Epac1 null mice have nephrogenic diabetes insipidus with deficient corticopapillary osmotic gradient and weaker collecting duct tight junctions

The cAMP‐mediator Epac1 (RapGef3) has high renal expression. Preliminary observations revealed increased diuresis in Epac1−/− mice. We hypothesized that Epac1 could restrict diuresis by promoting transcellular collecting duct (CD) water and urea transport or by stabilizing CD paracellular junctions to reduce osmolyte loss from the renal papillary interstitium.

In vitro studies using the Epac-stimulating cAMP analogue 8-pPCT-2′-O-Me-cAMP 10,11 suggest that Epac can stimulate transcellular urea transport, 5,12,13 maintain high aquaporin-2 (AQP-2) synthesis 5,14 and stimulate the insertion of AQP-2 in the principal cell (PC) apical membrane. 15 8-pPCT-2′-O-Me-cAMP has numerous 'off-target' actions. It can affect purinergic receptors or transmembrane transporters and indirectly activate PKA by inhibiting cAMP degradation. 16,17 To evaluate specific Epac1 actions, the present study compares Wt and Epac1 null mice. The Epac1 −/− mice have been described previously and do not express Epac1 in any tested tissue, including kidney. 7 Early data based on immunofluorescence suggested high Epac1 expression in the proximal tubules and the intercalated cells (IC) of the CD. 9 More recent, quantitative transcriptomic studies indicate low expression in proximal tubules and high expression in renal glomeruli and the CD, 6 whose PC express more RapGef3 than the ICs. 10 The reports that Epac1 contributes to the cAMP induction of urea transporters 5,11,12 imply that Epac1 can protect against urea-induced osmotic diuresis. If so, a high protein load 14,18 should induce more severe diuresis in the Epac1 −/− mice. We compared, therefore, the diuresis of Wt and Epac1 −/− mice on low (4%), standard (12.5%) or high (40%) protein diet, and determined if Epac1-deficient mice had altered expression of the major urea transporter (UT-A1). We determined also whether their AQP-2 expression was decreased based on reports that Epac1 may contribute to enhanced AQP-2 synthesis or stability. 5,14 The high Epac1 expression in renal glomeruli 6 suggests that Epac1 can affect glomerular function. We tested, therefore, if the Epac1 null mice had deficient glomerular filtration barrier, or had altered glomerular filtration rate (GFR) in the absence or presence of the tubuloglomerular feedback (TGF) 19,20 blocker furosemide. 19,21 Several Epac1 −/− mouse organs have increased basal in vivo microvascular permeability, 7,22 indicating that Epac1 can limit fluid extravasation. We hypothesized that Epac1 might act to conserve intravascular fluid also via renal actions. We therefore compared the diuresis of Wt and Epac1 −/− mice not only under basal conditions, but also after water loading (WL) when endogenous AVP is low, 23 after moderate water deprivation and after injection of the avpr2 receptor agonist desmopressin (dDAVP).
A number of observations suggest that the CD cell junctional complexes (JCs) could be targets for regulation by Epac1. Firstly, Epac1 is well known to strengthen cell junctions in vitro 24,25 and in vivo. 7 Secondly, Epac1 and its major downstream target E-cadherin have higher expression in the CD than in other parts of the rat nephron, as shown by deep sequencing of mRNA from dissected rat nephron segments. 6 The importance of E-cadherin for an intact CD barrier and ability to concentrate urine are demonstrated by mice with downregulated Grainyhead-like protein 2 homolog (GRHL2). They express less key junctional proteins and have diabetes insipidus as well as decreased renal medullary Na + content. 26 We therefore determined the Na + content in snap-frozen renal papillae from Wt and Epac1 −/− mice and studied the CD junctions by transmission electron microscopy (TEM).

