K⁺-ATP-channel-related protein complexes: potential transducers in the regulation of epithelial tight junction permeability

ATP-sensitive K⁺ channels (K⁺-ATP channels) are widely distributed throughout the tissues of the mammalian body. They contribute to important cellular actions by coupling the metabolic stage of a given cell to its electrical activity, including the regulation of vital functions, such as cytoprotection or release of hormones (Inagaki and Seino, 1998). K⁺-ATP channels are protein complexes composed of four pore-forming subunits and are members of the Kir6 family of inwardly rectifying K⁺ channels. They are supplemented by four auxiliary subunits that contribute sensitivity to sulfonylurea-type drugs and participate in the regulation of channel activity. These subunits are members of the sulfonylurea receptor (SUR) subfamily, which belongs to the superfamily of the ATP-binding cassette (ABC) proteins (Zingman et al., 2001).

So far, K⁺-ATP channels, like other ion channels, were considered to be passive transducers for ions, regulated by membrane potential, ligand binding or covalent modification (Zingman et al., 2001). The recent discovery of bifunctional protein complexes that combine enzymatic and ion conductance properties raises the possibility of alternative principles for the biological function of Kir6/SUR complexes (Banfi et al., 2000; Bienengraeber et al., 2000; Gadsby and Nairn, 1999; Gulbis et al., 2000; Runnels et al., 2001). Thus, the SUR subunit of the K⁺-ATP channel has been recognized as an ATPase that binds and hydrolyzes the Mg²⁺-ATP complex. Concomitantly, the confirmation of the SUR protein changes from prehydrolytic to posthydrolytic (Zingman et al., 2001). The chemical energy derived from ATP hydrolysis is apparently used to drive the SUR protein from its inactive (prehydrolytic, Kir6-pore closed as in resting conditions) to its active (posthydrolytic, Kir6-pore open) conformation. Similar mechanisms in other ABC proteins have led to the suggestion that SUR proteins change their conformation following changes in the ATP content and, therefore, were considered to be mechanochemical devices (Karpowich et al., 2001; Zingman et al., 2002b). Following this line of evidence, there might be additional SUR-based mechanochemical actions beyond channel gating.

Indeed, Kir6/SUR complexes are known to be involved in the regulation of gap junction permeability in astrocyte cultures (Velasco et al., 2000). The Ca²⁺-induced decrease in permeability between neighbouring cells is partially inhibited by drugs such as tolbutamide or glybenclamide, well-known blockers of the SUR-component of K⁺-ATP channel activity, which directly interact with the SUR subunit (Granda et al., 1998). Moreover, a comparable inhibition has been observed when using other gap junction uncouplers, such as oleic acid, arachidonic acid, endothelin-1, octanol and α-glycyrrhenic acid, suggesting a direct interaction of the SUR subunit with gap junction proteins (Granda et al., 1998).

Irrespective of the drug, the inhibitory effects require the presence of both, SUR and Kir6 proteins, since K⁺-channel openers such as diazoxide and pinacidil, which directly interact with the Kir6 subunit, are able to counteract the SUR-mediated inhibition of gap junction permeability. Interestingly, the Kir6 subunit of astrocytes Kir6.1, localizes predominantly to the very distal parts of astrocyte extensions, where astrocyte gap junctions are found (Thomzig et al., 2001).

Together, these data support the idea that Kir6/SUR protein complexes function as mechanochemical devices (Karpowich et al., 2001; Zingman et al., 2002b). So far, however, no direct data are available that support the physical interaction of the Kir6/SUR protein pair with other mechanochemical systems besides gap junctions. Epithelial cells are characterized by the formation of tight junctions that prevent uncontrolled exchange between the lumen of, for example, the gut and the subepithelial tissue. However, paracellelar permeability may occur and tight junctions are probably involved in its regulation.

Here, we used highly specific, affinity-purified antibodies against both pore-forming Kir6 and auxiliary SUR subunits to screen non-neuronal tissues for new and unexpected locations, which might provide new insights concerning the presumed mechanochemical role of Kir6 and SUR proteins. We found exclusive positioning of the Kir6.1-SUR2A protein pair at regulated but not at unregulated tight junctions, suggesting a new role for these proteins. In addition, we developed an in vitro model that allows us to study in detail the regulation of tight junction permeabilities through Kir6.1-SUR2A protein complexes.

