To analyze the mechanism for the therapeutic effects of tumor necrosis factor α (TNFα) inhibition in a murine model of systemic lupus erythematosus.
To analyze the mechanism for the therapeutic effects of tumor necrosis factor α (TNFα) inhibition in a murine model of systemic lupus erythematosus.
We used the (NZB × NZW)F1 (NZB/NZW) mouse model of interferon-α–induced lupus nephritis and treated mice with TNF receptor type II (TNFRII) Ig after TNFα expression was detected in the kidneys. Autoantibodies were measured by enzyme-linked immunosorbent assay (ELISA), and autoantibody- forming cells were determined using an enzyme-linked immunospot assay. Activation of splenocytes was analyzed by flow cytometry. Kidneys were harvested and analyzed using flow cytometry, immunohistochemistry, ELISA, Western blotting, and real-time polymerase chain reaction.
TNFRII Ig treatment stabilized nephritis and markedly prolonged survival. Autoantibody production and systemic immune activation were not inhibited, but the renal response to glomerular immune complex deposition was attenuated. This was associated with decreases in renal production of chemokines, renal endothelial cell activation, interstitial F4/80high macrophage accumulation, tubular damage, and oxidative stress. In contrast, perivascular lymphoid aggregates containing B cells, T cells, and dendritic cells accumulated unabated.
Our data suggest that TNFα is a critical cytokine that amplifies the response of the nephron to immune complex deposition, but that it has less influence on the response of the systemic vasculature to inflammation.
Tumor necrosis factor α (TNFα) is an important regulator of physiologic and inflammatory immune responses. Although TNFα antagonists have remarkable therapeutic benefits in several autoimmune diseases including rheumatoid arthritis, psoriasis, and Crohn's disease, they also have been associated with the development of antinuclear antibodies and even clinical systemic lupus erythematosus (SLE), multiple sclerosis, and other demyelinating diseases. TNFα is protective in the early stages of lupus in some mouse models but is overexpressed in the inflamed target organs of both mice and humans with lupus (1, 2). Consistent with both the protective and proinflammatory roles of TNFα, blockade of TNFα improved proteinuria in patients with treatment-refractory lupus nephritis but increased anti–double-stranded DNA (anti-dsDNA) and anticardiolipin autoantibody titers (3).
In this study, we determined the clinical effects of TNF receptor type II (TNFRII) Ig in a mouse model of lupus nephritis in which an increase in serum levels of TNFα and renal production of TNFα occurred concomitantly with the onset of renal disease. TNFα inhibition introduced after the development of autoantibodies and at the onset of clinical nephritis stabilized the nephritis and resulted in a long period of relapsing and remitting proteinuria that was associated with marked improvement in survival. Mechanistic studies revealed no effect of TNFα inhibition on autoantibody production or lymphocyte activation in the spleen or on renal autoantibody deposition; however, a marked decrease in renal periglomerular and interstitial accumulation of F4/80high renal macrophages was observed. Real-time polymerase chain reaction (PCR) analysis revealed a significant decrease in renal expression of chemokines CCL2, CCL5, and CCL9, the endothelial activation marker vascular cell adhesion molecule 1 (VCAM-1), genes involved in tissue remodeling, and the markers of proximal tubule damage lipocalin 2 and hepatitis A virus cellular receptor 1 (HAVCR-1) (kidney injury molecule 1; KIM-1). We conclude that TNFα inhibition decreases the renal inflammatory response to immune complex deposition. This decrease is associated with decreased accumulation of periglomerular and interstitial renal macrophages, which play a role in tissue remodeling, and decreased damage of renal tubular cells. In contrast, perivascular aggregates containing CD11chigh dendritic cells as well as B cells and T cells accumulate in the kidneys in a TNFα-independent manner and are not sufficient to induce terminal renal damage.
