Autocrine Production of IFN-γ by Macrophages Controls Their Recruitment to Kidney and the Development of Glomerulonephritis in MRL/lpr Mice¹

A hallmark of systemic lupus erythematosus (SLE)³ in humans and in mouse models is the production and deposition of anti-DNA Abs in the kidney. Murine lupus models based on knock-out technology have recently been described that maintain intact autoantibody production and glomerular deposition, but do not develop kidney inflammation or glomerulonephritis. NZB × NZW mice lacking the FcR γ-chain thus exhibit glomerular immune complexes (IC), but not glomerulonephritis, presumably by disruption of the FcR-mediated inflammatory cascade (1). We also observed protection from kidney disease in IFN-γ^+/− MRL/lpr mice, despite intact autoantibody production and IC deposition (2). A similar model of delayed disease has been described for monocyte chemoattractant protein (MCP)-1-deficient MRL/lpr mice (3).

IFN-γ is a key cytokine that controls Th1-dependent Ab production and is indispensable for functional immune responses (4, 5). The elevated IFN-γ expression in MRL/lpr mice (6, 7), the exacerbating effect on disease onset in NZB × NZW mice following treatment with this cytokine (8), and the amelioration of disease manifestations following inactivation of IFN-γ (8, 9) suggest an important role for this cytokine in lupus development. Nevertheless, the mechanism by which IFN-γ contributes to this autoimmune disease remains unclear. Deletion of IFN-γ or its receptor in MRL/lpr mice demonstrated that this cytokine is essential for autoantibody production (2, 10, 11, 12, 13), a result further confirmed in a myasthenia gravis model (14). Complete absence of IFN-γ is not required to ameliorate disease; severe reduction of IFN-γ, mediated by gene therapy, affects both autoantibody production and glomerulonephritis development (15). A presumably less efficient IFN-γ decrease, detected in a long-lived MRL/lpr mouse mutant, results in change in autoantibody isotype and enhances survival (16). An ∼50% decrease in IFN-γ levels in IFN-γ^+/− MRL/lpr mice results in greatly delayed disease onset (2). The elevated autoantibody and IC deposition levels in IFN-γ^+/− mice, similar to those of wild-type (wt) MRL/lpr mice, suggest that IFN-γ levels play a key role in the local inflammatory process.

Initiation of the kidney inflammatory process is thought to depend on autoantibody deposition, which ultimately leads to tissue destruction, but the underlying mechanism is not known. Analysis of MCP-1-deficient MRL/lpr mice showed that MCP-1 controls the inflammatory process in kidneys of lupus-prone mice by inhibiting macrophage migration and delaying disease onset, despite elevated IC deposits in the kidneys of these animals (3). Therefore, MCP-1 is required for local kidney inflammation, although IFN-γ may also exert an effect on this process.

To assess the role of IFN-γ in kidney inflammation, we studied factors that may influence this process in IFN-γ^+/− and in IFN-γ^−/− MRL/lpr mice following transfer of IFN-γ-producing monocyte/macrophages or T cells. The data showed that IFN-γ secretion by macrophages controls their recruitment to kidney interstitium, up-regulation of the adhesion molecules ICAM-1 and VCAM-1, as well as the overall inflammatory process. Furthermore, efficient IFN-γ production by macrophages is indispensable for disease progression in the presence of IC glomerular deposits.

We previously described the generation of IFN-γ^−/− MRL/lpr mice (2). In this study, we used IFN-γ^−/−, IFN-γ^+/−, and IFN-γ^+/+ mice. Mice were typed by PCR using appropriate oligonucleotides (2).

To detect IC deposits, cryostat kidney sections (6 μm) were fixed in chilled acetone and stained with a FITC-conjugated goat anti-mouse IgG (Southern Biotechnology Associates, Birmingham, AL). Immunostaining was also performed with biotin-conjugated Abs to CD4, CD8 (Caltag Laboratories, Burlingame, CA), VCAM-1 (BD PharMingen, San Diego, CA), and F4/80 (Serotec, Oxford, U.K.). MCP-1 was detected using a goat anti-mouse MCP-1 Ab (Santa Cruz Biotechnology, Santa Cruz, CA), followed by peroxidase (PO)-conjugated donkey anti-sheep/goat secondary Ab (The Binding Site, Birmingham, U.K.). Staining was performed as described (2). Endogenous peroxidase activity was blocked by incubation with PBS containing 0.03% H₂O₂. For staining with biotin-conjugated Abs, sections were incubated with a biotin blocking kit (DAKO, Glostrup, Denmark) and 20% normal mouse serum. Biotinylated Abs were developed with the StrepABComplex/HRP kit (DAKO). When anti-MCP-1 Ab was used, sections were blocked with 20% normal horse serum. Peroxidase activity was detected using FAST DAB (Sigma-Aldrich, St. Louis, MO).