To study the cellular and subcellular distributions of the Kir6.1 and SUR2A proteins, appropriate rat, pig and human tissues (mucosa of stomach, small and large intestine, and liver) were selected. The investigation was focused first on rat and human gastric glands. Antisera against SUR2A yielded a distinct spot- or string-like decoration of the surface of the gastric mucosa (Fig. 1a,b,e). The pattern suggests an antigen distribution restricted to small areas of the plasma membrane between neighbouring epithelial cells (Fig. 1a,b,e). Adjacent sections stained for Kir6.1 immunoreactivities showed identical patterns, suggesting colocalization or complex formation of Kir6.1 and SUR2A proteins (data not shown). Immunostaining for SUR2A and Kir6.1 was abolished by preincubation of antibodies with recombinant protein containing the cognate sequences (Fig. 1a insert and data not shown). Immunoreactivities were found at the intercellular junctional area of all epithelial cells in the gastric glands (Fig. 1b insert 1, see also Fig. 1a-e). Simultaneous incubation of the SUR2A antiserum with an antiserum against the gastric anion exchanger 2 (AE2), a highly specific marker of the basolateral plasma membrane of gastric parietal cells (Jöns et al., 1994), revealed no overlap (Fig. 1b). The SUR2A immunoreactivity was clearly separated from the AE2-positive lateral plasma membrane (Fig. 1b, insert 2). The free apical plasma membranes of the glandular epithelial cells and also the apical and canalicular membranes of the parietal cells (Jöns et al., 1994) were devoid of SUR2A (Fig. 1a,b,e) or Kir6.1 immunoreactivities (Fig. 1c,d). The distinct immunological distribution suggested a specific localization of Kir6.1 and SUR2A at epithelial cell-cell junctions. Simultaneous incubations with antisera against E-cadherin (Fig. 1c,d), p120 (data not shown), ZO-1 (zonula occludens protein 1) (Fig. 1f,g), or occludin (data not shown) further support this assumption. Double staining for Kir6.1 and E-cadherin revealed, besides the typical dot-like appearance of the E-cadherin-containing puncta adherentia at the basolateral plasma membrane, some overlap of both proteins within the zonula adherens (Fig. 1c,d). No E-cadherin staining was found at the most apical segment of the lateral plasma membrane, representing the tight junction area, whereas this subdomain was brightly decorated by the Kir6.1-specific antiserum (Fig. 1d, arrows). In addition, SUR2A immunoreactivity completely colocalized with ZO-1, a marker protein of epithelial tight junctions (Fig. 1e,f,g). Collectively, these data suggest a specific association of Kir6.1-SUR2A with tight junctions. Similar data were obtained when using tissue from the small intestine (human and rat) and from the large intestine (rat, data not shown). No obvious differences between the various species were observed.

The villi of the small intestine are lined by a simple columnar epithelium. The predominant cell type, the absorptive enterocytes are tall columnar cells with a brush border at their lumenal surface representing the microvilli. The mucus-secreting goblet cells are scattered among the enterocytes (Fig. 2a). The surface of the intestinal villi shows invaginations (Fig. 2a, large arrows), which are also visible in the immunofluorescence micrographs of the consecutive section (Fig. 2b-d, large arrows). The junctional complexes between adjacent epithelial cells of the small intestine are located in the most apical part of the lateral membrane, just beneath the microvilli (Fig. 2a, small arrows). A double immunolabeling of the consecutive section with Kir6.1 and ZO-1 antibodies revealed a perfect colocalization of both proteins, most probably highlighting the tight junctions (Fig. 2b-d, small arrows). Similar results were obtained with antibodies directed against SUR2A and ZO-1 (Fig. 2e-g), the arrows indicate the invaginations of the intestinal villi. Comparable to the gastric mucosa, Kir6.1-SUR2A complexes are exclusively localized to tight junctions in the small intestine.

This perfect colocalization of Kir6.1-SUR2A with ZO-1 at tight junctions was confirmed on the ultrastructural level (Fig. 3). The tight junctions of the intestinal columnar epithelium are located directly beneath the lumenal cell-surface, showing the multitude of the microvilli in cross section (Fig. 3a) and in full length (Fig. 3b). The Kir6.1 and ZO-1 immunodecoration, visualized by the 3,3′-diaminobenzidine-precipitation product is precisely located at the plasma membrane area of adjacent epithelial cells that represent the tight junction (Fig. 3a,b).

In the liver, bile acids conjugated to bilirubin and to biotransformed xenogens (drugs) are secreted into small, specialized spaces between adjacent hepatocytes, the bile canaliculi. They are sealed by tight junctions to prevent leakage of the canalicular contents into the blood. Cell fractionation techniques allow for the isolation of liver bile canaliculi and tight junctions at high biochemical purity (Tsukita and Tsukita, 1989).

