J Am Soc Nephrol 13: 86–95, 2002 Reactive Oxygen Species Alter in Podocytes: Induction of Granulocyte Macrophage–Colony- Stimulating Factor

STEFAN GREIBER, BARBARA MU¨ LLER, PETRA DAEMISCH, and HERMANN PAVENSTA¨ DT Department of Medicine, Division of Nephrology, University Hospital of Freiburg, Freiburg, Germany.

Abstract. It has been suggested that reactive oxygen radicals volved. The ROS scavengers dimethyl-thio-urea and pyrroli- (ROS) play a crucial role in the pathogenesis of proteinuria and done-dithio-carbamate strongly inhibited increased GM-CSF podocyte injury. It was investigated whether changes in gene production induced by ROS. GM-CSF release was also in- expression might account for ROS-induced podocyte dysfunc- duced when internal ROS production was triggered with tion. Differentiated podocytes were incubated with control NADH, whereas H2O2 had only a small effect. GM-CSF media or with exogenous ROS from the xanthine/xanthine- release by podocytes was also stimulated by lipopolysaccha- oxidase reaction for 4 h. A PCR-based suppressive subtractive ride (LPS), interleukin-1 (IL-1), and phorbolester (PMA). Di- hybridization assay was applied to isolate and clone mRNAs methyl-thio-urea significantly inhibited the LPS-, IL-1–, and that were differentially expressed by exogenous ROS. One PMA-induced GM-CSF production. Activation of the tran- differentially expressed clone was identified as the proinflam- scription factor nuclear factor–␬B (NF-␬B) but not activator matory granulocyte macrophage–colony-stimulating protein–1 was involved in the upregulation of ROS-induced factor (GM-CSF). Regulation of GM-CSF in podocytes was GM-CSF production. The data indicate that GM-CSF is dif- further studied by Northern analysis and enzyme-linked im- ferentially expressed by ROS in podocytes. ROS also partially munosorbent assay. Exogenous ROS caused a concentration- mediate the effects of PMA and IL-1 on podocyte GM-CSF dependent, Ͼ10-fold induction of GM-CSF mRNA after 4 h. A production. Because GM-CSF can enhance glomerular inflam- Ͼ50-fold increase in GM-CSF protein release in podocytes mation and induces mesangial proliferation, these data might that had been stimulated with ROS could be detected. Induc- provide further insight into the mechanisms of ROS-induced tion of GM-CSF protein was inhibited by actinomycin D, glomerular injury. which indicated that increased mRNA transcription was in-

The podocyte plays a crucial role in maintaining the perms- permselectivity (5,6). In these glomerular diseases, pretreat- elective function of the glomerular capillary wall (1). Under ment of animals with the antioxidant probucol or ROS scav- pathophysiologic conditions, the podocyte contributes to the engers markedly prevented foot-process effacement and pro- initiation and progression of a variety of glomerular diseases. teinuria. So far, the mechanisms by which ROS might Membranous nephropathy, minimal change disease, and focal contribute to podocyte damage are incompletely understood. segmental sclerosis in particular have all been related to pri- Because the majority of cellular processes are characterized by mary or secondary podocyte injury (2,3). Overproduction of changes in gene expression, we used a cell culture model of reactive oxygen radicals (ROS) has been found in glomerular differentiated podocytes to study changes in gene expression diseases in which the podocyte is the primary target cell of caused by ROS. Herein, we used a PCR-based suppressive glomerular injury, such as puromycin nephrosis, a model of subtractive hybridization (PCR-SSH) to identify genes in minimal change disease, Heyman nephritis, a model of mem- podocytes that are differentially changed by ROS. PCR-SSH is Ϫ Ϫ branous nephropathy, and the Mpv 17 ( / ) mouse, a model a method based on suppressive PCR that allows creation of for steroid-resistant focal segmental sclerosis (3,4). The release subtracted cDNA libraries for the identification of genes dif- of ROS leads to proteinuria by affecting glomerular endothe- ferentially expressed in response to a stimulus (7,8). PCR-SSH lial, and epithelial cells and disturbing normal glomerular differs from earlier subtractive methods by including a normal- ization step that equalizes for the relative abundance of cDNA within the target population. This modification enhances the Received February 10, 2000. Accepted July 12, 2001. probability to identify the increased expression of low-abun- Correspondence to Hermann Pavensta¨dt, Medizinische Universita¨tsklinik IV, dance transcripts and represents a potential advantage over Nephrologie, Hugstetter Str. 55, D-79106 Freiburg, Germany. Phone: ϩ49-761- 270-3492; Fax: ϩ49-761-270-3245; Email: [email protected] other methods, such as differential display PCR, for identifying differentially regulated genes (9). With the PCR-SSH tech- 1046-6673/1301-0086 Journal of the American Society of Nephrology nique, we demonstrate that the granulocyte macrophage–col- Copyright © 2001 by the American Society of Nephrology ony-stimulating factor (GM-CSF) is differentially expressed by J Am Soc Nephrol 13: 86–95, 2002 Reactive Oxygen Species Alter Gene Expression in Podocytes 87