For paraffin sections, kidneys were fixed in 5% formalin, 4-μm sections were deparaffinized and treated with 3% H₂O₂ in 70% methanol, followed by microwave treatment in 10 mM sodium citrate buffer. Biotin block and serum treatment was performed as above. ICAM-1 expression was detected using a biotin-conjugated Ab (BD PharMingen); MCP-1 was detected using goat anti-mouse MCP-1 Ab. CD3 was identified by a rabbit anti-human CD3 Ab, followed by PO-labeled goat anti-rabbit Ab (both from DAKO). PO-conjugated anti-proliferating nuclear Ag (PCNA; Santa Cruz Biotechnology) was used to stain paraffin sections according to manufacturer’s instructions, including blocking steps with H₂O₂ and 20% normal serum.

To evaluate the degree of F4/80, ICAM-1, and VCAM-1 expression, 10 random glomeruli and 10 tubular zones were evaluated from each section. Sections were scored on a scale of 1–4+; a score of 4+ was assigned to stained sections of a 4-mo-old MRL/lpr mouse with full-blown glomerulonephritis and maximal staining for F4/80, ICAM-1, and VCAM-1. Evaluation of degree of staining in sections from other mice were based on this mouse.

Kidney chemokines were detected by multiprobe RNase protection assay according to manufacturer’s instructions (BD PharMingen). Expression levels were quantitated as a function of L32 and GADPH housekeeping gene levels in a Storm 860 PhosphorImager and analyzed using ImageQuant software (Molecular Dynamics, Sunnyvale, CA).

Mice were given BrdU (0.8 mg/ml; Sigma-Aldrich) in drinking water, prepared freshly every 2 days, for a 9-day period.

BrdU expression in macrophages was determined by triple staining with FITC-conjugated anti-BrdU, PE-anti-Mac-1 (both from BD PharMingen) and Cy5-anti-F4/80 (Serotec) Abs, as described (2). To detect macrophage CCR2 expression, we used a biotin-conjugated mouse anti-human CCR2 (17) that cross-reacts with murine CCR2, followed by FITC-conjugated streptavidin (Caltag Laboratories) in combination with PE-anti-Mac-1 and Cy5-F4/80. For detection of macrophage IFN-γR expression, a biotinylated anti-IFN-γR α-chain Ab (BD PharMingen) was used in combination with PE-anti-MAC-1 and Cy5-anti-F4/80 Abs. We used these last two Abs and FITC-conjugated anti-I-A^k (BD PharMingen) to determine MHC class II expression levels. Stained cells were analyzed on a Coulter EPICS XL flow cytometer (Hialeah, FL).

To purify monocytes/macrophages, splenocytes were stained with FITC-labeled Abs to CD4, CD8, and B220 (BD PharMingen) and passed through a MACS separation column (Miltenyi Biotec, Bergisch Gladbach, Germany) designed to retain FITC-labeled cells. Recovered cells were further stained with a PE-anti-Mac-1 Ab and sorted using a Coulter EPICS sorter; a yield of 99% pure Mac-1⁺ cells was obtained. CD4⁺ and CD8⁺ lymphocytes were also sorted following double-labeling with FITC- and PE-conjugated anti-CD4 and anti-CD8 Abs, respectively. To sort CD4⁺ and CD8⁺ T cells as well as macrophages from adoptively transferred hosts, splenocytes were stained with a combination of FITC-labeled Abs to CD4 and CD8, and with PE-anti-Mac-1 and Cy5-anti-F4/80 Abs.

RNA was extracted from sorted CD4⁺ and CD8⁺ T cells (combined), macrophages, as well as from whole kidney. RT-PCR was performed using a OneStep RT-PCR kit (Qiagen, Hilden, Germany) using primers and PCR conditions previously described (2).

