IFN-γ is an Absolute Requirement for Spontaneous Acceptance of Liver Allografts


* Corresponding author: Philip F. Halloran, phil.halloran@ ualberta.ca


Experimental liver allografts undergo spontaneous acceptance despite undergoing rejection during the first few weeks post transplant. We explored the role of interferon-γ (IFN-γ) in the spontaneous acceptance of mouse liver allografts. Strain of mouse (CBA) liver allografts transplanted into normal BALB/c mice developed histologic changes typical of rejection that spontaneously regressed, permitting long-term survival of these allografts similar to that of syngeneic grafts. In contrast, CBA liver allografts in IFN-γ-deficient hosts manifested not only infiltration but also hemorrhage and necrosis, with no survival beyond 14 days. Despite differences in survival, local expression of cytotoxic T-cell genes in the transplant was not increased in IFN-γ-deficient hosts, but livers in interferon-γ-deficient mice (GKO) hosts displayed much less induction of major histocompatibility complex (MHC) class I and II expression. To determine whether the difference in survival was secondary to the direct effects of IFN-γ on the liver, we transplanted livers from IFN-γ-receptor-deficient mice into normal hosts. Liver allografts lacking IFN-γ receptors also developed hemorrhage and necrosis with minimal induction of MHC expression. Thus IFN-γ mediates a direct effect on rejecting liver allografts that reduces hemorrhage and necrosis, induces MHC expression, and is absolutely required for spontaneous acceptance.


strain of mouse


interferon-γ-deficient mice


major histocompatibility complex


natural killer


interferon-γ-receptor deficient




wild type


radiolabeled, antibody-binding assay


hypoxanthine phosphoribosyltransferase


Fas ligand




Spontaneous acceptance of major histocompatibility complex (MHC)-disparate liver transplants in the absence of immunosuppressive therapy is a striking observation in many experimental liver transplant models (1–3). Furthermore, spontaneously accepted liver allografts may induce acceptance of other organs including heart or kidney allografts without the need for immunosuppression (2,4). The mechanisms mediating spontaneous acceptance of livers and the ability of liver transplants to modify rejection of other transplants has implications for understanding the mechanisms regulating graft rejection and host-graft adaptation in general. Among the unique features of the liver as a transplant is its mass in relationship to a finite immune response. Spontaneous acceptance of other organs can occur when multiple organs are transplanted into one host (5). However, other features of the liver probably play a role, including its remarkable capability to regenerate after loss of mass, the properties of the portal circulation, and possible tolerogenic properties of liver antigen presenting cells (6).

One factor that could influence spontaneous acceptance is interferon-γ (IFN-γ), which acts via a specific receptor (IFN-γR) (7) to exert numerous effects in transplantation. These include direct effects on the graft such as induction of donor MHC expression (8), direct effects on the host alloimmune response including limitation of host CTL generation (9,10), and induction of systemic MHC expression in the host (11,12). In the early post transplant period, a major effect of IFN-γ in rejecting heart (9,10,13), and kidney transplants (14) is to protect against necrosis by an unknown mechanism, despite inducing high MHC expression (9). Thus kidney transplants lacking IFN-γ receptors develop necrosis and congestion, indicating that IFN-γ acting directly on the receptors of the allograft protects against failure of the microcirculation during acute rejection (15). Transcription factor IRF-1 is necessary for this protective effect (16). In addition, IFN-γ is necessary for initiating rejection of MHC class II-disparate skin allografts (17), perhaps because it must induce donor MHC class II for rejection to occur. The absence of IFN-γ accelerates heart graft rejection in mice lacking IL-2 (18), and a suppressor role for IFN-γ has been demonstrated in rejection of rat heart transplants (19).

How these effects of IFN-γ would operate in liver transplantation is unknown, and of considerable interest given the unique features of the liver. In the present studies we examined the systemic and local effects of IFN-γ during the spontaneous acceptance of liver allografts, using hosts deficient in IFN-γ production or grafts lacking IFN-γRs. Our results indicate that IFN-γ is absolutely required for spontaneous acceptance and for MHC induction in liver transplantation, acting at least in part by a direct effect on the liver itself.