Twelve-week-old female NZB/NZW mice (The Jackson Laboratory) were treated with a single intravenous injection of 2.2 × 108 particles of IFNα adenovirus (Ad-IFNα; Qbiogene) that induced proteinuria after a median of 36 days (range 24–161 days) and death after a median of 85 days. In 75% of the mice, proteinuria developed before day 56. Serum samples were obtained, and urine was tested weekly for proteinuria by dipstick (Multistix; Fisher Scientific), as previously described (4). Fully murine TNFRII Ig (a kind gift from Dr. David Shealy, Centocor) was administered intraperitoneally at a dosage of 50 μg (2 mg/kg) 3 times weekly, beginning on day 25 after the development of proteinuria was first observed. One group of 6 TNFRII Ig–treated mice was killed on day 86 after administration of IFNα, by which time approximately one-half of the control mice had died. Three groups of TNFRII Ig–treated mice (3–5 per group) were allowed to age and were killed between day 220 and day 240. Twelve control mice were killed on day 50 or day 86, and the remaining 16 mice were followed up for survival.
Previous studies from our group showed that administration of irrelevant monoclonal IgG2a had no effect on proteinuria or survival in NZB/NZW mice (ref.5 and unpublished observations), and that several other fusion proteins administered just before the onset of proteinuria are ineffective in the model of IFNα-induced lupus (6). Therefore, we used mainly untreated control mice in this study. A group of 5 IgG2a-treated control mice was included in one of the experiments. These mice developed fixed proteinuria at the same time as untreated controls and were killed on day 50. In a separate experiment, groups of 5 control mice were killed weekly after Ad-IFNα injection, and the perfused kidneys were harvested for real-time PCR analysis. All experiments using animals were carried out according to protocols approved by the Institutional Animal Care and Use Committee of the Feinstein Institute.
Serum levels of TNFα were measured using a commercial enzyme-linked immunosorbent assay (ELISA; AssayGate), as previously described (4).
ELISpot assays for anti-dsDNA antibodies were performed as previously described (4).
Spleen and peripheral blood mononuclear cells were analyzed for cell surface markers as previously described (7), using antibodies to CD4, CD8 (both from Caltag), and CD19. Activated CD4 T cells were identified by double staining with fluorescein isothiocyanate (FITC)–conjugated anti-CD4 and phycoerythrin (PE)–conjugated anti-CD69. Naive and activated/memory CD4 cells were identified by triple staining with FITC-conjugated anti-CD4, Cy-Chrome–conjugated anti-CD44, and PE-conjugated anti-CD62L. Naive B cell subsets were identified based on staining for CD19, IgM, IgD, CD21, and CD23 (7). Germinal center B cells were positive for Fas and peanut agglutinin, and plasma cells were positive for CD138. Single-cell suspensions were prepared from perfused kidneys and stained with antibodies to CD11b, CD11c, and F4/80 as previously described (8). Antibodies were obtained from BD PharMingen unless stated otherwise.
Hematoxylin and eosin–stained kidney sections were scored for renal damage, as previously described (7–9). Cryosections (6 μm) of kidney were stained as previously described (7), using FITC-conjugated anti-mouse IgG2a (SouthernBiotech) and PE-conjugated anti-mouse CD11b, CD11c, CD4, CD8, B220 (BD PharMingen), or F4/80 (Invitrogen). Images were captured using a Zeiss AxioCam digital camera connected to a Zeiss Axioplan 2 microscope.
RNA was purified from lysates of perfused kidneys, and real-time PCR was performed in triplicate, as previously described (8). The average value of the raw data for each sample (Ct value) was normalized to the internal control (housekeeping gene β-actin), as previously described (8). Normalized expression data were log2-transformed and scaled to the expression value in a single naive mouse, set at an arbitrary value of 1 (0 by log scale; ΔCt value for calculation 4.76).
