Oxidative Stress and Nuclear Factor κB (NF-κB) Increase Peritoneal Filtration and Contribute to Ascites Formation in Nephrotic Syndrome*

Water accumulation in the interstitium (edema) and the peritoneum (ascites) of nephrotic patients is classically thought to stem from the prevailing low plasma albumin concentration and the decreased transcapillary oncotic pressure gradient. However, several clinical and experimental observations suggest that it might also stem from changes in capillary permeability. We addressed this hypothesis by studying the peritoneum permeability of rats with puromycin aminonucleoside-induced nephrotic syndrome. The peritoneum of puromycin aminonucleoside rats displayed an increase in the water filtration coefficient of paracellular and transcellular pathways, and a decrease in the reflection coefficient to proteins. It also displayed oxidative stress and subsequent activation of NF-κB. Scavenging of reactive oxygen species and inhibition of NF-κB prevented the changes in the water permeability and reflection coefficient to proteins and reduced the volume of ascites by over 50%. Changes in water permeability were associated with the overexpression of the water channel aquaporin 1, which was prevented by reactive oxygen species scavenging and inhibition of NF-κB. In conclusion, nephrotic syndrome is associated with an increased filtration coefficient of the peritoneum and a decreased reflection coefficient to proteins. These changes, which account for over half of ascite volume, are triggered by oxidative stress and subsequent activation of NF-κB.

Water accumulation in the interstitium (edema) and the peritoneum (ascites) of nephrotic patients is classically thought to stem from the prevailing low plasma albumin concentration and the decreased transcapillary oncotic pressure gradient. However, several clinical and experimental observations suggest that it might also stem from changes in capillary permeability. We addressed this hypothesis by studying the peritoneum permeability of rats with puromycin aminonucleosideinduced nephrotic syndrome. The peritoneum of puromycin aminonucleoside rats displayed an increase in the water filtration coefficient of paracellular and transcellular pathways, and a decrease in the reflection coefficient to proteins. It also displayed oxidative stress and subsequent activation of NF-B. Scavenging of reactive oxygen species and inhibition of NF-B prevented the changes in the water permeability and reflection coefficient to proteins and reduced the volume of ascites by over 50%. Changes in water permeability were associated with the overexpression of the water channel aquaporin 1, which was prevented by reactive oxygen species scavenging and inhibition of NF-B. In conclusion, nephrotic syndrome is associated with an increased filtration coefficient of the peritoneum and a decreased reflection coefficient to proteins. These changes, which account for over half of ascite volume, are triggered by oxidative stress and subsequent activation of NF-B.
Nephrotic syndrome (NS) 3 is a multifactorial glomerular disease defined by massive proteinuria and hypoalbuminemia. Irrespective of its etiology, NS is commonly associated with renal retention of sodium leading to the generation of ascites and/or edema (1,2). Association of sodium retention with edema rather than with hypertension suggests that fluid flux across the capillary endothelium increases, and accordingly the capillary filtration capacity is 2-fold higher in nephrotic patients (3). Classically, this increase is thought to stem from hypoalbuminemia and a decreased oncotic gradient across the capillary wall. However, several observations militate against this theory: 1) the transcapillary gradient of oncotic pressure is almost unchanged in nephrotic patients (4,5); 2) during steroid-induced remission of nephrotic syndrome, natriuresis resumes and edema decreases before normalization of serum albumin levels (6); and 3) diuretic treatments reduce edema without significantly changing the oncotic pressure gradient (5). Consequently, a new concept has emerged according to which the asymmetry of volume expansion in NS results from alterations of the intrinsic properties of the endothelial filtration barrier, i.e. its water permeability and/or its reflection coefficient to proteins. However, this hypothesis has not been demonstrated experimentally, and neither the molecular basis of these capillary alterations nor their connection to proteinuria and/or hypoalbuminemia is elucidated. Several studies have highlighted the pathophysiological importance of reactive oxygen species (ROS) in nephrotic patients (7)(8)(9), and ROS induce endothelial barrier dysfunction (10) that leads to tissue edema in the context of inflammation or ischemia (11). The aims of this study were to characterize the intrinsic changes of the peritoneal barrier in nephrotic rats and the role of ROS in ascites formation.
injection. Control rats were treated with either NAC or JSH23 or water using the same protocol.
The rats were sacrificed at day 6 post-PAN injection, when sodium retention, proteinuria, and ascites are maximal (14). The greater omentum was immediately dry-frozen in liquid nitrogen and stored at Ϫ80°C. Liver and abdominal muscle samples were rapidly immersed in OCT, frozen in liquid nitrogen, and stored at Ϫ80°C. Blood was collected, and plasma was separated and stored at Ϫ20°C.