ROS in podocytes. GM-CSF is a cytokine that regulates the Screening of the Differentially Expressed survival, growth, and differentiation of hematopoietic progen- cDNA Library itor cells (10). In the kidney, GM-CSF exerts its effects pri- The subtracted cDNA library was cloned into a 3.9-kb PCRTM 2.1 marily on macrophages, where it stimulates tumor necrosis vector (TA-cloning kit; Invitrogen, San Diego, CA). Ultracompetent factor and interleukin-1 (IL-1) production of these cells. Both Escherichia coli (INV.␣F') were transformed and plated onto agar have been suggested to play a major role in the plates that contained 50 ␮G/ml ampicillin, 50 ␮G/ml isopropyl-␤-D- ␮ ␤ pathogenesis of glomerular inflammation and proteinuria thiogalactoside, and 50 G/ml 5-bromo-4-chloro-3-indolyl- -D-ga- (11,12). In addition, GM-CSF might serve as a critical signal lactopyranoside. Ninety-six transformed clones were regrown over- night in 100 ␮l of LB media in 96-well microtiter plates. A PCR with for macrophage migration into the glomerulus (13). the nested primer from the adapter sequences was then performed on 1 ␮l of each bacterial culture, to amplify the cloned cDNA inserts. Five microliters of the PCR reaction were then spotted onto two Material and Methods duplicate nylon membranes and cross-linked by ultraviolet irradiation. Cell Culture One nylon filter was then hybridized with a ␣32P-dCTP-labeled part Conditionally immortalized mouse podocytes were cultured as of the forward subtracted library, and the duplicate filter was hybrid- reported elsewhere (14). In brief, podocytes were maintained in RPMI ized with the reverse-subtracted library. Filters were then subjected to 1640 medium (Life Technologies, Eggenstein, Germany) supple- autoradiography. Positive clones that showed differential expression mented with 5% fetal calf serum (Boehringer Mannheim, Mannheim, on both blots were further analyzed by virtual Northern blotting: Germany), 100 kU/L penicillin, and 100 mg/L streptomycin (Life cloned inserts were excised from the vector by EcoRI digestion, Technologies). To propagate podocytes, cells were cultivated at 33°C separated by agarose gel-electrophoresis, and extracted from the gel on type I collagen (permissive conditions), and the culture medium (QIAEX; Quiagen, Heidelberg, Germany). A total of 25 ng of cDNA was supplemented with 10 U/ml recombinant interferon-␥ to enhance were then labeled with a ␣32P-dCTP by random priming and used to the expression of the T antigen. To induce differentiation, podocytes analyze 500 ng of double-stranded c-DNA from each control and were maintained on type I collagen (Biochrom, Berlin, Germany) at X/XO-treated podocytes by virtual Northern analysis. Virtual North- 37°C without interferon-␥ (nonpermissive conditions). Podocytes be- ern analysis shows that cDNA from clone 48 hybridized to a 1.2-kb tween passage 10 and 16 were used in all experiments. To examine the transcript that was strongly upregulated in cells that had been treated effects of ROS on podocyte GM-CSF mRNA expression or GM-CSF with X/XO (Figure 1). The insert was then sequenced with an auto- release, podocytes from one cell pool were plated at a cell density of mated ALF sequencer and then identified by a computer-based Blast 2 104 cells/cm in media that contained 5% fetal calf serum in six-well search of Genbank (15) as part of the mouse GM-CSF cDNA plates for Northern analysis or in 96-well plates for measurements of sequence. GM-CSF protein. Cells were switched to media that contained 1% fetal calf serum 24 h before the experiments and then exposed to various treatments. Northern Analysis of GM-CSF mRNA in Podocytes Differentiated podocytes were grown in six-well plates and re- ceived fresh media 24 h before the experiment. RNA was isolated by Generation of a Differentially Expressed cDNA Library acid phenol extraction (16). RNA (10 ␮g/lane) was size-separated in To screen for genes that are differentially expressed by ROS in agarose/formaldehyde gels and transferred to Hybond nylon mem- podocytes, a PCR-SSH approach (7,8) was used (PCR-Select; Clone branes (Pharmacia, Freiburg, Germany). GM-CSF cDNA was labeled ␣32 Tech, Palo Alto, CA). In brief, RNA was isolated from control cells and with -P-dCTP by use of a random-primer labeling kit (Stratagene, cells that had been stimulated with extracellular superoxide generated Heidelberg, Germany). Hybridization and washes were performed from the xanthine/xanthine-oxidase reaction (X/XO; 50 ␮M/5 or 50 according to the method of Church and Gilbert (17). After analysis of ␣32 mU/ml) for 4 h. cDNA synthesis from 1 ␮g of total RNA from each cell the GM-CSF signal, blots were rehybridized to a -P-dCTP–labeled population was achieved with the SMART PCR cDNA synthesis kit probe for the housekeeping gene GAPDH, to control for variation in (Clone Tech) and subsequent long-distance PCR. cDNA from ROS- loading and transfer among samples. stimulated cells was named “tester-cDNA,” and cDNA from control cells was named “driver-cDNA.” After column purification, PCR products Analysis of GM-CSF Protein by Enzyme-Linked were subjected to RsaI digestion to obtain shorter and blunted molecules. Immunosorbent Assay The tester cDNA was then divided into two portions, and each was Quantitative determination of GM-CSF protein release into the ligated to a different cDNA adapter. Driver cDNA was not ligated to an media by podocytes was performed with a mouse GM-CSF immuno- adapter sequence. Two hybridizations were performed. In the first, an assay (Quantikine; R&D Systems, Abingdon, UK), according to the excess of denatured driver-cDNA was added to each denatured tester manufacturers protocol. Podocytes were grown in 96-well plates for population. Hybridization kinetics led to equalization and enrichment of scheduled incubation with ROS or other stimulators of GM-CSF differentially expressed sequences. In the second hybridization, the two release, as indicated. Inhibitors and stimulators of GM-CSF release primary hybridization samples were mixed together without denaturing were added simultaneously, with the exception of actinomycin D, and denatured driver cDNA was added. The remaining subtracted tester which was added 1 h before the stimulation with ROS. cDNAs could now reassociate and form hybrids with different ends that corresponded to the sequences of the two adapters. The differentially expressed cDNA population of this “forward subtraction” was then Analysis of Nuclear Factor–␬B and Activator Protein– PCR-amplified with nested primers corresponding to the two different 1 Activation by Electrophoretic Mobility Shift Analysis adapter sequences. For further screening steps, a reverse-subtracted probe Activated nuclear factor–␬B (NF-␬B) and activator protein–1 with the original tester cDNA as a driver and the driver cDNA as a tester (AP-1) were assayed in nuclear extracts from podocytes, as described was generated. by Ogata et al. (18). Podocytes were treated with stimulants or vehicle 88 Journal of the American Society of Nephrology J Am Soc Nephrol 13: 86–95, 2002

