TNF receptor 2 induces IRF1 transcription factor-dependent interferon-β autocrine signaling to promote monocyte recruitment

SUMMARY

Endothelial-dependent mechanisms of mononuclear cell influx are not well understood. We showed that acute stimulation of murine microvascular endothelial cells expressing the receptors TNFR1 and TNFR2 with the soluble cytokine TNF, led to CXCR3 chemokine generation. The TNF receptors signaled through Interferon regulatory factor-1 (IRF1) to induce interferon-β (IFN-β) and subsequent autocrine signaling via the type I IFN receptor and the transcription factor STAT1. Both TNFR2 and TNFR1 were required for IRF1-IFNβ signaling and, in human endothelial cells TNFR2 expression alone induced IFN-β signaling and monocyte recruitment. In vivo, TNFR1 was required for acute renal neutrophil and monocyte influx after systemic TNF treatment, whereas the TNFR2-IRF1-IFN-β autocrine loop was essential only for macrophage accumulation. In a chronic model of proliferative nephritis, IRF1 and renal expressed TNFR2, were essential for sustained macrophage accumulation. Thus, our data identify a pathway in endothelial cells that selectively recruits monocytes during a TNF-induced inflammatory response.

INTRODUCTION

Inflammation is characterized by leukocyte recruitment and is choreographed largely by cytokines such as tumor necrosis factor alpha (TNF), which is widely recognized for its prominent role in inflammatory and autoimmune diseases (Ernandez and Mayadas, 2009). Many of the pro-inflammatory responses of TNF in vivo can be traced to its effects on the vascular endothelium and leukocyte infiltration (Bradley, 2008). TNF induces the expression of endothelial adhesion molecules that support leukocyte-endothelial interactions, and stimulates the local production of chemokines that promote leukocyte activation and transmigration into tissue (Bradley, 2008; Vassalli, 1992). Different chemokines control the movement of distinct subsets of immune cells into tissues (Luster et al., 2005). In cultured endothelial cells, acute TNF stimulation induces neutrophil chemoattractants such as CXCL8 (IL8), CXCL1 (Gro1) and CXCL2 (MIP-2) through the activation of NF-κB and MAPK (Bradley, 2008; Kuldo et al., 2005). In contrast, well-documented mononuclear chemokines, CXCL10 (IP-10), CXCL9 (Mig) and CCl5 (RANTES) are not significantly induced by TNF alone in many endothelial cell culture systems (Hillyer et al., 2003; Piali et al., 1998), despite the rapid induction of CXCL10 in the microvascular endothelium following systemic TNF stimulation (Ohmori et al., 1993).

TNF’s responses are relayed by two distinct receptors, TNFR1, constitutively present on virtually all cell types and TNFR2, expressed on leukocytes and endothelial cells (Bradley, 2008). TNFR1 has been widely studied and exhibits both proinflammatory as well as immunosuppressive roles (Vielhauer and Mayadas, 2007). The function of TNFR2 is less clear. In vitro, its role in endothelial cells is likely underestimated, as it is often undetectable under baseline conditions. Moreover, it more avidly binds the membrane bound versus the soluble form of the cytokine used routinely in in vitro assays (Grell et al., 1995; MacEwan, 2002). In vivo, TNFR2 is induced on the endothelium (Lucas et al., 1997; Vielhauer et al., 2005) and its deficiency in mice protects from the development of experimental proliferative glomerulonephritis (GN) (Vielhauer et al., 2005), and experimental cerebral malaria (Lucas et al., 1997). In both cases, resistance is attributed to a lack of endothelial TNFR2 and leukocyte infiltration. In contrast, in GN, TNFR1 deficiency results in early and transient protection but then leads to enhanced indices of renal injury associated with excessive renal accumulation of T cells that may be attributed to a delay in their apoptosis (Vielhauer et al., 2005). Studies in transgenic mice also support a role for endothelial TNFR2 in leukocyte accumulation. For example, TNFR2 causes perivascular inflammation associated with TNFR2 expression on the vascular endothelium in mice overexpressing a non-cleavable mutant form of TNF in astrocytes (Akassoglou et al., 2003).

TNF regulates cellular responses primarily by influencing gene transcription. There is evidence that TNF induction of IFN-β shapes the transcriptional and biological response to TNF (Fujita et al., 1989; Leeuwenberg et al., 1987; Tliba et al., 2003; Yarilina et al., 2008). TNF induction of IFN-β is dependent on IRF1 in vitro (Tliba et al., 2003; Yarilina et al., 2008) and is a mediator of lethal shock induced by TNF in vivo (Huys et al., 2009). Here, we described a pathway of TNFR2 induced IFN-β production and autocrine signaling in endothelial cells that leads to the generation of chemokines that promote monocyte interaction with the endothelium in vitro, and renal monocyte accumulation in vivo.

RESULTS

TNF induces expression of Type I interferon-response genes

We found that unlike cultured human endothelial cells, mouse heart microvascular endothelial cells (MHEC) constitutively express TNFR2 (encoded by Tnfrsf1b) RNA (data not shown), and protein on the cell surface (Figure 1A). The mediators that increase TNFR2’s expression remain largely undefined. Higher TNFR2 protein expression was detected after lipopolysaccharide (LPS) stimulation or cAMP elevation (Figure 1A). The latter is consistent with the presence of CREB-binding sites in TNFR2’s promoter (Santee and Owen-Schaub, 1996). To explore the contribution of TNFR2 to TNF mediated responses, MHEC from wild-type and Tnfrsf1b^−/− mice were treated with TNF for 4hrs and their transcription profile was assessed via genome-wide microarrays. As anticipated, TNF increased a number of proinflammatory genes. Among these, genes that are established targets of IFN-β represented a large fraction (38%; 120/314) (Table 1). CXCL9, CXCL10, and CCL5, which are well-described IFN inducible mononuclear cell chemoattractants (Lacotte et al., 2009) were the three most highly TNF induced genes, a finding not previously noted in other endothelial cell culture systems (Hillyer et al., 2003; Piali et al., 1998). These chemokines were markedly diminished when TNFR2 was absent (Table 1).