Kidney lysates were prepared by homogenization in a Dounce homogenizer using lysis buffer (RayBiotech) with Complete Protease Inhibitor (Roche). The homogenate was centrifuged at 14,000 revolutions per minute at 4°C for 20 minutes, and the pellet was discarded. The protein in the supernatant was quantified using the Bradford method, and proteins were run on a discontinuous sodium dodecyl sulfate–Tris glycine gel system (5% stacking, 12% resolving), followed by transfer to a nitrocellulose membrane. The membrane was then probed with anti–lipocalin 2 (ProSci) followed by donkey anti-goat IR800 Dye (Li-Cor). The lysates were also used for quantitation of mouse CCL9 and VCAM-1 using cytokine array (RayBiotech), and normalized lysates were used for highly sensitive quantitative VCAM-1 ELISA (R&D Systems), according to the manufacturer's instructions.
Renal oxidation was studied by immunolabeling, using an oxidation kit (Millipore) and IRDye 800–labeled secondary antibody (Li-Cor). Briefly, 8-μm frozen sections were fixed in ice-cold acetone for 15 minutes followed by washing with phosphate buffered saline (PBS). The proteins were then derivatized using dinitrophenylhydrazine (DNPH) at room temperature for 25 minutes followed by neutralization. After washing with PBS, the sections were blocked with 3% bovine serum albumin followed by serial incubations with anti-DNPH antibody for 1 hour and with IR800 Dye–labeled goat anti-rabbit antibody for 1 hour. The slides were then washed and scanned using a Li-Cor Odyssey instrument. As a negative control, a second section was processed in parallel for each sample, using the derivatization control solution.
Survival data were analyzed using Kaplan-Meier curves and log rank tests. Other comparisons were analyzed by Mann-Whitney test. P values less than or equal to 0.05 were considered significant.
We previously showed in unmanipulated NZB/NZW mice that renal expression of TNFα increases at the onset of proteinuria and decreases with induction of remission (8) (Figure 1A). In the current study, we demonstrated that after induction of lupus with IFNα, renal expression of TNFα increased beginning 4–5 weeks after induction (Figure 1B); this was associated with an increase in serum levels of TNFα (Figure 1C). When TNFRII Ig was administered starting 3–4 weeks after the initiation of IFNα-induced lupus, proteinuria either remained stable or followed a relapsing–remitting course, and survival was markedly prolonged (Figures 2A and B). This was associated with a decrease in renal damage scores (Figure 2C).
We previously reported that IFNα treatment induced serum autoantibodies to dsDNA in NZB/NZW mice within several weeks, and that this was associated with a marked increase in antibody-producing B cells in the spleen compared with naive controls (4). Serum levels of IgG autoantibodies to dsDNA, which were measured weekly throughout treatment, were the same in control mice and TNFRII Ig–treated mice (Figure 3A). Similarly, at the time when the mice were killed (on day 86 or day 200 after IFNα induction), there was no difference between TNFRII Ig–treated mice and control mice in terms of the number of IgG anti-dsDNA antibody–producing B cells per spleen, as measured by ELISpot (Figure 3B).
Flow cytometric analysis revealed no differences between IFNα-treated control mice and TNFRII Ig–treated mice (Figure 3C) or between IgG2a-treated mice and TNFRII Ig–treated mice (data not shown) with respect to accumulation of class-switched (IgM−/IgD−) B cells, early activation of CD4 T cells (as assessed by CD69 positivity), generation of memory (CD44+/CD62L−) CD4 T cells, or increase in splenic CD11b+ dendritic cells, all of which are features of IFNα-induced disease (4).
The onset of proteinuria in the NZB/NZW mouse model is associated with the expansion and activation of a dominant population of resident renal mononuclear phagocytes that have a resting phenotype of CD11b+CD11cintermediateF4/80high. In addition, nephritis onset is associated with the influx of a population of CD11b+/CD11chigh myeloid dendritic cells that accumulate in perivascular lymphoid aggregates (10) that also contain T cells and B cells. Using flow cytometry, we demonstrated a marked increase in the F4/80high macrophage population in IFNα-treated control mice compared with naive NZB/NZW mice; this increase was prevented by treatment with TNFRII Ig. In contrast, there was no difference in the percentage of renal CD11chigh cells between control and TNFRII Ig–treated mice (Figure 3D).