Isotype-specific anti-DNA Abs were determined by ELISA as described (2). The reactivity of PO-conjugated Abs against IgG1, IgG2, IgG2b, and IgG3 (Caltag Laboratories) was normalized for equivalent OD against corresponding myeloma isotype controls (Caltag Laboratories). Serum IFN-γ levels were quantified with a mouse IFN-γ ELISA kit (Endogen, Woburn, MA) following manufacturer’s instructions. Histological examination of lungs or kidneys was performed in a blind manner; glomerulonephritis was scored on a scale of 0–4+ and pyogranulomatous nephritis was evaluated as light, moderate, or severe. Blood urea nitrogen (BUN) levels were determined at autopsy using Azostix strips (Bayer, Elkhart, IN).

The Mann-Whitney U test was used to evaluate statistical significance. Two-tailed p values of <0.05 were considered significant.

Using IFN-γ^−/− and IFN-γ^+/− MRL/lpr mice as models, we attempted to define the factors that control kidney inflammation in lupus nephritis. IFN-γ deficiency prevents anti-DNA Ab production and disease onset (2). In contrast, IFN-γ^+/− MRL/lpr mice, characterized by an ∼50% reduction in IFN-γ production, develop elevated autoantibody levels, but are protected from glomerulonephritis development (2). IFN-γ^+/− mice thus constitute a useful model for the study of the pathogenic mechanism of kidney inflammation in SLE.

One characteristic of kidney disease in lupus-prone mice is extensive inflammatory cell infiltration. We observed a perivascular infiltrate, composed mainly of CD4⁺ cells, in the kidneys of IFN-γ^−/−, IFN-γ^+/−, and wt MRL/lpr mice (Fig. 1 A). CD8⁺ cells represented a smaller proportion of the infiltrating cells, with no apparent differences among the three types of mice examined (not shown).

Macrophages, considered to be leukocytes with destructive potential, migrate to the periglomerular and peritubular areas of MRL/lpr mouse kidney. Staining for macrophage markers F4/80 and Mac-1 showed that migration of these cells to the kidney interstitium in MRL/lpr mice reaches high levels at 4 mo of age, whereas 3-mo-old mice display minimal macrophage recruitment (Table I, p < 0.002; Fig. 1,B and not shown). Compared with wt MRL/lpr mice, 4-mo-old IFN-γ^−/− mice lacked kidney-infiltrating macrophages (Fig. 1,B; Table I), a finding compatible with the absence of IC deposits in the knock-out mice and in agreement with similar findings in IFN-γR-deficient mice (13). Macrophage accumulation was also dramatically reduced in IFN-γ^+/− MRL/lpr mice (Table I, IFN-γ^+/− vs IFN-γ^+/+, p < 0.002; Fig. 1 B). The absence of disease in these animals could thus be associated with reduced macrophage presence in kidney.

Because MCP-1 is a critical molecule in leukocyte migration to inflammation sites and lupus glomerulonephritis, we considered that reduction in IFN-γ levels may influence macrophage migration by affecting MCP-1 expression in kidney. We compared MCP-1, RANTES, and macrophage-inflammatory protein (MIP)-1β expression in kidneys of 4-mo-old IFN-γ^+/+, IFN-γ^−/− and IFN-γ^+/− MRL/lpr mice by immunohistochemistry in paraffin sections. RANTES was uniformly expressed by tubular cells, whereas MIP-1β was expressed solely by distal tubules (not shown). MCP-1, which is also present in tubules, shows a specific expression profile, with some tubular cells exhibiting intense MCP-1 production oriented toward the luminal brush border (Fig. 1,C). This expression pattern was similar in wt, IFN-γ^+/− and IFN-γ^−/− MRL/lpr mice. Although up-regulation of MCP-1 expression has also been detected in lupus-prone mouse glomeruli by in situ hybridization (18), glomerular MCP-1 expression was low in paraffin sections (Fig. 1,C). Staining of kidney cryosections showed pronounced glomerular MCP-1 expression in all three mouse types (Fig. 1 D). The variation in staining between cryo- and paraffin sections may depend on differences in accessibility of distinct kidney structures, depending on the technique used.

Multiprobe RNase protection assay was used to quantitate the expression of several chemokines. Kidney expression of RANTES, eotaxin, MIP-1β, and MCP-1 was similar in 3- and 4-mo-old MRL/lpr mice (Fig. 2,A). This concurs with immunohistochemical analysis, in which MCP-1 expression was similar in 3- and 4-mo-old mouse kidney (not shown) and with a previous report showing equivalent MCP-1 expression in 2- and 4-mo-old mice (3). One-month-old animals showed lower kidney MCP-1 levels compared with 3- and 4-mo-old animals. There was no difference in chemokine levels between IFN-γ^+/+ and IFN-γ^+/− animals (Fig. 2 A); kidney MCP-1 expression was lower in IFN-γ^−/− mice (by 33%) compared with IFN-γ^+/+ controls at 4 mo of age (p < 0.006, n = 10), but clearly elevated compared with 1-mo-old mice (p < 0.0001). This concurs with a previous report in which reduced MCP-1 production was detected in MRL/lpr mice subjected to gene therapy with cDNA encoding IFN-γR/Fc (15). IFN-γ-inducible protein-10 levels were determined using a different probe set (not shown); this chemokine was not up-regulated in MRL/lpr mouse kidney.