Materials and Methods


BALB/c (H-2d), CBA/J (H-2 k), and 129/SvJ mice (H-2b) were purchased from Jackson Laboratories, Bar Harbor, ME. IFN-γ-deficient mice were generated as described previously (20). Briefly, a normal allele in mouse embryonic cells was replaced with a defective gene using a targeted vector which introduced a termination codon after the 30 first amino acids of the mature IFN-γ protein. These stem cells were used to construct mice heterozygous for the disrupted gene, which were intercrossed and the progeny were selected for homozygosity. Breeding pairs of IFN-γ deficient mice with disrupted IFN-γ genes (GKO mice, H-2d on a BALB/c background) were a generous gift from Dr Tim Stewart (Genentech Inc South San Franscisco, CA) and a colony is maintained at HSLAS, University of Alberta, Edmonton, Canada. The original interferon-γ-receptor deficient (GRKO) (129/Sv/Ev, H-2b) mouse strain with disrupted IFN-γR α-chain genes was generated by gene targeting in murine embryonic stem cells (21). The gene was disrupted by inserting the neomycin-resistance gene into exon V, which encodes an extracellular domain. Homozygous 129/Sv/Ev mice were kindly provided to us through Dr Michael Aguet (University of Zurich, Zurich, Switzerland) and a breeding colony is housed in the HSLAS facility at the University of Alberta, Edmonton, Canada. All procedures were undertaken in accordance with Animal Care Protocols from HSLAS at the University of Alberta in accordance with the Canadian Committee on Animal Care.

Orthotopic liver transplants

Livers from male mouse donors were isolated and transplanted into male mice weighing 25–30 g according to a newly developed model of murine orthotopic liver transplantation with hepatic arterialization (Zhu et al in preparation). IFN-γ knockout mice (GKO, H-2d) and BALB/c (H-2d) were transplanted with CBA/J (H-2 k) livers. The IFN-γ receptor knockout (GRKO, H-2b) and (129/SvJ H-2b) livers were used as donors for CBA/J (H-2 k) recipients. For syngenic transplants, BALB/c livers were transplanted into BALB/c, and interferon-γ-deficient mice (GKO) livers were transplanted into GKO. Briefly, under isofluorane anesthesia, a transverse abdominal incision was followed by dissection and division of the portal and supra-hepatic veins, hepatic artery, and common bile duct, and removal of the liver. The host mice were similarly anesthetized and the donor liver implanted orthotopically with reanastomosis of the portal and supra-hepatic veins and hepatic artery. The common bile duct was reconstructed over a polyethylene stent. No immunosuppressive therapy was administered at any time during the experiments. Mice receiving liver grafts were monitored daily for evidence of graft failure (jaundice, decreased activity, and weight loss). Graft loss as a result of rejection was confirmed by histologic examination of tissues obtained at autopsy. Spontaneous acceptance was defined as liver allograft survival at 100 days following transplantation in the absence of immunosuppressive drug therapy. Statistical analysis of spontaneous acceptance rates in each of the strain combinations was performed using Kaplan-Meier statistical analysis (SPSS, Chicago, IL).

Evaluation of donor-specific tolerance

Mice with liver allografts surviving > 100 days were transplanted with syngeneic (BALB/c, H-2d), donor-specific allogeneic (CBA/J, H-2 k), and third-party allogeneic (129/SvJ, H-2b) skin grafts at the same time (22). Briefly, a full-thickness graft bed was prepared by surgically excising a 0.5 × 0.5-cm section of skin from the dorsum of each skin-graft recipient and removing all subcutaneous tissue to the deep fascia. A similar full-thickness skin graft was harvested from the dorsum of each skin-graft donor, placed on the graft bed, sutured in place with four 8–0 prolene sutures, and secured with gauze and adhesive tape to prevent shearing. The protective covering was removed at day 3 to allow monitoring for evidence of rejection. Grafts were scored as rejected when viable tissue could no longer be detected. Rejection or acceptance of skin grafts was confirmed by excision and histologic examination.


