 mg/ml gentamicin, and 1.5 mg/ml sodium bicarbonate at 37°C at a rate of 3 ml/min. The thorax was exposed by cutting along the midaxillary lines bilaterally, and a polyethylene PE50 catheter (Becton Dickinson, Sparks, MD) was introduced into the upper part of the inferior vena cava via an incision in the left atrium and tied into place. The inferior vena cava was ligated below the liver and immediately above the renal veins, diverting the flow of buffer toward the heart and out of the catheter. LN cells (18 × 106) from littermates were fluorescently labeled with PKH2 (Sigma, St. Louis, MO) according to the manufacturer’s instructions and mixed with 6 × 106 RBC labeled with PKH26. A sample of this input mixture was retained for further staining, and the remainder was injected into the buffer stream entering the portal vein, with collection from the catheter draining the liver for 10 min. Total perfusion time did not exceed 15 min in any experiment. Lack of retention of RBC by the liver was established by perfusion of the liver with a suspension of RBC at 1 × 106/ml for 15 min. The concentration of RBC in the efflux was unchanged, and there was no retention of RBC in the liver as demonstrated by confocal microscopy. This allowed calculation of the proportion of LN cells collected when a mixture of LN cells and RBC was infused into the portal vein: proportion of LN cells collected = [RBC % (input)/LN%(input)] × [LN % (output)/RBC% (output)].
The aorta and vena cava were cannulated below the renal vessels by a polyethylene PE50 catheter (Becton Dickinson), and a 24-gauge catheter (Critikon), respectively. The aorta and vena cava were both tied above the renal vessels, and the kidneys were perfused at 3 ml/min with the same medium used for liver perfusion. The mixtures of cells described above were perfused through the kidneys, and lymphocyte retention was calculated as described above for liver perfusion. Liver and renal perfusion was not conducted in the same animals.
For intraportal injections, the mice were anesthetized with inhaled metofane (Schering-Plough, Union, NJ). A 1.5-cm midline incision was made below the xiphesternum and extended below the left costal margin to the midaxillary line. The small intestines were displaced to expose the portal vein, which was injected with 0.5 × 106 cells in 150 μl of buffer using a 30-gauge needle. Pressure was applied to the injection site for 1 min with a sterile swab before closing the abdominal cavity.
FACS analysis
Confocal microscopy
Whole liver confocal microscopy shown in Fig. 4⇓, A and B, was performed on the livers postperfusion without any additional sectioning or fixation steps.
For visualization of interactions between endogenous Kupffer cells and lymphocytes flowing into the liver a >95% pure population of activated CD8+ T was injected into the portal vein of C57BL/6J mice. The activated CD8+ T cells were obtained from OT-1 mice, which are specific for an OVA peptide (SIINFEKL). LN cells were removed 2 days after i.p. injection of 1 μmol of specific peptide. CD8+ T cells were purified from the LN cells by negative FACS sorting for CD4+ (clone GK1.5), class II-positive (clone AF6-120.1), and IgM-positive cells (clone R6-60.2). The FACS-sorted CD8+ T cells were labeled with the fluorescent dye 5- and 6-carboxyfluorescein diacetate succinimidyl ester (CFSE; maximum emission, 515 nm; Molecular Probes), and 106 cells were injected into the portal veins of 6- to 8-wk-old C57BL/6J mice. The animals were sacrificed at 10 min, 90 min, 3 h, 6 h, 9 h, 14 h, and 16 h after cell injection. The livers were fixed with 1% paraformaldehyde, embedded in OCT compound (Sakura Finetek, Torrance, CA), and frozen in liquid nitrogen; 20-μm sections were cut; and endogenous biotin was blocked using an avidin/biotin blocking kit (Molecular Probes). Kupffer cells were labeled with avidin anti-F4/80 Ab (Serotec, Oxford, U.K.) and biotin-allophycocyanin (APC) (maximum emission, 660 nm; Molecular Probes). A Zeiss Axiovert 100 M (Jena, Germany) and LSM510 software were used to visualize the CFSE-labeled lymphocytes and APC-labeled Kupffer cells. To determine the nature of cell associations, our protocol was initially to identify a CFSE-positive lymphocyte using excitation at 488 nm and a 505- to 550-nm pore sized filter, secondly to establish that the whole cell was within the 20-μm section of tissue by taking serial optical sections at 1.5-mm intervals, and finally to retake optical sections at the same positions using excitation at 633 nm in combination with a 650-nm long pass filter to visualize any Kupffer cells in proximity to the CFSE-positive cell. The two series of images were then superimposed using Adobe Photoshop 4.0 (Adobe Systems, San Jose, CA).
