IFN-γ Alters the Pathology of Graft Rejection: Protection from Early Necrosis¹

Allograft rejection is associated with intense production of IFN-γ, which acts in the graft and on host cells. One manifestation of IFN-γ action on the graft is induction of MHC products in the parenchyma of the rejecting organ (1, 2) and in the host tissues (3, 4), due mainly to IFN-γ (1, 2). The effect of IFN-γ is usually considered to promote tissue injury, which it does effectively in autoimmunity (5) and in some transplant models. For example, IFN-γ is needed for rejection of established islet transplants by CD8 T cells in a TCR-transgenic model (6), for rejection of class II-disparate skin grafts (7), and aggravates chronic vascular injury in heart transplants (8, 9, 10). Yet despite the association of IFN-γ with inflammation and MHC regulation, mice with disrupted IFN-γ genes (GKO mice, BALB/c mice with disrupted IFN-γ genes) reject transplants briskly (11, 12, 13). IFN-γ deficiency does not prevent myocardial rejection in transplanted (Tx)⁴ mouse hearts (8), and mice lacking IFN-γ receptors reject islet transplants (14). The surprising efficiency of graft rejection in mice lacking IFN-γ has been attributed to the ability of IFN-γ to inhibit lymphocyte proliferation and CTL generation (13, 15). Heart allografts are rejected by IFN-γ-deficient hosts using either a CD4-dependent pathway or a novel CD8-dependent, CD4-independent pathway (16). The mechanism of rejection is not suppressed by anti-CD40 ligand. Moreover, IFN-γ plays a role in protecting the graft against early failure in concordant rat to mouse xenotransplants (17). Thus, in the process of rejection of allografts or concordant xenografts, IFN-γ displays diverse effects, many attributable to the immunoregulatory or effector activities of IFN-γ.

In previous studies, we examined the early functions of IFN-γ acting directly on vascularized organ transplants by studying the rejection of kidney allografts from donors that lack IFN-γ receptors (GRKO donors, mice with disrupted IFN-γ receptor genes). These transplants undergo massive necrosis beginning at days 5–7, which did not occur in allogeneic transplants with intact receptors for IFN-γ (18). The massive necrosis was apparently due to ischemia secondary to microvascular injury and congestion. However, in this model, we could not distinguish whether the requirement for IFN-γ receptors in the donor tissue was due to an effect before the transplant (e.g., a developmental effect) or during the rejection process.

In the present studies, we explored the role of host IFN-γ production in the pathology of acute rejection of vascularized heart and kidney allografts using hosts lacking IFN-γ. We monitored the effects of IFN-γ on the graft such as MHC induction (12). Compared with organs rejecting in hosts with wild-type (WT) IFN-γ genes (WT hosts), vascularized heart or kidney allografts rejecting in GKO hosts showed rapid development of necrosis, with congestion and small thrombi in veins, but with patent large vessels. The protective effect is probably due to IFN-γ produced in and acting on the graft, because it was not simulated by anti-IFN-γ Ab and rIFN-γ treatment only partially prevented the necrosis. Thus, the predominant early role of IFN-γ in rejection of vascularized organs such as kidney and heart is protection against early failure of the microcirculation and necrosis, probably by a direct action on the graft.

GKO mice were created by disrupting the IFN-γ gene, inserting a neomycin resistance gene and replacing one copy of the WT gene in embryonic stem cells by homologous recombination (19). These stem cells were used to construct mice heterozygous for the disrupted gene, which were intercrossed and the progeny were selected for homozygosity. Heterozygous BALB/c mice and BALB/c mice with intact IFN-γ genes were provided to us as a generous gift from Timothy Stewart (Genentech, South San Francisco, CA). The mice were maintained in the Health Sciences Laboratory Animal Services at the University of Alberta and were kept on acidified water. All experiments conformed to approved animal care protocols.