| Epac1 −/− mice have increased diuresis at normal and low urea load
The deletion of Epac1 did not impact kidney weight (average 262 mg in Wt and 263 mg in Epac1 −/− mice), kidney-to-BW ratio (10.3 mg g −1 in either Wt or Epac1 −/− ) or gross renal morphology (Table S1, Figure S1). However, cages housing Epac1 −/− mice were consistently 'wetter' than cages housing Wt mice under standard breeding conditions. Specifically, Epac1 null mice housed in metabolic cages with free access to standard (12.5%) protein chow and drinking water had more than twice as high diuresis as matched Wt mice (Table 1).
To test whether the increased Epac1 −/− mouse diuresis was related to deficient handling of urea, we compared the effect of low (4%) and high (40%) protein. The diuretic phenotype was strongest at 4% protein diet (Table 2A). The 40% protein | 3 of 16 SIVERTSEN ÅSRUD ET al. feeding led, as expected, 27 to increased diuresis in both Wt and Epac1 −/− mice, but it did not enhance the Epac1 −/− diuretic phenotype (Table 2B). Further, switching from 4% to 40% protein had opposite effects in Epac1-deleted and UT-A1/3-deleted mice 27 regarding water consumption, urine production and urine osmolarity ( Figure S2). The level of UT-A1/3 (Slc14a2) mRNA was similar in Wt and Epac1 −/− mice both 3 hours after WL and 6 hours after WL and dDAVP injection ( Figure S3A).
The increased diuresis of the Epac1 −/− mice at low or moderate protein load was matched by increased water intake (Tables 1 and 2A) resulting in similar retention of water for either strain. To rule out any effect of compulsory drinking behaviour, mice were denied access to drinking water. Dehydration was counteracted by intragastric infusion of 1.5-mL water (WL) at the onset of each experiment. We found a significantly higher 6-hour urine output for Epac1 −/− than for Wt (n = 4-5) and Epac1 −/− mice (n = 4-5) were first (A) fed 7 d 4% protein diet followed by 24 h determination of water intake and diuresis, and next (B) 7 d 40% diet followed by similar monitoring. The results are given as mean (SD), and excretion data are normalized per g body weight. Unpaired two-tailed Student's t test with Welch's correction when applicable, was used to determine statistical differences between the diuresis of Wt and Epac1 −/− mice, aa P = .0035, bbb P = .0007, c P = .031, d P = .018, e P = .037, f P = .039, gg P = .0051, h P = .044).

| Epac1 −/− mice have impaired urinary concentrating capacity under antidiuretic conditions, but intact dDAVP-induced increase of cAMP and AQP-2
The water-loaded Wt and Epac1 −/− mice had similar diuresis until the WL had been eliminated ( Figure 1A). Only thereafter did the Epac1 −/− mice have higher urine excretion than Wt mice ( Figure 1A,B). To test whether Epac1 −/− mice were subresponsive to AVP, they were given a bolus injection of dDAVP. Although both Wt and Epac1 −/− mice had a period of complete anuria after dDAVP + WL, the diuresis recovered slightly earlier for Epac1 −/− than for Wt mice ( Figure  1A,B). Analysis of the urine revealed that the Epac1 null mice had higher free water clearance than Wt mice ( Figure  1C). The subresponsiveness of the Epac1 null mice to AVP/ dDAVP could be due to deficient Avpr2-mediated increase of cAMP or of AQP-2. Both cAMP excretion ( Figure 1D,E) and expression of Aqp2 mRNA or AQP-2 protein ( Figure S3B-D) were similar in Wt and Epac1 −/− mice, under basal conditions as well as after dDAVP stimulation. F I G U R E 1 Epac1 −/− mice have higher diuresis and spare less free water than Wt mice under antidiuretic conditions. (A) Accumulative diuresis during the first 6 h after water loading (WL) alone or after injection of dDAVP. The horizontal hatched line represents the volume (1.5-mL H 2 O + 0.1-mL saline) loaded, n = 6-7 mice/group, * P = .016 at 5h, * P = .015 at 5.5 h and * P = .015 at 6 h. (B) The diuresis between 3 and 6 h after WL alone and between 5 and 6 h for WL + dDAVP, n = 5-6 mice/group, † P = .0058 for vehicle and * P = .045 for dDAVP. (C) Free water clearance (μL min −1 ), n = 6 mice/group, P = .69 for vehicle and * P = .023 for dDAVP. (D) cAMP excretion (pmol min −1 ), n = 5-7 mice/group, F(1, 20) = 0.043, P = .84 for interaction and ‡ P = .0003 for treatment (E) cAMP excretion (pmol min −1 ) for control mice, n = 5-7 mice/group, P = .58. Values are presented as mean ± SEM (panel A) or as both mean (SD) and mean (SEM) (panels B-E). Unpaired two-tailed Student's t test with Welch's correction when applicable (A-C, E) or two-way ANOVA with Tukey's adjustment for multiple comparisons (D) was used to determine statistical differences between Wt and Epac1 −/− mice in the various treatment groups. For further details, see Figure 9 and Materials and Methods