A mitochondria-selective dye (MitoTracker red, Molecular Probes) was used to monitor cell viability. This dye fluoresces at 599 nm in its oxidized state in healthy mitochondria, but if the mitochondrial membrane potential is lost, MitoTracker is converted to a reduced nonfluorescent molecule. Loss of mitochondrial potential is an early event in apoptosis, and the ability of MitoTracker to detect this was shown by double labeling Jurkat cells with CFSE and MitoTracker, and then incubating them with RPMI culture medium alone or with 100 μM etoposide (Sigma). To monitor the intrahepatic viability of activated CD8+ cells, they were dual stained with CFSE and MitoTracker before injection into the portal vein. Imaging of the lymphocytes retained in the liver was performed by confocal microscopy of 20-μm liver sections as described above.
Results
Hepatic accumulation of T cells on systemic T cell activation in normal mice
To test whether hepatic accumulation occurs in mice with a normal TCR repertoire, T cell activation was induced by the i.p. injection of 100 μg of an anti-CD3 Ab (clone 145-2C11). This resulted in distinct changes in T cell numbers in the LN and liver (Fig. 1⇓). In the LN by day 4 there was a significant reduction in the total cell number in anti-CD3-injected animals compared with the PBS-injected controls. The reduction in total T cell number was accompanied by a decrease in the percentage of CD4+ and CD8+ T cells. In the liver by day 4 there was a 5-fold increase in the total IHL number, and this returned to control levels by day 8. The increase in IHL number on days 4 and 6 was associated with a decrease in the percentage of CD4+ T cells, while on days 6 and 8 there was an increase in CD8+ T cells. The CD4−, CD8− IHL were also negative for NK1.1, TCRαβ, and TCRγδ. These results in normal mice are consistent with the previous reports from TCR transgenic mice and show an accumulation of T cells in the liver following systemic T cell activation. The data also suggest a preference for the retention of CD8+ over CD4+ cells.
Injection of anti-CD3 Ab caused peripheral T cell deletion and transient intrahepatic accumulation. C57BL/6J mice were injected i.p. with 100 μg of anti-CD3 Ab (clone 145-2C11) or PBS, and the LN and IHL were isolated and analyzed at various time points. The figure shows LN (A–C) and IHL (D–F) numbers and the percentages of CD4+ and CD8+ cells in PBS-injected controls (open bars) and anti-CD3 injected (filled bars).
No retention of resting LN T cells by normal liver
The liver receives most of its blood supply from the gastrointestinal tract via the portal vein (22), and we devised an experimental system in which this physiological perfusion was reproduced. In an anesthetized mouse, the portal vein and the superior vena cava were both cannulated, and the inferior vena cava was ligated below the hepatic vein, allowing the liver to be perfused with tissue culture medium at a constant flow rate of 3 ml/min, a rate chosen to approximate the normal blood flow (Fig. 2⇓). Some of the buffer, however, flows into tributaries of the portal vein and does not perfuse the liver. A simple comparison of input and output T cell populations would not allow us to correct for this. T cells lost down portal vein tributaries would not be collected in the efflux from the inferior vena cava and would be wrongly interpreted as being retained by the liver. To control for this potential cell loss, RBC were used as an internal standard. Lack of RBC retention was confirmed by two maneuvers. Firstly, after liver perfusion with physiological buffer containing RBC at a concentration of 0.5 million/ml for 10 min, the concentration of RBC in the buffer was unchanged. Secondly, a bolus of fluorescently labeled RBC was added to a stream of buffer perfusing a normal liver. After 10 min of perfusion no RBC were visible in the liver by confocal microscopy. A mixture of resting LN cells and RBC (labeled with the green fluorescent dye PKH2 and the red fluorescent dye PKH26, respectively) was infused into the portal vein buffer stream, and the cells leaving the liver via the hepatic vein were collected from the superior vena cava over 10 min.
Design of experimental system to detect T cell trapping in the liver. The portal vein of an anesthetized and heparinized mouse was cannulated (inflow catheter) as was the inferior vena cava through the heart (outflow catheter). The inferior vena cava below the liver was ligated isolating the liver, which was perfused with a physiological buffer at a constant rate of 3 ml/min. The cell populations to be tested were injected into the buffer stream of the inflow catheter in a volume of 200 μl, and the outflow from the liver was collected over a period of 10 min after injection.