To confirm and extend these findings, we performed immunofluorescence analysis of kidneys from control and TNFRII Ig–treated mice that were killed on day 86 or after day 220. Staining of kidneys from TNFRII Ig–treated mice (Figure 4A) was compared with staining of kidneys from control IFNα-treated mice (Figure 4B). In control mice, CD11bhigh cells were located in the periglomerular regions and were abundant in the interstitium as well as within perivascular and large perihilar lymphoid aggregates; in contrast, these cells were observed only in perihilar aggregates in TNFRII Ig–treated mice (Figures 4A and B, parts i and iii).
In accord with these results, F4/80high cells were virtually absent from the interstitium (Figures 4A and B, parts ii and iv) and periglomerular regions (Figure 4C) of kidneys from TNFRII Ig–treated mice compared with those from controls and were restricted to a ring around the perihilar aggregates (Figure 4D), a location at which they are also observed in unmanipulated nephritic NZB/NZW mice (11). In contrast, CD11chigh cells (Figures 4A and B, part v, and Figure 4D) were located in the perivascular aggregates of both control and TNFRII Ig–treated mice. These aggregates also contained abundant CD4 T cells (Figure 4E) and B cells (Figure 4F). Very few neutrophils were present in the inflamed kidneys of control mice, and no difference between control and TNFRII Ig–treated mice was observed (data not shown).
The onset of clinical nephritis in NZB/NZW mice is associated with renal expression of a set of inflammatory markers. Some of these are derived from F4/80high cells, some reflect endothelial activation, and some reflect tubular damage and tissue remodeling (10). To determine which of these processes is altered by TNFRII Ig, real-time PCR was performed for a subset of informative markers (8, 10) (Figure 5). The expression of CXCL13, which is a chemokine that is expressed early in the disease process before the onset of nephritis and is made by macrophages and dendritic cells predominantly within perivascular infiltrates (8), was not altered by TNFRII Ig treatment. In contrast, expression of the chemokines CCL2, CCL5, CCL9, and CXCL10 was not up-regulated in TNFRII Ig–treated mice, and CCL2 expression was still decreased even after day 220. The expression of the macrophage/dendritic cell surface markers ITGAM, Clec4a3, and Clec4e (mincle) was decreased in TNFRII Ig–treated mice relative to controls on day 86 but not after day 220. Expression of the endothelial marker VCAM-1 was decreased in TNFRII Ig–treated mice compared with controls even at late stages, as were the markers of tissue remodeling tissue inhibitor of metalloproteinases 1 and matrix metalloproteinase 14. Expression of the markers for tubular damage lipocalin 2 and HAVCR-1 (KIM-1) was decreased in TNFRII Ig–treated mice compared with controls on day 86, and lipocalin 2 expression was still inhibited after day 220. Finally, renal expression of IL-1β was not affected by TNFα blockade.
To confirm these data at the protein level, analysis of renal lysates was performed using semiquantitative ELISA, quantitative ELISA, and Western blotting for inflammatory markers that were previously shown to be significantly up-regulated in the late stages of renal disease (8). The results of cytokine array showed that lysates from TNFRII Ig–treated mice had lower levels of expression of CCL9 and VCAM-1 compared to controls (mean ± SD 0.79 ± 0.22 versus 1.43 ± 0.65 [P < 0.05] and 0.35 ± 0.17 versus 1.06 ± 0.37 [P < 0.01], respectively), and the results for VCAM-1 were confirmed using a high-sensitivity ELISA (Figure 6A). Treated mice also had lower levels of lipocalin 2 expression even after 220 days (Figure 6B). To confirm the overall decrease in renal damage in TNFRII Ig–treated mice, we performed quantitation of oxidized proteins in the renal lysates. Kidneys from TNFRII Ig–treated mice displayed a decrease in oxidation status as measured by densitometry/area (Figures 6C–F).