Immunocytochemical staining for Kir6.1 and ZO-1, a marker for canalicular tight junctions, resulted in an identical staining pattern, representing the distribution of hepatocyte tight junctions sealings (Fig. 4a). A similar colocalization was obtained with antibodies against Kir6.1 and SUR2A (data not shown). To provide further information on a direct molecular interaction between Kir6.1-SUR2A and occludin, immunoblot analysis and coimmunoprecipitations were performed using purified canalicular tight junctions and antibodies against Kir6.1, SUR2A and occludin. The antibodies against occludin, Kir6.1 and SUR2A gave signals at the expected 65 kDa, 43 kDa and 150 kDa, respectively (Fig. 4b). Occludin was also detected in precipitates obtained with Kir6.1 or SUR2A antibody (Fig. 4c). Precipitation with an antiserum against occludin or with a non-immune serum was used for positive or negative controls, respectively (Fig. 4c). In an additional immunoprecipitation experiment with antibodies against occludin, the ratio of precipitated or co-precipitated protein in comparison with the molecules still present in the supernatant after the immunoprecipitation were investigated (Fig. 4d). We found that the Kir6.1 and SUR2A subunit pair co-precipitated in almost equal amounts. The concentration of Kir6.1-SUR2A molecules remaining in the supernatant after the immunprecipitation were also similar to each other. The immunoprecipitated occludin molecules and the molecules in the corresponding supernatant were at the same ratio as co-precipitated Kir6.1-SUR2A molecules. These data suggest that, in rat liver the Kir6.1-SUR2A pair is associated with the tight junctional protein complexes. In addition, the presence of Kir6.1 message in the liver was confirmed by reverse transcriptase (RT)-PCR (Fig. 4e).

To address the question whether the Kir6.1-SUR2A complex is present in both tight junctions with regulated and also in those with non-regulated permeability, we studied their distribution in rat and human kidney epithelia and urinary bladder epithelia.

Kidney epithelial cells of proximal and distal tubules express regulated tight junctions that control the paracellular flux of water and ions. Prominent immunostaining was obtained for Kir6.1 (Fig. 5a) and SUR2A (data not shown), at positions characteristic for tight junctions in proximal and distal tubules in the rat kidney. However, collecting ducts clearly identified by aquaporin-2-positive chief cells, which are known to contain impermeable and non-regulated tight junctions, were devoid of Kir6.1 and SUR2A immunoreactivities (compare the consecutive sections Fig. 5a,b). The presence of tight junctions in tubules and in collecting ducts is confirmed by the presence of the ZO-1 protein in all types of renal epithelia (Fig. 5c).

In the urinary bladder, the tight junctions between the facet cells of the urothelium generate one of the tightest barriers of the body (Fig. 6a,b). To estimate the whole tissue section, the immunofluorescence pictures were electronically highlighted, by using the weak autofluorescence of the tissue (Fig. 6c,d). The tight junctions can be visualized by the presence of occludin (Fig. 6b). In accordance with the differential expression in kidney epithelia, tight junctions in bladder epithelia were devoid of Kir6.1 immunoreactivity (Fig. 6a; Table 1).

Food-intake-dependent increase in the glucose concentration in the intestinal lumen caused a coordinated alteration of the tight junction structure and consequently results in an upregulation of paracellular permeability (Pappenheimer et al., 1987; Madara, 1994; Tomita et al., 2004). An increase in paracellular flux is initiated when D-glucose is taken up by the sodium glucose transporter 1 (SGLT1) leading to an increase in the intracellular D-glucose concentration. The increased amount of D-glucose may initiate an ATP-dependent change in the conformation of the Kir6.1-SUR2A complex, probably leading to the increased permeability.

To investigate these D-glucose-dependent modulatory changes of tight junction permeability and to analyse the role of the Kir6.1-SUR2A complex in this regulatory process, an in vitro model system of the rat small intestine was developed (see Materials and Methods). To measure permeability, we used radioactive L-glucose, which is not transported by enantiomer-specific SLGT-1 (Rummel and Stupp, 1960). A rise in the D-glucose concentration in the intestinal lumen caused an increase in the accumulation of L-[¹⁴C]glucose in the lamina propria and tela submucosa, and was taken as an indicator of an increase in paracellular permeability.

Generally, only one defined segment of the small intestine was used per rat. First we established that the paracellular uptake of L-[¹⁴C]glucose was linear over 60 minutes (data not shown). For the analysis of paracellular permeability, each experimental series (consisting of three rats) included a base-line measurement of L-[¹⁴C]glucose uptake taken after 45 seconds with or without the indicated additives. The segments of the other rats were incubated for 60 minutes with or without the indicated additives. For calculation, L-[¹⁴C]glucose accumulation obtained after 45 seconds under the individual conditions was set as one.