At autopsy, fresh tissue samples were fixed in 10% buffered formaldehyde, paraffin-embedded, sectioned, and stained with hematoxylin and eosin. Pathological assessment of the liver sections was conducted by a pathologist (L.J.) who was blinded to the strain combinations used in the various experiments. Severity of liver allograft rejection was determined using the Banff grading system (23), which assigns cellular infiltration of portal triads, bile duct inflammation, and endothelial inflammation a score of 1–3: 1, minimal changes; 2, moderate changes; or 3, severe changes. Additional findings not included in the Banff scoring system were also studied. The extent of necrosis was scored as the percentage of parenchymal involvement (0, no necrosis present; 1, < 25% of the total parenchyma involved; 2, 25–50% of total parenchyma involved; 3, 50–75% of the total parenchyma involved and 4, > 75% of total parencyma involved). Mann–Whitney statistical analysis was used to compare the severity of histologic lesions seen in the strain of mouse (CBA) grafts transplanted into the WT mice vs. the GKO mice.


Hybridoma cell lines producing mouse monoclonal antibodies (mAb) 34–4-20S (anti H-2Dd), 25–9-17SII (anti I-Ad), 11–4.1 (Kk), 11– (I-Ak), 20–8-4S (KbDb) and AF6-120.1.2 (I-Ab), and rat monoclonals M1/ (anti H-2 antigens, all haplotypes) and M5/114.15.2 (anti I-Ab,d,q and I-Ed,k) were obtained from the American Tissue Culture Collection (Manassas, VA). Supernatants containing mouse monoclonals were purified by protein A chromatography. The supernatants from M1/ and M5/114.15.2 cell lines were precipitated using ammonium sulfate, purified through a DE52 anion exchanger column (Whatman, Hillsboro, OR), and concentrated by Amicon ultrafiltration (Beverly, MA). The protein concentration was determined by a modified Lowry method, adjusted to 1 mg/mL and kept frozen at −70 °C.

Radiolabeled antibody binding assay

A radiolabeled antibody-binding assay (RABA) was used to quantify MHC class I and II expression using allospecific mouse mAbs against MHC class I and II of H-2d (WT and GKO BALB/c), H-2 k (CBA/J), and H-2b (WT and GRKO 129/SvJ) haplotypes. This technique has been previously reported (24) and its semiquantitative characteristics have been previously described (25,26). Monoclonals anti-H-2Dd, anti-I-Ad, anti Kk, and anti-I-Ak mAb, anti-H-2KbDb mAb, and anti-I-Ab mAb were radiolabeled with [125I] iodide using the Iodogen method (Pierce, Rockford, IL) (27). Briefly, liver graft tissue obtained at various times following transplant was frozen in liquid nitrogen and stored at −70 °C. Liver tissue of individual mice was homogenized in 1 mL of PBS, washed in 10 mL of PBS, and centrifuged at 3000 r.p.m. for 20 min. The pellets were suspended in PBS at a concentration of 10 mg/mL. Liver tissue (2.5 mg) was centrifuged and resuspended in 100 µL of radiolabeled mAb (100 000 c.p.m.) and incubated on ice with agitation for 60 min. One milliliter of PBS was added to all the tubes and spun at 3000 r.p.m. for 20 min. The pellets were counted in a gamma counter and the nonspecific binding of a control negative tissue was subtracted. The results are expressed as specific c.p.m. bound by the tissue homogenate after subtracting the background c.p.m. absorbed by the negative control tissue. Based on standard curves, a two-fold change in specific c.p.m. bound in this assay corresponds to approximately a three-fold change in antigen output (28). However the results of such assays must be considered semiquantitative (29).

Indirect immunoperoxidase staining

The pattern of MHC expression was determined by indirect immunoperoxidase staining (26). Briefly, 4-µm-thick sections were cut from frozen liver allograft samples embedded in Tissue-Tek OCT (Skura Finetek, Torrance, CA) and mounted on poly l-lysine-coated glass microscope slides, fixed in acetone, and then incubated with normal goat serum. The slides were then incubated with rat antimouse MHC class I (M1), class II (M5), mouse MHC class I H-2D (host) and H-2K (donor), and class II I-Ad and I-Ak monoclonal antibodies or phosphate-buffered saline (PBS) as a control. The slides were then incubated with affinity-purified peroxidase-labeled goat anti rat or goat anti mouse IgG F (ab′)2 fragment (ICN, Costa Mesa, CA). The slides were then incubated with 3′3 diaminobenzidine tetrahydrochloride and hydrogen peroxide for a color reaction and counterstained with hematoxylin. Previous studies in our laboratory have not shown differences between staining obtained with PBS and isotype controls, both of which are negative when using rat anti mouse monoclonals M1 and M5. MHC staining was also confirmed by host-specific and donor-specific mouse monoclonals class I H-2D (host) and H-2K (donor), class II I-Ad and I-Ak, and goat anti mouse second antibody.