| Epac1 −/− mice have lower papilla and bladder urine osmolarity than Wt mice under antidiuretic conditions
The Epac1 −/− mouse water-sparing deficiency could be explained by a perturbed renal corticopapillary osmotic gradient. The length of the renal papillary tip as well as the thickness of the papillary base, medulla and cortex appeared similar in Epac1 −/− and Wt mice upon dissection. Moreover, the papillary blocks, dissected in the same way from Wt and Epac1 −/− mice, had similar average dry weight ( Figure 2A). The water content increased significantly from cortex to papilla for either Epac 1 −/− mice (74.3%-83.8%, ‡ P < .001) or Wt mice (73.8%-80.8%, † P = .004), but slightly more so (by about 9.5%) for Epac1 null mice than (about 7%) for Wt mice ( Figure 2B). More strikingly, the papillary Na + content of Epac1 −/− was only 53% of that in Wt ( Figure 2C). To evaluate whether the difference was related to less endogenous AVP stimulation in the Epac1 −/− mice, the plasma Na + concentration and osmolarity were determined. We found similar values for Epac1 −/− (152(SD6) mmol/L Na; 324(SD7) mosm Kg −1 ) and Wt mice (152(SD3) mmol/L Na; 328(SD8) mosm Kg −1 ).
To establish the correlation between papillary and urine content of osmolytes, we analysed also the urine aspirated from the bladder of overnight water-deprived mice just before the kidney sampling. The urine from Epac1 −/− mice had only 59% (SD18) as high urea concentration and 58% (SD14) as high osmolarity as Wt mouse urine ( Figure 3). The reduction of the urine concentration of the various osmolytes was similar to the reduction of Na + in the papilla (Figures 2 and 3).

| Epac1 −/− and WT mice have similar extravasation of albumin
In a previous study, Epac1 −/− mice had increased leakage from microvessels in skin, intestine and skeletal muscle as measured using the double isotope 125 I-HSA and 131 I-HSA injection method (see Ref. 7 for details). We found no difference in albumin permeability between Wt and Epac1 −/− mice. The Epac1 −/− mice also failed to increase permeability in response to ANP, which increases albumin permeability in Wt mouse skin, muscle and intestine 7 ( Figure 4A,B).

| The CD TJs are shorter and less dense in Epac1 −/− than in Wt mice
The corticopapillary osmotic gradient perturbation of Epac1 null mice (Figures 2 and 3) resembled that reported in mice with genetically induced CD TJ 'leakiness' associated with less E-cadherin expression. 26 The striking CD clustering of mRNA coding for Avpr2, Epac1 and the downstream Epac1 target E-cadherin 6 (Table S3) prompted a study of the CD junctions. To know if the Epac1 null mice had altered CD JCs, we compared Epac1 −/− mouse CD junctions with Wt controls using TEM.
Epac1 −/− and Wt CD had similar general paracellular organization (see Figure 5 for representative examples), with an apical TJ that merges into a JC. Morphometric analysis of 56 PC/PC junctions from six Wt mice and 51 PC/PC junctions from five Epac1 −/− mice revealed that compared to Wt the Epac1 −/− TJs were on average shorter, less dense (had higher electron transmittance) and associated with less electron-dense junctional material (Table S4, Figure 6A-C). A similar, but less strong, trend was found for the JC (Table S4). Table S4 lists also the percentage of each TJ appearing to be 'open', as subjectively, and blindly scored by independent operators (Table S4; see also Figure 5). The median value for this parameter was 0 for Wt and 38% for Epac1 null mouse TJs (Table S4). The ICs have been reported to have very high 28 or quite low 29 Epac1 expression. We found that intercalated/principal (IC/PC) TJs were longer and more electron dense than PC/ PC TJs, and less affected by Epac1 deletion ( Figure S4A-C, Table S5) than those in the PC/PC paracellular space (Table  S4).