Four mice of WT and GKO type each served as controls and were sacrificed without undergoing transplantation. Experimental mice were anesthetized with 60 mg/kg pentobarbital i.p. with supplemental doses given as needed for sedation. The chests of CBA/J male donor mice were opened and the hearts were excised and placed in cold lactate Ringer’s solution. The abdomen of the BALB/c male recipient was then incised and the donor CBA heart was anastomosed to the recipient using the standard heterotopic technique as described by Lower and Shumway (20). The mice were permitted to recover and graft function was assessed daily by palpation of the graft. The mice were assigned to one of three treatment groups: sham-operated controls (n = 14), cyclosporine (CsA) (Neoral, Novartis, Dorval, Canada) at a dose of 50–75 mg/kg/day by gavage (n = 9), or mycophenolate mofetil (MMF) (CellCept, Roche, Mississauga, Canada) at a dose of 40 mg/kg/day (n = 10). Mice were sacrificed at day 7 following anesthesia and cervical dislocation. Tissue was fixed in Formalin and paraffin, embedded for routine histology and slides, and stained with H&E and periodic acid-Schiff. Indirect immunoperoxidase staining for class I and class II expression was conducted using standard techniques.

The CBA/J male donor mice were anesthetized and the abdomen were opened through a midline incision. The left kidney was excised and preserved in cold lactate Ringer’s solution. The recipient male BALB/c mice, either GKO or WT, were similarly anesthetized and the right host kidney was excised. The donor kidney was then anastomosed heterotopically to the abdominal aorta and vena cava, without removing the host kidney. The mice were allowed to recover and were killed at day 7 or 10 following anesthesia and cervical dislocation. Tissue specimens were handled in a manner similar to the heart protocol. None of the kidney transplant recipients received immunosuppressive therapy. Mice with severe pyelonephritis at the time of harvesting were removed from the study.

Two pathologists assigned scores for the lesions observed in whole kidney sections (two sections per kidney) including cortex and outer medulla. Additional findings not included in the Banff scoring system (21) were also studied and scored as follows: For the extent of necrosis, the percentage of parenchymal involvement was recorded. Peritubular capillary (PTC) congestion, glomerulitis, tubulitis, and interstitial infiltrate were scored from 0 to 3 based on the percentage of parenchymal involvement (0, no changes; 1, <25% of the total parenchyma involved; 2, 25–75% of total parenchyma involved; and 3, >75% of the total parenchyma involved). Arteritis, arterial thrombosis, venulitis, and venous thrombosis were counted in each specimen and the mean number of lesions observed per kidney was calculated for each group. Thrombotic lesions were first assessed by H&E stain and the presence of fibrin in the thrombus was confirmed by Martius-Scarlet-Blue stain for fibrin.

Heart specimens were assessed (two sections per heart) for the presence of capillary congestion, interstitial hemorrhage, and hemorrhagic necrosis based on the percentage of parenchymal involvement. The extent of myocardial infiltrate was scored from 0 to 3 (0, no changes; 1, <25% of the total parenchyma involved; 2, 25–75% of total parenchyma involved; and 3, >75% of the total parenchyma involved). Arteritis, arterial thrombosis, venulitis, and venous thrombosis were counted in each heart specimen and the mean number of lesions observed per kidney was calculated for each group.

Hybridoma cell lines were obtained from American Type Culture Collection (Manassas, VA) and the cell lines producing mAb 25-9-17SII (anti-I-A^d), 34-4-20S (anti-H-2D^d), 11-5.2.1.9 (anti-I-A^k), 11-4.1 (anti-H-2K), M1/42.3.9.8 (anti-H-2 Ags all haplotypes), and M5/114.15.2 (anti-I-A^b,d,q and I-E^d,k) were maintained in tissue culture in our laboratory. The supernatants containing 25-9-17SII (anti-I-A^d), 34-4-20S (anti-H-2D^d), 11-5.2.1.9 (anti-I-A^k), and 11-4.1 (anti-H-2K) were purified by protein A chromatography; M1 and M5 were ammonium sulfate precipitated and then put 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. Radioiodination was performed using the Iodogen method (Pierce, Rockford, IL) (22).