Metabolic Studies-The animals were housed in individual metabolic cages with free access to food and water, starting 3 days before the onset of experimentation for acclimatization, and 24 h of urine was collected. Urine creatinine and protein concentrations and plasma protein concentrations were measured in an automatic analyzer (Konelab, Thermo, France). The amount of ascites was measured by moistening and weighing an absorbent paper.
Peritoneal Transport of Fluid-Body temperature of anesthetized rats was monitored at 37°C. A customized catheter was introduced into the peritoneal cavity and firmly attached to the abdominal muscle. 4 ml of saline solutions with different osmotic pressures (see below) equilibrated at 37°C were rapidly infused in the peritoneal cavity, the stomach was gently massaged, and aliquots of Ϸ200 l were taken at various times after infusion. We successively infused solutions of increasing osmolarity (340 -400 mOsm/liter, with 20 mOsm/liter steps). Between each dwell, the peritoneum was drained, rinsed with the forthcoming solution, and drained anew. Blood samples were collected before and at the end of the experiment.
The osmolarities were determined by osmometry (Microosmometer digital, Roebling). The water flux (J v in ml/min) across the peritoneum was calculated as, where V i is the volume of solution injected (in ml), Osm i and Osm c are the osmolarities of the injected and collected solutions, and t is the time between infusion and collection.
The infused solution contained 120 mM NaCl, 25 mM NaHCO 3 , 4 mM KCl, 1 mM CaCl 2 , 1 mM glucose, and mannitol up to the required osmotic pressure at pH 7.4. The osmotic gradient across the peritoneum was calculated as the difference between the osmotic pressure of the infused solution and that of the plasma. Osmolarities (Osm in mOsm/liter) were converted to osmotic pressure (⌸ osmo in mm Hg) using the following formula.
At 37°C, R ϫ T was taken as 18,583 liters⅐mm Hg/mol. Plasma protein concentration was determined on an automated analyzer (Konelab, Thermo, France), and the oncotic pressure was calculated according to the Landis-Pappenheimer equation for plasma proteins (15).

Calculation of Filtration Coefficients and Reflection
Coefficient to Proteins-The peritoneal filtration barrier is a complex structure made up of three layers: the capillary endothelium, the interstitium, and the mesothelium that lines the entire surface of the peritoneum (16). Water flux across the endothelium occurs mainly via paracellular pathways accounted for by small pores permeable to water and small solutes and by large pores permeable to water and macromolecules and by a transcellular pathway accounted for by aquaporin 1 (AQP1) (17). To estimate the contribution of these two pathways, we measured water fluxes before and after inhibition of AQP1 with mercury chloride. In these experiments, we tested three solutions of increasing osmolarity (340 -380 mOsm/liter, with 20 mOsm/ liter steps). After the first three dwells in the absence of HgCl 2 , 10 ml of saline containing 0.1 mM HgCl 2 were infused in the peritoneum and drained after 10 min. Thereafter, the peritoneum was rinsed with the 340 mOsm/liter solution, and the three measurements with solutions of increasing osmolarities containing 0.1 mM HgCl 2 were performed. In these experiments, the water flux across the peritoneal barrier (J v ) can be estimated as, where K 1 and K 2 are the filtration coefficients (the product of the hydraulic conductance L p and the filtration surface area S) of the HgCl 2 -sensitive and -insensitive pathways, respectively, ⌬P, ⌬⌸ onco , and ⌬⌸ osmo , are the gradients of hydrostatic, oncotic, and osmotic pressure across the barrier, respectively, and pl and man are the reflection coefficients to proteins and mannitol, respectively.
The slopes of the regression lines connecting J v to ⌬⌸ osmo in the absence and presence of HgCl 2 (S 1 and S 2 , respectively) are as follows.
The x axis intercepts of the regression lines determined under the same conditions (I 1 and I 2 , respectively) are as follows.
This system of equations cannot be solved to determine the four unknown variables (K 1 , K 2 , pl , and man ) because these four equations are not independent. However, if we fix the value of man , we can then calculate the three other variables as follows.