A sample of the RBC and LN cell mixture injected into the portal vein was analyzed by FACS (Fig. 3⇓A) and compared with the effluent population (Fig. 3⇓B).The effluent population contained unstained cells released by the perfused liver (in the lower left corner of Fig. 3⇓B), but the injected RBC and lymphocytes were clearly visible in the upper left and lower right corners, respectively. RBC provided a reference population for the calculation of lymphocyte retention. Perfusion of unstimulated LN cells through a normal liver resulted in very little lymphocyte retention, with 94% (±14%) of the injected cells passing into the vena cava (Fig. 3⇓I).
FACS analysis of fluorescently labeled cell populations before and after perfusion through the liver or the kidney. The input populations were a mixture of PKH26-labeled RBC and PKH2-labeled unactivated (A and E) or anti-CD3 activated (C and G) LN cells. Efflux from the liver (B and D) and kidney (F and H) contained unstained cells washed out of the perfused organ (lower left hand quadrant) in addition to the input RBC and LN cells. In the efflux samples there was reduction in the percentages of both the RBC and LN cells due to the presence of unstained cells. Relative to the RBC, there was no reduction in the percentage of LN cells in the efflux from the liver or kidney when unactivated LN cells are perfused (compare A with B and E with F). On perfusion of activated LN cells into the liver, there was retention of LN cells (compare C with D), but there was minimal retention on perfusion of activated LN cells into the kidney (compare G with H). I, Summary of the percent recovery of LN cells after perfusion through the liver or the kidney.
Retention of activated, but not apoptotic, LN cells by normal liver
In contrast to the recovery of essentially all the resting LN cells, perfusion of activated LN cells into a normal liver resulted in the recovery of only 42% (±18%) of the cells (Fig. 3⇑, C, D, and I). To test whether the selective retention of activated T cells was a specific property of the liver, parallel perfusion experiments were performed in kidneys. The aorta and inferior vena cava were cannulated below the renal vessels, and both vessels were ligated above the renal vessels. As in the case of liver perfusion, there was very little retention of resting LN cells perfused through the kidneys, with recovery of 93% (±7%) of the infused cells (Fig. 3⇑, E, F, and I). In contrast to the liver, there was also very little retention of activated LN cells by the kidneys, with 88% (±4.9%) of the cells recovered in the efflux (Fig. 3⇑, G–I). Hepatic retention of the lymphocytes was confirmed using whole liver confocal microscopy of the perfused livers. Fig. 4⇓, A and B, shows the fluorescently labeled cells trapped in the hepatic sinusoids.
Whole liver confocal microscopy after perfusion with activated LN cells, and FACS histograms of LN cells before and after perfusion through the liver. A, Whole liver confocal microscopy after perfusion with activated LN cells showing multiple fluorescent LN cells (magnification, ×64). B, Whole liver confocal microscopy after perfusion with activated LN cells showing that the LN cells were in the hepatic sinusoids (magnification, ×192). C, FSC profiles of resting input (whole line) and efflux (broken line) LN cells. Resting LN cells have a unimodal size distribution, which was unchanged after perfusion. D, FSC profiles of activated input (whole line) and efflux (broken line) LN cells. Anti-CD3 activated LN cells contained a population of FSChigh T cell blasts that was selectively removed from the efflux population. E, Annexin V staining of activated input (whole line) and output (broken line) LN populations, demonstrating that removal of annexin V-positive (apoptosing) cells was not the mechanism of T cell retention by the liver.
Lymphocyte activation is associated with an increase in size, and this can be detected by an increase in the forward light scatter (FSC) profile. The FSC profile of resting LN cells showed a unimodal distribution (Fig. 4⇑C, whole line), which was unchanged after hepatic perfusion (Fig. 4⇑C, broken line). In contrast, activated LN cells contained a population of cells with high FSC (Fig. 4⇑D, whole line). After hepatic perfusion there was selective loss of these FSC high activated cells (Fig. 4⇑D, broken line).Thus, hepatic retention was selective for the activated cells within the LN population. We tested whether the hepatic retention was selective for apoptosing T cells by identifying T cells in the early stages of apoptosis using annexin V staining. Surprisingly, there was no loss of annexin Vhigh cells from the output population (Fig. 4⇑E), showing that the hepatic retention was not selective for cells undergoing apoptosis.
