INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

SOLUTE AND WATER TRANSPORT across the peritoneal capillaries during peritoneal dialysis (PD) is best described by the three-pore model (26, 27). This model includes small pores (radius 30-40 Å), probably the interendothelial clefts, that allow the diffusion of low-molecular-weight solutes, and a much lower number of large pores (radius >150 Å), thought to correspond to the venular interendothelial gaps, that are involved in the transport of macromolecules such as albumin and IgG (12). The existence of a third, ultrasmall, type of pores (radius <3 Å), which allows the transport of water but not that of solutes, has been postulated based on computer simulations of peritoneal fluid transport during PD (27). Such ultrasmall pores could explain the dissociation between sodium and water transport observed during PD with hypertonic dwells (32); during the first hour, the dialysate-to-plasma ratio of sodium falls markedly, as the result of free water diffusion within the peritoneal compartment. Ultrasmall pores might also account for the osmotic activity of glucose in PD despite its small size (radius 2.9 Å) (14).

The recent identification of a family of proteins (called "aquaporins"), selectively expressed in water-permeable tissues, provides new insights in the molecular mechanisms involved in transcellular water transport (13). Its first member is aquaporin (AQP)-1 (or CHIP28), a 28-kDa channel-forming integral membrane protein identified first in erythrocytes and later in the proximal tubules and descending thin limbs of Henle's loop of the mammalian kidney (21). AQP-2 (or AQP-CD) is located in the apical membrane region of the principal cells of renal collecting ducts (20), where its expression is tightly regulated by vasopressin (7). Other aquaporins (AQP-3, -4, and -5) have been identified in the kidney and in other water-permeable tissues such as the choroid plexus, the salivary glands, and the lung (13).

Immunohistological studies in rats have demonstrated the presence of AQP-1 in the apical and basolateral membranes of endothelial cells lining continuous capillaries of a variety of tissues (22). Positive staining for AQP-1 has been recently reported in the endothelial cells of the peritoneum of four normal and uremic subjects (23). These data, together with recent functional data obtained in rats (5), suggest that AQP-1 could be the molecular counterpart of the ultrasmall pores. In the present study, we used several well-characterized antibodies and a large series of samples from various types of patients to investigate in detail the expression of AQP-1 in the human peritoneum by immunohistochemistry and Western blot analysis.

A decrease in net ultrafiltration is commonly observed in patients treated by PD. This ultrafiltration failure might be acute (such as in peritonitis) or chronic (such as in long-term PD) but has usually severe clinical consequences for PD patients (14). Because the capillary endothelium represents the major functional barrier for water permeability across the peritoneum (26), modifications in the expression of proteins that are specifically located at that level might be involved in the ultrafiltration failure. Expression of AQP-1 and of constitutive, endothelial nitric oxide synthase (eNOS) isozyme, another functional marker of the endothelium (16), was evaluated in control subjects, in uremic patients given either hemodialysis (HD) or PD, and in patients with various causes of peritoneal inflammation.

	METHODS

Top Abstract Introduction Methods Results Discussion References

Tissue samples. Surgical biopsies of the human peritoneum were obtained under sterile conditions from a total of 11 patients in end-stage renal disease. Four patients were treated by PD [male/female, 2/2; mean age, 24 yr (range, 3-47 yr); mean duration of PD, 13 mo (range, 4-30 mo)]. Biopsy was obtained at the time of renal transplantation (n = 3) or during catheter removal in the context of peritonitis (n = 1). End-stage renal disease in these four patients was attributed to bilateral vesicoureteral reflux, sarcoidosis, congenital nephrotic syndrome, and focal glomerulosclerosis, respectively. One of these patients suffered two episodes of peritonitis (first episode 18 mo before biopsy, no germ isolated; second episode 15 mo before biopsy, with isolation of Streptococcus acidominimus and Neisseria cinerea). Another patient had a clinical peritonitis in a context of cloudy and bloody dialysates, just before catheter removal and biopsy (no germ isolated). None of the PD patients had received intraperitoneal antibiotics at any time; peritonitis episodes were treated by a combination of oral and intravenous antibiotics. Seven patients were on HD [male/female, 3/4; mean age, 37 yr (range, 10-74 yr); mean duration of HD, 42 mo (range, 1-151 mo)]; biopsy was performed during renal transplantation (n = 5) or during insertion of a peritoneal dialysis catheter (n = 2). Control biopsies were obtained in five normal, living, related kidney donors [male/female, 1/4; mean age, 36 yr (range, 27-42 yr)] at the time of nephrectomy. All biopsies were obtained from the median, paraumbilical part of the parietal peritoneum. A similar density in terms of vascular structures was observed in biopsy samples routinely stained with hemalum-eosin. The patients were treated either at the St. Luc Academic Hospital (University of Louvain, Brussels, Belgium), the Johns Hopkins Hospital (Johns Hopkins Medical School, Baltimore, MD), or the Centre Hospitalier de Luxembourg (Luxembourg). All patients gave an oral informed consent.