A 60-minute incubation of a small intestine segment in a lumenal buffer (containing 25 mM D-glucose) increased the L-[¹⁴C]glucose accumulation by ∼60% compared with an incubation in buffer A (containing only 5 mM D-glucose) (Fig. 7a, compare bar I with bar II). Tolbutamide, a blocker of the SUR component of Kir6 ion channel permeability, increased L-[¹⁴C]glucose accumulation independently from a rise in the lumenal D-glucose concentration. This increase even exceeded the one obtained by 25 mM D-glucose (Fig. 7a, bar III). As a control, an incubation with the same lumenal buffer (5 mM D-glucose) but supplemented with 20 mM L-glucose instead of D-glucose had no significant effect because L-glucose is not taken up by the epithelial cells and, therefore, can not cause an increase in the intracellular ATP-concentration (Fig. 7a, see bar IV) (Rummel and Stupp, 1960). Thus, accumulation of L-[¹⁴C]glucose in the subepithelial tissue was not due to a specific transporter-regulated transcellular uptake but rather mediated by paracellular flux caused by changes in the permeability of tight junctions.

To substantiate the potential function of the Kir6.1-SUR2A complex in the regulation of tight junctions, we investigated whether the effect of 25 mM D-glucose stimulation can be modified by tolbutamide (3 mM) or by the Kir6-SUR activator diazoxide (3 mM). In this case the permeability increased after stimulation with 25 mM D-glucose alone was set as 100% (see the bracket in Fig. 7a, bars I and II). The stimulatory effect of 25 mM D-glucose is reduced by ∼50% in the presence of 3 mM diazoxide, whereas tolbutamide provoked a more than twofold increase in paracellular permeability (Fig. 7b, bars II and III).

The passage of small molecules, such as water, ions or nutrients, across various epithelia occurs by two distinct mechanisms. The transcellular pathway consists of membrane pumps, transporters and ion channels, whereas the paracellular pathway follows the intercellular space between two adjacent epithelial cells. Both routes are strictly regulated and, in some tissues, the transcellular pathway apparently controls the paracellular route by modifying the permeability of its dominant barrier, the tight junctions (Madara, 1987; Madara and Pappenheimer, 1987; Madara, 1994; Tomita et al., 2004).

Tight junctions are macromolecular complexes of at least twenty individual protein components. These include membrane proteins such as occludin, claudins and the junctional adhesion molecule (JAM), scaffolding proteins such as ZO-1, ZO-2 and ZO-3, as well as elements of the cytoskeleton and the vesicle trafficking machinery (Lapierre, 2000; Martin-Padura et al., 1998; D'Atri and Citi, 2002). Furthermore, the tight junction complex interacts with membrane lipids, generating an intramembrane barrier that prevents mixing of apical and basolateral membrane lipids. Intracellularly, tight junction complexes are linked to an apical perijunctional actomyosin ring (Fanning et al., 1998; Hirokawa and Tilney, 1982; Itoh et al., 1997).

As estimated from paracellular resistances, tight junction permeabilities vary in a wide range according to the functional requirements of the respective epithelium. For example the transepithelial resistance of proximal tubules (6-7 Ω×cm²) in the kidney is about 5×10⁴ times lower than that of the urothelium (300,000 Ω×cm²) (Powell, 1981). The proximal tubule is the place of major ion- and water-reabsorption in the kidney, requiring a low resistance of the paracellular pathway. By contrast, the bladder is designed to store the excreted and concentrated fluids of the kidney. Reabsorption of bladder fluid with its high content of urea would result in increased and potentially harmful plasma concentrations of urea (Born et al., 2003).

Tight junction permeability can vary even in the same tissue as known from the small intestine (Pappenheimer, 1987; Madara, 1994; Tomita et al., 2004). When, after a meal, glucose is taken up through the saturating Na⁺-dependent glucose transporter of the intestinal brush border, the increased intracellular D-glucose concentration - by a so far unknown mechanism - also increases tight junction permeability and, thereby, paracellular transepithelial fluxes of monomers, such as sugars and amino acids, by solvent drag (Pappenheimer, 1987; Madara, 1994; Turner, 2000; Turner et al., 2000b; Tomita et al., 2004). Apparently, the resistance of the paracellular pathway is regulated depending on local demands.