Retention preferentially affects CD8+ T cells
To test whether there was any specificity for the retention of either T cell subset on hepatic perfusion, the percentages of CD4+ and CD8+ cells in the input and efflux populations were determined by FACS analysis. Fig. 5⇓, A and B, shows the percentages of CD4+ and CD8+ cells of a resting LN cell population before and after liver perfusion, respectively. Consistent with the lack of retention of these cells, there was no difference in the percentages of CD4+ and CD8+ cells between the input and efflux populations. Fig. 5⇓, C and D, shows the corresponding percentages of CD4+ and CD8+ cells in the case of activated LN cells. For activated cells there was retention of both CD4+ and CD8+ cells, but with a strong preference for CD8+ T cells. Thus, the percentage of CD4+ cells was reduced from 34 to 19%, while the percentage of CD8+ cells was reduced from 38 to 5.8%. These experiments demonstrate that on passage of a population of T cells through the normal liver there is selective retention of activated T cells, that this retention is significantly greater for CD8+ than for CD4+ T cells, and that the specificity is not simply for cells undergoing apoptosis. In contrast, perfusion of activated LN cells through the kidneys does not result in a significant degree of retention of any T cell population (Fig. 5⇓, E–H).
FACS plots showing the percentages of CD4+ and CD8+ cells before and after perfusion through the liver or the kidney. A, Expression of CD4 and CD8 on the input resting LN cells. B, Expression of CD4 and CD8 on the resting LN cell population in FACS plot A after hepatic perfusion. There was no change in the CD4+ and CD8+ composition after hepatic perfusion. C, Expression of CD4 and CD8 on activated input LN cells. D, Expression of CD4 and CD8 on activated LN population in FACS plot C after hepatic perfusion. Note the reduction in the percentages of both CD4+ and CD8+ cells, with significantly greater reduction in the CD8+ cells. E, Expression of CD4 and CD8 on input resting LN cells. F, Expression of CD4 and CD8 on resting LN population in FACS plot E after hepatic perfusion. Analogous to the liver, there was no significant alteration in the percentage of CD4+ and CD8+ cells after renal perfusion. G, Expression of CD4 and CD8 on activated input LN cells. H, Expression of CD4 and CD8 on activated LN population in FACS plot G after renal perfusion. In contrast to the changes after hepatic perfusion, there was no significant difference in the percentages of CD4+ and CD8+ cells after renal perfusion of activated LN cells.
ICAM-1-deficient livers have reduced ability to retain activated T cells
The sinusoidal vascular bed of the liver has a combination of features, including the constitutive expression of ICAM-1, which may explain its T cell binding properties (23, 24). ICAM-1 on inflamed postcapillary venules has a well-defined role in maintaining firm adhesion of activated lymphocytes through ICAM-1/LFA-1 interactions (20, 21), but it is not known whether the ICAM-1 constitutively present in the liver has any function in lymphocyte adhesion. Because LFA-1 is up-regulated by T cell activation, we tested whether the ICAM-1 constitutively present in the normal liver had a role in the retention of activated T cells by comparing the livers of ICAM-1-deficient mice (25) with the livers of wild-type mice for their ability to retain activated lymphocytes. The ICAM-1-deficient livers were much less efficient at retaining activated LN cells, with 78 ± 12% of the perfused cells passing thorough the ICAM-1-deficient livers compared with 42 ± 12% for wild-type livers. Fig. 6⇓A shows that the trapping of large activated lymphocytes was particularly compromised, as these were present in the output population from the ICAM-1-deficient, but not from wild-type, mouse livers.
FACS plots comparing the ability of wild-type and ICAM-1-deficient livers to retain activated LN cells. A, Forward light scatter profiles of input cells (whole line, labeled 1), efflux from wild-type livers (broken line, labeled 2), and efflux from ICAM-1-deficient livers (light line, unlabeled). B, LFA-1 levels on activated CD4+ (light line) and CD8+ cells (dark line). C, CD4 and CD8 expression on the efflux cells from ICAM-1-deficient liver (compare with Fig. 5⇑C). D, Summary of CD4+ (striped bar) and CD8+ (filled bar) cell percentages of input population and for efflux populations from wild-type (wt) and ICAM-1-deficient (ICAM-1 KO) livers.