MCP-1 is considered a key chemokine for macrophage attraction to sites of experimentally induced inflammation and in MRL/lpr mouse kidney (3, 19, 20). Nevertheless, as shown above, MCP-1 up-regulation in the kidneys of IFN-γ^+/− MRL/lpr mice was not sufficient for macrophage attraction. To exclude the possibility that this was due to an intrinsic macrophage defect provoked by reduced IFN-γ production, we examined various functional aspects of these cells. Macrophages (Mac-1+, F4/80+) were represented in similar proportions in spleens of IFN-γ^+/+ and IFN-γ^+/− mice (12 ± 2 and 10 ± 1.2%, n = 6); there were no differences in size and cellularity between the spleens of the two mouse types. Macrophages from IFN-γ^+/+ and IFN-γ^+/− animals proliferated similarly, as assayed by BrdU incorporation (Fig. 2,B). Because CCR2 is the major MCP-1 chemokine receptor (21, 22, 23), we examined whether expression of this receptor was IFN-γ-dependent; there was no difference in CCR2 expression in spleen macrophages from IFN-γ^+/+ and IFN-γ^+/− mice (Fig. 2,C). Another indicator of macrophage activation is the MHC class II expression level, which is up-regulated in activated macrophages (5). We previously showed that macrophage I-A^k expression is reduced in MRL/lpr IFN-γ^−/− mice (2). In contrast, in this study, we find similar I-A^k expression in IFN-γ^+/+ and IFN-γ^+/− mouse macrophages (Fig. 2 D). We thus show that decreased IFN-γ production does not affect critical functional characteristics of macrophages that could interfere with their migration to inflammation sites.

ICAM-1 is an adhesion molecule involved in leukocyte attraction to inflammation sites; it is also implicated in the progression of lupus-associated glomerulonephritis (24). To assess the possible implication of this molecule in the delayed disease in IFN-γ^+/− and IFN-γ^−/− MRL/lpr mouse kidney, we examined ICAM-1 expression by immunohistochemistry in 4-mo-old mice. The data show elevated ICAM-1 expression in kidneys of 4-mo-old MRL/lpr mice kidneys, but not in those of 4- or 6-mo-old IFN-γ^−/− or IFN-γ^+/− animals (Fig. 3,A; Table I, IFN-γ^+/+ vs IFN-γ^+/−, p < 0.002), indicating that ICAM-1 expression depends on the degree of IFN-γ expression. In addition, the VCAM-1 adhesion molecule, which is up-regulated in SLE glomerulonephritis (24), appeared highly dependent on IFN-γ levels (Fig. 3,B; Table I, IFN-γ^+/− vs IFN-γ^+/+, p < 0.002). Finally, kidney ICAM-1 and VCAM-1 expression was very low in 3-mo-old MRL/lpr mice, whereas elevated expression was detected at 4 mo of age (Table I, p < 0.002 and p < 0.002, respectively). Because F4/80 cell migration takes place in this interval, we conclude that adhesion molecule up-regulation coincides with the appearance of F4/80⁺ cells in kidney.

F4/80⁺ macrophages have an important role in the progression of inflammation and lupus glomerulonephritis development. Macrophage recruitment to the kidney is blocked, even in the presence of heavy glomerular immune deposits in IFN-γ^+/− mice. Although MCP-1 is produced by tubular cells and glomeruli, the site of IFN-γ production during kidney inflammation is not clear. IFN-γ production by infiltrating cells has been reported under conditions of increased IL-12 expression (25). To identify the source of IFN-γ-producing cells responsible for macrophage attraction, we adoptively transferred purified monocyte/macrophages from IFN-γ^+/+ to IFN-γ^−/− animals.