Figure 1. CXCR3 chemokine expression in microvascular endothelial cells is dependent on autocrine IFN-β production.

(A) Left panels: Flow cytometric analysis of non-permeabilized wild-type (wt), Tnfrsf1a^−/− and Tnfrsf1b^−/− MHEC stained with TNFR2 antibody. Full histogram is control IgG staining. Dark grey line histogram is TNFR2 cell-surface expression. Right panels: Immunoblot analysis of protein lysates of wt MHEC treated with LPS or forskolin for the indicated times in hours. (B–D) Wt and Ifnar1^−/− cells were treated for 4 hrs with IFN-β, TNF or PBS and subjected to quantitative real-time PCR (qPCR) for Mx1 (related to the IFN type I response) and CXCR3 chemokines (B), adhesion molecules (C) or neutrophil chemokines (D). *p<0.05 between Ifnar1^−/− cells and similarly treated wt cells. ns : non-significant. (E) qPCR analysis for mRNA expression in wt MHEC treated with TNF for 4 hrs and treated with either a blocking anti-IFN-β antibody or an isotype control IgG. *p<0.05 between TNF with IgG control, and TNF with IFN-β blocking antibody. For (B)-(E) n = 3 to 5 independent experiments. qPCR data are expressed as mean mRNA fold-change compared to untreated wt. Error bars indicate SEM.

Table 1. TNF induced genes in wt and TNFR2 deficient mouse heart endothelial cells identified by transcriptional profiling.

The ten genes most highly upregulated by treatment of wt MHEC with TNF are shown. Mean logarithm of probe intensity of the two independent experiments is given for wt and Tnfrsf1b^−/− MHEC treated with PBS or TNF for 4 hours. Fold-change between PBS and TNF treated groups are shown for both genotypes. Shaded are genes known to be IFN-β inducible. Lower upregulation of genes including Cxcl10, Cxcl9, Ccl5, Cxcl11 and Mx1 is observed in Tnfrsf1b^−/− MHEC.

Rank	Gene Symbol	Gene Name	Probe intensity (Mean logarithm)				Fold- Change wt	Fold- Change Tnfrsf1b−/−
			MHEC wt		MHEC Tnfrsf1b−/−
			PBS	TNF	PBS	TNF
#1	Cxcl10	chemokine (C-X-C motif) ligand 10	6.19	11.89	6.33	11.00	297.65	106.10
#2	Cxcl9	chemokine (C-X-C motif) ligand 9	3.46	7.98	3.50	6.64	91.92	23.00
#3	Ccl5	chemokine (C-C motif) ligand 5	5.71	10.15	6.09	8.92	84.84	16.88
#4	Mx1	myxovirus (influenza virus) resistance1	4.62	8.85	4.41	7.85	69.19	31.45
#5	Ccrl2	chemokine (C-C motif) receptor-like 2	5.32	9.40	5.41	8.28	59.20	17.66
#6	Ubd	ubiquitin D	4.27	8.27	4.60	7.36	54.69	15.72
#7	Cxcl11	chemokine (C-X-C motif) ligand 11	4.13	7.79	3.98	6.70	38.85	15.23
#8	Ccl7	chemokine (C-C motif) ligand 7	8.84	12.49	9.29	12.11	38.67	16.83
#9	2510004L01Rik	RIKEN cDNA 2510004L01 gene	7.86	11.44	7.72	10.59	35.83	17.57
#10	Slc15a3	solute carrier family 15, member 3	3.53	7.09	3.88	6.38	35.23	12.21

To explore whether IFN-β signaling contributes to the transcriptional response to TNF we used MHEC lacking the type I IFN receptor (IFNAR, which is encoded by Ifnar1). First, we showed that MHEC treated with recombinant IFN-β expressed Mx1, a well-described IFN type I inducible gene and CXCL9 and CXCL10 in an IFNAR dependent manner (Figure 1B). IFN-β did not induce E-selectin or ICAM-1 (Figure 1C). Notably, recombinant TNF-α induced the expression of adhesion molecules, Mx-1 and chemokines, but only expression of Mx1, CXCL9, and CXCL10 required IFNAR (Figures 1B and 1C) as did CCL5 (data not shown). Thus, like the classical IFN responsive gene Mx1, TNF-induced expression of monocyte chemoattractants CXCL9 and CXCL10 requires IFNAR whereas TNF induction of neutrophil chemokines CXCL1 and CXCL2 did not require IFNAR (Figure 1D). IFN-β increased VCAM-1, and TNF-induced VCAM-1 expression was partially IFNAR dependent (Figure 1C). A highly selective IFN-β-depleting antibody markedly inhibited TNF-induced Mx1 and CXCL10, whereas E-selectin was unaffected (Figure 1E), suggesting that activation of IFN-β signaling after TNF stimulation was based on induction and secretion of IFN-β. The dependence on IFN-β alone is aligned with our observation that none of the microarray probes specific for IFN-α reached significant signal levels after TNF treatment (data not shown). Together, our data indicate that TNFR2 augments IFN-β gene expression and the induction of IFN-β-responsive mononuclear chemokines.