This technique has been previously reported (12), and its quantitative characteristics have been described elsewhere (23, 24, 25). Briefly, hearts and kidneys were homogenized, washed in PBS, and centrifuged at 3000 rpm for 20 min. The pellets were suspended in PBS at 20 mg/ml. Five milligrams of kidney and 10 mg of heart tissue were centrifuged and then suspended in 100 μl of radiolabeled mAb (100,000 cpm) and incubated on ice with agitation for 60 min. One milliliter of PBS was added to all of the tubes and spun at 3000 rpm for 20 min. The pellets were counted in a gamma counter, and the nonspecific binding of a negative tissue was subtracted. The results are expressed as specific cpm bound by the tissue homogenate after subtracting the background (the cpm absorbed by the negative control tissue.) The rate of rise in cpm underestimates the degree of change in Ag expression: each 2-fold change in cpm corresponds to about a 3-fold change in Ag input (25).

Fresh-frozen cryostat sections were fixed in acetone and then incubated with normal goat serum. The slides were then incubated with mAb anti-class I (M1) and class II (M5) or controls followed by affinity-purified peroxidase-labeled goat anti-rat IgG F(ab′)₂ (Organon Teknika, Scarborough, Ontario, Canada). The slides were incubated with 3,3′-diaminobenzidine tetrahydrochloride and hydrogen peroxide for the color reaction and then counterstained with hematoxylin.

We performed TUNEL of fragmented DNA on 3-μm sections of paraffin-embedded tissue (26, 27) to stain apoptotic cells. Briefly, sections were deparaffinized in xylene and hydrated through a series of alcohols. To inactivate endogenous peroxidase, the sections were immersed in 1% H₂O₂, then rinsed in distilled water. The sections were treated with proteinase K (20 μg/ml in PBS) and rinsed with PBS. The sections were air dried, then flooded with TDT buffer at room temperature. TDT buffer (30 mM Tris HCl (pH 7.2), 1 mM CoCl₂, 140 mM sodium cacodylate) containing 0.25 nmol/μl biotin-16-dUTP and 0.25 U/μl TDT was used to begin labeling, and after incubation at 37°C in a humidified chamber, the slides were washed twice in PBS. Ten percent BSA in PBS was used to block nonspecific staining followed by two washes in PBS. The slides were then incubated with the avidin-biotin complex and washed twice in PBS. 3,3′-Diaminobenzidine tetrahydrochloride substrate was used to visualize the reaction and the slides were counterstained with methyl green and mounted with Permount.

Total RNA was extracted from kidneys that were harvested on day 7. RNA was transcribed into cDNA using Superscript reverse transcriptase (Life Technologies, Burlington, Ontario, Canada). The DNA was amplified in a Perking-Elmer/Setups thermal cycler (Norwalk, CT) using Taq DNA polymerase and sequence specific primers for granzyme B, perforin, Fas ligand (FasL), NO synthase 2(NOS2), heme oxygenase-1 (HO-1), monokine inducible by IFN-γ (MIG), and hypoxanthine phosphoribosyltransferase (HPRT). The PCR products were Southern blotted and probed with radiolabeled oligonucleotide probes. The sequences of primers and probes are shown in Table I.

Mice were injected with ∼1 μg of anti-IFN-γ i.p. on day −1, day 0, and days 1–6 and were harvested on day 7. The anti-IFN-γ (R4-6A2) was obtained from American Type Culture Collection and grown in our laboratory, purified by ammonium sulfate precipitation, and then run through a DE52 column. The protein was concentrated down to ∼1 mg/ml.

The rIFN-γ was a generous gift from Genentech. The GKO knockout mice were injected with 3 μg i.p. on days 1, 3, and 5 after kidney transplant and harvested on day 7 or injected with 9 μg i.p. on days 2–6 and harvested on day 7.

WT and GKO KO mice received CBA donor kidneys on day 0 and were harvested on days 7 and 21. Serum dilutions were incubated with CBA spleen target cells for 30 min in a 37°C CO₂ incubator and then incubated with rabbit complement for 90 min at room temperature. Five percent eosin was added as well as 10% buffered Formalin, and the microtiter plates were read under an inverted microscope.