Two-month-old IFN-γ^−/− mice each received an i.v. injection of 10⁶ purified spleen monocytes/macrophages from IFN-γ^+/+ mice; recipient animals (n = 9) and controls (n = 7) were studied for 4 mo. Adoptively transferred macrophages produced IFN-γ in IFN-γ^−/− hosts, because at 4 mo posttransfer, IFN-γ mRNA was detected in macrophages, but not in T cells, of recipient animals (Fig. 4,A, left panel). IFN-γ production was further confirmed by the ELISA detection of this cytokine in the serum of transferred animals, at levels similar to those of BALB/c mice but much lower than of MRL/lpr mice (Fig. 4,A, right panel). IFN-γR expression by macrophages confirmed their potential responsiveness to IFN-γ (Fig. 4 B). Although some anti-DNA IgG1 production was detected in recipient mice compared with age-matched IFN-γ^−/− controls (13 ± 3.0 vs 5 ± 0.7 OD U), anti-DNA Ab generation was minimal compared with that of IFN-γ^+/+ mice (110 ± 33, n = 6). The autoantibody increase was detected only for the IgG1 isotype (considered nonpathogenic; Ref. 16); IgG2a, IgG2b, and IgG3 levels were undetectable in both adoptively transferred and control IFN-γ^−/− mice. The amount of IFN-γ produced was thus insufficient to provoke significant anti-DNA Ab production or a switch to Th1-dependent isotypes.

Concurring with these data, IC deposition in glomeruli of recipient mice was low compared with IFN-γ^+/+ animals and similar to IFN-γ^−/− controls (Fig. 4,C). MCP-1 expression quantified by RNase protection assay was similar in macrophage-transferred and IFN-γ^−/− control animals (40 ± 14, n = 8 vs 42 ± 17, n = 10 (Fig. 2,A)). Concurring with these results, immunohistochemistry experiments showed similar glomerular MCP-1 expression in recipient and control mice (Fig. 4,D). In overall agreement with these results, we observed reduced glomerular proliferation in kidneys of recipient mice compared with those of wt MRL/lpr mice, using an Ab to PCNA (Fig. 5,A); disease-protected IFN-γ^−/− controls and IFN-γ^+/− mice also showed minimal glomerular proliferation (Fig. 5 A).

Transfer of IFN-γ^+/+ monocyte/macrophages to IFN-γ^−/− mice was sufficient for migration of F4/80⁺ macrophages to the periglomerular and peritubular areas of the kidney, even in the absence of high autoantibody titers (Fig. 5,B) and despite the 33% reduction in kidney MCP-1 production (Fig. 2,A) in the host mice. Although IFN-γ-deficient macrophages may also have migrated to the kidney, IFN-γ^+/+ macrophages accumulate in this organ, because IFN-γ was expressed in host IFN-γ^−/− animal kidneys (Fig. 4,A, left panel). CD4⁺ cells were detected in the interstitial areas of kidneys in transferred mice, whereas CD8⁺ and B cells were rarely identified (not shown). In spleen, macrophages were found in similar proportions in adoptively transferred animals and IFN-γ^−/− controls (not shown). Immunohistochemistry experiments showed increased ICAM-1 and VCAM-1 production in IFN-γ^−/− recipient mouse kidney (Fig. 5,B; Table I), emphasizing the contribution of IFN-γ-producing macrophages in adhesion molecule up-regulation. In PCNA staining experiments, we detected no appreciable proliferation of the interstitial infiltrate in wt MRL/lpr mice or adoptively transferred IFN-γ^−/− animals (not shown). This suggests that the interstitial macrophage accumulation is due to migration and not to local proliferation. Finally, as an additional control, IFN-γ^−/− mice were adoptively transferred with IFN-γ^−/− macrophages (n = 5) and their kidneys were compared with those of control IFN-γ^−/− mice; there was no macrophage accumulation in the kidneys of mice adoptively transferred with IFN-γ^−/− macrophages (not shown). This confirms that the IFN-γ-producing capacity of macrophages is responsible for their accumulation in the kidney.

Histological analysis of kidneys showed absence of glomerulonephritis in IFN-γ^+/+ macrophage-recipient mice, as is the case for IFN-γ^−/− control mice. In the adoptively transferred animals, however, kidneys showed severe multifocal pyogranulomatous nephritis, with extensive perivascular and periglomerular accumulation of polymorphonuclear leukocytes and mononuclear cells. Degenerative glomerular lesions were observed, with atrophy and some completely fibrotic glomeruli (Fig. 6,A). Pyogranulomatous nephritis was severe in all transferred animals (n = 9, severity was classified on a scale of light, moderate, or severe disease), while nephritis was absent in all control animals (n = 7; Fig. 6 A).