Synergistic role for endothelial TNFR1 and TNFR2 in IFN-β production and autocrine signaling

A mechanism for our findings is that TNFRs induce IFN-β itself as this was abolished in the absence of either TNFR2 or TNFR1 (encoded by Tnfrsf1a), albeit residual expression was slightly higher in Tnfrsf1b^−/− MHEC (Figure 2A). Consistent with this, the induction of Mx1 and monocyte chemokines was significantly reduced in the absence of either TNFR (Figure 2B). The observed higher expression of Mx1, CXCL9, and CXCL10 in the Tnfrsf1b^−/− versus Tnfrsf1a^−/− cells may be secondary to residual IFN-β amounts in the former versus the latter. TNF-induced E-selectin, ICAM and VCAM (Figure 2C), neutrophil chemokines CXCL1 and CXCL2, and the neutrophil and monocyte chemoattractant CCL2 (Figure 2D) did not require TNFR2 when TNFR1 was present. IFN-β induces genes through interaction with IFNAR and activation of STAT transcription factors (Theofilopoulos et al., 2005). Protein blot analysis of TNF stimulated MHEC revealed rapid and sustained induction of STAT1 phosphorylation in wild-type cells that depended on IFNAR, TNFR1 a-nd TNFR2 (Figure 2E). Thus, both TNFR1 and TNFR2 induce the synthesis of IFN-β and subsequent IFN-β autocrine signaling through IFNAR.

Figure 2. TNF Induces IFN-β Production, STAT1 Phosphorylation, and Mononuclear Chemoattractants in Mouse Microvascular Cells that Depend on TNFR1 and TNFR2.

Wt, Tnfrsf1a^−/−, and Tnfrsf1b^−/− MHEC treated with PBS control (−) or TNF (+) for 4 hr were subjected to qPCR for mRNA expression of (A) IFN-β, (B) Mx1 and mononuclear chemokines, (C) adhesion molecules, and (D) neutrophil chemokines and Ccl2. N=3 to 5 independent experiments. *p<0.05 between wt and Tnfrsf1a^−/− or Tnfrsf1b^−/− MHEC after TNF treatment. ns : non-significant. Data are expressed as mean mRNA fold-change. Error bars indicate SEM. E) Western blot of phospho-STAT1 after TNF treatment of wt, Tnfrsf1a^−/−, Tnfrsf1b^−/− and Ifnar1^−/− cells for the indicated times in hours. Tubulin (tub) is the protein loading control. One of 4 representative experiments is shown.

TNF induces IFN-β expression through de novo synthesis of IRF1

Next, we explored the TNF-derived signals that induce IFN-β transcription, which for other stimuli involves members of the IRF family including IRF1, -3, -5 and -7 (Theofilopoulos et al., 2005). TNF-induced generation of IFN-β message was partially dependent on de novo protein synthesis, and was abolished by the NF-κB inhibitor Bay11-782 (Figure 3A). A 4-hour TNF stimulation of MHEC led to a significant increase in IRF1 and IRF5 mRNA (Figure 3B) but not IRF3 or IRF7. A deficiency in TNFR1 led to a reduction in IRF1 and -5 mRNA whereas TNFR2 deficiency primarily affected IRF1 (Figure 3B), the induction of which preceded that of IFN-β (Figure 3C). TNF stimulated IRF1 production was dependent on TNFR1 and partially on TNFR2 as assessed on protein blots (Figure 3D). IRF1 production was preserved in Ifnar1^−/− cells (Figure 3D) indicating that IRF1 is upstream of IFNAR. Studies in Irf1^−/− MHEC showed that IRF1 was essential for TNF mediated IFN-β, Mx1, CXCL9 and CXCL10 mRNA (Figure 3E), and STAT1 phosphorylation (Figure 3F). Thus, IFN-β transcription and IFN-dependent gene expression in MHEC upstream of IFNAR depends on IRF1.

Figure 3. TNF induces IRF1 mediated IFN-β autocrine signaling.

(A) Wild-type MHEC were pretreated with cycloheximide (CHX) or an NFκB inhibitor (Bay11-7082) followed by incubation with TNF (+) or PBS (−) for 4hrs and IFN-β mRNA levels were evaluated by qPCR. (B) qPCR analysis of IRF1, IRF3, IRF5 and IRF7 following a 4hr TNF treatment in wt cells (Left panel). qPCR data are normalized to GAPDH and Δ expressed as 2^−ΔCt to allow relative comparison between the different IRF isoforms. qPCR analysis of IRF1 and IRF5 mRNA expression in wt, Tnfrsf1a^−/− and Tnfrsf1b^−/− cells (Right panel). (C) Time course of IFN-β and IRF1 mRNA expression measured by qPCR in wt cells following TNF treatment. (D) Indicated MHEC were treated with TNF for the times shown and samples were subjected to western blot analysis for IRF1 (arrowhead). Actin is a protein loading control. One representative of 3 experiments is shown. (E) qPCR analysis of IFN-β and target gene expression 4 hrs after TNF treatment of wt and Irf1^−/− MHEC. (F) Wt and Irf1^−/− cells were treated with TNF for the indicated times in hours and STAT1 phosphorylation (arrowhead) was evaluated by western blot. One representative of 3 experiments is shown. N=3 independent experiments for each graph shown except (B) which is N=2. *p<0.05 between wt and the indicated knock-out MHEC after TNF treatment.

qPCR data are expressed as mean mRNA fold-change except where otherwise specified. Error bars indicate SEM.

TNFR2 activation in human endothelial cells induces IRF1, IFN-β and CXCR3 and supports monocyte recruitment