TNFα is a pleiotropic cytokine that has both protective and pathogenic functions in SLE. Lupus-prone NZB/NZW mice have a defect in TNFα production. In young mice, further depletion of TNFα exacerbates disease, whereas TNFα replacement is protective (1, 12). The relevance of this mouse model to human biology has been borne out by the finding that antinuclear antibodies develop in up to 50% of patients treated with TNFα inhibitors, antibodies to DNA develop in 15%, and clinical SLE (including SLE nephritis) develops in 1 of 500 patients treated with TNFα inhibitors (13, 14). The mechanism for the pathogenic role of TNFα blockers in the induction of autoimmunity has not been elucidated, but it is clearly a class effect of all TNFα blockers. The prevailing hypotheses include increased production of type I IFN (15), blockade of the antiinflammatory effects of TNFα (16), failure to adequately regulate activated T cells including Th17 T cells (17, 18), and failure to induce cytotoxic T lymphocytes that can kill autoreactive B cells (19).
In NZB/NZW mice older than age 28 weeks, TNFα replacement is no longer effective, and low-dose TNFα has even been shown to exacerbate disease in older NZB/NZW mice that have high serum levels of TNFα (1, 12). TNFα levels are similarly elevated in the serum of patients with active SLE (20), and TNFα is expressed in the kidneys of both mice (6) and humans with lupus nephritis (3). The renal source of this cytokine is not entirely clear, with infiltrating mononuclear phagocytes, T cells, and glomerular mesangial cells all having been suggested to produce TNFα. The expression of TNF receptors, particularly TNFRI, is markedly up-regulated in the glomeruli of patients with proliferative lupus nephritis (21), and soluble TNFRI is shed in the urine (22). Regardless of the precise source of TNFα in human lupus glomerulonephritis, a short course of infliximab has had remarkable and long-lasting therapeutic benefit in open-label studies in a small number of SLE patients with treatment-refractory nephritis. Nevertheless, general enthusiasm for this approach has been tempered by the observation of transient increases in autoantibodies, severe infusion reactions following multiple doses of infliximab, and concerns about concomitant treatment with other immunosuppressive agents. An increased rate of bacterial infection and one case of lymphoma were observed in patients receiving continuous anti-TNFα therapy (3).
In the current study, we show that treatment with TNFRII, which is analogous to etanercept, in a lupus model in which renal TNFα expression occurs concomitantly with nephritis onset protects the kidneys against damage and prolongs survival. Systemic lymphocyte activation, induction of class-switched anti-dsDNA autoantibodies, and renal immunoglobulin deposition, all of which occur in the first 4 weeks after the initiation of IFNα-induced lupus in this model, were not altered by TNFα inhibition, but the renal inflammatory response to immune complex deposition was dramatically reduced.
The onset of proteinuria in lupus-prone mice is associated with the expansion and activation of a dominant population of resident renal mononuclear phagocytes that have variably been referred to as intrinsic renal macrophages or resident renal dendritic cells, due to their mixed function as both antigen-presenting cells and phagocytic cells (23). At nephritis onset, these cells markedly up-regulate their expression of CD11b, and they accumulate in both the interstitium and the periglomerular space (10, 24). Proteinuria onset and disease progression in NZB/NZW mice are also characterized by the renal influx of CD11chigh myeloid dendritic cells that accumulate exclusively in large lymphoid aggregates that also contain B cells and T cells and are located in the hilum and perivascular regions (8). In this study, we demonstrated that the recruitment of renal F4/80high macrophages to the periglomerular region and the renal interstitium is TNFα dependent, whereas the accumulation of myeloid dendritic cells, B cells, and T cells in perivascular or perihilar lymphoid aggregates and the accumulation of F4/80high cells in a cuff around these aggregates occur independently of TNFα.
