Novel Endothelial Cell-Specific AQP1 Knockout Mice Confirm the Crucial Role of Endothelial AQP1 in Ultrafiltration during Peritoneal Dialysis

The water channel aquaporin-1 (AQP1) mediates about 50% ultrafiltration during a 2-hour hypertonic dwell in global AQP1 knockout (AQP1-/-) mice. Although AQP1 is widely expressed in various cell types including mesothelial cells, the ultrafiltration has been assumed to be mediated via endothelial AQP1 of the peritoneum. The partial embryonic lethality and reduced body weight in AQP1-/- mice may reflect potential confounding phenotypic effects evoked by ubiquitous AQP1 deletion, which may interfere with functional analysis of endothelial AQP1. Using a Cre/loxP approach, we generated and characterised endothelial cell- and time-specific AQP1 knockout (AQP1fl/fl; Cdh5-Cre+) mice. Compared to controls, AQP1fl/fl; Cdh5-Cre+ mice showed no difference in an initial clinical and biological analysis at baseline, including body weight and survival. During a 1-hour 3.86% mini-peritoneal equilibration test (mini-PET), AQP1fl/fl; Cdh5-Cre+ mice exhibited strongly decreased indices for AQP1-related transcellular water transport (43.0% in net ultrafiltration, 93.0% in sodium sieving and 57.9% in free water transport) compared to controls. The transport rates for small solutes of urea and glucose were not significantly altered. Our data provide the first direct experimental evidence for the functional relevance of endothelial AQP1 to the fluid transport in peritoneal dialysis and thereby further validate essential predictions of the three-pore model of peritoneal transport.


Introduction
The correct genomic targeting of the AQP1 floxed construct was obtained and characterised as shown in two representative clones by long-range PCR (Fig 2A) and Southern blot ( Fig 2B). Correctly targeted ES clones containing the AQP1 fl-neo allele were microinjected into C57BL/6 blastocysts to generate heterozygous AQP1 fl-neo mice. AQP1 fl-neo mice were crossed with FLPeR mice to remove the FRT-flanked neo cassette in their pups with an AQP1 fl/+ ; FLPeR + genotype (data not shown). AQP1 fl/+ ; FLPeR + mice were then backcrossed with C57BL/6 mice to generate AQP1 fl/+ mice that were negative of both neo cassette and FLPeR. As shown in Fig  1 and Fig 2C, PCR analyses with primers P3F/3R distinguished AQP1 fl-neo , AQP1 fl and AQP1del allele, the latter of which was obtained by Cre-mediated deletion of exons 2 and 3. With this pair of primers, no PCR product was amplified from the wild type (WT) allele (+) because of the lack of sequences of primer 3R. Primers 4F and 4R were used to differentiate heterozygous floxed AQP1 mice (AQP1 fl/+ ) from homozygous floxed AQP1 mice (AQP1 fl/fl ) (Fig 1 and  Fig 2D).
Intercrosses of heterozygous AQP1 fl/+ mice yielded 129 pups-33 wild-type (25.6%), 65 heterozygous (50.4%), and 31 homozygous (24.0%), which is according to the expected Mendelian ratio. Western blot ( Fig 2E) and immunohistochemistry (IHC) analysis ( Fig 2F) showed that the expression level and cellular distribution of the AQP1 gene in visceral peritoneum (VP) of AQP1 fl/fl mice were indistinguishable from those of AQP1 +/+ mice. Survival, Mendelian inheritance of the AQP1 fl allele and unaltered expression of AQP1 proteins are all indications of normal gene expression from the AQP1 allele with loxP-flanked exons 2 and 3. Targeting construct and screening strategies. A part of the wild type (+) allele of mouse AQP1 is shown with indicated exons (black boxes) and restriction enzyme sites. The targeting allele (fl-neo) is indicated with 3' and 5' targeting arms (thick lines), loxP / FRT sites and pro-and eukaryotic neomycin selection cassette (neo-gb2-PGK). The 5' probe and 3' probe for Southern blot located outside the targeting vector detect 11.1-kb (+) and or 7.1-kb (fl-neo) fragments from Nsi I-digested genomic DNA and 22.6-kb (+) and or 7.3-kb (fl-neo) fragments following Xho I-digestion genomic DNA, respectively. Mice carrying the floxed allele (AQP fl-neo ) were crossed to FLPeR mice for excision of the FRT-flanked neo cassette. The resulting floxed mice (AQP fl ) were crossed to Cdh5 (PAC)-CreERT2 (Cdh5-Cre) transgenic mice to excise exons 2 and 3 following tamoxifen induction, and then generate the AQP1 null allele (AQP1 del ) in endothelial cells. The P1F/1R, P2F/2R, P3F/3R and P4F/4R primers for PCR-based genotype analyses and the lengths of their responding PCR products are indicated.

