A Novel Mouse Model Reveals that Polycystin-1 Deficiency in Ependyma and Choroid Plexus Results in Dysfunctional Cilia and Hydrocephalus

Polycystin-1 (PC-1), the product of the PKD1 gene, mutated in the majority of cases of Autosomal Dominant Polycystic Kidney Disease (ADPKD), is a very large (∼520 kDa) plasma membrane receptor localized in several subcellular compartments including cell-cell/matrix junctions as well as cilia. While heterologous over-expression systems have allowed identification of several of the potential biological roles of this receptor, its precise function remains largely elusive. Studying PC-1 in vivo has been a challenging task due to its complexity and low expression levels. To overcome these limitations and facilitate the study of endogenous PC-1, we have inserted HA- or Myc-tag sequences into the Pkd1 locus by homologous recombination. Here, we show that our approach was successful in generating a fully functional and easily detectable endogenous PC-1. Characterization of PC-1 distribution in vivo showed that it is expressed ubiquitously and is developmentally-regulated in most tissues. Furthermore, our novel tool allowed us to investigate the role of PC-1 in brain, where the protein is abundantly expressed. Subcellular localization of PC-1 revealed strong and specific staining in ciliated ependymal and choroid plexus cells. Consistent with this distribution, we observed hydrocephalus formation both in the ubiquitous knock-out embryos and in newborn mice with conditional inactivation of the Pkd1 gene in the brain. Both choroid plexus and ependymal cilia were morphologically normal in these mice, suggesting a role for PC-1 in ciliary function or signalling in this compartment, rather than in ciliogenesis. We propose that the role of PC-1 in the brain cilia might be to prevent hydrocephalus, a previously unrecognized role for this receptor and one that might have important implications for other genetic or sporadic diseases.


Introduction
Autosomal Dominant Polycystic Kidney Disease (ADPKD) is one of the most common monogenic disorders, affecting approximately 1/500-1/1000 of the population [1]. The hallmark of the disease is bilateral renal cyst formation, although ADPKD is a systemic disorder affecting several other organs. In 85% of all ADPKD cases, germ line mutations in the PKD1 gene are responsible for the development of the disease, whereas in the remaining 15% of cases, the PKD2 gene is mutated. PKD1 and 2 are ubiquitously expressed and the generally accepted explanation for the formation of cysts in only some organs is the ''two-hit'' model. According to this model, a second somatic mutation affecting the normally inherited allele occurs in the renal and bile duct epithelia, resulting in the loss of function of either PKD1 or PKD2 and causing the expansion of clonal cysts. Haploinsufficiency is also believed to play a role in some of the systemic manifestation of the disease, such as some of the cardiovascular defects [1].
Complete loss-of-function murine models for either Pkd1 or Pkd2 have shown that the loss of either gene results in embryonic lethality associated with a number of different phenotypes, including renal cystogenesis, skeletal and cardiac abnormalities [2,3], vascular lesions such as haemorrhage and oedema [4,5], and placental defects [3,6], consistent with the ubiquitous expression of the two genes. Heterozygous mice are healthy and develop only a few renal cysts during their adult life, possibly due to a low rate of second-hit in the mouse kidney. In fact, conditional inactivation of the Pkd1 gene in the kidney results in massive cyst formation, with a very variable phenotype depending on the timing of inactivation [7,8].
Finally, mouse models carrying a reduction in Pkd1 expression [9], or transgenic overexpression of the Pkd1 gene [10] all result in renal cystogenesis, suggesting that the expression levels and/or appropriate subcellular localization of PC-1 are critical for its normal function.