| Epac1 −/− mice have intact glomerular filtration barrier and moderately elevated GFR
Rodent glomeruli have high Epac1 mRNA expression, 6 suggesting that Epac1 has a functional role in the glomerulus. We probed this possibility, first by testing whether Epac1 −/− mice could have a defect glomerular filtration barrier, a dysfunction of which would cause increased albumin leakage. 30,31 We found similar levels of albumin in Wt and Epac1 −/− mice urine, and whether the mice had been treated with dDAVP or not ( Figure 7).
We considered next if Epac1 could control the GFR, possibly by modulating the TGF loop at the glomerular end. This possibility was tested by comparing the renal clearance of creatinine and of [ 14 C]inulin in the absence and presence of the TGF-blocking agent furosemide. To know if the primary furosemide target (NKCC2 of the macula densa) was intact in Epac1 −/− mice, we compared first the effect of furosemide on Wt and Epac1 null mouse diuresis and Na + excretion.

| DISCUSSION
Our observation that Epac1 null mice have nephrogenic diabetes insipidus (Table 1A, Figure 1) with lowered papillary Na + content ( Figure 2C) and shorter and less dense CD TJs (Table S3, Figures 5 and 6) conforms to the hypothesis that the primary Epac1-dependent antidiuretic mechanism is to control the leakiness of the intercellular junction between CD PCs. Possible Epac1-dependent mechanisms modulating transcellular water and urea transport or glomerular filtration appear to be less significant to explain the diabetes insipidus phenotype. These observations provide new understanding of the biological significance of high renal Epac1 expression. [6][7][8] We discuss the possible Epac1-dependent mechanisms suggested in the introduction in more detail below. The diuretic phenotype with decreased conservation of free water ( Figure 1C) is a hallmark of deficient AVP action in the CD. 23,32 We tested several mechanisms that could lead to decreased AVP-mediated water conservation. We found no evidence that Epac1 −/− mice have a subresponsive Avpr2 receptor or excessive degradation of cAMP. The renal cAMP excretion after a general dDAVP stimulation increased at least as much in Epac1 −/− as in Wt mice ( Figure 1D). The secretory component (clearance-GFR) for cAMP was similar in Wt and Epac1 −/− mice ( Figure S5B), implying that Wt and Epac1 −/− mice have similar renal excretion of cAMP. The apparent cAMP resistance of Epac1 −/− mice to AVP effects  Table  S4, where they are highlighted in blue and marked with *. (B, D, F) Junctions from three representative Wt mice, highlighted in yellow and marked with * in Table S4. CD, collecting duct was, therefore, not explained by increased cellular cAMP extrusion.
The CD and connecting tubule (CNT) segments of the nephron have the highest expression of both Epac1 (RapGef3) and AVP receptor 2 (Avpr2), while the short (SDL) and long (LDLIM) descending limb and tAL segments of the nephron have moderate Epac1 (RapGef3) expression and no appreciable Avpr2 expression (Ref . 6 Table S3). Thus, Epac1 is likely to be activated by AVP mainly in the CD and CNT segments.
The action of Epac1 in the cAMP-stimulated CD could be to augment the known PKA-mediated actions such as induction or membrane insertion of aquaporins 2, 3 or UT-A1/3 or UT-A2. 5,13 Although Epac1 has been reported to promote AQP-2 expression upon long-term cAMP stimulation, 14 neither the Aqp-2 transcript nor the protein level differed between Wt and Epac1 null mice ( Figure S3B-D). Thus, direct Epac1-dependent regulation of AQP-2 expression is unlikely to account for the diuretic state.
The diuretic Epac1 −/− phenotype was not explained by deficient urea transport, since it was more severe at low (4%) than high (40%) protein diet (Table 2), and not associated with altered expression of the UT-A1/3 urea transporter ( Figure S3A). A lack of Epac1 does, therefore, not lead to diet/urea-induced osmotic diuresis in contrast to the expectation from an in vitro study 13 reporting Epac1 enhanced UT-A1/3 urea transporter expression.