We explored the effect of host IFN-γ on the pathology of rejecting heart or kidney allografts by assessing graft rejection in hosts lacking IFN-γ, compared with WT hosts. We adjusted the models so that the organ was not life-sustaining by leaving in place the normal host heart in heart recipients and the left kidney in kidney recipients. This prevented host death so that the evolution of the pathology could be studied. The normal host organs also served as controls. The hearts were studied at day 7 and the kidneys at days 5, 7, 10, and 21.

As controls for nonimmunologic effects of IFN-γ, we Tx syngeneic BALB/c WT hearts into two GKO recipients and one GKO heart into a GKO recipient (Table II). We also Tx one WT kidney into a WT recipient, three GKO kidneys into GKO recipients, and three WT kidneys into GKO recipients. None of these grafts showed rejection, congestion, hemorrhage, thrombosis, or infarction at day 7. Thus, the absence of IFN-γ did not affect the pathology of the isografts.

The CBA heart allografts (Table II and Fig. 1, A–D) in WT hosts at day 7 exhibited a typical mononuclear infiltrate (28), but with no arteritis, capillary congestion, interstitial hemorrhage, or necrosis (Fig. 1,A). The CBA heart allografts in GKO hosts showed severe capillary congestion and interstitial hemorrhage (Fig. 1,B) accompanied by mononuclear cell infiltration and vasculitis of large and medium-sized arteries (Fig. 1,C). Small thrombi were present in some veins, although most large arteries and veins were patent (Fig. 1,D). Arteritis and venous thrombosis were only found in grafts in GKO hosts, but did not seem to account for the extent of necrosis. Hemorrhagic necrosis affected up to 75% of the myocardium of transplants in GKO recipients (mean, 22%, Table II), but was absent in heart transplants in WT recipients.

Previous studies showed that heart allograft rejection in GKO hosts was resistant to the experimental immunosuppressive agent anti-CD40 ligand (16). This observation suggests two interpretations: that the damage is resistant to immunosuppression, or that the damage is resistant to anti-CD40 ligand in particular. To assess more conventional immunosuppressive approaches, we treated transplant recipients with immunosuppressive drugs, either CsA or MMF, by daily gastric lavage (Table II). In both WT and GKO recipients treated with CsA or MMF, Tx hearts showed less evidence of rejection (Table II). In particular, heart transplants in GKO recipients treated with CsA showed less capillary congestion, interstitial hemorrhage, necrosis, arteritis, venulitis, and venous thrombosis. MMF showed similar but milder trends at the dose used. Thus, the vascular lesions and necrosis in heart allografts in GKO hosts were immune mediated and sensitive to immunosuppressive therapy.

CBA kidney allografts in WT hosts at day 5 (Table III) showed mild tubulitis and mild to moderate interstitial infiltration of mononuclear inflammatory cells, with occasional arteritic lesions. There was minimal PTC congestion, glomerulitis, or necrosis. Venulitis was present in most small veins, but arterial and venous thromboses were absent.

On day 7 or day 10, CBA transplants in WT hosts showed increased tubulitis, interstitial infiltrate, and arteritis (Table III and Fig. 1 E). There was little glomerulitis or PTC congestion. Necrosis was increased compared with day 5 but most of the parenchyma was viable. Venulitis persisted but arterial and venous thromboses were absent.

By day 21, three of four CBA kidney allografts Tx into WT recipients showed severe rejection but were not necrotic (Fig. 1, F and G). Arteritis with severe endothelialitis (Fig. 1,F) was accompanied by extensive tubulitis and the epithelium of Bowman’s capsule (Fig. 1 G). The interstitial infiltrate was less than at days 7–10.