⌬⌸ onco was determined as described above. In the absence of estimates of ⌬P across the rat peritoneum, we used values derived from human measurements, i.e. 17.6 and 0.9 mm Hg for capillary and peritoneal pressure, respectively (3,18), yielding a ⌬P of 16.7 mm Hg. Importantly the capillary hydrostatic pressure does not change in nephrotic patients (3), and putative changes in peritoneal hydrostatic pressure associated with ascites have no major effect on ⌬P (19). man was taken as 0.02 (20). K 1 , K 2 , and pl were determined in each animal, and the results are the means from different animals.
DHE Staining-DHE was injected intravenously (5 mg/kg body weight) 2 h before sacrifice. Thereafter, pieces of liver and abdominal muscle were collected, frozen in OCT, and kept at Ϫ80°C until use. 5-m cryosections were transferred to Superfrost Goldϩ glass slides and rinsed twice with PBS. The slides were mounted and observed on a fluorescence microscope (ϫ40, Zeiss observer Z1, LSM710). Quantification of DHE labeling was performed using ImageJ.
RNA Extraction and RT-PCR-RNAs were extracted from the greater omentum using Tri Reagent solution, DNase I-treated, and purified on columns according to the manufacturer's instruction (Qiagen). Reverse transcription was performed using a first strand cDNA synthesis kit for RT-PCR (Roche Diagnostics) according to the manufacturer's protocol. Real time PCR was performed using a LightCycler 480 SYBR Green I Master qPCR kit (Roche Diagnostics) according to the manufacturer's protocol. Specific primers (Table 1) were designed using ProbeDesign (Roche Diagnostics).
Statistical Analysis-The results were expressed as means Ϯ S.E. from several animals. Comparison between groups was performed by unpaired Student's t test or by variance analysis, when comparing several groups together; p values Ͻ0.05 were considered significant.

Validation of the Method for Water Flux Measurements-To
find the optimal conditions to measure water fluxes under initial rate conditions (see "Experimental Procedures"), we evaluated the early kinetics of changes in the peritoneal fluid osmolarity following the injection of a solution with the highest osmolarity used in this study (400 mOsm/liter). The osmolarity of the collected fluid decreased linearly with time up to 30 s after injection (Fig. 1A). Consequently, we decided to collect the peritoneal fluid 20 s after injection to determine the water flux (J v ).
The technique exhibits two potential flaws that may lead to underestimating the osmotic gradient: 1) mannitol may leak toward the capillary compartment, and 2) incomplete draining of the peritoneal fluid at the end of dwell n may dilute the fluid infused during dwell n ϩ 1. Possibly because of the short duration of the successive dwells, mannitol diffusion had no significant effect on the osmotic gradient, as attested by the fact that we found no systematic decrease in J v throughout four successive dwells carried out with the same solution (Fig. 1B). The residual volume of peritoneal fluid remaining after drainage of dwell n (Ͻ0.25 ml) would decrease the effective osmolarity of peritoneal fluid during dwell n ϩ 1 by ϳ1 mOsm/liter. To limit this uncontrolled drift, the peritoneal cavity was rinsed with the n ϩ 1 solution between dwells n and n ϩ 1. Under such conditions, the calculated drift in osmolarity was Ͻ0.05 mOsm/liter and was neglected in the calculation of J v . Water Flux in the Peritoneum of Nephrotic Rats-J v varied linearly as a function of the gradient of osmotic pressure ⌬⌸ osmo , and the regression line was much steeper in PAN rats than in controls (in l/min/mm Hg; control: 0.26; PAN: 0.64; p Ͻ 0.001) (Fig. 2). The filtration coefficients of the trans-and paracellular water pathways (K 1 and K 2 , respectively), and the reflection coefficient to proteins ( pl ) were calculated assuming a reflection coefficient to mannitol of 0.02 (20), a mean gradient of hydraulic pressure of 16.7 mm Hg (see methods), and mean gradients of oncotic pressure of 15.6 and 12.7 mm Hg in control and PAN rats, respectively (as calculated from measured plasma protein concentrations). The results showed that 1) the HgCl 2 -sensitive pathway, i.e. the transcellular pathway, accounts for less than 3% of the total filtration coefficient; 2) K 1 and K 2 increased 2.9-and 1.5-fold, respectively, in PAN rats; and 3) pl decreased by ϳ90% in PAN rats ( Table 2).
Role of ROS-RT-qPCR analysis revealed a decreased expression of the superoxide detoxifying enzymes catalase and superoxide dismutase 3 (although not statistically significantly for the latter) in the greater omentum from PAN rats and overexpression of NADPH oxidase subunits (Fig. 3A). This suggests increased ROS levels in PAN rats. This was confirmed by histochemistry, which showed a weak but statistically significant increase in DHE labeling in the parietal and visceral peritoneum of PAN rats (Fig. 3B) and increased protein tyrosine nitration in the greater omentum (Fig. 3C).