The PKD1 gene encodes Polycystin-1 (PC-1, app. 520 kDa), a highly glycosylated plasma membrane receptor consisting of a large (app. 3000 aa) extracellular N-terminal portion with a novel combination of protein-protein interacting motifs [11,12], 11 transmembrane domains, and a short (198 aa) intracellular Cterminus for regulating signal cascades [13][14][15][16][17][18][19]. Full-length PC-1 can be cis-autoproteolytically cleaved at its GPS site (G-protein coupled receptor proteolytic site), generating a C-terminal fragment of approximately 150 kDa (CTF) and an N-terminal fragment of approximately 400 kDa (NTF), which remains tethered to the CTF [20,21]. This cleavage occurs ubiquitously and was shown to be essential for normal PC-1 function both in vitro and in vivo [20,21]. The PKD2 gene encodes Polycystin-2 (PC-2) protein, a TRP-like channel that interacts with the intracellular C-tail of PC-1 through a coiled-coil domain to form a functional complex [22], thus explaining the almost identical clinical phenotype of ADPKD1 and ADPKD2.
PC-1 localizes at cell-cell junctions in association with Ecadherin/b-catenin [23], at desmosomes [24], focal adhesions [25], and in the primary cilium [26][27][28]. In addition, two additional distinct cleavages have been reported to occur in the intracellular Cterminal domain of PC-1, resulting in the release of a membranefree fragment, which is able to translocate to the nucleus [29,30]. While the expression of PC-1 in all of these subcellular locations is not surprising for such a large, complex, and highly posttranslationally regulated receptor, it should be noted that there is a general consensus that studying the endogenous protein is challenging due to a series of factors including the low specificity of antibodies, low protein expression levels, and the high complexity of the protein. Characterization of the overexpressed protein in transgenic mice has been proposed as a way to gain insight into the protein's normal processing, localization, and function. However, the generation of a cystic phenotype in the transgenic mice models reported to date [10] demonstrates that this approach perturbs the physiological dosage/function of PC-1 and thus cannot be employed to study its normal function.
To overcome these limitations and facilitate the study of PC-1 in vivo, we have used targeted homologous recombination to generate an easily detectable epitope-tagged conditional allele of PC-1. We report here that our approach was successful in generating a fully functional and tagged endogenous PC-1 that can easily be tracked in vivo. We show strikingly strong staining in newborn brains in the ependymal and choroid plexus (CP) ciliated cells. Defective ciliary function in these compartments has been linked to defective secretion (by the epithelia of the CP) or movement (by the beating of the ependymal cilia) of cerebrospinal fluid (CSF), ultimately causing dilatation of the brain ventricular system and hydrocephalus [31][32][33][34][35][36][37][38]. Consistent with this and the localization of PC-1 in brain tissue, we have found that inactivation of the Pkd1 gene using either a ubiquitous Cre or a Cre expressed in the central nervous system results in hydrocephalus, a relatively common condition in humans.
Our data describe a previously unrecognized phenotype in the Pkd1 mutant mice and extend the understanding of the ciliary function of PC-1 to a part of the body other than the kidney.

Generation of epitope-tagged Pkd1 knock-in mouse lines
In the attempt to generate a mouse model that would facilitate the study of endogenous Polycystin-1, we have generated two new mice lines carrying epitope-tagged Polycystin-1 by inserting either HA or Myc tags in frame to exon 46 of the Pkd1 gene. In order to enhance the signal, several copies of both HA (3X) or Myc (5X) were inserted in the two targeting constructs ( Figure 1A and Figure  S1). In previous studies, we had found that similar modifications do not change the functional properties of PC-1 in a series of in vitro functional assays conducted on over-expressed protein 13, 19 and unpublished.
To avoid long-range sequencing reactions of the homologous arms of the targeting vectors, we used homologous recombination in bacteria (see Supporting Information Methods S1; 39) and generated a vector containing the sequences between Pkd1 exon 39 and Tsc2 exon 29 in which we stepwise inserted: I) recognition sites for Cre recombinase (loxP) in intron 44 and in the 39UTR; II) HA (or Myc) epitopes in frame to exon 46 followed by a STOP codon; III) multiple Myc epitopes downstream the 39 loxP site; IV) a neomycin cassette flanked by Frt sites in the 63 bp inter-region between the 39 UTR of the Pkd1 gene and the 39UTR of the Tsc2 gene ( Figure S1). The relevant regions carrying the modifications (i.e. Pkd1 intron 44, exon 46 and 39UTR, and the Pkd1/Tsc2 interregion) were fully sequenced.