Figure 1. Isolation of differentially expressed genes in reactive oxygen radical (ROS)–treated podocytes by PCR suppressive subtractive hybridization (PCR-SSH). PCR-SSH: reverse transcription was performed on RNA from control (driver) and ROS-treated (tester) podocytes. Tester cDNA was divided and ligated to two different adapters. In the subsequent hybridization, excess driver cDNA eliminates cDNA not differentially expressed in the tester cDNA. Differentially expressed tester I ϩ II cDNA eventually hybridizes as D molecules and is exponentially amplified by the subsequent PCR reaction with primers corresponding to two different adapter sequences. Differential screening: after cloning of the PCR reaction, which is enriched for differentially expressed cDNA, clones are spotted on duplicate filters. Several clones, including clone 48 (arrow), produced a stronger autoradiogram signal when hybridized to tester cDNA compared with driver cDNA. This finding suggests that these clones contain cDNA that represent differentially expressed mRNA in ROS-treated podocytes. Virtual Northern analysis: a clone 48 cDNA probe detects a differentially expressed 1.2-kb transcript in xanthine/xanthine-oxidase reaction (X/XO)–treated podocytes. Clone 48 is sequenced and identified as a 420-bp fragment of the granulocyte macrophage–colony stimulating factor (GM-CSF) gene by BLAST search. for 20, 60, or 120 min. Thereafter, cells were harvested for collection ate a cDNA library of genes that are differentially expressed by of nuclear extracts. Consensus oligonucleotides for NF-␬B (5'-AGT ROS in podocytes (Figure 1). One hundred clones from the TGA GGG GAC TTT CCC AGG C-3' and 5'-GCC TGG GAA AGT differentially expressed cDNA library were further analyzed by CCC CTC AAC T-3') were purchased from Santa Cruz (Heidelberg, differential screening, and 27 clones generated an increased Germany), and consensus oligonucleotides for AP-1 (5'-CGC TTG hybridization signal when hybridized to RNA from X/XO- ATG AGT CAG CCG GAA-3' and 5'-TTC CGG CTG ACT CAT treated podocytes compared with RNA from control cells (Fig- CAA GCG-3') were purchased from Promega (Mannheim, Germany). At least three different analysis were performed for each experimental ure 1). Out of these 27 clones, clone 48 detected a 1.2-kb setup: sample (nuclear extracts and 32P-labeled oligonucleotide transcript that was highly upregulated in podocytes that had probe), negative control (32P-labeled oligonucleotide probe and no been stimulated with ROS (Figure 1). This cDNA clone was nuclear extracts), and specific inhibition (nuclear extracts and 32P- sequenced and identified by BLAST analysis as a 420-bp labeled oligonucleotide probes and a 100-fold molar excess of unla- fragment of the mouse GM-CSF cDNA. beled oligonucleotide probes).

Statistical Analyses GM-CSF mRNA is Upregulated by X/XO after 4 Data are given as mean Ϯ SEM. Statistical analysis was performed and 12 h by one-way ANOVA for multiple comparisons (Bonferroni’s t test). P To further characterize the induction of GM-CSF mRNA by Ͻ 0.05 was considered to be significant. ROS, a time course of GM-CSF mRNA expression in podo- cytes was investigated. Induction of GM-CSF mRNA in podo- Results cytes was strongest detected at 4 h and was markedly declined GM-CSF is Induced by the X/XO Reaction but still detectable after 12 h after incubation of podocytes with in Podocytes X/XO (Figure 2A). No induction of GM-CSF mRNA was seen To identify genes that are induced by ROS, we used a when podocytes were incubated with xanthine or xanthine- PCR-based cDNA subtractive hybridization strategy to gener- oxidase alone (data not shown). We also stimulated podocytes J Am Soc Nephrol 13: 86–95, 2002 Reactive Oxygen Species Alter Gene Expression in Podocytes 89

with media that contained X/XO for only 30 min and then exchanged the media with fresh media that contained no X/XO. The time course of GM-CSF mRNA induction was not different from the cells that had been stimulated with X/XO for the whole incubation period, but the total increase in GM-CSF mRNA was reduced (Figure 2B). We also found that increase of intracellular superoxide production by diamide (19), caused a four- to fivefold induction of GM-CSF mRNA (data not shown).

Podocytes Release GM-CSF Protein in Response to ROS To investigate whether the increased GM-CSF mRNA levels in response to ROS results in a subsequent increase of protein release, GM-CSF protein concentration was measured in the media of ROS-treated and control podocytes by enzyme-linked immunosorbent assay. After 24 h of incubation with X/XO (50 ␮M/50 mU), GM-CSF levels increased from 0.5 Ϯ 0.3 to 156 Ϯ 20 pg/ml (n ϭ 30). Figure 3 shows the time and concentra- Figure 3. GM-CSF protein is time dependently expressed by ROS. tion dependency of 50 ␮M X and different concentrations of Podocytes were incubated with control media or media that contained ␮ XO (1 to 100 mU/ml) on GM-CSF release. A threshold con- X (50 M) and XO (1 to 100 mU) for indicated times. Data are mean Ϯ SEM from six experiments. *, P Ͻ 0.05 versus control. centration of 5 mU/ml XO was required to induce GM-CSF release in podocytes. After4hofincubation with X/XO, a first significant GM-CSF release was observed. After8hofincu- bation, XO concentrations (5 mU) induced a maximal GM- increased GM-CSF release, whereas the effect of hydrogen ␮ CSF release that stayed elevated over controls even after 24 h peroxide (H2O2, 100 M) was less pronounced (Figure 4). of incubation (Figure 3). Effects of NADH and H2O2 were dose-dependent (data not ␮ To test the hypothesis that intracellularly generated ROS can shown), with NADH 5 mM and H2O2 100 M, respectively, induce GM-CSF release, additional experiments were per- resulting in the strongest increase in GM-CSF protein release. formed with the use of NADH (5 mM), a substrate for the podocyte NAD(P)H oxidases (20). NADH (5 mM) strongly