Human umbilical vein endothelial cells (HUVEC) had barely detectable surface levels of TNFR2 (Figure 4A). This low expression of TNFR2 and its known suboptimal engagement with soluble TNF routinely used in vitro likely underestimates TNFR2’s contribution to TNF dependent functions in these cells. Thus we overexpressed TNFR2 in HUVEC (Figure 4A) as this leads to spontaneous receptor clustering and ligand-independent signaling (Gaeta et al., 2000). TNF treatment of GFP-transduced HUVEC led to enhanced expression of adhesion molecules E-selectin, ICAM-1 and VCAM-1 while TNFR2 transduced HUVEC exhibited an increase only in ICAM-1 and VCAM-1 (Figure 4B). TNFR2 overexpression significantly induced transcripts of CXCL9 and CXCL10 whereas soluble TNF treatment of HUVEC led to a smaller increase in these two chemokines (Figure 4C). Secreted protein amounts of CXCL10 and another mononuclear chemokine, CCL5, were only measurable in TNFR2 overexpressing cells (Figure 4D) whereas both TNF stimulated and TNFR2 transduced cells secreted CXCL1 (Figure 4D). TNFR2 overexpression increased IFNβ, Mx-1, IRF1 and IRF7 albeit IRF7 induction was abolished in TNF treated MHEC lacking IFNAR (data not shown) suggesting that IRF7 is downstream of IFNβ-IFNAR signaling. It is notable that TNF treatment of HUVEC led to the induction of IRF1 despite the lack of IFN-β expression (Figure 4E). Thus IRF1 induction is not sufficient for upregulation of IFN-β mRNA as shown in other cell lines (Fujita et al., 1989), and implies that TNFR2 in HUVEC engages additional pathways necessary for IFN-β production. TNFR2 overexpression in human dermal microvascular endothelial cells (HDMEC) also induced Mx-1 and Cxcl10 (Figure S1), suggesting that this pathway operates in large vein and microvascular human endothelial cells. The induction of CXCL10 mRNA in the absence of significant IFN-β or Mx1 in HUVEC and HDMEC stimulated with TNF suggests that canonical TNFR1-dependent signals may also contribute to this process.

Figure 4. TNFR2 overexpression in human umbilical vein endothelial cells induces IRF1, IFN-β and CXCR3 chemokines and supports monocyte recruitment.

HUVEC were transduced with lentivirus expressing control GFP alone (GFP) or Flag-tagged TNFR2 (TNFR2). (A) Left panel: Non-permeabilized cells were stained with anti-human TNFR2 antibody and subjected to flow cytometric analysis. Grey is control isotype IgG, black and green histograms represent GFP and TNFR2 transduced HUVEC, respectively. Right panel: TNFR2-Flag HUVEC were permeabilized, stained with an anti-Flag antibody and analyzed by confocal microscopy. White arrow indicates TNFR2-Flag staining at the cell surface. (B) Cells transduced with either GFP or TNFR-Flag were treated with PBS or TNF (+TNF) and subjected to flow cytometry analysis as in A) using anti-E-selectin, ICAM-1 or VCAM-1 antibodies. One representative of 3 is shown. (C–E) qPCR analysis of Cxcl9 and Cxcl10 mRNA (C), quantification of secreted chemokines by ELISA (D) and qPCR analysis of IFN-β, Mx1, IRF1 and IRF7 mRNA levels in HUVEC transduced with either GFP or TNFR2-Flag and incubated without (−) or with (+) TNF. ND = not detectable. N=3 independent experiments. See also Figure S1. (F) Neutrophils or mononuclear cells (10–13% were monocytes) were perfused across GFP- or TNFR2-transduced HUVEC. Cells that arrested and transmigrated were enumerated. Data is mean ± SD from 3 independent experiments. Left and middle panels show representative phase-contrast images. Mononuclear cells adhered (arrow) and transmigrated (arrowhead). Right panel: Immunofluorescence staining for cells positive for CD14⁺ (monocyte marker, arrowhead) interacting with the endothelium. qPCR data are expressed as mean mRNA fold-change. Error bars indicate SEM. *p< 0.05 compared to GFP(−)transduced samples.

Next, we assessed the functional consequences of TNFR2 expression in HUVEC for leukocyte recruitment under flow conditions. For this, purified human neutrophils or the mononuclear fraction from human blood were drawn across GFP- or TNFR2- transduced HUVEC. TNFR2, but not GFP overexpression led to significant adhesion and transmigration of mononuclear cells. Only CD14⁺ monocytes present in the mononuclear fraction adhered and transmigrated across TNFR2-HUVEC (Figure 4F). Cxcl10 and its receptor CXCR3 promote monocyte recruitment in mice in vivo (Zhou et al., 2010). Although expression of CXCR3 on circulating human monocytes is debatable (Katschke et al., 2001; Taub et al., 1993), endothelial TNFR2 also induces CCL5 (Figure 4D, Table 1), a well-known monocyte chemoattractant. T cell interactions were not detected, which is in agreement with reports that T cells are unable to bind and transmigrate in the absence of apically perfused CXCL12 (SDFα)(Cinamon et al., 2001). TNFR2 induced VCAM-1 (Figure 4B) a known ligand for α₄β₁ that supports monocyte adhesion and transmigration (Luscinskas et al., 1996). A blocking α₄ antibody abrogated monocyte accumulation (data not shown) suggesting that the TNFR2 induced monocyte interactions with HUVEC are dependent on VCAM-1. TNF stimulated HUVEC also support monocyte accumulation (data not shown) as reported by others (Luscinskas et al., 1996) and have detectable transcripts for mononuclear chemokines despite the absence of measurable IFN-β (Figure 4C, E). This infers that TNFR1 can induce mononuclear chemokines through a canonical, IFN-β independent pathway. In contrast to monocytes, human neutrophils failed to specifically interact with TNFR2-HUVEC (Figure 4F) that may reflect the absence of E-selectin on these cells (Figure 4B), but did accumulate on TNF treated GFP-HUVEC (data not shown).