The two constructs, named HALM and MLHA ( Figure S1) are expected to generate a HA-tagged and a Myc-tagged PC-1, respectively ( Figure 1A). In addition, upon recombination with Cre, a truncated form of PC-1 lacking the C-terminus should be generated ( Figure 1A). Furthermore, a second tag was inserted in the 39UTR sequence so that upon Cre recombination it would become in frame to label the truncated PC-1 protein and generate a Myc-tag (Pkd1 DCMyc ) or a HA-tag (Pkd1 DCHA ), respectively, allowing us to track the mutant protein. Both constructs were used to generate ES cells carrying the desired modifications and the positively selected clones were used to generate chimeras and subsequently the Pkd1 HAneo/+ and Pkd1 Mycneo/+ lines from which the neo cassette was removed to generate Pkd1 HA/+ and Pkd1 Myc/+ lines (see Supporting Information).
We first examined whether we were able to detect tagged endogenous PC-1. To this end, heterozygous Pkd1 HA/+ or Pkd1 Myc/+ mice were intercrossed and homozygous Pkd1 HA/HA or Pkd1 Myc/Myc embryos were generated. Total lysates from the embryos or from murine embryonic fibroblasts (MEFs) isolated at day E11.5 were generated and analyzed for PC-1 expression either by western blot or by immunoprecipitation. In Figures 1B and C we show that endogenous full-length PC-1 (,520 kDa) as well as the CTF cleavage product (,150 kDa; 20) could be detected both in total lysates and after immunoprecipitations with anti-HA in the Pkd1 HA/HA and with anti-Myc antibodies in the Pkd1 Myc/+ or Pkd1 Myc/Myc cells and tissues ( Figures 1B and C).
The epitope-tagged Pkd1 knock-in mouse lines contain fully functional PC-1 protein To exclude the possibility that the inserted modification might generate a hypomorphic allele, we tested if the tagged form of PC-1 was able to rescue embryonic lethality in Pkd1 knock-out mice. To this end, we crossed our knock-in/conditional knock-out mice lines with a mouse line carrying ubiquitous Cre recombinase ( Figure 1A). We observed that inactivation of the Pkd1 gene generated a truncated tagged protein lacking the C-terminal domain ( Figures 1D and E) and resulting in a phenotype very similar to the one described for all other knock-out lines, including oedema, haemorrhage, and cystic kidneys ( Figure S2) by day E15.5 [2][3][4][5][6]40]. However, Pkd1 DC/DC appear to have a milder phenotype as compared to the complete null mouse model, since in Pkd1 DC/DC lines analyzed more than 90% of Pkd1 DC/DC mice survived to E15.5 and approximately 60% survived to day E16.5 (Table S1) on a pure C57BL/6 background, whereas complete null [40] mice die at day E12.5 in a C57BL/6 background (Chiaravalli, Wodarczyk, Germino and Boletta, unpublished).
Next, we performed rescue experiments by crossing heterozygous Pkd1 HA/+ and Pkd1 Myc/+ mice with the heterozygous Pkd1 knock-out line Pkd1 +/DC . Pkd1 HA/DC and Pkd1 Myc/DC pups were born at the expected mendelian ratio (Table 1), were viable and fertile and did not show any appreciable difference in renal, hepatic, or pancreatic cyst formation at 18 months of age as compared to Pkd1 +/2 mice (not shown). In addition, the Pkd1 HA/HA and the Pkd1 Myc/Myc double homozygote mice did not show any renal or extra-renal defects even with aging (.18months). These data strongly suggest that the modifications inserted did not generate a grossly hypomorphic allele.