The CBA kidneys in GKO hosts at days 5 and 7 or 10 (Table III) were markedly different from their WT counterparts, showing more PTC congestion, glomerulitis, parenchymal necrosis (Fig. 1,H), and partially occluding fibrin thrombi in some small veins (Fig. 1 I). There was minimal arterial thrombosis, and most large veins were patent. The partially occluding fibrin thrombi in small veins were not sufficient to account for the degree of necrosis. The relationship of the thrombosis to the necrosis is unclear. The degree of glomerulitis, tubulitis, infiltrate, arteritis, and venulitis was similar to kidneys rejecting in WT mice. The extensive necrosis at day 7 rendered many kidneys in GKO hosts difficult to grade. Grades were based on the viable areas. Due to the massive necrosis, we did not follow GKO kidneys to day 21.

As noted earlier, syngeneic kidney transplants into GKO mice showed no rejection lesions, and the host organs of WT and GKO recipients were normal.

TUNEL assays were performed to determine the nature of the massive cell death in transplants in GKO and WT recipients at day 7. TUNEL-positive cells were absent in tubules, despite massive epithelial cell death in the GKO kidneys and despite the presence of numerous TUNEL-positive cells in the interstitium. The number of TUNEL-positive interstitial cells was similar in GKO and WT hosts. The TUNEL-positive interstitial cells appeared to be infiltrating lymphocytes, but apoptosis of endothelial cells of peritubular capillaries cannot be excluded. We used Southern blotting of graft DNA to search for a DNA “ladder” as evidence of apoptosis but no ladder was observed. These observations are compatible with the morphologic assessment that the massive parenchymal cell death in GKO hosts was necrosis, not apoptosis.

In control kidneys, class I was expressed only on arterial endothelium and class II only on the interstitial cells. Rejecting transplants in WT hosts displayed intense class I and II expression on the basolateral membranes of the tubules, glomerular mesangium, and arterial endothelium. Both class I and class II were also expressed in the interstitial infiltrate in WT hosts (Fig. 2, A and E).

Expression of class I and II was much less in the rejecting kidneys in GKO hosts, even allowing for the extent of necrosis (Fig. 2, B and F). Class I and II expression was largely absent from parenchymal cells, but was also less in the interstitial infiltrating cells. Class I and II was strongly induced in the host kidney of WT recipients (Fig. 2, C and G) but not of GKO recipients (Fig. 2, D and H).

The findings in the heart transplants were similar. In WT recipients, the rejecting heart allografts showed donor class I expression in the cardiac muscle cells and host class I and II in the interstitial infiltrate. The host hearts and kidneys also showed increased MHC expression. Rejecting hearts in GKO recipients showed little class I and II staining in the parenchymal cells, weak expression in the interstitial infiltrate, and no MHC induction in the host hearts or kidneys (data not shown).

Using RABA (12), we confirmed the host and donor MHC expression in homogenates from the transplants into WT and GKO recipient hearts (Fig. 3,A) or kidneys (Fig. 3,B). The MHC expression is shown for a sham-treated control, the Tx organ, and the host organ. In the Tx heart or kidney (Fig. 3, ▪), donor MHC was increased in the WT hosts: class I in the heart and class I and II in the kidney. There was no increase in donor MHC expression in the rejecting transplants in GKO recipients. Host-type MHC was also increased in the rejecting transplants in both WT and GKO recipients due to leukocyte infiltration. Host-type MHC was increased in the host hearts or kidneys (Fig. 3, ▧) of WT hosts but not in GKO hosts.

We studied whether the genes characteristic of CTL were expressed at higher levels in kidneys rejecting in GKO hosts by RT-PCR (Fig. 4). During rejection, WT and GKO kidneys showed similar increases in expression of FasL, perforin,and granzyme B mRNA compared with control kidneys.

We also studied the expression of two IFN-γ-regulated genes, NOS2 and MIG, and of endothelial protective gene, HO-1 (29). MIG mRNA was massively induced in kidneys rejecting in WT hosts but was absent in kidneys rejecting in GKO hosts (Fig. 4). The expression of NOS2 and HO-1 was increased in the rejecting CBA kidneys in either WT or GKO hosts. However, the kidneys rejecting in GKO hosts showed relatively more expression of HO-1 and less of NOS2 than those in WT hosts.