Generation and Characterisation of an Endothelial Cell-specific AQP1 Knockout Mouse
Subsequently, AQP1 fl/fl mice were crossed with transgenic Cdh5 (PAC)-CreERT2 mice (Cdh5-Cre) carrying the Cre recombinase gene driven by the vascular endothelial cadherin Cdh5 promoter [27]. The AQP1 fl allele was converted to a AQP1 null allele (AQP1 del ) in a tamoxifen-inducible manner in early adult mice by EC-specific Cre-mediated excision of exons 2 and 3. In addition, AQP1 fl/fl ; Cdh5-Cre + mice survived at a similar rate (1/53) as control AQP1 fl/fl ; Cdh5-Cremice (0/57) during the period of 1-14 days following the last day of tamoxifen treatment until functional analyses.
EC-specific Cre-mediated deletion efficiency was investigated by immunocytochemistry (ICC) staining of ECs isolated from lung or VP (Fig 3), and immunohistochemistry (IHC) staining of VP sections (Fig 4A and 4B). CD31-stained cells in both groups showed similar immunoactivity. In the absence of Cre recombinase, AQP1 was strongly expressed in the cultured ECs, but it was remarkably diminished in ECs with Cre recombinase expression. Quantification of the relative CD31-stained endothelial area in the VP sections showed no significant difference between the AQP1 fl/fl ; Cdh5-Cremice and AQP1 fl/fl ; Cdh5-Cre + mice (Fig 4B, left). This confirms that the endothelial area was not significantly altered by the deletion of AQP1 Representative Western blot (left) showed AQP1 protein expression in the lung ECs from control (AQP1 fl/fl ; Cdh5-Cre -) and endothelial AQP1-deficient mice (AQP1 fl/fl ; Cdh5-Cre + ). Equal loading (10 μg of protein from each sample) was verified using an anti-β-actin antibody. Scanning densitometric analysis of AQP1 protein expression (AQP1 protein/β-actin) was done using Image J (right). Data are expressed as a ratio relative to the control group. Each bar indicates mean ± S.E. ÃÃÃ P < 0.001.
Comparison of the signal intensity obtained from lung and EC DNA samples extracted from the AQP1 fl/fl ; Cdh5-Creand AQP1 fl/fl ; Cdh5-Cre + mice indicates a high degree of specific excision of exon 2 and 3 in the ECs (Fig 4C). Detection of AQP1 expression in isolated endothelial cells from lung or visceral peritoneum (VP) by immunofluorescence. The representative isolated endothelial cells (ECs) from lung were immunostained for the EC marker CD31 (green, with Hoechst-stained blue nuclei) and AQP1 (red) in AQP1 fl/fl ; Cdh5-Creand AQP1fl/fl; Cdh5-Cre + mice (upper two panels). ECs showed a similar staining for the CD31, whereas ECs from AQP1 fl/fl ; Cdh5-Cre + mice showed a markedly decreased staining for the AQP1 compared to those in AQP1 fl/fl ; Cdh5-Cremouse. From the same mice, comparable results were observed in isolated ECs from visceral peritoneum (VP) (lower two panels). Calibration bar: 50μM.  Western blot analysis also demonstrated the efficiency of AQP1 deletion in lung ECs ( Fig  4D). Compared to AQP1 fl/fl Cdh5-Cremice (n = 4), the level of AQP1 protein was diminished in AQP1 fl/fl Cdh5-Cre + mice (n = 11) by 89.0%.