Figure 2. GM-CSF mRNA is time dependently expressed in ROS- treated podocytes. (A) The time course of GM-CSF mRNA induction

by ROS was studied in control podocytes or in podocytes that had Figure 4. GM-CSF production is also stimulated by H2O2 and acti- been incubated with X/XO (50 ␮M/50 mU/ml) for indicated times. vation of endogenous NAD(P)H-oxidases. Podocytes were incubated (B) The time course of GM-CSF mRNA induction by ROS was with control media or media that contained X (50 ␮M)/XO (50 mU), ␮ studied in control podocytes or in podocytes that had been incubated H2O2 (100 M), or NADH (5 mM), respectively, for 24 h. Data are with X (50 ␮M/ml)/XO (50 mU/ml) for 30 min. mean Ϯ SEM from 12 experiments. *, P Ͻ 0.05 versus control. 90 Journal of the American Society of Nephrology J Am Soc Nephrol 13: 86–95, 2002

At NADH concentrations Ͼ10 mM, GM-CSF release began to was required for increased GM-CSF release. When podocytes decline. were preincubated with actinomycin D (0.1 ␮g/ml) before stimulation with X/XO, increased GM-CSF release was re- ROS Scavengers Dimethyl-Thio-Urea and Pyrrolidone- duced from 172 (15 pg/ml) to control levels (2 Ϯ 1 pg/ml) (n Dithio-Carbamate Inhibit the GM-CSF Release ϭ 12). Figure 5 shows the effect of actinomycin D on GM-CSF Induced by ROS release from podocytes. To further investigate whether X/XO-induced effects were attributable to the generation of ROS, we used oxygen radical IL-1, Phorbolester, and Lipopolysaccharide Increase scavengers to inhibit the effects of ROS. Incubation of podo- GM-CSF Release in Podocytes cytes with the oxygen radical scavenger dimethyl-thio-urea Next we examined whether GM-CSF release is regulated by (DMTU, 10 mM) inhibited GM-CSF release from podocytes agents that have been implicated to participate in glomerular by 81 Ϯ 5% (n ϭ 12), whereas equimolar urea that was used inflammation such as IL-1 (11), phorbolester (PMA) (22), and as a nonscavenger control had no effect on X/XO-induced lipopolysacharide (LPS) (12,23). Stimulation of GM-CSF was GM-CSF release (Figure 5). Another thiol compound and induced by as little as 1 pg/ml IL-1. The effects of IL-1 radical scavenger, pyrrolidone-dithio-carbamate (PDTC), has plateaued at 100 pg/ml. PMA caused an significant increase in been shown to inhibit activation of the GM-CSF release at doses Ն10 to 9 M: the maximal effect was NF-␬B, which is involved in the regulation of CSF-1 in other seen at 10 to 7 M. The maximal effect of PMA on GM-CSF cell lines (21). Incubation of podocytes with PDTC (50 ␮M) release was about twice the effect of IL-1. LPS, when used in completely inhibited the X/XO-induced GM-CSF release from concentrations from 0.1 to 100 ␮g/ml, also had a dose-depen- podocytes (n ϭ 12) (Figure 5). dent effect on GM-CSF release: the maximal effect was seen at a dose of 1 ␮g/ml (data not shown). Increased GM-CSF Production by ROS Requires New Figure 6 shows that IL-1 (50 pg/ml), PMA (0.1 ␮M), and RNA Transcription LPS (1 ␮g/ml) induced a time-dependent increase in GM-CSF To further delineate the regulation of GM-CSF production release in podocytes (n ϭ 6). A significant increase of GM- by ROS in podocytes, we tested whether mRNA transcription CSF release was seen after2hofincubation with PMA and IL-1 and after4hofincubation with LPS. With all three

Figure 5. ROS scavengers dimethyl-thio-urea (DMTU) and pyrroli- done-dithio-carbamate (PDTC) inhibit the GM-CSF release induced by X/XO. Podocytes were incubated with control media or media that contained X (50 ␮M)/XO (50 mU) with or without inhibitors (DMTU, Figure 6. Phorbolester (PMA), interleukin-1 (IL-1), and lipopolysac- 10 mM; urea, 10 mM; and PDTC, 50 ␮M). In an additional set of charide (LPS) cause a time-dependent increase in GM-CSF release. experiments, podocytes were incubated with an inhibitor of RNA Podocytes were incubated in control media or media with PMA (0.1 transcription, actinomycin D (0.1 ␮g/ml), as indicated in the Materials ␮M), IL-1 (50 pg/ml), and LPS (1 ␮g/ml) for indicated times before and Methods section. Data are mean Ϯ SEM from 12 experiments. *, GM-CSF release into the media was measured. Data are mean Ϯ SEM P Ͻ 0.05 versus X/XO. of six experiments. *, P Ͻ 0.05 versus control. J Am Soc Nephrol 13: 86–95, 2002 Reactive Oxygen Species Alter Gene Expression in Podocytes 91 agents, GM-CSF release was still elevated compared with controls, even after 24 h.