Assessment of gene expression

Granzyme A, granzyme B, perforin, Fas ligand (FasL), and hypoxanthine phosphoribosyltransferase (HPRT) expression were assessed by TaqMan real-time PCR. Total RNA was extracted from individual liver graft tissue obtained at various times post transplant. Briefly, liver tissue, stored in liquid nitrogen at −70 °C, was homogenized in 4 m guanidinium isothiocyanate, and the RNA was pelleted through a 5.7-m CsCl2 cushion. RNA was transcribed into cDNA and amplified in ABI Prism 7700 sequence detection system the using sequence-specific primers and probes listed in Table 1. The cDNA was amplified in a multiplex system using HPRT as the control gene. The data was analyzed using the sequence detector software (Applied Biosystems, Foster City, CA.)

Table 1.  Sequence of real-time polymerase chain reaction primers
Genes Primer sequence


Spontaneous acceptance of liver transplants and the role of IFN-γ

We transplanted syngeneic and allogeneic livers into wild type (WT) and GKO hosts. Long-term survival of syngeneic BALB/c livers transplanted into BALB/c mice (n = 20) was 45% at 100 days post transplant (Figure 1). The majority of graft loss in these syngeneic transplants was the result of common bile duct complications. Thus during the early post transplant period, two mice died of bile duct leaks (day 5 and day 11), and a third mouse died at day 12 from multiple liver abscesses. Later in the post transplant period, most grafts were lost because of biliary duct obstruction secondary to ischemia resulting in biliary duct stenosis. When allogeneic strain of mouse (CBA) livers were transplanted into WT mice (n = 33) the survival rate was 42.4% at 100 days post transplant, which was not statistically different from syngeneic transplants (p = 0.74). Thus in the WT hosts few if any grafts were lost to rejection, beyond the rate expected from technical difficulties as observed in the control syngeneic transplants.

Figure 1.

Survival of CBA/J (H-2 k) liver allografts in WT (BALB/c, H-2d) and GKO (H-2d) hosts compared with syngeneic grafts. Syngeneic [BALB/c into BALB/c] grafts survived > 100 days in 45.0% of mice (n = 20). The majority of graft loss in syngeneic transplants was the result of common bile duct complications. Survival of liver allograft wild-type hosts (CBA into BALB/c) was 42.2% (n = 33) at 100 days post transplantation which was not statistically different from the survival rate observed in the BALB/c syngeneic transplants (p = 0.71). There was no spontaneous acceptance of liver allografts in GKO hosts [CBA into BALB/c IFN-γ–/–] (n = 15); no graft survived beyond 14 days post transplantation. All control syngeneic GKO into GKO (n = 4) liver transplants survived beyond 100 days post transplant. GKO = interferon-γ-deficient mice; CBA = strain of mouse; WT = wild type.

In contrast, no CBA liver allograft transplanted into a GKO mouse survived longer than 14 days (n = 15) (Figure 1, p < 0.0001; CBA allografts into WT vs. GKO mice). Most of these failures occurred between day 7 and day 14. To determine if the loss of liver allografts in the GKO hosts was immunologic, we transplanted syngeneic GKO livers into GKO hosts. All syngeneic GKO grafts survived beyond 100 days (n = 4).

We transplanted syngeneic (BALB/c), donor-specific (CBA), and third-party (129/SvJ) skin grafts onto hosts with spontaneously accepted liver allografts to evaluate the specificity of liver allograft tolerance. Both the syngeneic (BALB/c, H-2d) and donor-specific allogeneic (CBA/J, H-2 k) skin grafts were accepted and survived long-term (> 100 days) (n = 7), whereas the allogeneic third-party (129/SvJ, H-2b) skin grafts were rejected by all the mice that had spontaneously accepted liver allografts.