Antibodies. Two polyclonal anti-AQP-1 antibodies were used. The first anti-AQP-1 antibody, raised in New Zealand White rabbits against purified nonglycosylated human erythrocyte AQP-1, was further affinity purified, as described by Nielsen et al. (21). This antibody is specific for the 4-kDa cytoplasmic COOH-terminal domain of AQP-1 (21) and has been extensively characterized in human and rat kidney, with immunoreactivity located in the apical and basolateral membrane domains of proximal tubule and descending thin limbs of Henle's loop epithelial cells (6, 21). The second anti-AQP-1 antibody, raised in rabbits with an albumin-conjugated synthetic peptide corresponding to the NH₂ terminus of AQP-1, was further affinity purified (30).

Other reagents and supplies. Peroxidase-labeled goat anti-rabbit IgG and goat anti-mouse IgG were from Kirkegaard & Perry Laboratories (Gaithersburg, MD), and rabbit IgG and mouse IgG avidin-biotin peroxidase complex (ABC) kits were from Vector Laboratories (Burlingame, CA). Electrophoresis reagents were from Bio-Rad (Melville, NY), and enhanced chemiluminescence reagents were from Amersham (Arlington Heights, IL). Other reagents and supplies were from Sigma Chemical (St. Louis, MO), J.T. Baker (Phillipsburg, NJ), National Diagnostics (Atlanta, GA), Boehringer (Mannheim, Germany), EM Sciences (Fort Washington, PA), Polysciences (Warrington, PA), and Pierce Chemical (Rockford, IL).

Electrophoresis and immunoblotting. Membrane extracts were prepared from eight peritoneal biopsies (normal subjects, n = 4; uremic patients on PD, n = 2; uremic patients on HD, n = 2), as well as from the three autopsy samples. The samples were obtained after washing in ice-cold neutral buffered salt solution and <2 h storage on ice. Membrane fractions were prepared as described previously in detail (6). After washing in ice-cold PBS, pH 7.2, the peritoneal samples were finely minced with a scalpel in ice-cold homogenization buffer [300 mM sucrose, 25 mM HEPES made to pH 7.0 with Tris, containing the protease inhibitors 1 mM 4-(2-aminoethyl)-benzenesulfonyl fluoride, 1 mM benzamidine, 10 µg/ml leupeptin, 1 µg/ml pepstatin A, 1 µg/ml aprotinin, and 1 µg/ml chymostatin, pH 7.2], and then homogenization was carried out in the cold using a Potter apparatus. The homogenate was centrifuged at 1,000 g for 20 min at 4°C to remove nuclei and cell debris. The supernatant was further centrifuged at 80,000 g for 30 min at 4°C. The pellet (whole cell membranes) was suspended in the ice-cold homogenization buffer, and protein concentrations were determined with the bicinchoninic acid protein assay (Pierce Chemical) using BSA as standard. The samples were then stored at 80°C until use. The extracts from normal human kidney and bovine aortic endothelial cells, used as positive control, were prepared as described (6, 33). Three biopsy samples were not used for electrophoresis because of a very low yield of extraction.