To determine whether intact IFN-γ production by T cells could initiate the inflammatory process in the kidney and recruit F4/80⁺ cells, we adoptively transferred purified IFN-γ^+/+ T cells to IFN-γ^−/− mice. Recipient mice were 2-mo old at the time of transfer of 10⁶ CD4⁺ and 10⁶ CD8⁺ T cells. Transferred and control animals were followed for 4 mo after T cell transfer. As detected by RT-PCR of purified CD4⁺ and CD8⁺ splenocytes at 4 mo posttransfer, adoptively transferred T cells expressed IFN-γ (not shown). The results of IFN-γ^+/+ T cell transfer to IFN-γ^−/− mice showed that the kidney interstitium of these mice remained free of F4/80⁺ cells, T cells, or other types of lymphocytes, and showed low ICAM-1 and VCAM-1 levels (Fig. 5,C; Table I). This indicates that IFN-γ^+/+ T cells, different from IFN-γ^+/+ macrophages, are not sufficient to initiate local interstitial kidney inflammation (Fig. 5, B and C; Table I). Host mice developed heavy perivascular infiltrates composed of CD4⁺ T cells (not shown), as was the case for IFN-γ^−/− mice (Fig. 1,A). T cell-transferred mice showed no signs of glomerulonephritis (not shown) and no glomerular proliferation (Fig. 5,A), as was also the case for nontransferred controls. In contrast, transferred mice developed diffuse severe acidophilic pneumonia, characterized by intraalveolar accumulation of strongly eosinophilic macrophages with heavy CD3⁺ cell infiltration compared with IFN-γ^−/− controls (Fig. 7); IFN-γ^−/− animals reconstituted with IFN-γ^+/+ macrophages presented occasional peribronchial T cell infiltration (Fig. 7). We also examined the serological effects of IFN-γ^+/+ T cell transfer in IFN-γ^−/−-recipient mice, and found that transfer alone did not reconstitute anti-DNA autoantibody production (12 ± 0.5 OD U in IFN-γ^−/− transferred vs 7 ± 0.5 OD U in IFN-γ^−/− controls). As lymph node and spleen hypercellularity and hyperproliferation persist in IFN-γ^−/− mice (2), the number of transferred IFN-γ^+/+ T cells may be insufficient to give rise to anti-DNA autoantibody production. Alternatively, break of tolerance to DNA may require intact IFN-γ production by both T cells and macrophages.

Adoptive transfer of IFN-γ^+/+ macrophages or T cells into IFN-γ^−/−-recipient mice shows that intact IFN-γ production by macrophages is responsible for their attraction to the kidney interstitium and establishment of the inflammatory process. Nonetheless, glomerulonephritis did not develop in the absence of glomerular deposits in IFN-γ^−/− mice. To determine the role of macrophage-produced IFN-γ in glomerulonephritis development, we transferred IFN-γ^+/+ macrophages (10⁶ cells/mouse) into 2-mo-old IFN-γ^+/− mice, which develop glomerular autoantibody deposits, but not glomerulonephritis. We followed these mice (IFN-γ^+/− (M)) for 4 mo posttransfer and observed that all transferred mice developed severe diffuse proliferative glomerulonephritis with hypercellular glomerulus, increased mesangial cells, and thickening of capillary walls (Fig. 6,B). Transferred mice were sacrificed when they became moribund (n = 4) or at 4 mo posttransfer (n = 2), while all controls remained alive at this time. Glomerular damage was significantly more severe in transferred mice than in IFN-γ^+/− controls (Table II); concurring with the glomerulonephritis data, transferred animals showed elevated BUN levels at death. The degree of glomerular damage and BUN levels in IFN-γ^+/− (M) mice was similar to 6-mo-old IFN-γ^+/+ controls. F4/80⁺ cell migration was restored, as was adhesion molecule up-regulation in mice reconstituted with IFN-γ^+/+ macrophages (not shown). The data indicate that glomerulonephritis development in IFN-γ^+/− mice is dependent on the transfer of IFN-γ^+/+ macrophages, which potentiated the inflammatory pathway and resulted in glomerular damage. Glomerular autoantibody deposition and intact IFN-γ production by macrophages are thus two events sufficient for glomerulonephritis development.