Histologic analysis of rejecting liver allografts

Using a modification of the Banff grading system for liver allograft rejection, we observed higher scores for inflammation in grafts harvested from the GKO mice (H-2d) compared with the WT mice (BALB/c, H-2d) (Table 2) (Figure 2). At days 5, 7, and 10 post transplant, CBA (H-2 k) liver allografts in the GKO mice manifested greater infiltration of portal triads, inflammation of veins, and invasion of bile duct epithelium by lymphocytes. There was no hemorrhage or parenchymal necrosis in the WT recipients (Figure 2A,B) but there was extensive necrosis with some congestion and hemorrhage in livers being rejected by the GKO hosts (Figure 2C,D). The parenchymal necrosis was concentrated in zones 2 and 3 of the hepatic acinus, with sparing of zone 1, the peri-portal areas. This pattern indicates ischemic necrosis.

Table 2.  Pathology of liver transplants in WT (BALB/c H-2d) and GKO (H-2d) hosts
CBA into WTCBA into GKO 
pathologyDay 5Day 7Day 10Day 5Day 7Day 10p-value
  1. Significant difference between WT and GKO was calculated at day 10 using Mann–Whitney test.

  2. Five mice in each group.

  3. WT = wild type; GKO = interferon-γ-deficient mice; CBA = strain of mouse.

Portal triad inflammation1.2 ± 0.22.2 ± 0.42.0 ± 0.32.4 ± 0.23.0 ± 0.03.0 ± 0.00.02
Venous inflammation1.4 ± 0.22.2 ± 0.41.4 ± 0.22.8 ± 0.23.0 ± 0.02.8 ± 0.40.01
Bile duct inflammation1.0 ± 0.02.0 ± 1.61.6 ± 0.22.2 ± 0.22.0 ± 0.32.4 ± 0.20.06
Necrosis0003.2 ± 0.43.2 ± 0.23.6 ± 0.20.005
Figure 2.

Pathology of liver allografts in WT and IFN-γ-deficient mice. (A,B) CBA/J (H-2 k) liver allografts in WT mice (BALB/c, H-2d) at day 10 post transplant (A, ×100, B, ×200). (C,D) Liver allografts into IFN-γ-deficient mice (GKO, H-2d) at day 10 post transplant (C, ×100, D, ×200). The arrows indicate areas of necrosis with loss of hepatocytes. (E) Liver allografts in WT mice (BALB/c) at > 100 days post transplant showing mild cellular infiltrate with normal hepatic architecture (E, ×200). (F) WT (129/SvJ, H-2b) liver allografts in CBA hosts at day 10, showing acute rejection (F, ×200). (G,H) GRKO (H-2b) liver allografts into CBA hosts at day 10 post transplant, showing infiltrate (G, ×200) and severe necrosis of areas of parenchyma (→; H, ×200). GRKO = interferon-γ-receptor deficient.

Liver allografts in the WT hosts surviving > 100 days following transplantation demonstrated resolution of the inflammation, leaving only foci of mononuclear cells with no necrosis or hemorrhage (Figure 2E). There was no disturbance of the portal triad architecture.

Evaluation of the direct effects of IFN-γ on liver allografts

We studied whether in protecting the liver against necrosis the IFN-γ was acting on the graft or the host. We compared rejecting liver allografts from mice lacking IFN-γ receptors (GRKO, H-2b) with those from wild-type mice with intact IFN-γ receptors (129/SvJ, H-2b) that were transplanted in CBA hosts (H-2 k). If the protective effects of IFN-γ are the result of regulation of the host immune response, then the pattern of acceptance of grafts lacking IFN-γ receptors should be similar to that of WT grafts. If the effect is on the liver transplant itself, then grafts from donors lacking IFN-γ receptors should manifest increased destruction resembling that in GKO hosts.

WT (129/SvJ) allografts demonstrated infiltration typical of acute rejection, with no necrosis at day 10 post transplant (Figure 2F). Interferon-γ-receptor deficient (GRKO) liver allografts at day 10 post transplant demonstrated increased cellular infiltrate with necrosis of the parenchyma (Figure 2G,H). The grading of the pathology is shown in Table 3. Thus the phenotype of rejecting allografts lacking IFN-γ receptors resembled that of allografts transplanted into hosts lacking IFN-γ, with greatly increased necrosis. Unlike the latter, there was less increase in the infiltrating inflammatory cells in portal triads in the GRKO livers compared with the WT livers.

Table 3.  Pathology of liver transplants in 129 (H-2b) and GRKO (H-2b) hosts at day 10
Pathology129 into CBA
(n = 5)
(n = 7)
  1. Significant difference between 129 and GRKO was calculated using the Mann–Whitney test.