Current hypotheses suggest that the paracellular resistance depends on the intracellular perijunctional actomyosin ring. Actomyosin contraction could alter ZO-1, and trigger modification of occludin and claudin function, two processes that may be related to paracellular fluid movements (Kojima et al., 1999; Van Italie and Anderson, 2004). In addition, the activation of myosin light-chain kinase results in the contraction of the perijunctional actomyosin ring, as well as in increased permeability of epithelial and endothelial tight junctions (Goeckeler and Wysolmerski, 1995; Hecht et al., 1996; Turner, 2000; Yamaguchi et al., 1991). These data gave rise to the idea that, paracellular transport is controlled by a `purse-string' mechanism (Florian et al., 2002; Bement et al., 1993), emphasizing the suspected role of the cytoskeleton in the control of tight junction permeability.

Here, we report that the Kir6.1-SUR2A protein pair is found in the epithelial cells of stomach, small and large intestine, liver and proximal and distal tubules of the kidney (Table 1). In all these tissues Kir6.1-SUR2A is exclusively localized at positions characteristic of tight junctions. This focal localization is surprising for Kir6-SUR pairs, which are widely accepted to subserve channel function. In general, ion channels are distributed to more or less extended areas of the cellular plasma membrane, which is a reasonable localization to depolarize or hyperpolarize the corresponding cell.

However, there are cases where a highly distinct distribution of K⁺ channels apparently subserves an unconventional function. One example is the cerebellar pinceaux. Here, the channels are believed to attribute a positive charge to a restricted extracellular compartment (Laube et al., 1996), thereby locally increasing the transmembrane potential of the Purkinje-cell axon hillock. In this case, the septate junctions, a distinct intercellular barrier system, are supposed to prevent uncontrolled charge spread.

Tight junctions represent a similar system. They are composed of several interconnected rows of proteins or protein complexes, best visualized in freeze-fracture electron micrographs. Usually they are considered as mechanical barriers, preventing the uncontrolled leakage of fluids and their components from one compartment (vessel or gut lumen) to another (interstitium). However, they could also act as electrical barriers, building up a potential between the interstitium and the gut lumen. A possible role of the Kir6.1-SUR2A protein pair in tight junctions could be to control the charge of an extracellular microdomain defined by the outer protein rows of the tight junction complex. Following this idea, a modified charge concentration within the extracellular area of the tight junction could be involved in their opening.

However, modified charge concentration within the extracellular area of the tight junction can not explain the regulation of paracellular fluxes. Such a regulation might involve a coupling between extracellular conditions (i.e. increase in nutrients) and changes in intracellular metabolic processes.

So far, the molecular composition of epithelial K⁺-ATP channels in the gastrointestinal tract is not completely understood (Warth and Barhanin, 2003). Other Kir6-SUR-based K⁺-ATP channels, such as the Kir6.1-SUR2B complexes, are widely accepted as the molecular constituents of endothelial and smooth muscle K⁺-ATP channels, whereas the SUR2A subunit is mostly expressed in heart and skeletal muscle. The Kir6.1-SUR2A protein pair, therefore, represents a so far unconventional combination of K⁺-ATP channel subunits. Interestingly, co-expression experiments of this pair of subunits have failed to induce K⁺-conductance (Aguilar-Bryan et al., 1998). This finding is consistent with the fact that Kir6 and SUR messages and proteins are found in epithelial cells of the ileal mucosa without corresponding channel activity (Burleigh, 2003). These data, together with its localization restricted to tight junctions, support the idea that the Kir6.1-SUR2A pair fulfils a biological function quite distinct from just channelling ions through membranes.

This is further supported by the fact that, although expressed in several epithelial cell types, the colocalization of Kir6.1 and SUR2A subunits with tight junction proteins does not represent a general feature. Indeed, this pair is only found in tight junctions with regulated permeability - not in constitutively impermeable ones such as in the collecting duct of the kidney or the facet cells of the bladder urothelium. This indicates a direct involvement of the Kir6.1-SUR2A complexes in the molecular mechanisms regulating tight junction permeability.

In line with the above considerations, SUR proteins belong to the ABC family, whose members act as mechanochemical devices (Karpowich et al., 2001; Zingman et al., 2001). Binding and hydrolysis of ATP by the SUR protein results in an altered conformation, which usually induces channel-gating. In regulated tight junctions, the Kir6.1-SUR2A protein pair could influence the paracellular resistance through direct interaction with members of the tight-junction-protein complex, as deduced from our coimmunoprecipitation experiments. The altered conformation of the SUR protein may be transferred to other tight junction proteins, finally regulating paracellular permeability (Karpowich et al., 2001; Zingman et al., 2002a).

To provide further support, we established an in vitro model, using morphologically defined segments of the rat small intestine. This model system enables us to work with the best-studied and most plausible example of an acute physiological regulation of tight junctions, the intestinal response to lumenal D-glucose (Pappenheimer, 1987; Madara, 1994; Tomita et al., 2004).