Immunohistochemistry and immunoelectron microscopy. For light microscopical analysis, tissue blocks were prepared from a total of 14 biopsies of human peritoneal samples (5 control subjects, 6 uremic patients on HD, 3 uremic patients on PD), as thoroughly described previously (6). Briefly, after washing in ice-cold PBS, tissues were fixed in 4% paraformaldehyde in PBS (pH 7.4) at 4°C for 4-6 h. After overnight rinsing in three changes of PBS, blocks were dehydrated in graded ethanols and embedded in paraffin. Six-micrometer sections were cut, dewaxed, and rehydrated in a graded series of ethanols. Endogenous peroxidase was blocked by 0.3% hydrogen peroxide for 30 min at room temperature. Nonspecific antibody staining was blocked by incubation with 10% normal goat or horse serum in PBS for 20 min at room temperature in a humidified atmosphere. All subsequent antibody incubations were carried out for 45 min at room temperature in a humidified chamber. Sections were incubated with the primary antibody (anti-AQP-1, anti-AQP-2, anti-eNOS antisera at 1:250; anti-cytokeratin at 1:300) diluted in PBS containing 2% BSA. After three washes of 5 min each in PBS-Tween 20 (0.02%), slides were incubated with biotynylated goat anti-rabbit or horse anti-mouse IgG (Vector Laboratories), washed twice for 5 min each in PBS-Tween 20 (0.02%), and once for 5 min in PBS and then incubated for 45 min with the avidin-biotin peroxidase complex (Vectastain Elite, Vector Laboratories). After one wash of 5 min in PBS followed by two washes of 5 min each in Tris buffer solution, antibody localizations were visualized using aminoethylcarbazole or diaminobenzidine as substrate for peroxidase color development. Sections were mounted in Aquamount (Polysciences) and viewed under a Nikon FXA-Microphot equipped with Nomarsky optics. The specificity of the immunolabeling for AQP-1 was confirmed by the following controls: 1) incubation without primary or secondary antibody, 2) incubation with preimmune rabbit serum, and 3) incubation with nonimmune rabbit serum. The specificity of the immunolabeling for eNOS was demonstrated by incubation with nonimmune mouse IgG (Vector Laboratories).

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Western blot analysis. The affinity-purified antibody raised against the NH₂-terminal part of AQP-1 detected a major band at 28 kDa in membrane extracts from all human peritoneal samples tested, as well as more diffuse bands between 35 and 50 kDa (Fig. 1A, lanes 1-7). This pattern of reactivity for AQP-1 on immunoblots is identical to that observed in membranes isolated from human kidney cortex (Fig. 1A, lane 8); the 28-kDa band represents the nonglycosylated form of the protein, whereas the 35- to 50-kDa bands correspond to glycosylated AQP-1 isoforms (30). Despite the difference in the source of the tissue, the signal pattern for AQP-1 was similar in peritoneal samples obtained at autopsy (Fig. 1A, lanes 1-3) or by biopsy (Fig. 1A, lanes 4-7). The level of AQP-1 expression was also similar in control subjects (Fig. 1A, lanes 1, 4, and 5), uremic patients (lanes 6 and 7), and patients with other conditions (lanes 2 and 3).

View larger version (96K): [in this window] [in a new window]   View larger version (24K): [in this window] [in a new window]   Fig. 1.   Expression of aquaporin (AQP)-1 and endothelial nitric oxide synthase (eNOS) in human peritoneum. A: representative immunoblot for AQP-1 (affinity-purified antibody raised against NH2-terminal domain of human AQP-1, diluted 1:1,000) in membrane extracts from human peritoneum (lanes 1-7) and human kidney cortex (lane 8). Lanes 1 (M61), 2 (M76), and 3 (M48) are autopsy samples from nonuremic patients (loading, 40 µg protein/lane); lanes 4 and 5 (control subjects), 6 (uremic patient on peritoneal dialysis with peritonitis), and 7 (uremic patient on hemodialysis) are biopsy samples (loading, 5 µg protein/lane). A major band at 28 kDa is detected in all samples, as well as more diffuse bands of between 35 and 50 kDa. These bands are identical to those seen in kidney, used here as a positive control for AQP-1 expression; they correspond to unglycosylated and glycosylated isoforms of AQP-1, respectively. Blot was exposed for 2 s. Gel was 12% acrylamide. B: immunoreactivity for eNOS in samples (lanes 1-7) loaded as in A. A band at 135 kDa, identical in size to that seen in bovine aortic endothelial cells (positive control for eNOS expression, lane 8) is identified with variable intensity in lanes 2, 3, and 6. Blot was exposed for 4 min (lanes 1-7) or 10 s (lane 8). Note that a more contrasted print has been used on lane 3 to show signal. Gel was 7.5% acrylamide. C: absolute optical densities of 28-kDa (AQP-1) and 135-kDa (eNOS) bands shown in A and B, respectively (lanes 1-7). Lanes 1-3 are autopsy samples (40 µg protein/lane), whereas lanes 4-7 are biopsy samples (5 µg protein/lane). Irrespective of origin of sample, expression of AQP-1 was similar in all extracts; in contrast, eNOS expression was upregulated in lanes 2 and 6 and, to a minor extent, lane 3.