We administered mouse rIFN-γ posttransplant to GKO mice to see whether it would prevent the massive necrosis of the graft (Table IV). In experiment 1, we injected GKO mice with 3 μg of IFN-γ at days 1, 3, and 5 before studying the kidneys at day 7. This dose increased the host MHC in the transplant (presumably in the infiltrating cells), induced host MHC in the GKO host kidney, and donor MHC in the transplant kidney. This was confirmed by immunostaining (data not shown). Nevertheless, we did not prevent necrosis, congestion, or thrombosis of the grafts.

In experiment 2, we injected 9 μg of rIFN-γ daily from days 2–6 (45 μg total). This dose induced host MHC expression in the transplant and in host kidneys and donor MHC in the transplants. The rejecting kidneys in hosts that received this higher dose of rIFN-γ showed less necrosis, congestion, and thrombosis, indicating partial protection (Table IV).

We evaluated the effect of rIFN-γ on the expression of NOS2, MIG, and HO-1. In experiment 1, NOS2mRNA was increased by rIFN-γ, but MIG was not, and HO-1 was not reduced (data not shown). In experiment 2, NOS2 and MIG mRNA were increased by the high dose of rIFN-γ (Fig. 5 A). Nevertheless, the increases fell far short of the expression in rejecting transplants in WT recipients. The HO-1 was not normalized by rIFN-γ.

Thus, rIFN-γ in GKO hosts was able to induce MHC expression, but even at high doses induction of MIG and other aspects of gene expression and protection from necrosis were only partial. Thus, systemic administration of rIFN-γ has a limited ability to create the effects of high local IFN-γproduction in the graft.

We studied whether injections of anti-IFN-γ could make transplants into WT hosts resemble those in GKO hosts. We used a protocol that neutralizes the systemic effects of IFN-γ on MHC expression (3). The anti-IFN-γ at a low or high dose neutralized the induction of host MHC in BALB/c host kidneys in WT hosts rejecting the kidney allografts. However, the anti-IFN-γ did not induce necrosis and did not prevent host or donor MHC induction in the transplant (Table V).

Anti-IFN-γ also failed to prevent MIG and expression in the rejecting kidneys although it did decrease NOS2expression (Fig. 5 B).

Thus, anti-IFN-γ in WT hosts blocked the induction of MHC expression systemically but failed to cause the graft to resemble grafts in GKO mice. This indicates that systemic anti-IFN-γdid not neutralize the local IFN-γ effects in the transplant itself (e.g., induction of MHC and MIG expression), reflecting the intensity of local IFN-γ production in the graft.

Five mice in each group were tested at days 7 and 21 for cytotoxic activity against CBA and BALB/c target cells (control). All mice had specific cytotoxic activity against CBA target cells at day 7, mean titer of 1 in 624 ± 555 in WT mice and 1 in 482 ± 342 in GKO mice (NS, p = 0.64). Thus, all hosts mounted donor-specific cytotoxic Ab responses by day 7 but were not different between GKO and WT hosts. At day 21, the titers in WT mice increased to 1 in 3757 ± 386 (p < 0.001).

IFN-γ has unique roles early in acute rejection of MHC-incompatible kidney and heart allografts which change the course of allograft rejection. IFN-γ induces MHC and many other genes in the graft and protects the parenchyma against necrosis and the microcirculation against failure, accompanied by fibrin thrombi partially occluding some small veins. In heart transplants, IFN-γ also reduces arteritis. IFN-γ leaks from the graft to act systemically on the tissues of the host. The ability of IFN-γ produced by host cells to minimize necrosis in the grafted parenchymal tissue despite florid rejection may explain why IFN-γ is needed in some tolerance protocols (13, 30, 31). The excessive damage to allografts in GKO hosts was immune mediated: it was absent in syngeneic heart or kidney transplants and was reduced by immunosuppressive drugs. The effects of IFN-γ deficiency appear ∼5–7 days after transplantation, simultaneous with the immune response, and involves the microcirculation, unlike spontaneous vascular failure which occurs early and involves large arteries and veins. The protective effect of IFN-γ could reflect actions on the host immune response or directly on graft cells to make them resistant to the host effector mechanisms’ response. However, for reasons outlined below, it is likely that the action is on the IFN-γ receptors in the graft.