Biochemical analysis of PC-1
Next, in order to control that the tags added onto the PC-1 Cterminus do not affect its ability to interact with other proteins, we investigated the interaction between PC-1 and PC-2, which is known to be mediated by the PC-1 C-terminus [22].
First, protein lysates from Pkd1 HA/HA and control wild type MEFs were immunoprecipitated using an anti-PC-2 antibody followed by immunoblotting with anti-HA. In Figure 2A, we show that PC-2 is able to immunoprecipitate endogenous, HA-tagged PC-1 both in its full-length and CTF forms. Next, we performed the reverse experiment and used anti-HA antibodies to immunoprecipitate endogenous, tagged PC-1 from mouse embryonic fibroblasts. Western blot analysis show that PC-2 is immunoprecipitated along with PC-1 in a specific manner, since an IP using anti-HA antibodies in lysates from wild-type MEFs did not show any PC-2 immunoprecipitation, despite the fact that equal amounts of PC-2 are expressed by both cell lines ( Figure 2B).
Next, in order to compare the expression levels and biochemical processing of the tagged protein, we used recently developed antibodies directed against murine PC-1 [21] to immunoprecipitate PC-1 from wild-type and Pkd1 Myc/Myc fibroblasts and embryos to compare the expression levels. In Figures 2C and D we show that equal levels of full-length PC-1 were detected in Pkd1 +/+ and Pkd1 Myc/Myc fibroblasts and embryos. Likewise, the extent of cleavage at the GPS site and the amount of NTF cleavage product remaining tethered to the CTF fragment were indistinguishable between the wild-type and the tagged proteins ( Figure 2D). Interestingly, the CTF cleavage product observed in the Pkd1 Myc lines migrated at a slightly slower rate in SDS-page due to the larger size of the protein generated by the tags distinguishing the tagged from the untagged form of the protein in a single lane. We therefore took advantage of this feature to perform a direct comparison of the expression levels of the tagged versus untagged protein using Pkd1 Myc/+ embryos. Figure 2E shows that immunoprecipitation with an anti-PC-1 C-terminal antibody followed by immunoblotting reveals that there are equal amounts of tagged versus untagged CTF in these lysates. All of these data taken together strongly suggest that the tagged protein behaves like the endogenous untagged protein.

Characterization of PC-1 Expression and Distribution in the Kidney
We first investigated the expression of PC-1 in the kidney. Studies in mice have previously shown that PC-1 expression decreases during the postnatal period [21,41]. We therefore assessed the expression of PC-1 at different time points starting in newborn kidneys and up to P50. As shown in Figure 3A, PC-1 appears to be highly expressed in the newborn kidney at day P2 and it appears that there is a very sharp time window between days P4 and P10 in which PC-1 expression levels drop drastically. Immunoprecipitation using a much higher amount of total protein lysate (10 mg) showed the presence of PC-1 in the adult kidney (P20) ( Figure 3B), although the protein appears to be expressed at much lower levels.
We therefore performed immunofluorescence studies in newborn kidneys (P2) from homozygous Pkd1 HA/HA mice using an anti-HA antibody. We detected the presence of bright single spots on the apical side of epithelial cells at the base of cilia, as evidenced by counter-staining with anti-acetylated alpha-tubulin. A similar bright staining pattern was also present in the absence of acetylated tubulin counterstaining. These results are consistent with previously reported studies [27,28] further confirming that the tagged form of PC-1 behaves like the endogenous protein ( Figures 3C-E). Staining of control, wild-type sections in identical conditions showed that the staining is specific since anti-HA antibodies were not able to detect cilia in the wild-type mice (not shown). In addition, very similar staining was observed by staining Pkd1 Myc/Myc embryos using an anti-Myc antibody.