Discussion
Ultrafiltration (UF) is an important cause for peritoneal dialysis (PD) failure [1]. The threepore model of peritoneal transport predicts that the water channel AQP1 is the ultrasmall pore located in the endothelial cells and mediates about 50% of ultrafiltration during a 2h-dwell. This concept has been validated experimentally in global AQP1 knockout mice (AQP1 -/-) [6]. The present study bypassed the drawbacks of potential confounding phenotypic effects from ubiquitous deletion of AQP1 gene, using a well-characterised inducible endothelial cell-specific AQP1 knockout mouse model, to investigate the central functions of endothelial AQP1 in UF during PD.
To achieve cell-specific and /or conditional gene knockout, first we generated the AQP1 fl/fl mice. It showed that the insertions of loxP sites do not interfere with expression of the floxed gene and normal protein production (Fig 2E and 2F). Then, the floxed AQP1 alleles were converted to the deleted AQP1 alleles in a tamoxifen inducible manner by endothelial-specific not in small arterioles (*). Sections from AQP1 fl/fl ; Cdh5-Cre + mice showed similar CD31-labelled endothelial area as those from control AQP1 fl/fl ; Cdh5-Cremice. In contrast, sections from AQP1 fl/fl ; Cdh5-Cre + mice showed a nearly~80.19% decrease in AQP1-labelled endothelial area compared to controls. ***P< 0.001 versus control group. Calibration bar: 100μM in (A). (C) PCR analysis of AQP1 null allele (del). DNA preparations from lung tissue and lung endothelial cells (ECs) were analysed. The floxed (fl) and null (del) alleles were distinguished by PCR with primer 3F and 3R. Cdh5-CreF/R primers were used to detect the Cre transgene. (D) Quantitative measurement of AQP1 gene expression in total protein fractions from lung ECs. Cre-mediated excision of exons 2 and 3. The resulting AQP1 fl/fl ; Cdh5-Cre + mice revealed a high efficient EC-specific deletion of the AQP1 in tissues (Figs 3 and 4) with strongly expressed AQP1 such as lung and PM [3,16].
In contrast to global AQP1 knockout mice with impaired survival, lower body weight [16,17], AQP1 fl/fl ; Cdh5-Cre + mice survived at a similar rate as control AQP1 fl/fl ; Cdh5-Cremice during the period of tamoxifen treatment until functional analyses, and exhibited similar clinical and biological parameters at baseline as AQP1 fl/fl ; Cdh5-Cremice (Table 1). Body weight is known to be an important parameter for interpreting experimental outcomes including peritoneal UF calculation [3,18]. Therefore, the endothelial cell-and time-specific AQP1 knockout (AQP1 fl/fl ; Cdh5-Cre + ) mice will provide a more specific experimental model for functional investigation of endothelial AQP1. To our best knowledge, this is the first report on an inducible endothelial cell-specific AQP1 knockout mouse model.
Although the 1-hour 3.86% mini-PET has been validated to be a simple, fast and more practical method for assessing AQP1-related solute-free water transport [23], no specific data have been provided in a mouse PD model yet. In addition, until now no data have showed the association between endothelial AQP1 function and its related clinical relevance for UF in PD. The present study addressed the question, using 1h 3.86% mini-PET in the well-characterised tamoxifen-inducible endothelial cell-specific AQP1 knockout mouse model.
These data confirm the hypothesis that deletion of endothelial AQP1 significantly decreases water transport during the 3.86% mini-PET, without significant changes in small solute transport parameters. It has also been suggested that the deletion of exons 2 and 3 in AQP1 gene may putatively destroy the symmetrical membrane-spanning "hour-glass" structure of mouse AQP1 protein, which has a great functional impact on the resulting AQP1 channel protein [26]. To demonstrate the UF capacity of AQP1, Ni et al. characterised water and solute transport in global AQP1 knockout mice (AQP1 -/mice) during a 2-hour unphysiological high glucose (7%) dialysate dwell [3]. In comparison with WT littermates, AQP1 -/mice had no sodium sieving and an about 50% decrease in Net UF and 70% decrease in the initial, solute-free UF. Smit et al. quantified water transport in patients during the 3.86% mini-PET [25]. Compared to normal UF patients, UF failure patients showed a 42% decrease in Net UF and a 56% decrease in FWT at the end of the 3.86% mini-PET.
Our data provide the first direct evidence for the crucial role of AQP1 expressed in endothelium during 3.86% mini-PET and thereby validate not only essential predictions of the threepore model, but also the essential role of endothelial AQP1 in ultrafiltration and free water transport during peritoneal dialysis.
Modulation of endothelial AQP1 might provide a promising therapeutic target for preventing ultrafiltration failure in peritoneal dialysis. These findings may also contribute to an improved understanding of AQP1-endothelial biology, such as angiogenesis, wound healing, tumour spread and organ regeneration.