Stimulation of GM-CSF Release in Podocytes by PMA, IL-1, and LPS is Partially Mediated by ROS In another set of experiments, we tried to determine whether the induction of GM-CSF release by PMA, IL-1, and LPS was ROS-dependent. Figure 7 shows the effect of ROS scavengers DMTU (10 mM), PDCT (50 ␮M), and N-acetylcystein (NAC, 10 mM) on PMA-, IL-1–, and LPS-induced GM-CSF release by 64 Ϯ 1, 59 Ϯ 2, and 58 Ϯ 2%, respectively. Addition of PDCT to podocytes reduced PMA-, IL-1–, and LPS-induced GM-CSF by 74 Ϯ 3, 78 Ϯ 4, and 85 Ϯ 1%, respectively. Glutathione peroxidase represents an important intracellular defense mechanism against ROS. Glutathione peroxidase does not directly scavenge superoxide but is necessary in the sub- sequent detoxification of hydrogen peroxide, which is gener- ated by the superoxide dismutase from the superoxide anion. Addition of NAC increases intracellular glutathione, which is the substrate for the glutathione peroxidase during the detoxi- Figure 8. Inhibition of RNA synthesis blunts induction of GM-CSF by fication of hydrogen peroxide (24). NAC suppressed the in- PMA, IL-1, and LPS. Podocytes were incubated in control media or crease of GM-CSF release in response to IL-1 and PMA but media with PMA (0.1 ␮M), IL-1 (50 pg/ml), or LPS (1 ␮g/ml) with only marginally attenuated the effects of LPS (Figure 7). or without actinomycin D (AmD, 0.1 ␮g/ml) or dexamethasone (Dex, To test whether induction of GM-CSF release by PMA, 1 ␮M), as indicated, for 24 h before GM-CSF release into the media IL-1, and LPS also requires new mRNA synthesis, stimulation was measured. Data are mean Ϯ SEM of 12 experiments. *, P Ͻ 0.05 of podocytes with these agents were performed in the presence versus no inhibitor. of actinomycin D. Figure 8 demonstrates that actinomycin D completely suppressed the stimulation of GM-CSF release caused by PMA, LPS, and IL-1. Dexamethasone (1 ␮M), ROS, PMA, IL-1, and LPS Induce NF-␬B Activation which not only represses AP-1– and NF-␬B–mediated tran- in Podocytes scriptional activation (25,26) but also may serve as an antiox- Because indirect evidence from inhibitor experiments sug- idant, also significantly suppressed the effects of PMA, IL-1, gested that activation of NF-␬B as well as AP-1 might be and LPS on GM-CSF release (Figure 8). involved in the activation of GM-CSF, we directly analyzed activation of these transcription factors in response to different stimuli. Electrophoretic mobility shift analysis showed that podocytes endogenously contained a small amount of active NF-␬B in the nucleus without any stimulus. Figure 9 shows that treatment with ROS, PMA, IL-1, and LPS for 20, 60, or 120 min increased NF-␬B/DNA binding activity (n ϭ 4). The addition of 100-fold excess of unlabeled NF-␬B–oligonucleo- tide probe as a competitor eliminated the NF-␬B/DNA binding from treated and untreated cells. ROS as well as the other agents tested elicited an early, two- to threefold increase of NF-␬B/DNA binding activity that was detectable after 20 min of stimulation. Increase NF-␬B/DNA binding activity was vis- ible for ϳ120 min. After 240 min, NF-␬B/DNA-binding ac- tivity had returned to nearly control levels in ROS- and PMA- treated cells, whereas upregulation seemed to be more prolonged in LPS- and IL-1–treated cells (data not shown).

ROS Do Not Induce AP-1 Activation in Podocytes Figure 7. Induction of GM-CSF release by PMA, IL-1, and LPS is A basal AP-1/DNA binding activity was detectable in un- mediated by reactive oxygen species. Podocytes were incubated in control media or media with PMA (0.1 ␮M), IL-1 (50 pg/ml), and treated podocytes (Figure 10). An increased AP-1/DNA bind- LPS (1 ␮g/ml) with or without ROS scavengers, as indicated (DMTU, ing activity compared with control cells was detectable in 10 mM; PDTC, 50 ␮M; and NAC 10 mM) for 24 h before GM-CSF PMA-, IL-1–, and LPS-treated podocytes after 60 min (n ϭ 4). release into the media was measured. Data are mean Ϯ SEM of 12 In contrast, X/XO and H2O2 did not induce an increase in experiments. *, P Ͻ 0.05 versus no inhibitor. AP-1/DNA binding activity (n ϭ 5, Figure 10). 92 Journal of the American Society of Nephrology J Am Soc Nephrol 13: 86–95, 2002

␬ Figure 9. ROS, IL-1, LPS, and PMA induce nuclear factor– B Figure 10. ROS, IL-1, LPS, and PMA induce AP-1 binding activity in ␬ ␮ (NF- B) binding activity in podocytes. (A) Effect of H2O2 (250 M), podocytes. Effect of H O (250 ␮M), LPS (1␮g/ml), X/XO (50 ␮ ␮ 2 2 LPS (1 g/ml), X/XO (50 M/50 mU/ml), PMA (100 nM), and IL-1 ␮M/50 mU/ml), PMA (100 nM), and IL-1 (50 pg/ml) on AP-1 ␬ (50 pg/ml) on NF- B activation in podocytes. C indicates unlabeled activation in podocytes. C indicates unlabeled oligonucleotide probe. Ϯ probe. (B) Summary of experiments. Data are mean SD from four (B) Summary of experiments. Data are mean Ϯ SD from four to five Ͻ independent experiments. *, P 0.05 versus control. independent experiments. *, P Ͻ 0.05 versus control.

Discussion ROS are not only produced by infiltrating neutrophils and so far has yet to come. One reason for the weak efficiency of monocytes but are also generated by mesangial cells and podo- ROS scavengers in reversing established glomerular injury cytes during glomerular damage. For example, in the Heymann may be that ROS itself change several signaling cascades of nephritis model of glomerulonephritis, cytochrome b558, a podocytes, which will then maintain podocyte injury by mech- major component of the NADPH oxidoreductase complex, is anisms distinct from ROS. Therefore, because ROS might localized in podocytes, and cytochrome b558 subunits have directly change podocyte function for the longer term, it is of been demonstrated in cultured human podocytes (27). In addi- interest to identify ROS-mediated effects of podocyte function. tion, an increase in glomerular XO activity, due to a conversion This study demonstrates that GM-CSF mRNA protein can of xanthine dehydrogenase to the oxidase form, seems to be be induced in podocytes and characterizes the regulation of responsible for ROS production in this model of glomerular GM-CSF by ROS and other proinflammatory agents. Induction damage (28). Recent studies have provided further evidence of GM-CSF mRNA was not immediate, because it required Ͼ1 that oxidative damage is involved in the pathophysiology of h to be detectable, but was relatively long-lasting, given that human renal disease; Gro¨ne et al. (29) demonstrated oxida- elevated GM-CSF mRNA was still detectable 12 h after podo- tively modified proteins in podocytes, mesangial cells, and cytes had been exposed to exogenous ROS. There was a basal membranes in kidney biopsies of patients with membra- time-dependent activation of GM-CSF release in podocytes in nous nephropathy. In patients with focal segmental glomeru- response to ROS, which was already evident with 5 mU/ml losclerosis, increased expression of the NAD(P)H-oxidase pro- XO. Activation of GM-CSF release in podocytes was triggered tein cytochrome b558 was found (30). However, despite the from both exogenous X/XO and endogenous ROS. Activated