Overnight water-deprived Epac1 −/− mice had not only a more dilute bladder urine than Wt mice (Figure 3), but also a similarly decreased renal medullary Na + and presumably osmotic gradient ( Figure 2C). Epac1 therefore appears to be required to maintain a steep medullary osmotic gradient to prevent water loss under antidiuretic conditions. A similar osmotic gradient deficiency was observed in mice with lowered CD paracellular TJ resistance due to CD-specific deletion of the transcription factor GRHL2. 26 The similar decrease of Na + and osmolarity in the renal papilla and urine in GRHL2-and Epac1-targeted mice ( 26 and Figures  2 and 3), as compared to Wt mice, argues that water can equilibrate between the papillary tissue and the pre-urine in the CD lumen. Such equilibration would not be expected if Epac1 were critically required for aquaporin synthesis or insertion. In line with the observation above, there was no difference between Epac1 −/− and Wt mice regarding AQP-2 expression ( Figure S3). Isolated GRHL2-deficient PC cells have severely decreased paracellular electrical resistance. 26 This observation provides a link between paracellular 'leakage' and the in vivo phenotype of decreased papillary osmotic gradient and increased diuresis, 26 but it does not tell if the PC TJ permeability can be under physiological control.
We found that the PC/PC TJs were less electron-dense and had less associated junctional material in Epac1 −/− than Wt mice (Figures 5 and 6; Table S4). Epac1 is well known to tighten JCs, both in vitro (see Ref. 25 for recent review) and in vivo. 7 It may also mediate the cAMP-induced stabilization of the retinal blood brain barrier TJs. 33 The simplest explanation of the present Epac1 null mouse diuretic phenotype is, therefore, that Epac1 maintains the corticopapillary osmotic gradient by strengthening the CD paracellular TJs.
The diuretic Epac1 −/− phenotype described here, and preliminarily reported, 9 is characterized by dilute urine under antidiuretic conditions, as well as near unaltered Na + excretion (Table 2). Therefore, it differs strongly from the recently described Epac1 −/− osmotic diuresis phenotype, caused by increased Na + excretion ascribed to decreased proximal tubule (PT) expression of the sodium-hydrogen exchanger 3. 34,35 Although a PT-localized primary phenotype is plausible in view of an early study showing strong immunohistochemical staining for Epac1 in the PT brush border, 28 it is not supported by more recent quantitative mRNA expression data, which show insignificant RapGef3 expression in the PT (Table S3, 6 ). There is no obvious explanation for the dramatic difference of renal phenotypes between the two RapGef3 'knock out' mice. The mice used in the present study have increased Further, endothelial cells isolated from the Epac1 −/− mice used here have increased junctional permeability when monolayers are tested in vitro 36 implying that the permeability phenotype is caused by the lack of Epac1 in the endothelial cells and not by an altered in vivo environment in the Epac1 −/− mice. The high RapGef3 mRNA expression in rat glomeruli 6 suggests that Epac1 can impact glomerular functions. Epac1 appeared not to be required to maintain an intact glomerular filtration barrier since the Epac null mice did not have albuminuria (Figure 7). The Epac1 null mice had increased clearance of creatinine and cAMP (Table 3, Figure 8C and Figure S5), which could be explained either by increased tubular excretion 37,38 or increased GFR. The lower [ 14 C]inulin clearance of Wt than Epac1 −/− mice, in the absence, but not in the presence, of the TGF inhibitor furosemide ( Figure 8D) suggests that Epac1 may restrain the GFR via TGF modulation.
In conclusion, Epac1 appears to prevent diabetes insipidus mainly by maintaining the corticomedullary osmotic gradient. This may at least in part be achieved through Epac1mediated control of paracellular CD 'leakage'.

| Ethical approval
The animal experiments were approved by the Norwegian Animal Research Authority and conducted according to the European Convention for the Protection of Vertebrates Used for Scientific Purposes, Norway. These regulations comply with the UK policies and regulations. 39 The Animal Care and Use Programs at University of Bergen are accredited by AAALAC international.