To gain insight into the pathology of SLE-associated nephritis, we examined the inflammatory pathway in IFN-γ^+/− MRL/lpr mice, which are protected from glomerulonephritis, and in Ab-free IFN-γ^−/− mice adoptively transferred with IFN-γ^+/+ monocyte/macrophages or T cells. The following conclusions are drawn as a result of this study: 1) impaired IFN-γ production in IFN-γ^+/− MRL/lpr mice results in a remarkable decrease in interstitial F4/80⁺ macrophages, but does not affect the elevated MCP-1 expression in the MRL/lpr mouse kidney, 2) adhesion molecules may be essential for macrophage accumulation, because ICAM-1 and VCAM-1 expression is substantially reduced in IFN-γ^+/− MRL/lpr mouse kidneys, 3) IFN-γ production by macrophages, but not by T cells, is sufficient to initiate adhesion molecule up-regulation in the kidney and further macrophage recruitment, even in the absence of glomerular autoantibody deposits, and 4) in addition to anti-DNA autoantibody deposition, intact IFN-γ production by macrophages is necessary for glomerular destruction.

Several studies implicate MCP-1 in the attraction of inflammatory cells to sites of tissue destruction, and up-regulation of this chemokine has been linked with advanced lupus disease (18). A study of MCP-1^−/− MRL/lpr mice showed a substantial delay in the induction of glomerulonephritis due to absence of macrophages in kidney, although anti-DNA Ab production and deposition in glomeruli remained intact (3). IFN-γ^+/− MRL/lpr mice constitute a similar model, as they develop anti-DNA Abs, but not disease. Our data show that kidneys from these mice are devoid of macrophages, indicating that reduction in IFN-γ is crucial for progression of the inflammatory process. Perivascular infiltrate was nevertheless present in these mice as well as in IFN-γ^−/− kidneys, indicating that kidney inflammation and lymphocyte migration are initiated in the absence of immune deposits, perhaps as a result of an overactive immune response in lpr mice. Because IFN-γ is thought to induce increased MCP-1 production (26), we examined whether the reduced macrophage migration into IFN-γ^−/− and IFN-γ^+/− mouse kidneys was due to reduced MCP-1 expression. MCP-1 was up-regulated in kidneys of both types of mice. Conversely, MCP-1 appears not to interfere with IFN-γ production, as suggested by recent data showing that absence of MCP-1 interferes with Th2, but not Th1 polarization, and that MCP-1^−/− mice produce normal IFN-γ levels (27). Because MCP-1 was up-regulated equally in kidneys of wt as well as IFN-γ^+/− MRL/lpr mice, which lack interstitial macrophage accumulation, we deduce that IFN-γ contributes to macrophage accumulation by a mechanism independent of MCP-1 expression.

The reduced ICAM-1 and VCAM-1 levels in IFN-γ^+/− mouse kidneys as compared with those of wt MRL/lpr mice suggest that the mechanism for regulating macrophage recruitment by IFN-γ is related to the adhesion process. Although adhesion molecules may be up-regulated by chemokines (28), this appears not to be the case for IFN-γ^+/− and IFN-γ^−/− mice exhibiting high levels of several chemokines in kidney. It thus appears that up-regulation of these molecules depends on IFN-γ production by interstitial macrophages. Experiments showing that transfer of IFN-γ^+/+ T cells was unable to restore interstitial kidney inflammation or ICAM-1/VCAM-1 up-regulation strengthen the view that macrophage-produced IFN-γ regulates the expression of these adhesion molecules by tubular and mesangial cells. On the basis of these data and the simultaneous up-regulation of adhesion molecules and F4/80⁺ cell accumulation in the kidney, it can be envisaged that macrophage-produced IFN-γ enhances adhesion molecule up-regulation; this in turn consolidates macrophage accumulation in the kidney interstitium. The ability of macrophages to initiate the inflammatory cascade through their IFN-γ secretory potential may be due to their migration into the interstitium, whereas T cells, also an abundant IFN-γ source, remain perivascular. Only macrophage-produced IFN-γ thus leads to adhesion molecule up-regulation. Previous studies showed that, following IFN-γ stimulation, tubular and mesangial cells up-regulate ICAM-1 and VCAM-1, among other molecules (29, 30, 31). ICAM-1 is required for leukocyte infiltration and ischemic renal injury (32). Furthermore, ICAM-1 and VCAM-1 are up-regulated in the kidneys of lupus model mice as well as in human disease (24). In MRL/lpr mice, ICAM-1 deficiency has been implicated in increased survival in two studies; in one, absence of pneumonia was reported (33); in the second, a delay was shown in glomerulonephritis onset (34). Nevertheless, absence of ICAM-1 did not result in strong inhibition of interstitial inflammatory cell recruitment, as is the case for IFN-γ^+/− mice. Although direct evidence is not available, the additive effect of ICAM-1 and VCAM-1 down-regulation may impede macrophage accumulation in kidney and effectively delay disease onset; up-regulation of other molecules involved in macrophage chemoattraction or adhesion may also be IFN-γ-dependent.