Evidence of a TNFR2-IFN-β autocrine loop in the kidney following acute TNF stimulation

The physiological significance of the TNF induced IFN-β signaling pathway was evaluated in vivo 2hrs after an intravenous injection of soluble TNF, which did not result in changes in heart rate, blood pressure or peripheral blood counts albeit neutrophil counts were higher (Table S1). STAT1 was strongly activated in the kidney of TNF treated wild-type mice and was markedly diminished in similarly treated Ifnar1^−/− mice (Figure 5A). It was also decreased in both Tnfrsf1a^−/− and Tnfrsf1b^−/− animals to a similar extent implying that in vivo, the contribution of TNFR1 and TNFR2 to IFN-β signaling are equivalent. Moreover, a reduction in phospho-STAT was observed in TNF treated Irf1^−/− animals (Figure 5A) indicating that in vivo, as in vitro, TNF-induced IFN-signaling requires IRF1. TNF strongly induced IRF1 generation in kidneys of wild-type animals whereas it was absent in Tnfrsf1a^−/− mice, partially reduced in Tnfrsf1b^−/− animals and preserved in Ifnar1^−/− animals (Figure 5B). This demonstrates that IRF1 is downstream of the TNFRs and upstream of IFN-β signaling. IFN-β signaling was evident in glomerular endothelial cells as they were positive for pSTAT by immunohistochemistry (Figure 5C). Finally, renal CXCL10 protein amounts increased following TNF treatment of wild-type mice, and were dependent on TNFR1, TNFR2 and IFNAR (Figure 5D). Endothelial cells rather than macrophages are the likely source of this chemokine as macrophages do not express appreciable amounts of CXCL10 following a 2hr TNF stimulation in vivo (Ohmori et al., 1993) or in vitro (Yarilina et al., 2008).

Figure 5. TNFR2 induced IFN-β autocrine signaling is required selectively for renal monocyte recruitment following acute systemic TNF treatment.

Indicated mice were given an i.v. injection of TNF (+) or PBS (−) and kidneys were harvested 2hrs later. STAT1 phosphorylation (A) and IRF1 protein expression (B) were analyzed by western blot of total protein extracts from kidneys. The dashed lines indicate the removal of intervening lanes of the same gel. For (A) and (B), 1 of 3 representative experiments is shown. Bands optical density was quantified and the average fold induction ± SEM normalized to control wt PBS sample is given for each treatment. (C) Phospho-STAT1 immunohistochemistry staining of kidneys of indicated mice treated (+) or not with TNF. Phospho-STAT1 positive staining in glomerular endothelial cells is indicated (arrows) and semiquantitative scores of staining (means ± SEM) are given from 3 independent experiments. (D) Western blot analysis of Cxcl10 protein expression in kidney lysates. Quantification was performed as in (A–B). (E) In PBS (−) or TNF (+TNF) treated mice, renal macrophage and neutrophil recruitment was assessed by immunohistochemistry. Representative pictures show interstitial and periglomerular macrophages (arrows) (upper panel) and glomerular neutrophils (arrows) (middle panel). Results were quantitated and expressed as mean percentage ± SEM normalized to wt + TNF (set at 100). N=3–4 mice per group except for IRF^−/− mice which is n=2. *p< 0.05 between wt treated with TNF (wt +) and indicated knock-out mice. See also Table S1. (F) Flow cytometry analysis of membrane-bound TNF (mTNF) in human neutrophils stimulated with LPS for 30 min. Quiescent (−) or LPS-prestimulated (+) human neutrophils were incubated with MHEC isolated from indicated mice for 2hrs. Mx1 mRNA expression was then quantified by qPCR using specific murine primers. Results are expressed as mean mRNA fold-change. Error bars indicate SEM. n = 3

TNF induced a significant influx of monocyte and neutrophil influx in wild-type animals (Figure 5E), while T cell accumulation was minimal (data not shown). Monocyte recruitment was reduced in Tnfrsf1a^−/−, Tnfrsf1b^−/−, Ifnar^−/− and Irf1^−/− mice. In contrast, neutrophil accumulation, which was absent in Tnfrsf1a^−/− animals, was largely preserved in mice lacking TNFR2, IFNAR or IRF1 (Figure 5E). Thus, the TNF-induced IRF1-IFN-β autocrine loop preferentially mediates monocyte recruitment. To explore the hypothesis that membrane bound TNF engages endothelial TNFR2 and promotes monocyte infiltration, we administered soluble TNF in TNF deficient mice that lack both soluble and membrane TNF. This resulted in neutrophil recruitment but did not induce mononuclear cell infiltration (Figure 5E), inferring that the endogenous membrane bound form of TNF is required for mononuclear influx. Next, we investigated whether neutrophils are a potential source of membrane bound TNF (mTNF). LPS stimulated human neutrophils expressed mTNF (Figure 5F) as previously reported (Wright et al., 2011). LPS or PBS treated neutrophils were extensively washed and incubated with wild-type, Tnfrsf1a^−/− or Tnfrsf1b^−/− MHEC. Two hours later, Mx1 expression was evaluated using primers for murine Mx1. This cross-species approach is feasible as hexameric human TNF, which may resemble membrane-anchored TNF, binds murine TNFR2 with equal efficiency as murine TNF (Bossen et al., 2006). PBS treated neutrophils did not induce Mx1 in wild-type MHEC. In contrast, LPS pretreated human neutrophils induced Mx1 in wild-type but not in Tnfrsf1a^−/− or Tnfrsf1b^−/ MHEC. (Figure 5F). Thus membrane anchored TNF on neutrophils can potentially induce IFN-β in endothelial cells through TNFR1 and TNFR2.

Renal derived TNFR2 and IRF1 support mononuclear cell recruitment following anti-glomerular basement membrane nephritis

Chronic proliferative GN induced by anti-glomerular basement membrane (GBM) sera in mice results in renal injury that is dependent on TNF (Timoshanko et al., 2003) and IFNAR (Fairhurst et al., 2009). Following induction of anti-GBM nephritis, Irf1^−/− mice exhibited a reduction in macrophage and neutrophil accumulation as well as histopathological indices of disease (Figure S2). Unexpectedly, proteinuria in Irf1^−/− mice was reduced at early (day 7) but not late time points (day 14 and 21) (Figure S2). The discrepancy in proteinuria and leukocyte infiltration may be attributed to a role for IRF1 and associated IFN-β signaling in directly maintaining the glomerular filtration barrier (Satchell et al., 2007).