The principal finding in the heart and kidney allografts in GKO hosts was necrosis, probably ischemic. In the hearts there were some arterial and venous thrombi, although it was not clear that these were sufficient to account for the necrosis. The large arteries were patent in the kidney transplants in GKO hosts. In the kidneys there were some nonoccluding fibrin thrombi in small veins, which did not seem to explain the ischemic necrosis. Moreover, venous occlusion produces hemorrhage in kidney, which was generally not seen in these kidneys. It is more likely that the necrosis reflects damage in the microcirculation.

The protective effects of IFN-γ against ischemic necrosis appear to be conditional on massive local production of IFN-γin the graft. IFN-γ production by T cells is highly dependent on engagement of the TCR and rapidly decays when the TCR contact with an Ag-bearing cell is lost (32). Thus, the intense local production of IFN-γ in the rejecting allograft presumably reflects host T cells engaging MHC on donor cells or on host cells presenting donor peptides. Indeed, given that IFN-γ production requires TCR engagement, it is probable that some key effects are on cells contiguous to the T cell, making complete inhibition of local IFN-γ effects by Ab very difficult. We were unable to simulate the effects of IFN-γdeficiency by giving anti-IFN-γ to WT hosts, probably because the Ab did not reach sufficient concentrations in the graft. However, we did fully block the systemic MHC induction, which we believe is due to leakage of IFN-γ from the graft. We restored some of the effects of IFN-γ in GKO hosts by rIFN-γ administration. This probably indicates that these changes (e.g., MHC and MIG induction, graft protection against necrosis) reflect the large quantities of IFN-γproduced by host cells in rejecting grafts in close proximity with graft cells (33). Perhaps the local continuous administration of Ab or of rIFN-γ by osmotic minipumps into the transplant artery will be able to manipulate the local conditions effectively, but technical difficulties have limited the application of this method in the small vessels of the mouse. The fact that anti-IFN-γ did not simulate, and rIFN-γdid only partially prevent, the consequences of IFN-γdeficiency in a local inflammatory site, may be an important conceptual point in interpreting other consequences of IFN-γdeficiency: local production by inflammatory cells in contact with the graft endothelium and parenchyma may be difficult to fully simulate or disrupt by systemic interventions.

Although IFN-γ affects many aspects of the host immune response, including cytokine production, Ab production, and CTL generation (13, 16), deviations in the host immune response are unlikely to account for the phenotype of rejecting allografts in GKO hosts. The principal basis for this conclusion is that the phenotype of kidney allografts in GKO hosts is very similar to that of kidney allografts lacking IFN-γ receptors in normal hosts, indicating that the key action of IFN-γ is on the graft IFN-γ receptors (18). It is well established that the absence of IFN-γ deviates the alloimmune response (15, 16), and we have confirmed these effects. For example, we have shown that CTL generation in response to an allograft is increased in hosts lacking IFN-γ, and kidney allografts in GKO and WT hosts elicit brisk cytotoxic alloantibody responses but with less IgG2a in GKO hosts (our unpublished results). However, these effects do not explain the present observations. For example, cellular infiltration was abundant in kidney allografts with or without IFN-γ, with similar expression of CTL gene expression. Moreover, the overall contribution of CTL mechanisms to allograft rejection remains unproven (34). Thus, although GKO hosts undoubtedly have altered effector mechanisms, we believe that this is not the main factor in the accelerated destruction of vascularized allografts in GKO hosts.

Despite its ability to protect the allograft from early massive necrosis, IFN-γ often participates in graft injury in the weeks that follow. For example, transient immunosuppression temporarily prevents thrombosis and infarction in allografts in GKO hosts (e.g., the present studies). After withdrawal of immunosuppression, grafts in GKO hosts show less chronic vascular rejection, indicating that IFN-γ accelerates graft vessel disease (8, 35, 36). Similarly in the special circumstance of nonvascularized class II-disparate skin allografts, IFN-γ is actually essential for graft rejection (7, 37). Thus, the net effect of IFN-γ will depend on the details of the system studied: whether the graft is vascularized, whether there are strong antigenic differences, whether temporary immunosuppression is used, and the time posttransplant.