PC-1 localizes to ependymal and choroid plexus cilia
We next examined the expression of PC-1 in several organs at P20. Tissues from homozygous Pkd1 HA/HA as well as control wildtype littermates were isolated and used in immunoprecipitation experiments using anti-HA antibodies. As shown in Figure 4A, expression in the lung is much higher than in any other organ. PC-1 expression was also detected in the brain, heart, and spleen. As noted above, expression of PC-1 in P20 kidneys appeared to be much lower than in any other organ, although detectable ( Figure 3B) We thus analysed if expression in the brain and lung was subject to postnatal regulation similar to what was observed for the kidney. We found that PC-1 expression in brain appears to be diminished at day P20 compared to P2, while expression in the lung appears to be constant throughout the life of the mouse ( Figure 4B).
We next focused on PC-1 distribution in brain, since this receptor was never characterized in this compartment. To this end, we analyzed the distribution of PC-1 in brain by immunofluorescence studies on early postnatal tissue (P2), where expression appears to be reasonably strong ( Figure 4B and data not shown). Analysis of brain sections from Pkd1 HA/HA and Pkd1 Myc/Myc mice as well as littermate wild-type control mice revealed the presence of PC-1 in ciliated structures. Counterstaining with anti-acetylated alpha tubulin confirmed immunolocalization of PC-1 in all types of cilia observed in the lateral and third ventricles [31], including mono-as well as multiciliated choroid plexus cells and multiciliated ependymal cells. Identical results were seen by analyzing the distribution of the HA-or Myc-tagged protein (Figures 4C-E and Figure S3). Neither anti-HA nor anti-Myc antibodies detected any positive signal in wild-type tissues, demonstrating the specificity of the antibody ( Figure S3). Furthermore, staining with anti-HA or anti-Myc antibodies in the absence of anti-acetylated alpha-tubulin counterstaining revealed an identical staining pattern, demonstrating that the signal is not generated in a non-specific manner by the intense staining of cilia (not shown).
At higher magnification, we discovered profound differences in the distribution of PC-1 depending on the type of cilia analyzed. In grouped cilia present in the ependyma, PC-1 was concentrated    either at the tip or along the axoneme of the cilia ( Figure 4C). In primary cilia of the choroid plexus ( Figure 4D), PC-1 was clearly present along the ciliary axoneme with spotty staining both in the HA and in the Myc lines, resembling the distribution observed for IFT particles. We also observed bundles of cilia in the choroid plexus exhibiting a shorter and more compact structure, as described previously [32]. In these types of cilia, PC-1 appeared to be present at the base of cilia ( Figure 4E). These results suggest that PC-1 might exert different functions on different cilia. Alternatively, since the tags are inserted into the C-terminus of PC-1 and we are therefore able to detect all of the cleavage products containing the C-tail, it is also possible that the different cleavage products are distributed differently in the cilia.
Pkd-1 Knock-Out Mice Show Signs of Mild Hydrocephalus by E16.5 Since we found strong ciliary staining both in ependymal and choroid plexus cilia, we wondered whether the absence of Polycystin-1 in these structures would cause an overt phenotype or if PC-1 function in this compartment is redundant and compensated for by other factors, given that a defect at this level was never reported before in Pkd1 mutant mice. To answer this question we analyzed the brains of Pkd1 DC/DC mice compared to the wild-type littermates at day E16.5. As shown in Figure 5, the brains of wild-type and Pkd1 DC/DC mice appear macroscopically different, with the two separated frontal lobes ( Figure 5A) and several dilated vessels ( Figure 5A, arrows; n = 8). Lateral views show that the brains of Pkd1 DC/DC mice appeared more curved, with the olfactory bulbs (ob) pointing down due to the dome-shape of the brain (Figures 5B and C, ob; n = 8). Morphological analysis of sections showed a mild dilatation of the lateral ventricles and a more prominent dilatation of the third ventricle, with a considerable defect in choroid plexus (CP) morphology ( Figure 5D-K, n = 5). Interestingly, the aqueduct of Sylvius does not show obvious signs of stenosis, excluding the possibility of ventricle dilatation as a secondary cause of aqueduct stenosis (not shown). The images shown in figure 5A-C are representative of an identical phenotype observed in eight distinct mutant mice (100% penetrance) and completely absent from eight distinct littermate wild type controls withdrawn at the same time under identical conditions, thus excluding the possibility that the defects observed might be secondary to problems in dissection. Likewise, the images in figure 5D-K are representative of five distinct mutant and wildtype littermate controls subject to cross-sections showing an identical phenotype.