compelling evidence for the contribution of ROS to the pro- NAD(P)H-oxidases or diamide, as well as H2O2, all increased gression of experimental glomerular disease and especially endogenous ROS, although the latter compound had a much podocyte injury, the clinical benefit derived from these insights smaller effect. We could also demonstrate that LPS and IL-1, J Am Soc Nephrol 13: 86–95, 2002 Reactive Oxygen Species Alter Gene Expression in Podocytes 93 as well as activation of the protein kinase C-pathway by PMA, Of interest is also the prolonged time course of GM-CSF induced GM-CSF and that these effects were also partially mRNA increase in response to even short-term stimulation mediated by ROS: the ROS scavenger DMTU diminished with ROS, which indicates that a short-lived exposure of GM-CSF production by only 50% to 60%. Surprisingly, NAC podocytes to ROS induces relatively long-lived sequel. This suppressed the increase of GM-CSF in response to IL-1 and observation might help to explain why there has so far been no PMA but had only marginal effects on LPS-related effects. convincing clinical evidence for the usage of oxygen radical One explantation for this observation is that LPS might induce scavengers in human glomerular diseases: because the short- downstream effectors other than ROS such as tumor necrosis lived ROS cause relatively long-lived alterations of transcrip- factor–␣, given that it has been described in human umbilical tional patterns, a treatment with ROS scavengers might come vein endothelial cells that were cocultivated with LPS-treated too late because (1) the transcriptional cascade has been al- peripheral blood mononuclear cells (31). ready initiated and (2) even a short period of suboptimal Because the increase in protein release was completely inhib- scavenging during dosage intervals might be sufficient to reini- ited by actinomycin D, our data indicate that regulation of GM- tiate the altered transcription pattern. CSF expression by ROS involves translation of protein from What is the possible contribution of GM-CSF to glomerular newly transcribed mRNA. Transcriptional regulation of genes by injury? Local glomerular macrophage proliferation has been ROS has been attributed to enhanced expression and/or DNA considered as an important mechanism of macrophage accu- binding of transcription factors in response to ROS. Numerous mulation during the development of severe nonimmune renal transcription factors have been identified to be ROS-sensitive, injury, such as in the rat remnant kidney. In this model, a tight including fos, jun, myc, erg-1 NF-␬B, heat shock protein, and T association of local macrophage proliferation within areas of cell factor/stem cell factor (32). GM-CSF production was nearly renal injury, such as glomerular segmental lesions and a cor- completely inhibited with PDTC in response to PMA, IL-1, and relation between local macrophage proliferation and progres- LPS. Because PDTC has been shown to suppress activation of sive renal injury, has been demonstrated. Podocyte injury has NF-␬B (33,34), these data point to the fact that activation of this been suggested to initiate and maintain the progression of transcription factor might be involved in the regulation of GM- glomerulosclerosis in the rat remnant kidney; thus, cytokine- CSF in podocytes. Indeed, promoter analysis of the murine CSF-1 mediated interaction between podocytes and macrophages may gene did demonstrate the presence of a NF-␬B binding site (35). contribute to the pathogenesis of podocyte injury in this model Direct analysis of NF-␬B/DNA binding by electrophoretic mobil- (3). Podocyte release of GM-CSF may therefore play a crucial ity shift analysis in podocytes revealed an increase after stimula- role in the inflammatory events by functionally activating tion with ROS, LPS, IL-1, and PMA, which supports the hypoth- mature leukocytes at an inflammatory side, inhibiting their esis that NF-␬B is involved in the regulation of GM-CSF migration away from the focus, and enhancing the proliferation production in podocytes. Transactivation of the phorbol-respon- and differentiation of progenitor cells. In addition, GM-CSF sive transcription factor AP-1 by oxygen radicals might have also release by podocytes may not only alter functions of macro- been involved in the regulation of GM-CSF in podocytes. This phage but also modulate cellular properties of glomerular en- hypothesis would have been supported by the fact that dexameth- dothelial cells in a paracrine fashion. GM-CSF possesses an- asone, which is known to blunt AP-1 as well as NF-␬B–mediated giogenic activity in vivo, and it stimulates proliferation and transcriptional activation, significantly suppressed the effects of repair of mechanically injured cultured endothelial cells. Hat- PMA as well as IL-1 and LPS on GM-CSF release. In electro- tori et al. (40) demonstrated that increased GM-CSF release phoretic mobility shift assays, a moderate increase in AP-1 activ- might also contribute to glomerular injury in other nonimmu- ity in podocytes was observed after stimulation with LPS, IL-1, nologic models of renal disease. In a model of lipid-induced and PMA. In contrast, AP-1 activity was not elevated over con- glomerular injury, they could demonstrate by immunohisto- trols in podocytes stimulated with X/XO or H2O2, which indicates chemistry and in situ hybridization that, before macrophage that AP-1 is not involved in the regulation of ROS-induced infiltration in the glomerulus, a significant upregulation of GM-CSF production in podocytes. In several cultured cells, ROS GM-CSF was observed in glomerular podocytes and mesangial have been shown to activate AP-1 (for review, see Ref. [36]). On cells. the other hand, several studies have reported that ROS-mediated Data from renal biopsy studies have also demonstrated a activation of NF-␬B is not necessarily associated with activation possible role of increased GM-CSF expression for the patho- of AP-1: in smooth-muscle cells, ROS generation induced by genesis of glomerular disease such as mesangial proliferative cisplatin leads to an activation of NF-␬B but not AP-1 (37), and glomerulonephritis: in situ hybridization and immunohisto- ROS-induced expression of epithelial Naϩ channels in lung epi- chemistry data have clearly shown a positive correlation be- thelial cells is associated with NF-␬B but not AP-1 activation (38). tween GM-CSF protein expression and glomerular prolifera- In addition, in alveolar macrophages, the activity of NF-␬B, in tion, macrophage infiltration, and the degree of proteinuria contrast to that of AP-1, was activated after stimulation with LPS (41). (39). It has been discussed that there is no single redox paradigm In a recent investigation by Huang et al. (13), the require- into which AP-1 can be fitted. O2-related molecules do not just