  2. GRKO = interferon-γ-receptor deficient.

Portal triad inflammation1.8 ± 0.42.1 ± 0.90.43
Venous inflammation2.0 ± 02.6 ± 0.80.07
Bile duct Inflammation1.8 ± 0.42.6 ± 0.50.03
Necrosis0.4 ± 0.53.0 ± 1.20.01

MHC expression in liver allografts

We measured donor MHC expression in the grafts using a radiolabeled antibody-binding assay described previously (30). CBA liver grafts transplanted into WT hosts showed induction of donor (H-2 k) MHC class I expression (Figure 3), with a more variable induction of donor class II expression. In contrast, neither class I nor class II was induced in liver grafts transplanted into the GKO hosts.

Figure 3.

MHC Class I and II expression in liver allografts by radiolabeled Ab-binding assay. Radiolabeled-antibody binding assay measurements of donor MHC class I and class II in liver allografts transplanted into WT (BALB/c H-2d), and GKO (H-2d) hosts. Donor MHC class I and II expression remains at basal levels in grafts that are rejected in GKO hosts. *Significant difference compared with normal CBA. **Significant difference between WT and GKO. MHC = major histocompatibility complex; Tx = transplant.

We confirmed these differences in MHC expression by indirect immunoperoxidase staining, using rat monoclonal antibodies specific for mouse MHC class I and II. In the control sections from rejecting transplants, peroxidase-positive cells were observed in the infiltrate in the rejecting transplants, both in the WT and GKO hosts (Figure 4A,B). MHC class I was increased on the endothelial cells and sinusoidal cells of grafts into the WT hosts, as well as on the cellular infiltrate (Figure 4C). In contrast, there was little staining for MHC class I on the endothelial or parenchymal cells or the infiltrating cells in grafts transplanted into the GKO hosts (Figure 4D).

Figure 4.

MHC Class I and II expression in rejecting liver allografts at day 10 post transplant. (A,B) Control staining of WT (BALB/c, H-2d) and GKO (H-2d) hosts. Note peroxidase positive cells which were in the infiltrate of the rejecting livers but were absent from the normal livers. (C,D) MHC donor class I (M1) staining of the parenchyma of the graft into WT and GKO hosts. (E,F) MHC class II (M5) staining of grafts in WT and GKO hosts (×200). Stained with peroxidase-labeled goat anti rat as second antibody.

Class II expression in normal mouse liver is confined to occasional interstitial cells called dendritic cells (11,31,32), Class II was increased in livers rejecting in the WT hosts (Figure 4E) in discreet positive cells in sinusoids, which have previously been shown to be Kupffer cells (31). These were absent in liver transplants in the GKO hosts, whose class II expression remained in the basal state (Figure 4F).

The MHC staining was confirmed using donor-specific mouse monoclonals anti class I (Kk) and anti class II mouse monoclonals (I-Ak). Again the livers rejecting in BALB/c hosts showed diffuse increased staining of endothelium, Kupffer cells, and the sinusoidal face of the hepatocytes for class I; and increased staining of individual Kupffer cells for class II. This increased staining was absent in the GKO hosts, whose MHC staining remained in the basal state. Host-specific monoclonals against Dd and I-Ad showed staining only of the infiltrating cells in the rejecting livers in the BALB/c mice but much less staining of the infiltrate in the GKO mice.

For the 129 WT liver allografts in the CBA hosts, the RABA demonstrated strong induction of donor MHC class I and moderate induction of donor class II antigens (Figure 5). In contrast, GRKO grafts had no induction of donor MHC class I or class II expression, remaining at basal expression.

Figure 5.

Radiolabeled-antibody binding assay of MHC class I and II expression in WT and GRKO grafts. (A) Measurement of donor MHC class I expression demonstrated strong induction of donor MHC class I antigens in wild-type 129/SvJ (H-2b) grafts transplanted into CBA/J (H-2 k) hosts at post transplant day 10. GRKO (H-2b) grafts had no induction of donor MHC class I expression. (B) Donor MHC class II antigens were also increased at day 10 post transplant in WT grafts, but not in GRKO grafts. *Significant difference compared with normal CBA, **Significant difference between WT and GKRO.