We previously reported that renal derived TNFR2 promotes macrophage accumulation at day 21 following induction of nephritis. In contrast, opposing effects of TNFR1 in early and late stages of GN were observed, reflecting both the proinflammatory and suppressive functions of TNFR1 (Vielhauer et al., 2005). Here, we examined glomerular neutrophil influx, as well as a time course of macrophage accumulation in WT, Tnfrsf1a^−/− and Tnfrsf1b^−/− mice following induction of anti-GBM nephritis. Tnfrsf1a^−/− mice had a delay in neutrophil and macrophage infiltration, whereas Tnfrsf1b^−/− animals exhibited a sustained reduction in accumulation of both leukocyte subsets throughout the disease course (Figure S2). Consistent with this, only Tnfrsf1b^−/− and not Tnfrsf1a^−/− mice had a significant reduction in several parameters of renal damage (proteinuria, fibrinoid necrosis and crescents) (Figure S2). Next, we assessed the role specifically of renal derived TNFR2 in leukocyte infiltration. The experiment was designed to also address whether the observed absence of interstitial macrophages in the Tnfrsf1b^−/− mice reflects a curtailed neutrophil influx, or a direct role for TNFR2 in macrophage recruitment. That is, we prepared bone marrow chimeras using WT, Tnfrsf1a^−/− and Tnfrsf1b^−/− mice as recipients of bone marrow harvested not from wild-type mice, but from mice deficient in the common γ-chain (Fcer1g^−/−) (i.e. lacking murine activating FcγRs for IgG) that express human FcγRs, FcγRIIA and FcγRIIIB (hFcγRs-Fcer1g^−/−) on neutrophils. The reason this approach was chosen is because following anti-GBM nephritis, glomerular neutrophil influx in hFcγRs-Fcer1g^−/− mice clearly precedes macrophage accumulation, and is dependent exclusively on neutrophil hFcγR recognition of IgG-immune complexes (Tsuboi et al., 2008). Thus, we anticipated that our study design would lead to neutrophil influx in the bone marrow chimeras independent of the TNFR present in the recipients. We found that WT mice reconstituted with Fcer1g^−/− bone marrow were completely protected from developing renal injury (Figure S2), thus confirming the importance of FcγRs in circulating cells in neutrophil accumulation and disease induction. WT, Tnfrsf1a^−/− or Tnfrsf1b^−/− mice with hFcγRs-Fcer1g^−/− bone marrow exhibited comparable neutrophil accumulation at day 7. Macrophage influx was minimal at this time point. At day 14, a dramatic increase in periglomerular and interstitial macrophage accumulation was observed in both chimeric WT and Tnfrsf1a^−/− mice. In contrast, no increase in macrophage accumulation was observed in chimeric Tnfrsf1b^−/− mice. Proteinuria and histopathological indices of disease were consistently reduced only in the reconstituted Tnfrsf1b^−/− mice. (Figure 6 A, B). Thus, TNFR2 on intrinsic renal cells supports monocyte recruitment in a manner that is independent of the preceding neutrophil influx, and is the primary mediator of renal injury in GN. Our findings may have correlates in human conditions. We observed that TNFR2, absent in normal human kidney, was induced on the glomerular endothelium of patients with anti-GBM glomerulonephritis (Figure 6C). Additional staining of intrinsic renal cells is also observed and may reflect expression of TNFR2 on podocytes.

Figure 6. TNFR2 induced IFN-β autocrine signaling promotes monocyte recruitment following anti-GBM nephritis.

(A–B) Wild-type (wt), Tnfrsf1a^−/− and Tnfrsf1b^−/− mice reconstituted with hFcγR-Fcegr1^−/− bone marrow were subjected to anti-GBM nephritis. A) Proteinuria is graphed. Each data point represents one mouse, and the solid line is the median of the group. The dashed line indicates the mean of proteinuria in WT mice with Fcegr1^−/− bone marrow at day 14. Indicated parameters were assessed following immunohistochemistry for neutrophils and macrophages, and Periodic Acid-Schiff (PAS) staining (fibrinoid necrosis) (N=4–5 per time point). The number of neutrophils per 100 glomeruli, or in 10 hpf (40X) of the interstitium was quantitated. F4/80-positive area in 5 hpf (20X) is given and at day 0 was 396±150. Data are expressed as mean ± SEM. nd = not determined. *p<0.05 compared to wt. B) Representative pictures of F4/80 stained macrophages and Periodic Acid-Schiff (PAS) stained kidney sections of the indicated animals at day 14 (original magnification X20). Arrow: Macrophage influx. Black arrow: fibrinoid necrosis, blue arrow: crescents. See also Figure S2. C) Confocal microscopic analysis of normal kidney (top panels) and a renal specimen from a patient with anti-GBM glomerulonephritis (bottom panels). Tissue was stained with antibody to TNFR (red, left) and an endothelial marker CD31 (green, middle). The merged image (right) shows CD31 and TNFR2 colocalization (arrow) along the peripheral capillary walls. Original magnification ×60.

DISCUSSION

The importance of TNF in inflammatory responses and immune regulation are unquestionable, but the mechanisms by which this cytokine triggers specific cellular responses are not well defined. Moreover, the individual role of TNFR2 is largely debated. Through a comprehensive approach encompassing transcriptional profiling and biochemistry, our studies revealed that acute TNF stimulation induced a strong IFN type I response in endothelial cells through a IRF1-IFN-β-IFNAR-STAT1 regulatory pathway that selectively enhanced the expression of mononuclear chemokines. Optimal generation of IFN-β leading to mononuclear chemokine production relied on TNFR2. In MHEC, TNFR2 plays a supportive role in triggering the IFN-β autocrine loop. However, when overexpressed in endothelial cells, or in vivo, under conditions of more physiological TNFR2 activation (e.g. by membrane anchored TNF), TNFR2 plays a primary role in monocyte recruitment. In vivo, renal macrophage accumulation was dependent on TNFR2 both following acute TNF stimulation and anti-GBM nephritis. On the other hand, TNFR1, although essential for monocyte influx following systemic treatment with soluble TNF, is insufficient to support this function in the context of anti-GBM nephritis. The receptor requirements for neutrophil accumulation depend on the primary initiating stimulus. In the case of systemic soluble TNF TNFR1 predominates while under conditions of in situ immune complex deposition FcgRs primarily operate.