Recruitment of inflammatory cells to the lupus kidney in response to immune complex deposition depends on several factors. In NZB/NZW mice, engagement of Fc receptors on circulating myeloid cells by glomerular immune complexes is required for renal damage to occur (25). Fc receptor crosslinking on monocytes induces TNFα production (26), providing a direct link between renal immune complex deposition and TNFα production by renal mononuclear phagocytes. Complement activation and consequent activation of inflammatory cascades also contribute to renal damage. Endothelial cell activation by immune complexes and cytokines results in expression of integrins that help to capture circulating cells. The onset of proteinuria in NZB/NZW mice is associated with up-regulation of the endothelial cell–derived adhesion molecules E-selectin, P-selectin, and VCAM-1 (8). Elaboration of chemokines that attract cells expressing the corresponding receptors facilitates transmigration of inflammatory cells into the renal parenchyma and precedes or coincides with the onset of clinical nephritis. In NZB/NZW mice, we have observed renal expression of chemokines that can attract macrophages and dendritic cells (8). Similarly, in MRL/lpr mice, expression of a similar set of chemokines in glomeruli is an early feature of lupus nephritis that precedes inflammatory cell infiltration (27). Of these chemokines, CCL2 has definitively been shown to be pathogenic (28) and, as shown here, is one of several chemokines whose up-regulation is dependent on TNFα.
We further show that renal IL-1β production is not decreased by TNFα blockade. IL-1β is produced by both intrinsic renal cells and macrophages/dendritic cells and has a proinflammatory function in acute immune complex–mediated renal inflammatory disease, being involved in crescent formation and inflammatory cell recruitment. In the model of acute nephrotoxic nephritis, leukocyte-derived IL-1β expression is required for the induction of glomerular TNFα expression and for the subsequent recruitment of macrophages to the kidneys (29). Our results are consistent with these findings. Similarly, CXCL13, which is highly expressed by CD11chigh dendritic cells in lymphoid aggregates (8), is not decreased by TNFα inhibition.
As shown here, TNFα inhibition markedly decreased the recruitment of periglomerular and interstitial F4/80high macrophages that amplify tissue injury (10, 24). Periglomerular macrophages may be attracted by locally produced chemokines or may be activated by inflammation mediators to proliferate in situ. In addition, because effluent blood flow from the glomerulus provides the sole blood supply for peritubular capillaries, glomerular hypertension and hypertrophy compromise peritubular blood flow, resulting in hypoxia, tubular activation (including chemokine secretion), and tubular epithelial cell death. This induces resident macrophage activation and peritubular inflammation and is associated with tissue remodeling, fibrosis (30), and finally, irreversible renal damage. Damaged tissue, cytokines, and other inflammation mediators amplify the inflammatory process.
We have previously shown in NZB/NZW mice that several markers of inflammation are up-regulated during renal disease, normalize upon remission induction, and remain at low levels even late after remission induction, when glomerular damage is recurring but interstitial damage has not yet recurred (8). Using protein analysis, we showed that the expression of CCL9, VCAM-1, and lipocalin 2 was decreased in TNFRII Ig–treated mice compared with controls. Lipocalin 2 is a highly sensitive marker for renal tubule damage whose expression increases before any increase in the serum creatinine level, and it has proinflammatory functions (31, 32). Furthermore, oxidative damage is also decreased in TNFRII Ig–treated kidneys, confirming attenuation of end-stage renal disease. In contrast, TNF inhibition does not alter the systemic inflammatory response, as evidenced by failure to regulate lymphocyte activation or to prevent the accumulation of renal perivascular and hilar lymphoid infiltrates. These infiltrates may be attracted by soluble circulating mediators and vascular activation/injury, independently of glomerular immune complex deposition and/or the resulting peritubular hypoxia, and do not appear to be sufficient to mediate end-stage renal failure.
The results of this study show that TNFα is a critical cytokine in the renal effector response to glomerular immune complex deposition, and that regulation of this response is highly therapeutic even in the IFNα-induced lupus model, in which systemic inflammation is difficult to control (6). Judicious inhibition of TNFα may be a potential therapeutic option for rapid control of renal damage in patients who have active lupus nephritis with increased renal expression of TNF and its receptors.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Davidson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Bethunaickan, Sahu, Tang, Huang, Ramanujam, Davidson.
Acquisition of data. Bethunaickan, Sahu, Liu, Tang, Huang, Edegbe, Tao, Ramanujam, Madaio, Davidson.
Analysis and interpretation of data. Bethunaickan, Sahu, Liu, Tang, Huang, Tao, Madaio, Davidson.