ment of GM-CSF for leukocyte-mediated glomerulonephritis interact with AP-1 itself—they regulate other molecules in the was investigated in GM-CSF Ϫ/Ϫ knockout mice by use of pathway. Therefore, there is no consensus models of heterologous and homologous anti-GBM glomeru- whether AP-1 mediates or counters oxidative stresses (36). lonephritis. GM-CSF Ϫ/Ϫ mice showed a significant reduction 94 Journal of the American Society of Nephrology J Am Soc Nephrol 13: 86–95, 2002 in proteinuria and neutrophils influx in the heterologous model, 11. Tam FW, Smith J, Cashman SJ, Wang Y, Thompson EM, Rees as well as reduction of crescent formation, glomerular accu- AJ: Glomerular expression of interleukin-1 receptor antagonist mulation of CD4ϩ T cells, and serum creatine in the homolo- and interleukin-1 beta genes in antibody mediated glomerulone- gous model compared with control animals. This is of partic- phritis. Am J Pathol 145: 126–136, 1994 ular interest, because crescents are best understood as a 12. Messmer UK, Briner VA, Pfeilschifter J: Tumor necrosis factor alpha and lipopolysaccharide induce apoptotic cell death in bo- mixture of glomerular epithelial cells, most likely of parietal vine glomerular endothelial cells. Kidney Int 55: 2322–2337, origin, and macrophages (42). The results of our study now 1999 point to the podocyte as a likely source of GM-CSF in this type 13. Huang XR, Kitching AR, Tipping PG, Holdsworth SR: Delayed of glomerular diseases. type hypersensitivity and crescentic glomerulonephritis require In conclusion, we have demonstrated that podocytes can be GM-CSF whereas neutrophil mediated glomerular injury re- the source of increased GM-CSF production in response to quires both GM-CSF and G-CSF. J Am Soc Nephrol 10: 512A, several stress factors and that increased GM-CSF production is 1999 at large mediated by ROS. This finding, in combination with 14. Mundel P, Reiser J, Borja AZ, Pavenstadt H, Davidson GR, Kriz the investigations of other researchers who elucidated the role W, Zeller R: Rearrangements of the cytoskeleton and cell con- of GM-CSF in glomerular diseases, further underlines the tacts induce process formation during differentiation of condi- critical importance of ROS and the podocyte in the initiation tionally immortalized mouse podocyte cell lines. Exp Cell Res 236: 248–258, 1997 and maintenance of glomerular injury. 15. Benson DA, Boguski MS, Lipman DJ, Ostell J, Ouellette BFF: GenBank. Nucleic Acids Res 26: 1–7, 1998 Acknowledgments 16. Chomczynski P, Sacchi N: Single step method of RNA isolation Podocyte cells were a generous gift from Peter Mundel, Albert by acid guanidinium thiocyanate-phenol-chloroform extraction. Einstein College of Medicine, New York, NY. This study was sup- Anal Biochem 162: 156–159, 1987 ported by Deutsche Forschungs gemeinschaft PA 483/5-1. 17. Sambrook J, Fritsch EF, Manitias T: Molecular Cloning: A Laboratory Manual. 2nd ed. Cold Spring Harbor, New York, References Cold Spring Harbor Laboratory, 1989 1. Laurens W, Battaglia C, Foglieni C, De Vos, R, Malanchini B, 18. Ogata N, Yamamoto H, Kugiyama K, Yasue H, Miyamoto E: Van Damme B, Vanrenterghem Y, Remuzzi G, Remuzzi A: Involvement of protein kinase C in superoxide induced activation Direct podocyte damage in the single nephron leads to albumin- of NF-␬B in human endothelial cells. Cardiovasc Res 45: 512– uria in vivo. Kidney Int 47: 1078–1086, 1996 521, 2000 2. Rennke HG: How does glomerular epithelial cell injury contrib- 19. Kosower NS: Kosower EM: Diamide: An oxidant probe for ute to progressive glomerular damage? Kidney Int 45: S58–S63, thiols. Methods Enzymol 251: 123–133, 1995 1996 20. Greiber S, Mu¨nzel T, Ka¨stner S, Mu¨ller B, Schollmeyer P, 3. Kriz W: Evolving role of the podocyte in chronic renal failure. Pavensta¨dt H: NAD(P)H oxidase activity in cultured human Kidney Blood Press Res 20: 180–183, 1998 podocytes: Effects of adenosine triphosphate. Kidney Int 53: 4. Binder CJ, Weiher H, Exner M, Kerjaschki D: Glomerular over- 654–663, 1998 production of oxygen radicals in Mpv 17 gene-inactivated mice 21. Satriano JA, Shuldiner M, Hora K, Yun X, Zihe S, Schlo¨ndorff causes podocyte foot process flattening and proteinuria. Am J D: Oxygen radicals as second messengers for expression of the Pathol 154: 1067–1075, 1999 monocyte chemoattractant protein. JE/MCP-1, and the monocyte 5. Nath KA, Fischereder M, Hostetter TH: The role of oxidants in colony stimulating factor, CSF-1, in response to tumor necrosis progressive renal injury. Kidney Int 45: S111–S115, 1994 factor-alpha and immunoglobulin G. J Clin Invest 92: 1564– 6. Johnson RJ, Lovett D, Lehrer RI, Couser WG, Klebanoff SJ: 1571, 1993 Role of oxidants and protease in glomerular injury. Kidney Int 22. Ganz MB, Abu-Nader R, Saxena R, Grond J: Protein kinase C 45: 352–359, 1994 beta II isoform is up-regulated in human proliferative glomeru- 7. Diatchenko L, Laz YFC, Campbell AP, Chenik A, Moquadam F, lonephritis. Exp Nephrol 5: 225–232, 1997 Huang B, Lukyanov S, Gurskaya N, Sverdlov ED, Siebert PD: 23. Haas C, Ryffel B, Le Hir M: Lipopolysaccharide-induced glo- Suppression subtractive hybridization: A method for generating merulonephritis develops in the absence of interferon gamma differentially regulated or tissue specific cDNA probes and li- signalling. Exp Nephrol 4: 222–230, 1996 braries. Proc Natl Acad Sci USA 93: 6025–6030, 1996 24. Staal FJT, Roederer M, Herzenberg LA: Intracellular thiols reg- 8. Gurskaya NG, Diatchenko L, Chenik A, Siebert PD, Khaspekov ulate activation of nuclear factor ␬B and transcription of human GL, Lukyanov KA, Vagner LL, Ermolaeva OD, Lukyanov AA, immunodeficiency virus. Proc Natl Acad Sci USA 87: 9943– Sverdlov ED: Equalizing cDNA subtraction based on selective 9947, 1990 suppression of polymerase chain reaction: Cloning of Jurkat cell 25. Scheinman RI, Cogswell PC, Lofquist AK, Baldwin AS Jr: Role transcripts induced by phytohemaglutinin and phorbol 12-myris- of transcriptional activation of I kappa B alpha in mediation of tate 13-acetate. Anal Biochem 240: 90–97, 1996 immunosuppression by . Science 270: 283–286, 9. Liang P, Pardee AB: Differential display of eukaryotic messen- 1995 ger RNA by means of the polymerase chain reaction. Science 26. Jonat C, Rahmsdorf HJ, Park KK, Cato AC, Gebel S, Ponta H, 257: 967–971, 1992 Herrlich, P: Antitumor promotion and antiinflammation: Down- 10. Herrmann F, Brugger W, Kanz L, Mertelsmann R: In vivo modulation of AP-1 (Fos/Jun) activity by hor- biology and therapeutic potential of hematopoietic growth fac- mone. Cell 62: 1189–1204, 1990 tors and circulating progenitor cells. Semin Oncol 19: 422–431, 27. Neale TJ, Ullrich R, Ohja P, Poczewski H, Verhoeven AJ, 1992 Kerjaschki D: Reactive oxygen species and neutrophil respira- J Am Soc Nephrol 13: 86–95, 2002 Reactive Oxygen Species Alter Gene Expression in Podocytes 95