In general, an increase in intracellular glucose with a concomitant elevation of ATP concentration favours the prehydrolytic conformation of the SUR protein with closed K⁺-ATP channels (see Introduction). Intestinal brush-border cells contain Kir6 and SUR proteins, but never show cannel activity (Burleigh, 2003). In these cells, lumenal glucose import through the sodium-dependent transporter also results in an increased ATP and a correspondingly decreased ADP concentration. Again, this constellation favours the prehydrolytic conformation of the SUR protein. But instead of keeping K⁺-ATP channels closed, in this case, apparently the SUR protein mechanically targets the tight-junction-protein complex, increasing its permeability and thereby elevating paracellular fluxes of sugars and other small nutrients (Turner et al., 2000a). This might allow the organism after an occasionally carbohydrate-rich meal to recruit nutrients even when the transcellular pathways are already saturated.

According to our new hypothesis, Kir6.1-SUR2A complexes are directly involved in the regulation of tight junction permeability. Consequently, this permeability can be influenced by drugs known to alter Kir6.1 and SUR2A function specifically.

The sulfonylurea compound tolbutamide and the K⁺-ATP channel activator diazoxide were shown to provoke significant effects. The increase in tight junction permeability observed after physiological stimulation with 25 mM D-glucose alone was enhanced more than two fold with tolbutamide and was decreased by approximately 50% with diazoxide. In addition, tolbutamide alone - without increase of the lumenal D-glucose concentration - showed a strong stimulatory effect on the paracellular flux. Thus, pharmacological modulation of the Kir6.1-SUR2A protein pair, indeed, results in an altered paracellular permeability of the intestinal epithelial barrier.

The specific effects demonstrated in our model system shed some light on the possible mechanism of a well-known but often neglected side effect of anti-diabetic therapy. Shortly after starting treatment with sulfonylureas, some patients suffer from severe diarrhoea (Groop and Neugebauer, 1996), which strongly affects their quality of life. Fortunately, diarrhoea often stops immediately when the therapy is stopped. This side effect could be explained if Kir6.1-SUR2a complexes were involved in the regulation of intestinal tight junction permeability. Acting agonistic to an increased D-glucose concentration, sulfonylureas usually provoke channel closure. In intestinal epithelial cells sulfonylureas would increase tight junction permeability, thereby mediating sulfonylurea-induced diarrhoea.

Taking together these information, the intestinal mucosa might become a suitable and promising model to analyse potentially new functions of Kir6-SUR complexes apart from just forming K⁺-ATP channels. It should allow further investigations of the molecular mechanisms underlying the controlled regulation of paracellular permeability and the design of anti-diabetic drugs without deleterious side effects, such as long-lasting diarrhoea.

Monoclonal antibodies against ZO-1 and E-cadherin were purchased from BioTrend (Köln, Germany) or Transduction Laboratories BD Biosciences (Heidelberg, Germany). The polyclonal antiserum against occludin was obtained from Zymed Laboratories (San Francisco, CA). Antisera against aquaporin-2 (Klussmann et al., 1999) (kindly provided by E. Klussmann, Forshungs Institute of Molecular Pharmacology, Berlin, Germany), Kir6.1 (Thomzig et al., 2001), and AE2 (Jöns et al., 1994) have been described elsewhere. An antiserum against a fusion protein of the less conserved C-terminus of SUR2A (4507-4638 bp, presented as recombinant fusion protein) was raised in rabbits. Antibodies were made monospecific by crossabsorption and affinity-purification.

Specificity of antibodies was verified in ELISA assays, western blots and immunoflorescence experiments by antigen blocking (Fig. 1a) and immunostaining of SUR2A-transfected COS cells (Brune et al., 2000). Secondary anti-rabbit or anti-mouse antibodies from goat or donkey, labelled with either Oregon-Green or Texas-Red, were purchased from Molecular Probes, Europe BV (Leiden, Netherlands). For electron microscopy, biotinylated goat anti-rabbit and anti-mouse secondary antibodies were visualized with DAB using the Elite ABC system (Vector Laboratories, Burlington, Canada).

A mixture of protein A/G sepharose was purchased from Santa Cruz Biotechnology, IC Chemikalien, (Ismaning, Germany). L-[¹⁴C]glucose was obtained from Perkin Elmer (Boston, MA).