To further understand the role of IFN-γ in macrophage attraction, we sought to identify the source of IFN-γ required for kidney inflammation. Our results show that in IFN-γ^−/− MRL/lpr mice, adoptively transferred IFN-γ^+/+ macrophages migrate to kidney peritubular and periglomerular zones, initiate adhesion molecule up-regulation, and induce further macrophage migration. Although IFN-γ-producing macrophages are identified in the kidney interstitium, we cannot exclude that IFN-γ^−/− macrophages are also recruited in this organ. That macrophage recruitment occurs in the absence of IC deposits indicates that the inflammation process can be disengaged from IC deposition, which is necessary for glomerulonephritis development. Macrophages accumulate in the kidney interstitium, coinciding with the sites of adhesion molecule up-regulation, suggesting that they are the primary cells involved in inflammation. This result is further supported by data showing that adoptive transfer of IFN-γ^+/+ T cells to IFN-γ^−/− mice does not reconstitute the inflammatory process, as is the case of adoptively transferred IFN-γ^+/+ macrophages. The IFN-γ deficiency of macrophages does not appear to affect the overall activation of T cells. Indeed, in IFN-γ^−/− mice, reduced MHC class II Ag expression by macrophages may cause reduced autoantibody production, but does not affect hypergammaglobulinemia, CD4⁺ and CD8⁺ T cell proliferation (2), or the extent of perivascular infiltration compared with IFN-γ^+/+ mice. The severe lung inflammation observed in the IFN-γ^−/− mice adoptively transferred with T cells supports the view that transferred T cells become activated. Furthermore, CD3⁺ cell accumulation in the lung suggests that although IFN-γ is an important inflammatory molecule, distinct IFN-γ-producing cells may exacerbate inflammation in different organs.

Macrophage recruitment to the kidney is considered an important factor in initiation of tissue destruction (3). In this study, we show that impaired IFN-γ production inhibits macrophage recruitment to the kidney in MRL/lpr mice, indicating that the level of this cytokine is critical for disease onset and progression. In another study, Tnf hemizygosity induces lupus development in the otherwise lupus-resistant NZB mouse strain, presumably by influencing early B cell autoimmune responses (35). The degree of expression of these two cytokines thus appears critical for modulation of lupus development.

Adoptive transfer of IFN-γ^+/+ macrophages in IFN-γ^+/− mice, which have glomerular anti-DNA deposits, reconstitutes glomerular disease; this indicates that the degree of IFN-γ production by macrophages is a critical factor in glomerulonephritis development. Based on these findings, we propose a model describing the roles of IFN-γ and MCP-1 in macrophage recruitment to the kidney of lupus-prone mice. MCP-1 may thus be up-regulated in the kidney due to IC deposition or to an overactive immune response. Elevated MCP-1 production provokes migration of macrophages, whose ability to secrete large amounts of IFN-γ leads to their adherence and residence in the kidney by up-regulating adhesion molecules and perhaps other inflammatory factors. These destructive cells then contribute to further inflammation and macrophage recruitment. Under conditions of suboptimal IFN-γ production, as is the case for IFN-γ^+/− mice, MCP-1-attracted macrophages do not initiate the adhesion process, probably due to reduced IFN-γ secretion into the interstitial microenvironment; the macrophages thus do not accumulate in the kidney interstitium. This course of events concurs with the view that leukocyte migration is initially regulated by selectin-dependent rolling, followed by chemokine-mediated attraction; finally, the recruitment process is completed by up-regulated adhesion (28).

Understanding the mechanism of the inflammation pathway in lupus nephritis, as well as the molecules involved in this process, merits study due to the potential for therapeutic intervention even after autoantibodies are generated. This type of treatment may be useful in the treatment of human SLE, because autoantibody production precedes diagnosis. The results of this study suggest that targeting key inflammatory factors such as MCP-1, in combination with adhesion molecules or local IFN-γ production in the kidney, may constitute promising strategies for treatment of autoimmune disease without affecting critical IFN-γ-mediated immune responses.

We thank A. C. Carrera for suggestions, A. N. Theofilopoulos for useful discussions, L. Gómez for help with animal handling, and C. Mark for editorial assistance.