CTL gene expression

We compared the expression of a number of genes associated with CTL activity (granzyme A and B, FasL and perforin) in rejecting liver allografts in the WT and GKO mice at days 5, 7, and 10 post transplant by real-time PCR analysis (Figure 6). Expression of all of these genes was massively elevated post transplant in grafts transplanted into WT or GKO hosts, peaking at day 5 and declining at days 7 and 10. However, the levels of granzyme A, granzyme B, and Fas L mRNA were significantly lower in the GKO hosts at days 5 and 7. The differences were too great to be attributable to the necrosis in the liver allografts in the GKO hosts, and were surprising in view of the heavier infiltrate of lymphocytes in the GKO transplants. Perforin mRNA levels were similar in grafts transplanted into the WT and GKO hosts.

Figure 6.

Real-time RT-PCR of granzyme A, granzyme B, perforin, and FasL expression in liver allografts from WT and GKO mice (H-2d) at days 5, 7, and 10 post transplant. The mRNA levels are expressed as fold increase over a normal donor CBA (H-2 k). *Significant difference between WT and GKO. FasL = Fas ligand.


These studies of liver allograft rejection in hosts lacking IFN-γ and in normal hosts with livers from donors lacking IFN-γ receptors indicate that IFN-γ action on the graft is an absolute requirement for the spontaneous acceptance of liver allografts. There were no survivors if IFN-γ was not present, whereas if IFN-γ was present the survival of MHC incompatible livers was similar to that of the syngeneic controls, i.e. complete spontaneous acceptance. Comparison of liver rejection in WT and GKO hosts establishes that IFN-γ affected several aspects of the early rejection phenotype: it reduces hemorrhage, necrosis, and cellular infiltration while inducing high MHC expression. Despite the aggressive rejection in the GKO hosts, mRNAs for genes associated with T-cell effector activity were not increased in the GKO recipients. The phenotype of rejecting liver allografts lacking IFN-γ receptors was similar in that parenchymal necrosis was increased and MHC expression was not induced. The necrosis was dependent on the alloimmune response, coinciding in time with rejection and being absent in syngeneic grafts. The results point to a major effect of IFN-γ being on the graft itself, inducing donor MHC expression and preventing ischemic necrosis, while leaving open the likelihood of significant direct effects on host immune regulation.

The ability of IFN-γ to prevent graft necrosis during rejection via graft IFN-γ receptors is observed in several types of vascularized organ allografts. Excessive necrosis has been observed in kidney allografts in GKO hosts (30) and in kidneys from donors lacking the IFN-γRs (15). This effect was also seen in kidneys lacking transcription factor IRF-1 (33), suggesting that this IFN-γ-regulated transcription factor acts distally to IFN-γ receptors (15,33). A potentially related effect has been reported in concordant xenografts when the host lacks IFN-γ (14). IFN-γ does not act across unrelated species, suggesting that lack of host IFN-γ action on the graft may contribute to the vascular destruction in discordant xenografts. However, administration of recombinant IFN-γ to GKO hosts may not readily prevent necrosis, as shown in our earlier studies (30). This may reflect the fact that production of IFN-γ in the rejecting graft is massive and possibly paracrine, acting on contiguous cells, and may be difficult to simulate by systemic IFN-γ administration. We have also attempted to prevent liver allograft necrosis by injecting rIFN-γ intraperitoneally in GKO recipients but as yet with limited success (unpublished results). We investigated this in our kidney model and found that the administered anti IFN-γ was not capable of neutralizing IFN-γ produced in the graft {14478}. Others have also found that anti IFN-γ failed to alter survival {15248}.

The destruction of rejecting livers in IFN-γ-deficient hosts resembled ischemic necrosis despite patent vessels, suggesting failure of the microcirculation. This suggests that IFN-γ sustains the microcirculation, by an action that requires IFN-γ receptors in the graft. The nature of this regulation is under investigation: nitric oxide synthase, indoleamine 2,3-dioxygenase (IDO), IFN-γ-regulated chemokines, and many other potential mechanisms must be considered. The microcirculation is a target of rejection in some models (34,35), but the immunologic mechanism of injury is not clear. We have ongoing studies in hosts with disruptions of various effector mechanisms to resolve this. For example, renal transplants lacking IFN-γ receptors undergo accelerated rejection with typical necrosis in hosts lacking B cells and immunoglobulin (36), indicating microvascular injury by a cell-mediated mechanism. This is probably T-cell mediated, but a role for natural killer (NK) cells, or for T cells with NK receptors, is an intriguing possibility. Natural killer receptors engaging allogeneic MHC can inhibit effector mechanisms (37), suggesting that the reduced MHC induction could facilitate graft destruction. The observation that cells bearing NK receptors promote rejection in mice lacking CD28 signaling supports the need for further study of NK receptors (38).