Bile canalicular membranes were isolated as described (Tsukita and Tsukita, 1989), with minor modifications. In short, 12-week-old Wistar rats were anaesthetized with diethylether and immediately decapitated. The liver was quickly removed and soaked in ice-chilled phosphate-buffered saline (PBS). All subsequent steps were carried out at 4°C. The liver was minced and treated with hypotonic buffer (1 mM NaHCO₃ and 2 μg/ml leupeptin, pH 7.5) for 30 minutes. The samples were homogenized in two volumes of hypotonic buffer with a dounce homogenizer. The homogenate was diluted with hypotonic buffer to 400 ml, filtered through four layers of gauze and centrifuged at 1500 g for 10 minutes. The pellets were resuspended in hypotonic buffer, filled up to 13.1 ml and mixed with 96.3 ml of a 55% (weight/volume) sucrose solution, resulting in a final sucrose concentration of 48.5%. This solution was divided into six tubes and overlaied with 2 ml of a 42.9% sucrose solution and centrifuged for 60 minutes at 100,000 g. Bile canaliculi were recovered at the interface between 42.9% and 48.5%. The bile-canaliculi-enriched membrane fraction was diluted with ten volumes of hypotonic buffer and centrifuged at 4000 g for 30 minutes.

For solubilization, the pellet was homogenized again in immunoprecipitation buffer, consisting of 1% Triton X-100, 150 mM NaCl, 10 mM Tris (pH 7.4), 1 mM EDTA, 1 mM EGTA, 0.2 mM sodium vanadate, 0.2 mM PMSF and 0.5% Nonidet P-40 and recentrifuged at 20,000 g for 30 minutes.

Of the respective antibody, 3 μl were added to 1 ml of membrane extract, corresponding to 300 μg of the total amount of membrane protein. After incubation with antibody for 16 hours at 4°C, 20 μl of a protein A/G-sepharose mixture (50% in immunoprecipitation buffer) was added and the suspension incubated for an additional 4 hours, followed by three washing steps with the same buffer. The resulting pellet was resuspended in electrophoresis sample buffer, boiled for 10 minutes and subjected to SDS-PAGE analysis.

Electrophoretically separated proteins were transferred to nitrocellulose membranes and incubated in PBS containing 10% milk powder to block unspecific protein-binding sites. Membrane strips were incubated with antisera in an appropriate dilution of PBS-10% milk powder. Washing steps were performed with PBS containing 0.05% Tween 20. The bound immunoglobulins were incubated with horseradish-peroxidase (HRP)-labelled antisera raised against rabbit or mouse IgG and chemoluminescence was visualized by using the ECL detection system (Amersham Braunschweig, Germany).

Immunofluorescence microscopical analysis followed protocols described earlier (Jöns et al., 1994). Briefly, small pieces of rat, pig and human tissue (approximately 1-2 mm³ in volume) were rapidly frozen in melting isopentane (cooled with liquid nitrogen), freeze-dried and then vacuum-embedded in Epon (Drenckhahn and Franz, 1986). Indirect immunofluorescence was performed with 1-μm-sections. Epon was removed with sodium methoxide solution (Drenckhahn and Franz, 1986) and sections were incubated with monoclonal or polyclonal primary antibodies. Bound immunoglobulins were visualized with anti-mouse or anti-rabbit secondary antibodies labelled with Oregon-Green or Texas-Red, as indicated.

For transmission electron microscopy, frozen rat duodenal tissue was cut in 25-μm-sections on a cryostate and rinsed several times in PBS. Sections were treated with 1% sodium borohydride in PBS to remove excessive aldehyde groups, followed by washing steps with PBS containing 0.3% Triton X-100 and 0.05% phenylhydrazine to reduce the endogenous peroxidases. Sections were subsequently washed three times in PBS, followed by incubation for 36 hours at 4°C with the primary antibody in a solution of 10% normal goat serum (NGS) in PBS containing 0.1% sodium azide and 0.01% thimerosal. Sections were washed in PBS for 20 and 40 minutes, preincubated for 1 hour in PBS-A (PBS supplemented with 2 mg/ml BSA), and incubated for 16 hours at room temperature in the secondary antibody solution (biotinylated goat anti-rabbit IgG or rabbit anti-mouse IgG in PBS-A containing 0.1% sodium azide). After another two washing steps in PBS and incubation with avidin-biotin complex (Elite A+B, Vector Laboratories; 1:1000 in PBS-A) for 6 hours at room temperature, sections were washed three times in PBS, followed by preincubation in 3,3′-diaminobenzidine (0.5 mg in 1 ml 50 mM Tris pH 7.6), containing 10 mM imidazole, for 15 minutes. Bound antibodies were visualized by adding H₂O₂ (0.015% final concentration) for 10 minutes. Finally, sections were washed several times in PBS.