View larger version (106K): [in this window] [in a new window]   Fig. 2.   Expression of AQP-1 and AQP-2 in human peritoneum. Immunoreactivity for AQP-1 (anti-AQP-1 antiserum raised against COOH terminus, diluted 1:1,000) in membrane preparations from human kidney cortex (lane 1) and human peritoneum (lane 2). A major band at 28 kDa is seen in kidney and peritoneum. Slight shift in mobility of 28-kDa band observed between kidney and peritoneal samples is attributed to relative abundance of AQP-1 in these extracts. No signal is seen when blot with same peritoneal extract is probed with anti-AQP-2 antiserum (dilution 1:1,000) (lane 3) or with preimmune serum for AQP-1 (dilution 1:1,000) (lane 4). Forty micrograms of protein were loaded in each lane, and after transfer, blot was stripped to be incubated with different antibodies/sera. Blot was exposed for 30 s (lanes 1 and 2), 2 min (lane 3), or 15 s (lane 4). Gel was 12% acrylamide.

Immunohistochemistry. The peritoneal membrane consists of two anatomical layers: the interstitial tissue containing peritoneal capillaries and lymphatic vessels and the mesothelium supporting microvilli and cilium (8). On light microscopy, the global architecture is similar in uremic patients and normal controls. However, it appears that the mesothelial layer is often damaged in peritoneal biopsies (8), a fact that was also true for our samples. Examination of our sections stained with hemalum-eosin or with a monoclonal antibody against cytokeratin (a marker of the mesothelial cells, see Ref. 3) showed, indeed, that the mesothelium was often lost at the edge of the section, with only a few isolated cells remaining at their anatomical location (data not shown). The main factor involved in that damage is probably the mecanical abrasion during the biopsy procedure itself (which is also suggested by the relative preservation of the mesothelium in infolded areas of the peritoneum) (data not shown). In addition, a number of conditions, such as ascites or PD itself, are known to alter the morphology and the integrity of the mesothelium (8).

View larger version (139K): [in this window] [in a new window]   Fig. 3.   Immunohistochemical localization of AQP-1 in human peritoneum. A-D: representative serial sections from a 38-yr-old normal human subject, stained with hemalum-eosin (A and C) or with affinity-purified antibody raised against COOH-terminal AQP-1 (dilution 1:100) (B and D). AQP-1 is very specifically expressed in endothelial cells of peritoneal capillaries and venules (B). In some vascular profiles, immunoreactivity for AQP-1 appears to be more intense in apical membrane region (B, inset). Endothelial reactivity for AQP-1 is restricted to capillaries, venules, or small veins (v), since peritoneal arterioles and small arteries (a) are negative (D). Mesothelium (ciliated cells) is indicated by arrowheads. Original magnification: A and B, ×150; inset, C, and D, ×300. E-G: serial sections from a 38-yr-old control subject stained with anti-AQP-1 antiserum (dilution 1:100) (E), anti-AQP-2 antiserum (dilution 1:250) (F), or preimmune rabbit serum (dilution 1:100) (G). Labeling of endothelial cells in this longitudinal section of a peritoneal venule is specific for AQP-1; there is no cross reactivity with AQP-2, and no signal is observed when sections are incubated with preimmune serum. Arrowheads point to mesothelium (partially damaged on this section). Original magnification, ×150. H: reactivity for AQP-1 in a uremic patient on hemodialysis (30-yr-old woman). AQP-1 expression is restricted to endothelial cells of a peritoneal venule (v). An adjacent arteriole (a) is negative. Original magnification, ×300. I: AQP-1 reactivity in a uremic patient on peritoneal dialysis (10-yr-old girl). Reactivity is located in endothelial cells of a peritoneal venule. Note that some erythrocytes within lumen are stained. Original magnification, ×300.