In MHEC, TNFR1 was needed for stimulation of TNF-responsive genes across all the gene clusters including adhesion molecules, chemokines and IFN-β responsive genes. On the other hand, TNF-induction of IFN-β also required TNFR2, suggesting that TNFR2 acts synergistically with TNFR1 for the generation of IFN-β and downstream signaling. TNFR2 and IFN-β signaling were not essential for expression of adhesion molecules and neutrophil chemokines, thus establishing the importance of TNF-induced IFN-signaling in the transcription of a subset of proinflammatory genes. Although the exact mechanism of synergy of TNFR2 with TNFR1 in the context of soluble TNF was not elucidated in our studies, the observation that TNFR2 ligand independent activation in human endothelial cells can generate IFN-β and CXCL10 indicates that this receptor has the capacity to induce, by itself, IFN-β and downstream signaling. Moreover, TNFR2 overexpression is sufficient to upregulate some adhesion molecules such as VCAM-1. Consistent with these findings, TNFR2 expressing HUVEC supported monocyte adhesion and transmigration. These results may explain the observation that TNFR2 transgene expression alone, in the absence of TNF or TNFR1 can lead to multi-organ inflammation (Douni and Kollias, 1998).

Our studies in both endothelial cell culture systems suggests that one critical role of TNFR2 is the induction of IRF1 gene expression. Limited signaling pathways have been identified downstream of TNFR2 and include NFκB signaling, activation of the tyrosine kinase Bmx-Etk and MAPK-JNK (Ernandez and Mayadas, 2009). Whether these signals are involved in IRF1 generation remains to be determined. The induction of IFN-β in response to virus or endotoxin is largely mediated by IRF3 and IRF7, in collaboration with the NFκB and AP-1 family of transcription factors (Doyle et al., 2002; Honda et al., 2006; Sakaguchi et al., 2003; Sharma et al., 2003). Interestingly, in our work, IFN-β is transcriptionally upregulated by TNF through a process requiring de novo synthesis of IRF1. IRF1 has previously been implicated in the ability of TNF to induce IFN-β in macrophages (Yarilina et al., 2008) but this did not require new protein synthesis. Thus, the mechanisms involved in TNF-induced IFN-signaling vary by cell-type and might reflect variances in the relative contribution of TNFR2 in distinct cell types.

Acute TNF stimulation induced STAT1 phosphorylation, CXCL10 expression, and mononuclear cell accumulation in the renal parenchyma, and these events were markedly TNFR2, IRF1 and IFNAR dependent. The stronger effect of TNFR2 deficiency on induction of CXCL10 and pSTAT1 in vivo than in MHEC likely reflects the suboptimal activation of TNFR2 by soluble TNF in tissue culture conditions. In vivo, endothelial TNFR2 could potentially be engaged by recruited neutrophils expressing membrane anchored TNF. In addition to availability of membrane bound TNF, our observation that TNFR2 in endothelial cells is upregulated by LPS or cAMP, suggest that these or other locally generated mediators could influence the TNF dependent repertoire of endothelial chemoattractants in vivo. Indeed, the observed variability in TNF induced CXCR3 chemokine expression in different microvascular beds (Ohmori et al., 1993) may be rooted in differing endothelial TNFR2 expression in distinct anatomic loci. We found that TNFR2 was significantly induced on the glomerular endothelium of patients with anti-GBM nephritis. TNFR2 expression was also induced on intrinsic renal cells. Podocytes have been shown to express TNFR2 in the context of renal allograft rejection (Al-Lamki et al., 2001; Hoffmann et al., 2009) and TNFR2 on these cells promote chemokine production and monocyte migration in vitro (Bruggeman et al., 2011). Thus, we speculate that TNFR2 in podocytes may also engage an IFN-β autocrine loop to generate mononuclear chemokines. The unexpected strong induction of mononuclear chemokines in endothelial cells by TNF-TNFRs in our culture systems indicates that the endothelium may play a central role in supplying local chemoattractant cues. In addition to monocytes, the endothelial TNFR2-IFN-β autocrine pathway may also be relevant to the recruitment of T cells such as the CXCR3-positive Th1 cells (Taub et al., 1996; Xie et al., 2003; Luster et al., 2005).

Our data indicate that TNF induced IFN-β production results in autocrine signaling in endothelial cells that in turn plays a central role in monocyte recruitment during inflammation and may potentially contribute to TNF-mediated pathology in a number of immune-mediated conditions in which a pleiotropic role for IFN type I is recognized. These include systemic lupus erythematous (SLE) (Bennett et al., 2003), rheumatoid arthritis (van der Pouw Kraan et al., 2007), multiple sclerosis (Comabella et al., 2009), sepsis (Decker et al., 2005) and psoriasis (Yao et al., 2008). We speculate that TNFR2 or intermediate molecules such as IRF1 could represent promising targets for therapeutics with potentially safer pharmacological profiles than the current regimen of TNF blockade (Ernandez and Mayadas, 2009).

EXPERIMENTAL PROCEDURES

Mice

The following mice were used in our study, details of which are in Supplemental experimental procedures. Wild-type C57Bl6 (wt), mice deficient in TNFR1 (Tnfrsf1a^−/−), TNFR2 (Tnfrsf1b^−/−), TNFα (Tnf^−/−), IFNAR (Ifnar1^−/−), or IRF1 (Irf1^−/−), and mice transgenically expressing human FcγRIIA and FcγRIIIB on neutrophils, in the absence of murine activating Fcγ receptors (hFcγRs-Fcer1g^−/−). All genetically engineered mice are on a C57Bl/6 background. Further details are in the Supplemental Experimental Procedures.