ICAM-1-deficient livers have lost their specificity for CD8+ T cells
The binding of activated T cells to ICAM-1 on hepatic endothelium explains the selectivity of the liver for activated T cells, but does not immediately explain the hepatic selectivity for CD8+ T cells. We examined the expression of the ICAM-1 counter-receptor, LFA-1, on anti-CD3 activated CD8+ T cells (Fig. 6⇑B, thick line) and found that it was ∼3-fold higher than that on anti-CD3-activated CD4+ cells (Fig. 6⇑B, thin line). If the high LFA-1 levels on CD8+ T cells were to account for the selective CD8+ T cell retention, we would expect this selectivity to be lost in ICAM-1-deficient livers. Perfusion of the activated LN cells shown in Fig. 5⇑C through an ICAM-1-deficient liver confirmed this prediction (compare Fig. 5⇑D with Fig. 6⇑C). A summary of the CD4+ and CD8+ frequencies of the input and the effluent cells from wild-type and ICAM-1-deficient livers is shown in Fig. 6⇑D. In the ICAM-1-deficient livers there was a small residual bias toward CD8+ T cell retention. This may have been due to adhesion through vascular adhesion protein-1, which is expressed on liver sinusoids and is known to preferentially mediate CD8+ T cell binding (26, 27).
Retention of CD8+ cells is mediated by both Kupffer cells and sinusoidal endothelium
The above data show the hepatic retention of activated CD8+ cells and identify a major role for hepatic ICAM-1 in this process. Lymphocytes entering the liver traverse a vascular bed with a branched structure, a low blood flow rate, and a very mobile population of Kupffer cells (28). In this complex microenvironment, ICAM-1 is expressed on sinusoidal endothelial cells and Kupffer cells (29). To identify the cellular interactions responsible for the retention of activated T cells, in vivo activated CD8+ T cells were labeled with the fluorescent marker dye CFSE and injected into the portal veins of anesthetized mice. Fig. 7⇓A shows three lymphocytes 10 min after portal vein injection, and two of these lymphocytes are clearly in association with Kupffer cells. Fig. 7⇓, B and C, show further interactions of injected CD8+ T cells with resident Kupffer cells at 90 min and 3 h, respectively. At 10 min after injection, a total of 78 CD8+ T cells entirely within the 20-μm section were examined, and 50 (64%) were in association with Kupffer cells. At time points after 10 min, all the retained lymphocytes were associated with Kupffer cells (Fig. 7⇓D). We conclude that while a substantial fraction of the infused T cells were initially retained in association with Kupffer cells, a significant minority was retained without interacting with Kupffer cells. Other candidate cells are sinusoidal endothelium, hepatocytes, ito cells, and NK-T cells. ICAM-1 is important in hepatic T cell retention, and of the candidate cells, only sinusoidal endothelium is known to express this adhesion molecule. The T cells not interacting with Kupffer cells have therefore probably been retained through an initial interaction with sinusoidal endothelium.
Covisualization of endogenous Kupffer cells (red) and portal vein-injected lymphocytes (green) in thick (20-μm) liver sections by confocal microscopy. CFSE-labeled activated CD8+ T cells (1 × 106) were injected into the portal vein of C57BL/6J mice. Optical sections (1.5 μm) were taken to ensure that the whole lymphocyte was within the tissue section, and then Kupffer cells (red) were visualized using F4/80-avidin, and a secondary biotin-APC. A, Ten minutes after portal vein injection, the majority of CD8+ T cells are in contact with Kupffer cells. By 90 min (B) and later time points (C; 3 h) 100% of the lymphocytes were associated with Kupffer cells. D, Graph showing the percentage of lymphocytes in contact with Kupffer cells at different time points after portal vein injection.
CD8+ T cells apoptosis after liver retention
Dual labeling with CFSE and MitoTracker allows the identification of apoptosing cells, as shown in Fig. 8⇓, A–D. Jurkat cells labeled with CFSE (green) and MitoTracker (red) were incubated with a DNA-damaging agent (etoposide) and visualized on glass coverslips. At 10 min and 6 h after culture MitoTracker signal was visible (Fig. 8⇓, A and B), but was almost completely lost by 8 h (Fig. 8⇓C) and was entirely gone at 10 h (Fig. 8⇓D). Jurkat cells cultured in the absence of etoposide remained MitoTracker positive for >24 h (data not shown). CFSE and MitoTracker double-labeled, activated CD8+ cells are visible in hepatic sinusoids 10 min after injection into the portal vein and were MitoTrackerbright (Fig. 8⇓E). These cells remained MitoTrackerbright at 9 h postinjection (Fig. 8⇓F), but at 14 h postinjection 19% of the CD8+ T cell population has lost the MitoTracker signal. At 18 h the injected CD8+ T cells were no longer visible.