Because ROS scavengers reduce glomerular damage and proteinuria in nephrotic animals (21), we tested the effect of NAC treatment on the urinary excretion of proteins in PAN rats. NAC altered neither the kinetics of proteinuria following PAN injection nor its intensity. Accordingly, it did not alter PANinduced hypoalbuminemia (Fig. 4, A and B), indicating that the transcapillary oncotic gradient was similar in control PAN and NAC-treated PAN rats. However, the volume of ascites was markedly reduced in NAC-treated PAN rats (Fig. 4C), suggesting an effect of ROS on the peritoneal permeability. Accordingly, NAC treatment prevented the accumulation of ROS (Fig.  3B) and the increase in J v in the peritoneum of PAN rats (Fig.  4D). Calculations showed that NAC treatment reversed the PAN-induced changes in K 1 , K 2 , and pl ( Table 2).
Role of NF-B Pathway-Because NF-B is a target of ROS, we evaluated whether NF-B is activated in the peritoneum of nephrotic rats and contributes to the increase in peritoneum permeability. Immunofluorescence revealed expression of the p50 active subunit of NF-B in the parietal (Fig. 5A) and visceral peritoneum (not shown) of PAN rats but not in that of controls. RT-qPCR demonstrated overexpression of RANTES and TNF␣, two targets of NF-B, in the greater omentum (Fig. 5B). Activation of p50 was associated with increased phosphorylation of IKK and IB␣ in the greater omentum and decreased abundance of IB␣, suggesting increased degradation (Fig. 5C). The activation of NF-B was prevented by NAC treatment (Fig.  5D). Inhibition of NF-B by treating PAN rats with JSH-23 (Fig.  5D) prevented the increase in J v (Fig. 6A) and reduced the vol-   ume of ascites (Fig. 6B), whereas it altered neither the proteinuria nor the hypoproteinemia (Fig. 6, C and D). AQP1 Expression-PAN rats displayed a 3-fold increase in the expression of the water channel AQP1 mRNA in the omentum (Fig. 7A) and increased AQP1 immunolabeling of the visceral (Fig. 7B) and parietal peritoneum (not shown). NAC and JSH23 treatments prevented the overexpression of AQP1 (Fig.  7, A and B).

Discussion
It has been reported that the peritoneum of nephrotic patients displays increased capillary filtration (3). The present study confirms this finding in the PAN model of NS and demonstrates that enhanced filtration capacity stems from ROSinduced activation of NF-B. In addition, it shows that NS is associated with a decreased reflection coefficient to proteins and with an increased filtration coefficient of both the para-and transcellular pathways of water transport.
The filtration coefficient K is the product of the hydraulic conductance (L p ) and the surface area (S) of the filtration barrier. An increase in K may therefore stem from an increase in L p and/or in S. The surface area of the filtration barrier is difficult to estimate, but we did not observe any macroscopic change, such as hypervascularization of the omentum, which would have suggested a surface area increase in PAN rats. In addition, the concomitant variations in K and AQP1 expression in the peritoneum suggest that the K increase originates from an increase in L p rather than in S, at least for the transcellular pathway. The results also show that the peritoneum of PAN rats displayed a decreased coefficient of reflection to proteins. This indicates that the disease is associated not only with an increase  in the number of pores but also with structural changes that render pores more permeable to proteins.
We observed a marked overexpression of NOX4 and its stabilizing partner p22 phox , whereas overexpression of NOX2 was borderline significant. In opposition to NOX2, which generates superoxide, NOX4 directly releases hydrogen peroxide (22). This isotype specificity of NADPH oxidase induction in the peritoneum of nephrotic rats likely accounts for the weakness of DHE labeling, because DHE is oxidized by superoxide but not by hydrogen peroxide. The NOX4-p22 phox complex is also unique among NADPH oxidases in that its activity is constitutive and is therefore determined by the NOX4 expression level (23). The stimulus for peritoneal NOX4 overexpression during nephrotic syndrome remains to be determined. However, once it is triggered, overexpression of NOX4 might be maintained through positive feedback regulation because we showed that ROS activate NF-B, which in turn may induce NOX4 (24). NF-B may also exert a positive feedback on hydrogen peroxide levels by promoting SOD production (25).