| Targeted disruption of the Epac1 gene to produce C57BL/6J Epac1 −/− mice
The C57BL/6JBomTac mouse strain is described at http:// www.tacon ic.com/wmspa ge.cfm?parm1 =764. Epac1 null (Epac1 −/− ) mice were generated using the Cre-LoxP system as described previously. 7,40,41 In short, loxP sites were inserted by homologous recombination into the gene encoding Epac1 (RapGef3) flanking exons 7-10. These exons encode the CBD-B domain, which codes for the only functional cAMP site of Epac1. 42,43 In addition, a frame shift was introduced distal to the deletion in order to minimize the probability of expression of truncated Epac1 fragments through unscheduled translation. C57BL/6JBomTac mice from Taconic, Denmark were used to backcross the recombined chimeric mice for at least 10 generations. The stability of the deletion was confirmed by genotyping for each new generation.

| General animal handling
The mice used in the present study were about 3-4 months old females, weighing 18-30 g. The wild-type C57BL/6J BomTac mice (Wt) and their Epac1 null counterparts (Epac1 −/− ) were littermates, or they were born and kept in neighbour cages of the Epac1 −/− mice. They were generally housed 4-5 mice together in IVC-II cages (Sealsafe ® IVC Blue Line 1284L, Tecniplast) and maintained under standard housing conditions at 21°C ± 0.5°C, 55% ± 5% humidity, 12-hours artificial lightdark cycle (150 lux). The Epac1 null mice had the same weight as their Wt littermates, from which they were indistinguishable F I G U R E 9 Overview of the experiments conducted to test the effect on urine production and composition of protein diet, dDAVP and furosemide. A, Mice were fed low (4%) protein diet for 7 d and the urine sampled for 24 h thereafter. One day later, the mice received an acute per-oral (p. o.) intubation of 1.5-mL water, and the urine was collected continuously (as voided) for 6 h thereafter. The animals were next switched to high-protein (40%) diet, and 7 d thereafter, subjected to the same experiments as when fed low-protein diet. B, Mice fed standard (12.5%) protein diet were transferred to metabolic cages and their urine collected for 24 h under basal conditions. Seven days later, the mice received a 1.5-mL water load (WL) and 0.1-mL saline i.p., and their urine was collected continuously for 6 h thereafter. The experiment was repeated after 7 d, but with injection of dDAVP instead of saline. C, Mice, inserted sc. on the back 3 d previously with a mini-osmotic pump loaded with [y 14 C]inulin, were water loaded as described above, and the urine collected for 3h. Two days later, the experiment was repeated, but with a furosemide injection at the time of water loading based on visual observation under standard housing conditions. The Wt and Epac1 null mice had also similar blood pressure. 7 For most experiments, the mice were provided with standard rodent chow containing 12.5% protein (Special Diet Services RM1, 801151, Scanbur BK) and water ad libitum. For experiments designed to manipulate the standard urea load, the animals received low (4%) protein or high (40%) protein chow from Custom Diet Services, Special Diet Services, Witham, UK. Since most experiments were conducted in single-housed metabolic cages, each animal was acclimatized for 24 hours in a metabolic cage for 3-4 days before the onset of each experiment.
Anaesthesia was with 1.5%-2% Isoba ® vet. Isoflurane (Schering-Plough Animal Health) under continuous flow of O 2 (1 L m −1 ) and N 2 O (1 L m −1 ). The anaesthetized animals were placed on a heating pad at 37°C to limit any drop of the body temperature. The dDAVP (1-ng g −1 body weight) or furosemide (40-µg g −1 body weight) was injected intraperitoneal (i.p.) in 0.1-mL 150-mmol/L aqueous saline (NaCl) solution during anaesthesia shortly before the intragastric intubation with 1.5-mL water (see Figure 9 below for details). Cardiac blood (0.4 mL) was aspirated into a 0.5-mL syringe with 0.1-mL citrate-dextrose solution by cardiac puncture of animals killed with CO 2 , and thereafter centrifuged for 3 minutes at 500 rpm to yield plasma. Plasma samples were stored at −80°C and urine samples at −20°C for further analysis.
The animals showed no symptoms of distress or illness during the experiments. Two Epac1 −/− mice appeared debilitated during a 24-hours pre-experimental stay in metabolic cages. They were therefore excluded from any experiment. Two mice (one Wt and one KO) had badly fitted drinking bottles during the 24-hours experiment, resulting in contamination of urine. They were excluded from that experiment, but not from the subsequent water-loading experiments.