A—D, Loss of MitoTracker in Jurkat cells early in apoptosis. Jurkat cells were double stained with CFSE (green) and MitoTracker (red), cultured with 100 μM etoposide, and visualized on glass coverslips. At 3 h (A) and 6 h (B) Jurkat cells were MitoTracker positive (shown by the yellow color produced by superimposition of red and green). By 8 h (C) the MitoTracker signal was almost completely lost, and it was entirely gone by 10 h (D). Jurkat cells cultured in the absence of etoposide remained MitoTracker positive for >24 h. E—G, Demonstration of CD8+ T cell apoptosis after liver retention using MitoTracker. Activated CD8+ T cells were loaded with CFSE and MitoTracker and injected into the portal vein. At 6 (E) and 9 (F) h the T cells were all MitoTracker positive, but 19% had lost the MitoTracker signal by 14 h (G), and no CFSE or MitoTracker signal was detectable at 16 h.
Discussion
The activation of CD8+ T cells in vivo results in a large expansion in T cell numbers (30, 31). Most of the responding cells subsequently disappear. Although T cell activation and deletion in response to superantigens are accompanied by clear evidence of apoptosing T cells in the lymphoid organs, this is not always the case during peripheral T cell deletion due to the injection of specific antigenic peptides into TCR transgenic mice (1, 32, 33, 34). This raises the question of whether T cells leave lymphoid tissues and undergo apoptosis somewhere else. We found that during peptide-induced peripheral deletion of CD8+ T cells in vivo, apoptotic T cells accumulated in the liver. This has been confirmed in two other TCR transgenic systems (2, 3) and provoked the idea that the liver is a specialized site for the trapping of activated T cells (6), a hypothesis tested and supported in the present study. To determine whether this phenomenon was a property of normal T cells rather than an artifact peculiar to TCR transgenic animals, we induced T cell activation and peripheral deletion in normal mice by anti-CD3 injection. The main features of T cell activation, peripheral deletion, and transient liver lymphocytosis were reduplicated in these normal mice. Furthermore, there was a bias toward CD8+ T cells in the cells trapped in the liver.
The finding of apoptotic T cells in the liver could be explained by several alternative mechanisms. The liver could selectively retain activated T cells, in which case either a specific death signal might be delivered in the liver, or the cells might simply be immobilized in an environment lacking essential survival signals. Alternatively, the liver may remove cells that are already committed to apoptosis on the basis of changes in their cell membranes, such as the loss of membrane polarity with the exposure of phospholipid head-groups characteristic of the inner membrane leaflet in healthy cells. An example is phosphatidylinositol, which is expressed on cells at an early stage of apoptosis and may be engaged by a specific phosphatidylinositol receptor on macrophages, leading to phagocytosis of the apoptotic cell (35).
T cells accumulating in the liver during AICD are very similar to the resident CD8+ and DN IHL in unmanipulated mice (11, 36). The origins and immunological relationships of these IHL are controversial. Although we interpret them as end-stage cells in a continuous process of trapping and destruction of activated peripheral T cells, this is not the only hypothesis. The main alternative is that the IHL arise as a result of local, thymus-independent differentiation. There is some evidence for this view. First, IHL are present, albeit in reduced numbers, in congenitally athymic (nude) mice, and the population regenerates in adult thymectomized, irradiated, bone marrow-reconstituted (AT × BM) chimeras (15, 16). This evidence suggests that at least some IHL are thymus independent, but there is less direct evidence for their intrahepatic development. Circumstantial evidence includes the observation that DN cells appear in the livers of AT × BM chimeras earlier than they appear in the LN, and the detection of recombination-activating gene-1 and recombination-activating gene-2 mRNA by RT-PCR in the livers of these chimeras (but not in normal livers) (16). A c-Kit-expressing precursor cell population has been described in normal livers, with the ability to differentiate into myeloid and erythroid lineages (13), but the relevance of these cells to IHL is uncertain. In summary, while the liver probably contains thymus-independent T cells, it is not certain whether they differentiate there or somewhere else.
These two interpretations of IHL as end-stage thymus-derived T cells or as extrathymic T cells are not mutually incompatible. The liver lymphocytes are a complex mixture of conventional T cells, NK-1.1+ T cells, and NK cells, with minor populations of B cells and TCRγδ cells (4, 14). The exact proportions of cells found in each subset are dependent on the technique used to isolate the cells (34), which could account for some of the discrepancies in the literature. The data can accommodate the interpretation that a subset of IHL are indeed thymus independent, while other cells are derived from conventional peripheral T cells by the trapping of CD8+ cells during T cell deletion. The data in this report do not bear on the question of thymus-independent differentiation, but provide very direct evidence that the normal liver is able to selectively retain activated CD8+ T cells that reach it via the portal vein, the vessel through which 80% of the liver’s blood supply arrives in vivo.