The transient protective effect of IFN-γ could represent IFN-γ-induced resistance of endothelium to destruction by Ab, consistent with conclusions in concordant xenografts (17). Cytotoxic alloantibody levels were similar in GKO and WT mice at day 7, making differences in alloantibody production unlikely. The evidence for a role for alloantibody in the lesions in GKO hosts is circumstantial: alloantibody is present at day 7, and vasculitis, interstitial hemorrhage, and infarction are suspicious for Ab-mediated destruction (38). Alloantibody against class I can mediate early acute rejection in mice (39, 40, 41, 42), and alloantibody against class I mediates an uncommon but severe form of acute rejection in human transplantation, with prominent vascular damage, specifically Ab against MHC molecules (38, 43). Alloantibody could contribute to the CD4 T cell dependency of heart allograft rejection in GKO hosts, although a second CD8-dependent mechanism also exists (16).

Although many genes are regulated by IFN-γ, none at present explains the protection against immune-mediated necrosis. IFN-γ-regulated genes in the graft include MHC genes, NOS2, endothelial protective genes such as HO-1, and MIG, and/or IFN-γ-regulated chemokines. MHC induction could be the protective entity. The fact that we induced some donor MHC expression at the intermediate doses of rIFN-γwithout inducing protection argues somewhat against this. However, perhaps the quantity of MHC induced may not have been sufficient. HO-1 is not the protective entity: HO-1 mRNA was actually more abundant in the kidneys in GKO hosts. HO-1 is known to be induced by tissue ischemia, suggesting that ischemic damage in the kidneys in GKO hosts induced HO-1 expression. NOS2 is reduced in allografts in GKO hosts, making it a candidate for the protective effect of IFN-γ. Previous studies of NO and of NOS2 support either a protective or an aggressive role in rejection (44, 45, 46), in keeping with the diverse sources and roles of NO and NOS2. Similarly, resistance to early graft loss paralleled MIG expression in the grafts. MIG was abundant in WT, severely reduced in GKO, and relatively unchanged by rIFN-γor anti-IFN-γ treatment. However, MIG has been incriminated in promoting skin graft rejection across class II differences (47). It is difficult to see how the absence of MIG expression could render graft susceptibility to necrosis when the graft lacks IFN-γRs, because the host cells can make MIG. For example, MIG is reduced but still abundant in rejecting kidneys lacking IFN-γRs due to expression in host inflammatory cells. In summary, the gene that mediates protection against necrosis must depend on local IFN-γ production in the graft acting on IFN-γ receptors and transcription factor IFN regulatory factor-1 in donor cells, but the mediator of protection is not apparent.

The ability of IFN-γ to prevent immune-mediated necrosis in rejecting tissues could have practical implications, where mild rejection occurs and resolves. In allotransplantation, IFN-γ might be considered to protect the vessels against thrombosis in early immune injury in sensitized patient, and may be useful in some tolerance protocols (48). There are also implications for xenotransplantation, since IFN-γ is essential for the survival of concordant xenografts (17). Because IFN-γ is species specific, xenografts from distant species (e.g., pig to primate) will not be protected because host IFN-γ cannot trigger donor IFN-γRs. This could contribute to the accelerated vascular rejection of discordant xenografts. Indeed, the rejection of allografts by IFN-γ-deficient hosts in the present study resembles accelerated vascular rejection of xenografts in some respects. However, IFN-γ is a double-edged sword which can aggravate vascular injury after the first few weeks (8, 9, 10). The experience that type I IFN can induce rejection in humans is also cause for caution (49). Thus, the challenge will be to capture the protective action of IFN-γ therapeutically while avoiding its harmful effects (9, 17, 50, 51, 52, 53, 54).

We are grateful to Dr. Timothy Stewart of Genentech for providing the GKO breeders. We also thank Pamela Publicover for secretarial assistance and Angeline Batocchio for technical assistance.