View larger version (60K): [in this window] [in a new window]   Fig. 4.   Immunohistochemical localization of eNOS in vascular endothelium of human peritoneum. A and B: a representative section of peritoneum of a 73-yr-old uremic patient with acute peritonitis was incubated with a monoclonal antibody against human eNOS (dilution 1:250). A specific immunoreactivity for eNOS is located in endothelial cells lining capillaries and small venules (A) and, in contrast to AQP-1, in endothelium of arterioles and small peritoneal arteries (B). Original magnification, ×400.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

Our results demonstrate in a large series of subjects that AQP-1 is expressed at a remarkably stable level in all peritoneal tissue samples obtained from both normal subjects and uremic patients given either HD or PD. These findings are based on unequivocal Western blot analysis relying on two well-characterized antibodies raised against different domains of human AQP-1. They are confirmed by the very selective histochemical location of AQP-1 in the endothelial cells of nonfenestrated peritoneal capillaries, venules, and small veins. In contrast, AQP-2, the vasopressin-regulated aquaporin, is not expressed in the peritoneum. The eNOS isozyme was found in the endothelium of capillaries, venules, and arterioles. The expression of eNOS was markedly increased in extracts from patients presenting various causes of peritoneal inflammation.

Localization, as well as staining intensity for AQP-1, was virtually identical in normal and uremic subjects whether treated by HD or PD. These observations suggest that neither uremia nor PD alters the normal expression of AQP-1 within the capillary endothelium of the peritoneum. Immunoreactivity for AQP-1 was denser in the apical membrane of the endothelial cells lining some vascular profiles (Fig. 3B, inset), whereas an equivalent labeling of both apical and basolateral membranes was observed in others (Fig. 3E). High apical reactivity for AQP-1 might represent more concentrated antigenic sites (i.e., molecules of AQP-1) in this area. Ultrastructural studies show that the luminal surface of the capillary endothelium contains numerous invaginations that form plasmalemmal vesicles or "caveolae" in the cytosol of the cell (4, 24). A significant amount of AQP-1 is located in these caveolae in rats (5). It is thus conceivable that the strong apical signal includes this immunoreactivity located just beneath the apical membrane. Alternatively, a clustering of AQP-1 labeling in the apical area of some endothelial cells could be because of local conditions, for instance, by interaction with the negatively charged glycocalyx.

The inconsistent mesothelial staining for AQP-1 in our study is certainly related to the poor preservation of the mesothelium, a common finding in peritoneal biopsies (8). A distinct immunoreactivity for AQP-1 has been observed in adequately preserved rat mesothelium (5), but because the mesothelium does not represent a significant functional barrier for water transport in PD (26), the functional importance of AQP-1 at that level remains obscure.

The lack of immunoreactivity for AQP-2 within the peritoneum was expected, since AQP-2 is known to be regulated by vasopressin, whose action is limited to the distal part of the nephron (7). With the knowledge of the relatively high degree of homology between aquaporins in general, this negative staining further demonstrates the lack of cross reactivity and therefore the specificity of the antibodies. The putative expression of other types of aquaporins (AQP-3, AQP-4) in the peritoneum, which has been suggested by RT-PCR data (1), needs to be specifically addressed when antibodies against these aquaporins are appropriately characterized.

Western blot analysis was used to confirm biochemically the expression of AQP-1 within human peritoneal tissues. Using an affinity-purified antibody, we detected in the membrane extracts of all peritoneal samples tested a major band at 28 kDa, as well as diffuse bands between 35 and 50 kDa. This pattern is identical to that obtained for membrane extracts of human kidney, used here as a positive control for AQP-1 expression. The 28-kDa band represents the unglycosylated AQP-1, whereas the diffuse bands correspond to glycosylated isoforms of the protein (21, 30). Similar to what has been observed on immunostaining, the pattern of expression of the two AQP-1 isoforms on immunoblot was virtually identical in control subjects, uremic patients, and nonuremic patients.