| Urine collection and WL
In one series of experiments, the urine of animals given chow with various protein content was collected for 24 hours via a tube connected to the funnel-shaped bottom of the metabolic cage. The experiments were conducted in individual MMC10 metabolic cages (Hatteras Instruments, Inc) and started at 09.00 am. The urine bladders were emptied by bladder massage immediately before placing the animals in the metabolic cages.
In another series, each mouse was anaesthetized just after the bladder had been emptied, and then subjected to acute water loading (WL, 1.5-mL H 2 O) by intragastric intubation through a 38-mm metal gavage with a silicon tip (AgnTho's AB), and placed in a metabolic cage. The urinary bladder was emptied by external massage before the urine was collected (see Materials and Methods and legend to Figure 9 for further details). In this way, errors caused by individual differences of recent water intake and residual urine content were minimized. The urine that was spontaneously voided thereafter was collected into a siliconized beaker and manually recovered by pipetting. The volume of each recovered urine sample was determined, before adding it to a pre-weighed tube. The sum of the volumes determined fit well with the weight increase of the collecting tube.

| Experiments using high protein dietinduced urea loading
To test the ability of the mice to handle urea loading, they were first fed low protein diet for 1 week in ordinary cages, and thereafter for 24 hours in individual metabolic cages. The mice were next water loaded as described above, returned to the metabolic cage and the urine voided was collected during a period of 6 hours. Thereafter, 0.2-mL whole blood was collected from a puncture in the facial vein for analysis of serum creatinine and urea. The next week, the same mice were switched to high protein diet, and thereafter treated as described for the low protein fed mice.

| Determination of GFR and microvascular permeability
The GFR was determined as the renal clearance of [ 14 C]inulin as described in, 44 except that the infusion pumps (Alzet ® osmotic minipump model 1007D, DURECT Corporation) were placed subcutaneously between the withers rather than intraperitoneally. Each pump was filled with 25 μCi [ 14 C]inulin in 150-mmol/L aqueous saline (NaCl) solution, and delivered 0.5 ± 0.1 μL h −1 . They were primed overnight in the 0.9% saline solution at 37°C. Experiments were conducted at least 2 days after pump implantation. The amount of [ 14 C]inulin was determined in the urine and in whole blood collected from the facial vein in 10-μL capillary pipettes (Minicaps, Hirschmann-Laborgeräte). For terminal experiments, the blood (2 × 50 μL) was collected by heart puncture. Samples of urine (10 µL), plasma (28 µL) or whole blood (50 µL) were mixed thoroughly with 1-mL 2% SDS (20 g SDS L −1 ) and next with 10-mL scintillation liquid (Emulsifier-safe™, Perkin-Elmer, Inc) before being counted in Tri-Carb 2900TR Liquid Scintillation Analyzer (Perkin-Elmer, Inc). To minimize any non-specific chemiluminescence, the vials were left for 1 day in | 13 of 16 SIVERTSEN ÅSRUD ET al.
the dark before starting the counting, and each vial was counted 3 × 10 minutes. The counting efficiency of [ 14 C] inulin was determined by recounting each vial after a small volume with a known high [ 14 C]inulin concentration had been added to each vial.
A previous study comparing Wt and Epac1 −/− mice 7 showed that Epac1 and ANP acted oppositely on the microvascular permeability for albumin in tissues like skin, fat, muscle and intestine, with known capacity to store excess intravascular fluid. Similar data for the kidney were obtained, but not reported. They are reported here since they could tell whether Epac1 deletion or ANP affects the extravascular renal accumulation of serum albumin. The mice were anaesthetized with 1.5%-2% isoflurane in air (Isoba vet, Schering-Plough Animal Health) administered through a nose cone. Human serum albumin (HSA) labelled with either 125 I or 131 I was injected sequentially in the jugular vein, respectively, 35 and 5 minutes prior to termination of the experiment. Blood samples were obtained before the mice received a lethal iv injection of saturated KCl, and tissue samples harvested. The right and left kidneys were removed, rinsed for blood, the radioactivity determined and spillover from 131 I to 125 I corrected for. The kidneys were dried until stable dry weight was obtained. Distribution volumes were determined as plasma equivalent volumes (ie cpm per g tissue divided by cpm per mL plasma obtained at the end of the experiment for each mouse) and referenced to dry tissue weight. For further details, see Ref. 7.