In plasma, albumin is a potent anti-oxidant agent through its dual action as a ROS scavenger and a Cu 2ϩ chelating agent (26). Thus, the hypoalbuminemia observed in PAN nephrotic rats likely increases ROS levels in blood, but it cannot account for peritoneal oxidative stress. Sphingosine 1 phosphate (S1P), a bioactive lipid with increasingly recognized functions, especially on capillary endothelia (reviewed in Ref. 27), could provide a possible link between hypoalbuminemia and endothelial  The values are means Ϯ S.E. from seven or eight rats. ***, p Ͻ 0.001 as compared with controls. C, urinary excretion of proteins before (day 0) and 1-6 days after PAN injection (arrow) to untreated and JSH23-treated rats. Protein excretion is expressed relative to creatinine excretion, and the data are means Ϯ S.E. from six rats. D, plasma protein concentration in the JSH23treated control and PAN rats. The values are means Ϯ S.E. from six to eight rats. Gray zones correspond to the mean Ϯ S.E. domain of variation in JSH23-untreated control (lower zone) and PAN rats (upper zone). ***, p Ͻ 0.001 as compared with the corresponding controls.
oxidative stress: plasma albumin facilitates the release of S1P from red blood cells and its delivery to the endothelium (28), and S1P inhibits NADPH oxidase and ROS production in vascular cells (29). In addition, S1P released from red blood cells maintains a low vascular permeability (30,31). Thus, hypoalbuminemia might decrease S1P availability in the endothelium and thereby induce oxidative stress and increase vascular permeability.
The effects of ROS on NF-B signaling are complex and cellspecific: ROS can alter either the activation of NF-B in the cytosol or its binding to DNA in the nucleus, and both activating and inhibitory effects on the expression of target genes have been reported (reviewed in Ref. 32). Our finding that nephrotic syndrome is associated with phosphorylation of IKK␣/␤ and IB␣ and with increased p50 nuclear labeling suggests that ROS activate NF-B in the cytosol via the canonical pathway. This is at variance with the present consensus that H 2 O 2 activates NF-B independently of IKK through tyrosine 42 phosphorylation of IB␣ (33)(34)(35).
Because AQP1 is not a known direct target of NK-B, its induction requires an intermediate NF-B-induced mediator. TNF␣ is a known target gene of NF-B (36), which is induced in the peritoneum of nephrotic rats and has been reported to increase capillary hydraulic conductance in vitro (37). However, this acute effect cannot account for the transcriptional induction observed in the present study. Moreover, TNF␣ represses rather than induces AQP1 expression (38,39). Another candidate is COX2 (cyclooxygenase 2), a target gene of NF-B (40). Prostaglandin E2, a product of COX2, increases microvascular permeability (41), and activation of the COX2/PGE2 pathway increases AQP1 mRNA expression in human umbilical vein endothelial cells (42).
The reversion of K 1 , K 2 , and pl to control levels in response to NAC reduced the volume of ascites by 60% but did not eliminate it. This suggests that 60% of ascites stems from changes in the intrinsic filtration properties of the peritoneum, whereas the remaining 40% is due to changes in the driving force for water filtration, i.e. the decreased oncotic pressure of plasma. Calculation shows that the observed changes in K 1 , K 2 , and pl in the peritoneum of PAN rats per se would increase water filtration by 42.6 l/min, whereas the 6 mm Hg decrease in the oncotic gradient by itself would increase the water flux by 16.4 l/min. According to these calculations, changes in the intrinsic properties of the filtration barrier and in the oncotic pressure of plasma, respectively, account for 72 and 28% of the water flux increase, which is consistent with our experimental findings.
In summary, our results show that PAN-induced nephrotic syndrome in rats is associated with a marked increase in the water permeability and a decrease in the reflection coefficient to proteins of the peritoneal barrier. These changes, which are triggered by oxidative stress and subsequent activation of NF-B, account for approximately two-third of the volume of ascites.
Author Contributions-K. U., G. B., and A. D. performed the experiments; A. E. analyzed data; K. U., G. B., M. F., B. V., and A. D. conceived the experiments; A. E. and A. D. wrote the manuscript. All authors reviewed the results and approved the final version of the manuscript.  PAN rats (black bars). The values were normalized by Rps23 expression, and the data are means Ϯ S.E. from five or six rats. *, p Ͻ 0.01 as compared with corresponding controls. Bottom panel, AQP1 immunolabeling (green) of visceral peritoneum from untreated (C), NACtreated, and JSH23-treated control and PAN rats. Similar results were observed on parietal peritoneum. Scale bars, 10 m.