Although the IFN-γ-inducible molecules that mediate the protective effect of IFN-γ on liver allografts are not known, one candidate is donor MHC molecules in the graft. MHC induction accompanies rejection and precedes spontaneous acceptance, and lack of IFN-γ-induced MHC expression accompanies necrosis. High levels of MHC expression can potentially neutralize and divert immune effector mechanisms from destroying the graft during the early post transplant period, perhaps by releasing soluble donor MHC class I antigens. Liver allografts produce soluble donor MHC class I antigens (39–41), which have potential for immune modulating effects (42,43). Perhaps the lack of IFN-γ-induced MHC expression on the parenchymal cells leaves the microvascular endothelium to receive the full force of effector mechanisms. The massive parenchymal MHC induction might act as a sink to divert or buffer the effector mechanisms. This situation may have similarities with donor blood transfusion, in which MHC class I and II induction is increased but rejection is reduced (44).

The fact that IFN-γ mediates protection against necrosis by a direct action on the graft does not contradict the potential importance of the direct effects of IFN-γ on T-cell homeostasis. IFN-γ promotes activation-induced death and regulates immunodominance in CD8 effector T cells (45,46). CD8+ effector cells are regulated differently in GKO hosts, developing independently of CD4+ cells and CD40–CD40 ligand interactions (47). Thus the phenotype of liver allografts in GKO and GRKO mice may reflect multiple IFN-γ mediated mechanisms. For example, liver allografts in GKO hosts displayed necrosis and increased infiltration, whereas GRKO allografts showed necrosis without increased infiltration. Given that the hosts for GRKO grafts, unlike GKO hosts, have normal IFN-γ production, this observation at face value suggests that the extent of infiltration is controlled by host IFN-γ production acting directly on host lymphocytes, in keeping with its known homeostatic functions. (The lower expression of granzyme A, granzyme B, and FasL at some time points in rejecting livers in GKO hosts vs. WT hosts is also compatible with altered T-cell homeostasis.) However, given the differences in genetic background between the GKO and GRKO experiments, this conclusion must remain tentative pending GRKO and GKO transplants across the same histocompatibility differences.

Spontaneous acceptance of liver allografts may be dependent on the fact that, after antigen-specific activation, CTL are intrinsically programmed to undergo contraction independent of antigen clearance (48). This contraction may prevent the destruction of organs when a massive viral infection cannot be cleared (49). The liver transplant is not unique: kidney transplants in mice frequently undergo survival after a rejection episode (50). Spontaneous acceptance of organ transplants is not the absence of rejection, but the ability of the organ to endure infiltration and immune effector mechanisms to permit survival of the tissue and repair of injury as rejection involutes. The greater tendency of the liver to undergo spontaneous acceptance probably reflects the addition of the intrinsic advantages of the liver (mass, regeneration after injury, dual blood supply, special liver antigen-presenting cells) to general properties of T-cell-effector systems.

The phenotype of grafts rejecting in GKO hosts may turn out to be mediated by multiple discreet IFN-γ-regulated mechanisms rather than by a single mechanism. An impressive number of distinct activities can now be assigned to IFN-γ during rejection, each acting at a specified time and place but often with opposite net effects. Among the early effects are protection from necrosis, MHC induction, induction of chemokine expression (MIG, IP-10, I-Tac), antagonism with IL-4 in ‘TH1-TH2’ type regulation, induction of rejection of class II mismatched skin grafts, induction of enzyme indoleamine 2,3-dioxygenase (51), and homeostasis of effector cell populations. Later in the course, IFN-γ promotes arterial deterioration in some models in which early rejection is suppressed (52,53). (Spontaneously accepted liver transplants did not develop such lesions in our study, at least at 100 days, presumably because of the completeness of the tolerance state.) In this sense, the separate study of IFN-γ receptors in the graft and the host may prove very productive in dissecting the complex pathogenesis of organ graft rejection.