A series of unique features of the hepatic sinusoids result in slow blood flow and may contribute to lymphocyte retention. The blood velocity in the sinusoids is ∼7 times lower than that in postcapillary venules and intermittently stops completely due to occlusion of the sinusoidal lumen by Kupffer cells (28). This, in combination with the tortuous structure of the sinusoids, could allow continuous contact between lymphocytes and hepatic sinusoidal endothelium and render unnecessary lymphocyte rolling, which is an essential prerequisite for T cell binding on postcapillary venule endothelium. In support of this view, functional studies showed little or no role for selectins in neutrophil accumulation in the liver (37). The branching structure of the sinusoids also favors lymphocyte retention, because the obstruction of a sinusoidal space should not result in an increase in pressure upstream of the blockage. The distribution of adhesion molecules in the sinusoids is also unique. Hepatic sinusoids do not stain for CD62P and CD34, which are constitutively expressed on all other endothelium. Unlike other vessels, sinusoids constitutively express abundant ICAM-1 without the need for induction by inflammatory cytokines (29, 38, 39). They also express vascular adhesion protein-1 (VAP-1), which is present on the high endothelial venules of LNs and in the liver (41). In high endothelial venules, VAP-1 mediates selective adhesion of CD8+ T cells.
Although lymphocyte rolling may be unnecessary in the liver, firm adhesion of lymphocytes requires specific molecular interactions. In liver inflammation, ICAM-1 has been shown to mediate neutrophil adhesion in sinusoids (37). However, it was unknown whether the ICAM-1 constitutively expressed on noninflamed sinusoidal endothelium has any functional role. In the present study the difference in the efficiency of T cell trapping between the normal liver and the ICAM-1-deficient liver clearly identifies ICAM-1 as a major component in the retention of activated CD8+ T lymphocytes (Fig. 6⇑A). In addition, the preferential retention of CD8+ rather than CD4+ T cells may also be explained on this basis, because the bias toward CD8+ cell retention is largely lost in ICAM-1-deficient livers (Fig. 6⇑D). We propose that this is due to the higher levels of LFA-1 on activated CD8+ T cells (Fig. 6⇑B) (42). Changes in LFA-1 affinity may have an additional role in the retention of activated T cells by hepatic ICAM-1, but this remains to be studied. There is some preferential retention of activated T cells, even in ICAM-1-deficient livers, and VAP-1 may account for these effects (27, 40).
Sinusoidal endothelium and Kupffer cells both express ICAM-1 and are therefore both candidates for the cell type that immobilizes activated CD8+ cells by an ICAM-1-dependent mechanism. Kupffer cells could also recognize T cells in the early stages of apoptosis, using their phosphatidylinositol receptor, which recognizes phosphatidylinositol head-groups exposed on the surface of cells that are beginning to lose the polarity of their plasma membranes (35). To determine the importance of Kupffer cells in the trapping process, we visualized them by confocal microscopy of thick sections of normal livers perfused with fluorescence-labeled T cells. Two-color confocal imaging of lymphocytes and Kupffer cells 10 min after portal vein injection of the T cells showed that the initial interaction of the ∼60% of retained lymphocytes was with Kupffer cells, and this increased to 100% by 90 min. This strongly suggests that Kupffer cells and sinusoidal endothelial cells are both important in the ICAM-1-dependent trapping of activated CD8+ cells. Kupffer cells are very mobile, and the increase in Kupffer cell-CD8+ T cell association with time may due to the production of chemotactic factors by activated CD8+ cells.
The monitoring of CD8+ T cell viability demonstrates that for at least 9 h after liver retention CD8+ T cells have not undergone apoptosis. By 14 h apoptosis has started in a significant proportion of the retained CD8+ cells, and by 16 h they are no longer visible. The time course in interesting, because apoptosis induced by Fas ligation is relatively fast and is often detectable within 3–5 h. The longer time course observed by us is more compatible with apoptosis induced by TNF-α, by cytokine withdrawal, or by a combination of the two. Identification of the mechanism of CD8+ T cell removal is of interest because during a CD8+ T cell response to liver infection, such as chronic viral hepatitis, prolonging the survival of activated CD8+ T cell may enhance viral eradication. The very direct demonstration of apoptosis using MitoTracker also eliminates concerns that previous data showing the apoptosis of hepatic lymphocytes may have been due to lymphocyte apoptosis during the digestion or purification steps (1).