| Determination of the osmolarity and the concentration of NaCl and selected other ions in the renal papilla interstitium and in bladder urine
Eleven female Epac1 −/− mice and four Wt mice, deprived of drinking water overnight (total 10 hours), were anaesthetized with 1.5% isoflurane (IsoFlo ® Vet 100%, ABBOT Laboratories Ltd) in 100% O 2 . Bladder urine was collected by suprapubic puncture and collected in Eppendorf tubes for measuring osmolarity (Wescor vapor pressure osmometer model 5600) and the concentration of urea (mixed-mode chromatography) and cations (ion chromatography). Blood was collected by heart puncture and both kidneys were then rapidly removed through a midline incision and snap-frozen in Eppendorf tubes in liquid nitrogen. The kidney poles of the shock-frozen kidneys were cut-off perpendicularly to the long axis with a scalpel. The remaining central 2-to 3-mm thick tissue block containing the papilla was then immediately dissected on a chilled glass plate into a pyramidalshaped tissue block with the papillary tip as the apex and the cortex as the basis. From this block, three consecutive tissue samples were cut-off while still semi-frozen separating the pale papilla from the darker medulla and cortex from the remaining medulla.
The tissue samples were immediately transferred into pre-weighed Eppendorf tubes, weighed and dried at 40°C for estimation of water content. The dried tissue samples were then equilibrated in 500-µL Milli-Q water in the same Eppendorf tubes for at least 3 days to measure total tissue cation content. Sodium, ammonium, potassium, magnesium and calcium in diluted urine (1:201), plasma (1:101) and undiluted eluate from renal papilla, medulla and cortex were baseline separated in a 10 minutes 7.5-60 mmol/L MSA gradient by a Dionex IonPac CS 12A analytical column (8.5 μm, 2 × 250 mm, P/N 046075) at a flow rate of 0.2 mL min −1 using a Dionex Integrion HPIC System equipped with a CDRS 600 (2 mm) Cation Electrolytic Suppressor and a high pressure EGC 500 methane sulfonic acid eluent generator cartridge. Urea in urine was separated in an acetonitrile-water gradient using a Thermo Scientific™ Acclaim™ Trinity P2 column (100 mm, catalogue number 085434) connected to a Thermo Scientific UltiMate 3000 Rapid Separation system and detected by a Corona Veo Charged Aerosol Detector.

| Laboratory methods
Urinary and plasma cAMP was determined as described previously. 45,46 The urea concentration in urine was determined by the QuantiChrom™ Kit (DIUR-500; Bio-Assay systems) according to manufacturer's protocol. To achieve accurate determination of urea in mouse plasma, the samples were diluted fivefold in H 2 O and partially deproteinized by spinning through Ultracel-30kDa membrane filters (Millipore), before being assayed. Osmolality was determined using a Wescor 133 vapor pressure osmometer with SS-033 filter discs (Wescor Inc) according to the manufacturer's protocol. Urine Na + and K + were determined using ion-selective electrodes (analysed by the Laboratory for Clinical Biochemistry at Haukeland Hospital in Bergen, Norway).
Creatinine in urine was determined either by an enzymatic creatinase assay (CREA Plus kit; Boehringer Mannheim), using a Roche Cobas Bio Analyzer (Roche) or based on HPLC. 20 Only the latter method, with the modifications described below, produced reliable determinations in mouse plasma. Briefly, plasma samples (20 μL) were deproteinized by adding 2-mL acetonitrile, vortexed and centrifuged for 15 minutes at 13 000 g at 4°C. After evaporation to dryness, the samples were dissolved in 120 µL of 5-mmol/L sodium acetate, pH 5 (HPLC mobile phase), and 15 µL of 20 000×g supernatant injected into a 100 × 1.45-mm PRP-X200 cation exchange column (Hamilton) equilibrated with HPLC mobile phase to separate creatinine, which is positively charged at