Generation of Inducible Endothelial Cell-specific AQP1 Knockout Mice
The Cre/loxP system was employed to generate the conditional knockout mice. A targeting vector that encompasses 17.2 kb of the AQP1 locus and a loxP / FRT site-flanked pro-and eukaryotic neomycin selection cassette (Neo-gb2-PGK) was designed (Gene Bridges GmbH, INF 584, Heidelberg, Germany) (Fig 1A). The PGK-gb2-neo template encodes the neomycin/ kanamycin resistance gene which combines a prokaryotic promoter (gb2) for expression of kanamycin resistance in E.coli with a eukaryotic promoter (PGK) for expression of neomycin resistance in mammalian cells. Embryonic stem cells (ES cells, 129/Ola) were transfected with the linearised targeting construct, and subjected to G418 selection (300 μg / ml). The surviving clones were screened for homologous recombination by long-range PCR with LATaq DNA polymerase (TaKaRa, Otsu, Japan) and the PCR primers are listed in Table 3.
The PCR-positive clones were then confirmed by Southern blot by using 339bp and 415bp fragments of the intron sequence as a 5' or 3' probe, respectively. These two fragments were created by PCR from wild type 129/Ola genomic DNA by primers probe 5F/5R and probe 3F/ 3R (Table 3) and then labelled with alkaline phosphatase using AlkPhos Direct Labelling and Detection system with CDP-star (Amersham). 15 μg of genomic DNAs were digested with Nsi I or Xho I, electrophoresed, transferred to a nylon membrane, hybridised with the labelled probes. Southern blot analysis was also preformed with a neo probe to exclude additional random integrations of the targeting vector DNA.
The correctly targeted ES clones were expanded and injected into C57BL/6 blastocysts (Biotech Lab., Interfacultary Biomedical Faculty, University Heidelberg, Germany) to yield chimeric founder mice. Germline transmission was achieved by mating male chimeric founders with C57BL/6N female mice (The Jackson Laboratory, Maine, USA). The chimeric mice with proven germ line transmission ability (AQP1 fl-neo ) were bred with FLPeR (B6.129S4-Gt (ROSA) 26Sor tm1(FLP1)Dym /RainJ) mice (The Jackson Laboratory, Maine, USA) to remove the FRTflanked neo expression cassette, and then backcrossed to C57BL/6 mice to exclude the FLP allele and produce floxed AQP1 mice (AQP1 fl/fl ). The inducible endothelial cell (EC)-specific AQP1 knockout mouse line was generated by crossing AQP1 fl/fl mice with Cdh5 (PAC)-CreERT2 (Cdh5-Cre) mice 22 to obtain control AQP1 fl/fl ; Cdh5-Crelittermates and AQP1 fl/fl ; Cdh5-Cre + mice. To inactivate the AQP1 gene, 5-week-old mice (both groups) received intraperitoneal injections or IP injections of tamoxifen (2 mg of tamoxifen, Sigma, T5648; tamoxifen 20mg/ml in Miglyol 812) for 5 consecutive days to induce Cre/loxP mediated gene deletion. All treated mice were not used until two weeks following the last day of tamoxifen treatment.