These data support a general model of the role of the liver in the clearance of activated CD8+ T cells. Activation of these cells during an acute immune response results in multiple cycles of cell division and massive clonal expansion, even in normal mice (30, 31, 41). For the duration of the response, these cells dominate the lymphoid system and are abundant in the blood. However, it is our hypothesis that the cells are subject to continuous clearance. In each circulatory cycle, around 20% of the blood passes through the intestine and via the portal vein to the liver, where activated CD8+ T cells expressing a high level of LFA-1 are selectively retained by ICAM-1-dependent adhesion, primarily to sinusoidal endothelial cells. These CD8+ T cells undergo apoptosis, due either to the action of intrahepatic killing mechanisms or to the lack of survival signals. Under normal circumstances, the apoptotic bodies are endocytosed, probably by Kupffer cells (43, 44). Although this process is rapid and efficient, the expansion and subsequent clearance of CD8+ T cells during systemic virus infection are so massive that transient liver lymphocytosis may be observed.
Why is this mechanism specific for CD8+ T cells? And is there a homologous trapping site for activated CD4+ cells? We have searched through many tissues in a TCR transgenic model in which CD4+ T cells undergo antigenic peptide-driven peripheral T cell deletion, but have to date not identified a specific trapping site (D. P. Metz, W. Z. Mehal, J. Huleatt, and I. N. Crispe, unpublished observations). One possibility is that there is no such specific site because none is needed. The available evidence suggests that clonal expansion of Ag-specific CD4+ cells in response to a priming immunization is much more modest than the massive expansion observed in CD8+ cells (45). These cells may be cleared by endocytic mechanisms intrinsic to lymphoid tissue.
The striking ability of allogeneic liver transplants to induce tolerance across MHC disparities even when the host has been immunologically primed may be an epiphenomenon related to this CD8+ T cell clearance mechanism (8). Liver allografts show an early influx of T cells, but after a time interval these cells disappear, leaving the liver immunologically stable (46). We interpret this as an example of T cell trapping by the process documented here, followed by apoptosis of the trapped cells by mechanisms we have yet to elucidate.
If, as we believe, a specific trapping and deletion mechanism for activated CD8+ T cells exists in liver, the problem arises of the clearance of intrahepatic pathogens. One might predict that there would be difficulty clearing pathogens for which the main effector mechanism is CD8+ cell-mediated cytotoxic killing of pathogen-infected cells. Two examples suggest that this is so. Both malaria parasites and hepatitis C virus are subject to immune defense by CD8+ T cells and yet establish chronic infections in the liver (47, 48, 49). However, the liver is not a universally privileged tissue with respect to CD8+ T cell responses. Infection with hepatitis A virus in humans is always cleared, and infection with hepatitis B is eradicated in the majority of cases. The reason for the effectiveness of these immune responses is not understood, but one component may be that infection with these agents results in the production of antiviral and proinflammatory cytokines such as IFN-γ and IL-18 by hepatic NK and NK-T cells (50, 51). We have already shown that IL-18 potentiates the cytotoxicity of resident liver lymphocytes. As in many other situations (52) inflammation may switch the normally tolerogenic environment of the liver to an environment that favors the delivery of CD8+ T cell immune responses.
This line of reasoning may hold the key to understanding pathogen-associated chronic inflammatory disease of the liver such as chronic active hepatitis in the context of hepatitis B and C viral infections. Inflammation may be the price we pay for switching the local environment away from tolerance and allowing CD8+ T cells the opportunity to attack resident liver pathogens. In this context, immunosuppression may be a two-edged sword, and the way to more sophisticated and effective treatment may be through an understanding of the mechanisms of intrahepatic trapping and destruction of activated CD8+ T cells.
Footnotes
-
↵1 This work was supported by National Institutes of Health Grant 1-RO1-AI37554 and the Liver Center at Yale University. W.Z.M. is the recipient of a Howard Hughes Medical Institute Physician Postdoctoral Fellowship.
-
↵2 Address correspondence and reprint requests to Dr. Wajahat Z. Mehal, Section of Immunobiology, Yale School of Medicine, P.O. Box. 208011, BML 458, New Haven, CT 06520-8011. E-mail address: crispebemail.med.yale.edu
-
↵3 Abbreviations used in this paper: LN, lymph node; AICD, activation-induced cell death; DN, CD4- and CD8-negative T cells, double negative; IHL, intrahepatic lymphocytes; NK-T, NK T cells; FSC, forward scatter; CFSE, 5- and 6-carboxyfluorescein diacetate succinimidyl ester; VAP-1, vascular adhesion protein-1; APC, allophycocyanin.
- Received April 14, 1999.
- Accepted June 28, 1999.
- Copyright © 1999 by The American Association of Immunologists
References
In this issue