Sections were postfixed with 1% osmium tetroxide in PBS for 30 minutes, dehydrated with graded ethanol [70%, 90%, 96%, 100% (including block staining with 2% uranyl acetate in 70% ethanol for 10 minutes at 4°C)] and propylene oxide, and embedded in araldite between two sheets of Aclar plastic (Ted Pella, Redding, CA). Ultrathin tissue sections (60 nm) were cut with a diamond knife and collected on copper grids. The sections were counterstained with uranyl acetate and lead citrate and examined with a transmission electron microscope (EM 900; Zeiss, Oberkochen, Germany).

Total RNA was extracted from rat liver tissue using the guanidinium thiocyanate, phenol-chloroform extraction method (Chomczynski and Sacchi, 1987). DNA-free RNA was obtained by treatment with RNase-free DNase I (Boehringer, Mannheim, Germany) for 30 minutes at 37°C. After phenol-chloroform extraction and ethanol precipitation the probes were stored at -80°C. The first-strand reactions contained 50 ng oligonucleotide dT15-18, 400 units of superscript II reverse transcriptase and 100 nM DTT in 20 μl 1× transcription buffer (all from Gibco, Germany) and were incubated with 2 μg RNA and 200 μM dNTPs (Pharmacia, Germany) for 1 hour at 42°C. Following first-strand synthesis, two units of RNase H were added and the reactions were incubated at 37°C for 30 minutes. The cDNA was purified with the PCR Purification Kit (Qiagen).

PCR was performed with 2 μl of the RT reaction per 25 μl. All reactions were performed in quadruplicates. The primer used anchor and arbitrary primer (Roth, Karlsruhe, Germany) in different combinations. Reactions were performed in 25 μl with 25 mM MgCl₂, 2.5 μl 10× reaction buffer, 2.5 units Ampli Taq Gold (all from Perkin Elmer), 200 μM dNTPs (Pharmacia), 20 pmol per anchor-primer 5′-GGA GCT GGA CGA GAA ACC TTC CAT C-3′ (spanning bases 1246-1271) and 10 pmol of arbitrary-primer 5′-GTC GTC CTG CCC TCA TGA TTC TGA TG 3′ (spanning bases 1440-1466). The thermal cycler was set for 1 cycle of 1 minute at 95°C, 90 seconds at 40°C and 45 seconds at 72°C, followed by 39 cycles of 30 seconds at 94°C, 90 seconds at 40°C, 45 seconds at 72°C. The cycling was followed by 5 minutes at 72°C. For electrophoretic separation the whole PCR reaction product was precipitated by ethanol. For the analysis of the PCR-products the CleanGel DNA analysis Kit (Pharmacia Biotech, Germany) and the DNA silver staining Kit (Pharmacia Biotech) were used according to the manufacturer's instructions.

Adult Wistar rats (weighing ∼300 g) were anaesthetized with diethylether and immediately decapitated. A 15-cm-segment of jejunum, starting ∼2 cm caudal to the ligament of Treitz, was carefully separated from the mesenterium, connected to a two-way valve and thoroughly rinsed with buffer A (130 mM NaCl, 3 mM KH₂PO₄, 2 mM K₂HPO₄, 5 mM D-Glucose, 10 mM Hepes, 1 mM CaCl₂, 1 mM MgCl₂, pH 7.4, 37°C). The intestinal segment was tightly closed with a dialysis tubing-clip and filled with the appropriate buffer. As a starting point, the incubation was performed for 45 seconds, the endpoint was determined after a 60-minute incubation in a 37°C water bath containing buffer A.

All experiments were performed in triplicates. The incubation buffer contained varying concentrations of D-glucose, L-glucose, diazoxide and tolbutamide, alone or in combination. All incubation-buffers were saturated with O₂ and supplemented with 5.5 μCi L-[¹⁴C]glucose/25 ml. The incubation was stopped by washing with buffer A. The intestinal segment was cut into six identical parts that were immediately homogenized in 2 ml H₂O and centrifuged at 20,000 g for 30 minutes at 4°C. The supernatant was collected and the protein content was determined with Roti®-Quant according to the manufacturer's instructions (Roth, Karlsruhe, Germany). An aliquot of the supernatant corresponding to 500 μg of protein was analysed for radioactivity with a scintillation counter (LS 6500 Multi-Purpose Scintillation Counter Beckman Coulter).

The authors thank J. Roeper for illuminating discussions, A. Bräuer for help with the PCR-analysis, and I. Mitschke for helpful comments on the manuscript. Large parts of this publication belong to the MD thesis of D.W.