Preparation and treatment of endothelial cell cultures

Primary MHECs isolation and characterization, and HUVEC and HDMEC isolation are described in Supplemental experimental procedures. MHECs were treated with murine TNF (20ng/ml) or IFN-β (250U/ml), and collected for RNA and Western blot. 20ug/ml neutralizing anti-IFN-β antibody or IgG isotype control was added at the same time as TNF. Purified human neutrophils from fresh blood were stimulated with LPS with gentle shaking. 30 min later, neutrophils were collected, washed extensively with PBS to remove LPS and co-cultured with MHEC for 2hrs. HUVEC and HDMEC were stimulated with 20ng/ml human TNF.

Preparation of human peripheral blood neutrophils and mononuclear cells

Human polymorphonuclear neutrophils (PMNs; >95% pure) were isolated from anticoagulated whole blood drawn from healthy volunteers as described (Luscinskas et al., 1996). The mononuclear fraction was taken from the buffy coat and washed with DPBS plus 2% FBS. 10–13% of the total cell population was monocytes as assessed by FACs analysis with an anti-CD14 mouse IgG2a monoclonal antibody (Invitrogen).

DNA constructs and lentiviral transduction

Details for construction of TNFR2 lentivirus is in Supplemental Experimental Procedure. HUVEC were infected with 0.45μm filtered supernatants of virus expressing TNFR2 or GFP. After 12hrs, cells were cultured in fresh HUVEC media and the transduced cells were used for experiments 3.5 days later.

Live cell analysis of neutrophil and monocyte recruitment under shear flow conditions

1×10⁶ neutrophils or mononuclear cells were drawn across TNFR2- or GFP- transduced HUVEC plated on fibronectin coated coverslips in a flow chamber at 0.75 or 0.20 dynes/cm² respectively for 2 min and then increased to 1.0 dynes/cm² and adhesion and transmigration (TEM) were monitored with a Nikon TE2000 inverted microscope attached to a video camera and recorder as previously described (Luscinskas et al., 1996). Cells that were stationary or displaced less than one cell diameter for 30s were defined as arrested. The percent TEM was calculated as (total transmigrated leukocytes) ÷ (total number of adherent and transmigrated cells)×100. For antibody blocking, mononuclear cells were pretreated with functional blocking murine anti-human α4 monoclonal antibody (20μg/ml; clone HP2/1) for 30 min. For immunostaining, cells were pre-incubated with a non-functional blocking anti-CD14 (Invitrogen) for 30 min prior to their use in the flow assay.

Flow cytometric (FACs) analysis, immunofluorescence and immunoblotting

FACS analysis was conducted on endothelial cells detached with trypsin/EDTA. MHEC TNFR2 was evaluated using a biotinylated anti-mouse TNFR2 antibody and streptavidin-PE reagent. HUVEC were fixed and permeabilized with 0.1% saponin and stained with anti-flag and Alexa488-anti-mouse secondary antibody to analyze TNFR2-flag expression. For human renal tissue, cryosections were stained with antibody to human TNFR2 (Rabbit polyclonal, Abcam) and CD31 (Mouse monoclonal antibody, Dako) and processed for immunofluorescence using appropriate secondary antibodies. Cell lysates prepared from endothelial cells or tissue using RIPA buffer was subjected to SDS-PAGE and western blot analysis following standard protocols and antibodies detailed in Supplemental Experimental Procedures.

Cytokine expression analysis by ELISA

HUVEC were transfected with control GFP or TNFR2 expressing lentivirus and conditioned media was obtained at 96 hr post-transfection. Expression of CXCL1, CXCL10, CCL5 were examined using the human cytokine Multi-Analyte ELISArray kit (Super Array) according to the manufacturer’s protocol.

Transcriptional profiling and real-time quantitative PCR

Confluent MHEC isolated from WT and Tnfrsf1b^−/− mice were treated for 4 hours with PBS vehicle alone or 20ng/ml recombinant soluble TNF. RNA was subsequently isolated and subjected to gene array analysis as detailed in Supplemental Experimental Procedure.

Quantitative real-time PCR. Real-time PCR was performed as described in Supplemental experimental procedures. Except when specified otherwise, results were analyzed using the 2^−ΔCt method (Livak and Schmittgen, 2001) and differences were expressed in mean fold-change ± SEM compared to wild-type PBS control which is set at 1.

Acute, systemic TNF treatment of mice and tissue histochemistry

Mice were given an i.v. injection of murine TNF (5μg) in endotoxin free PBS. 2hrs later, mice were euthanized and kidneys were isolated and decapsulated. Tissues samples were either homogenized in RIPA buffer containing protease and phosphatase inhibitors and processed for western blot analysis, or fixed in neutral formalin and paraffin embedded for immunohistochemistry. Details of immunohistochemistry are in Supplemental Experimental Procedures.

Induction and analysis of anti-GBM nephritis and generation of bone marrow chimeras

Anti-GBM nephritis was induced and analyzed essentially as previously described (Vielhauer et al., 2005) and detailed in Supplemental Experimental Procedures. Bone marrow transplantation: Bone marrow obtained from hFcγRs-Fcer1g^−/− mice was transferred into lethally irradiated WT, Tnfrsf1a^−/− and Tnfrsf1b^−/− mice. One month after reconstitution, anti-GBM nephritis was induced.

Statistical analysis

Data are presented as the mean ± SEM. Statistical differences were assessed using unpaired two-sided t-test or analysis of variance (ANOVA) followed by a Bonferroni’s multiple comparison test when more than two groups were compared. P values less than 0.05 were considered significant.