Mouse Genotype Analysis by PCR
The locations of all PCR primers used for genotyping are indicated in Fig 1. All PCR primer sequences and their related amplification sizes are listed in Table 3. Mouse genomic DNA samples were prepared from tail tips by following standard protocols.

Peritoneal Transport Studies
All animal procedures were approved by the University of Heidelberg Animal Care Committee and were performed on mice according to the guidelines of the Institute of Laboratory Animal Science of the University of Heidelberg.
To minimise animal suffering and distress: (i) mice were treated with the shorter dwell time (1 hour) compared to the published data [3]; (ii) at the end of the dwell, mice were anaesthetised with Ketamine (100 mg/kg i.m, Bremer Pharma GMBH, Warburg, Germany) and Xylazine (5 mg/kg i.m, Ecuphar GMBH, Germany), (iii) at the end of the peritoneal transport study (the dialysate samples were recovered from the peritoneal cavity [7] and blood samples were taken from the posterior vena cava), mice were enthanised by cervical dislocation. Peritoneum and lung samples were processed for further investigation, such as immunohistochemistry (IHC) analysis, immunocytochemistry (ICC) analysis and isolation of endothelial cells.
During the 1-hour dwell, mice were monitored every 15 minutes and were enthanised by cervical dislocation if they were in a state of respiratory distress. There were no unexpected deaths and all mice survived during the 1-hour dwell. All parameters from the plasma and dialysate were assessed using Roche/Hitachi cobas c311 in the Medical Research Centre, University Medical Centre Mannheim, Mannheim, Germany. Sodium plasma and dialysate levels were measured with ISE indirect Na, K, CI for Gen.2, glucose levels with Glucose HK in haemolysate Gen.2, and urea with Urea/BUN, respectively. Dialysate creatinine levels were measured with the Jaffe method in the central laboratory of Heidelberg University Hospital, University of Heidelberg, Heidelberg, Germany.
The following calculations were performed [7, 23]: 1. Total net UF (NUF) NUF (μl) = (VD60-VD0), where VD0 and VD60 are the volume (μl) of the recovered dialysate at 0 and 60 minutes of the dwell time. V0 was equal to 1.985 ml [an average volume recovered from 6 8-week-old mice (4 males and 2 females), immediately after 2 ml IP injection with 3.86% glucose-based dialysate] 2. Dialysate / plasma sodium concentration (mmol/L) ratio at 0 minute (Na0D/P) and at 60 minutes of the test (Na60D/P). Plasma sodium concentration at the start (Na0P) was an average concentration obtained from 6 mice of each group without processing the 3.86% mini-PET.