tory burst cytochrome b558 are produced by kidney glomerular promoter for the colony-stimulating factor-1-encoding gene. cells in passive Heymann nephritis. Proc Natl Acad Sci USA 90: Gene 102: 165–170, 1991 3645–3649, 1993 36. Marshall HE, Merchant K, Stamler JS: Nitrosation and oxidation 28. Gwinner W, Plasger J, Brandes RP, Kubat B, Schulze M, Regele in the regulation of gene expression. FASEB J 14: 1889–900, H, Kerjaschki D, Olbricht CJ, Koch KM: Role of xanthin oxidase 2000 in passive Heymann nephritis in rats. J Am Soc Nephrol 10: 37. Nie Z, Mei Y, Ford M, Rybak L, Marcuzzi A, Ren H, Stiles GL, 538–544, 1999 Ramkumar V: Oxidative stress increases A1 adenosine receptor 29. Gro¨ne EF, Malle E, Stocker R, Gro¨ne HJ: Hypochlorite modified expression by activating nuclear factor kappa B. Mol Pharmacol proteins in glomeruli of human membranous nephropathy. JAm 53: 663–669, 1998 Soc Nephrol 8: 537A, 1997 38. Rafii B, Tanswell AK, Otulakowski G, Pitkanen O, Belcastro- 30. Lee HS, Kim YS: Identification of malondialdehyde-altered pro- Taylor R, O’Brodovich H: O2-induced ENaC expression is as- teins (MAP) and superoxide generating NADPH-oxidase system sociated with NF-kappa B activation and blocked by superoxide in human renal biopsies. J Am Soc Nephrol 8: 620A, 1997 scavenger. Am J Physiol 275: L764–L770, 1998 31. Lindner H, Holler E, Ertl B, Multhoff G, Schreglmann M, 39. Pepperl S, Dorger M, Ringel F, Kupatt C, Krombach F: Hyper- Klauke I, Schultz-Hector S, Eissner G: Peripheral blood mono- oxia upregulates the NO pathway in alveolar macrophages in nuclear cells induce programmed cell death in HUVECs and may vitro: role of AP-1 and NF-kappa B. Am J Physiol 280: L905– prevent repair: Role of cytokines. Blood 89: 1931–1938, 1997 32. Pahl HL, Baeuerle PA: Oxygen and the control of gene expres- L913, 2001 sion. Bioessays 16: 497–502, 1994 40. Hattori M, Nicolic-Paterson DJ, Miyazaki K, Isbel N, Lan H, 33. Schreck R, Rieber P, Baeuerle PA: Reactive oxygen intermedi- Atkins R, Kawaguchi H, Ito K: Mechanisms of glomerular ates as apparently widely used messengers in the activation of the macrophage infiltration in lipid-induced renal injury. Kidney Int NF-kappa B transcription factor and HIV-1. EMBO J 10: 2247– 55: S47–S50, 1999 2258, 1991 41. Matsuda M, Shikata K, Makino H, Sugimoto H, Ota Z: Glomer- 34. Toledano MB, Leonard WJ: Modulation of transcription factor ular expression of macrophage colony stimulating factor and NF-kappa B binding activity by oxidation-reduction in vitro. granulocyte macrophage stimulating factor in patients with var- Proc Natl Acad Sci USA 88: 4328–4332, 1991 ious forms of glomerulonephritis. Lab Invest 75: 403–412, 1996 35. Harrington MA, Edenberg HJ, Saxman S, Pedigo LM, Daub R, 42. Cattell V: Macrophages in acute glomerular inflammation. Kid- Broxmeyer HE: Cloning and characterization of the murine ney Int 45: 945–952, 1994