Taken together with the present expression study, several lines of evidence suggest that AQP-1 is the molecular counterpart of the ultrasmall pore of the human peritoneum. 1) The distribution of AQP-1 in the endothelium of capillaries is consistent with the predicted topology of the pore. The continuous endothelium of the peritoneal capillaries is the most important barrier for solute transport during peritoneal dialysis; within that structure, the ultrasmall pore must be transcellular, i.e., an intramembrane protein (27). Both immunostaining and immunoblotting data presented here support that very point. 2) AQP-1 is a constitutively expressed and specific pore for water. AQP-1-mediated water permeability has been demonstrated by microinjections of an in vitro transcribed RNA in Xenopus oocytes (25), and an even more direct demonstration of osmotic water permeability of AQP-1 was provided after reconstitution of highly purified AQP-1 within proteoliposomes (36). 3) Recent ultrastructural information about the pore formed by AQP-1 (34) fits the postulated ultrasmall pore size. In vivo, AQP-1 monomers are assembled in the plasma membrane as tetramers, each individual monomer transporting water independently through a pore less than 6 Å in diameter. Such a small size had been postulated in the three-pore model to explain the effectiveness of glucose and other low-molecular-weight solutes as osmotic agents in PD. 4) Elegant functional studies conducted in rats have shown that peritoneal water permeability is significantly inhibited by HgCl₂ (5). Because mercurials are potent inhibitors of most of the aquaporins, the latter study provided in vivo evidence for a role of water channels in water transport during PD. Thus, based on its distribution, water permeability, and structure, AQP-1 is likely to be the molecular counterpart of the predicted ultrasmall pore of the peritoneum.

Peritonitis, as well as other causes of peritoneal inflammation, is consistently associated with a decrease in the net ultrafiltration (14, 15). Based on Western blot analysis, the level of AQP-1 expression was not affected in several patients in our series presenting with clinical and/or histological evidence of peritoneal inflammation. There is thus no obvious correlation between peritoneal AQP-1 expression and water permeability. This stability of AQP-1 expression differs strikingly from the variable expression of eNOS, used here as another functional marker of the endothelium (Fig. 1C). We investigated the expression of eNOS in the peritoneum for two main reasons. First, although eNOS is a "constitutive" nitric oxide synthase isozyme, its expression might be modulated. Analysis of the 5'-flanking region of the human NOS3 gene has revealed regulatory elements that include several AP-1, AP-2, nuclear factor 1, shear stress, and sterol-responsive Cys-regulatory elements (28). These elements are probably involved in the upregulation of eNOS transcript abundance by physiological and pathophysiological factors, such as exercice (29) or hypoxia (2). It is thus conceivable that situations linked to inflammation (hypoxia, stress) might somehow modulate eNOS expression, thereby participating in the increased NO levels observed in case of peritonitis (35). Second, as compared with other endothelial markers such as the von Willebrand factor, the expression of eNOS at the protein level can reliably be assessed by use of specific monoclonal antibodies (18). Immunoblot analysis demonstrated in the present study that eNOS was mostly detected in peritoneal extracts from three patients with a clinical history of ascites and/or peritonitis. Histological examination of two of these samples disclosed signs of peritoneal inflammation. These findings thus suggest a correlation between high expression of eNOS and peritoneal inflammation. This putative correlation is in keeping with recent reports of high peritoneal levels of nitrates during peritonitis (9) and increased levels of eNOS expression in animal models of cirrhosis with ascites (19).

If we assume that AQP-1 represents the ultrasmall pore and that its basal water permeability remains unaffected, the stability of AQP-1 expression renders unlikely that a quantitative loss of ultrasmall pores is responsible for the ultrafiltration failure encountered in cases of peritoneal inflammation. In contrast, an increased expression of eNOS during peritonitis is probably associated with nitric oxide hyperproduction (31, 35) and an attendant nitric oxide-induced vasodilation. Thus the ultrafiltration failure could be related to a hemodynamically mediated increase in glucose absorption, with an ensuing decrease in the osmotic gradient. Finally, it is also conceivable that structural alterations of AQP-1 at the molecular level might play a role in the ultrafiltration failure. Examples of such putative structural alterations include advanced glycosylation of peritoneal proteins, related to prolonged exposure of the peritoneum to high glucose solutions (14), or oxidation of critical residues, secondary to increased reactive oxygen species in case of peritoneal inflammation.