Free water transport (FWT):
FWT ðmlÞ ¼ NUF ðmlÞ À UFSP ðmlÞ According to the established three-pore model of PD, UF occurs either through capillary small pores and accompanied by a solute (e.g. sodium, urea), or transcellular fluid transport that is selective for water only and occurs through specialised AQP1 water channels [22]. Therefore, free water transport (FWT) can be calculated with the La Milia method [23] by using sodium concentration values in dialysate and plasma, as well as NUF.
It has been known that, during the first hour of a hypertonic dwell, free water transport is maximal (since the glucose dialysate concentration is at its peak) and that the diffusive transport of sodium from blood to dialysate is very low (sodium concentration is close to its equilibrium across the peritoneal membrane and thus its gradient for diffusion is very low) [23]. In these conditions, nearly all the sodium transport is due to convection through small pores and thus the ratio of sodium removal (NaR) to plasma water sodium concentration gives the value of UF through the small pores (UFSP). It is then easily possible to calculate FWT by subtracting UFSP from total NUF.
Because the 3.86% mini-PET of La Milia et al. only provides more accurate information on free water transport when a correction for sodium diffusion is made [24], a correction for all dialysate sodium (NaD) is made with the mass transfer area coefficient of creatinine as described by Westra et al. [28].
NUF, FWT and sodium sieving were used as indexes for transcellular water transport by AQP channels, and glucose D60/D0 and urea D60/D0 ratio were measured to evaluate the peritoneal transport for small solutes [23].
To evaluate the microvascular endothelial area or endothelial AQP1 area in mouse VP immunostained for CD31 and AQP1, VP slides from AQP1 fl/fl ; Cdh5-Cremice were applied with the same excitation, gain and exposure settings as those of AQP1 fl/fl ; Cdh5-Cre + mice. Processed images were imported to ImageJ where the intensity threshold was set automatically and analysed for units of positive pixel area. The relative endothelial area (the ratio between the CD31-stained area and the entire VP area, in %), or the relative endothelial AQP1 area (the ratio between the AQP1stained area and the CD31-stained area, in %) was measured. All slides were masked to prevent bias and imaged at 20× magnification. No less than three fields of view (FOV) were obtained per slide, no less than 10-15 FOV were obtained per animal and five animals were evaluated.

Primary Culture of Endothelial Cells
ECs isolated from lung [29] or VP [30] were cultured from 8-week-old mice by a modified method. Briefly, the lung or VP were minced and digested in collagenase A solution (1 mg/ml, Roche Diagnostics GmbH, Mannheim, Germany) at 37°C for~60 minutes. The cellular digest was filtered through sterile 40 μm nylon mesh and centrifuged at 800 g for 10 minutes. The cell suspension in 10% FCS-DMEM (Dulbecco's modified Eagle's medium containing 10% foetal calf serum) was incubated with Dynabeads (M-450, sheep-anti-rat IgG, Dynal, Life Technologies, CA, USA) for 30 minutes at 4°C. The bead-bound cells were recovered, washed with 10% FCS-DMEM and FCS-free DMEM, and then digested for 5 to 10 minutes at 37°C with trypsin/ EDTA (GIBCO, Life Technologies, CA, USA) to release the beads. The bead-free cells were centrifuged in 10% FCS-DMEM and then resuspended in EC culture medium.

Image Acquisition
For the immunoassay, images for cells and tissue slides were acquired using the Nikon Eclipse 80i upright automated microscope with a Nikon Plan Fluor 20x NA 0.50 objective (Nikon GmbH, Düsseldorf, Germany). Image acquisition and processing were controlled by NIS Elements BR 3.0 software. The images were visualised and processed with Image J.

Western Blot
Tissue or EC samples were homogenised in lysis buffer. Approximately 40μg of tissue protein from VP or 10μg EC protein from lung was separated by 12% SDS-PAGE, transferred to a PVDF Membrane (WesternBreeze Chromogenic Western Blot, Invitrogen, Life Technologies, CA, USA), and incubated with 1:500 rabbit anti-AQP1 (Biozol, Hamburg, Germany) overnight (4°C). Samples were treated with alkaline phosphatase-conjugated anti-rabbit secondary antibody, and immunoreactivity was detected by using the WesternBreeze1 Chromogenic Kit according to the manufacturer's specifications (WesternBreeze Chromogenic Western Blot, Invitrogen, Life Technologies, CA, USA). The intensity of the resulting bands was quantified by densitometry using Image J.

Statistical Analyses
The statistical data were presented as the means ± S.E. Statistical significance was evaluated with the Student's t-test for pairwise comparison of data. Differences were considered significant when a p value was < 0.05. In the figures, significance levels are shown as asterisks (" Ã " for p < 0.05, " ÃÃ " for p < 0.01, " ÃÃÃ " for p < 0.001).