Conditional inactivation of Pkd1 in brain results in postnatal hydrocephalus
To follow the development of hydrocephalus at later time points we crossed our Pkd1 Flox mice with a line carrying Cre recombinase in a restricted expression pattern to overcome the embryonic lethality of the Pkd1 DC/DC mice, but with a broad expression pattern in the brain to allow targeting of all the regions of the CNS. We used the Nestin-Cre mouse line previously employed to inactivate Stumpy, another ciliary protein reported to cause hydrocephalus when inactivated [33,42]. First of all, we analyzed the expression levels of Polycystin-1 in the different brain regions at P7 and found that the protein is expressed equally and abundantly in all regions ( Figure 6A and not shown). Next, we compared the expression levels of the protein in Pkd1 Myc/DCMyc in the presence or absence of Cre recombinase, and found that inactivation of the Pkd1 gene and reduction of the PC-1 protein occurring at equal rates in all of the brain regions examined ( Figures 6A and B). Most importantly, analysis of brain sections derived from Nestin-Cre conditional mouse models revealed the presence of hydrocephalus with dilation of the third ventricle at day P8 ( Figure 6C-H, n = 2), which was subsequently followed by the dilation of the lateral ventricles, showing a mild but typical triventricular phenotype by day P10 (Figure 6I-J, n = 3). As previously reported, the Nestin-Cre transgene is also active in the kidneys [33]. In agreement with this, the Pkd1 Flox/DC ;Nestin-Cre mouse developed renal cystogenesis and died of kidney failure around P12 (not shown). This prevented us from determining if a more severe phenotype would develop at later time points. In all of these experiments the Nestin-Cre transgene was employed as a heterozygous allele, since homozygosity of a similar transgene was previously reported to cause hydrocephalus itself [43]. In addition, Pkd1 Flox/+ ;Nestin-Cre brains were used as negative controls (Figures 6C, E, G and I) and none developed any sign of hydrocephalus, excluding the possibility that the transgene alone induced this phenotype in the heterozygous state. The ventricular dilations observed are very similar to the phenotype present in the null embryo and were again found with 100% penetrance (n = 5).

Cilial morphology and apico-basal polarity in the ependyma and choroid plexus
Hydrocephalus is a condition associated with dilatation of the brain ventricles caused by accumulation of the cerebrospinal fluid which is either: I) secreted in excess by the choroid plexus; II) not properly propelled through the aqueduct and ventricles due to a failure of the ependymal cilia beating or stenosis of the aqueduct; III) improper re-absorption by the archnoid villi [31,34,37]. In previous mouse models of hydrocephalus due to mutations of ciliary proteins a dual mechanism of hydrocephalus formation due to both lack of proper cilia beating and defective fluid secretion, possibly regulated by cilia themselves, was proposed [34,37]. We therefore analysed the morphology of cilia in E16.5 brains and found no differences in the shape of different cilia types on the choroid plexus ( Figures 7A-H), nor in their relative numbers (not shown). In addition, electron microscopy analysis showed the normal 9+2 structure of the cilia of ependymal cells of the third ventricle ( Figures 7M-N). These data are in agreement with the previously proposed role of Polycystin-1 in ciliary function rather than in ciliogenesis [27,28]. Since defects in the apicobasal polarity of epithelial cells has been reported in the renal cystic epithelium of ADPKD patients, we wondered whether major defects in the morphology of the choroid plexus epithelium might account for the phenotype observed. Staining with Aquaporin-1, a marker of apical membrane, or beta-catenin, a marker of cell-cell junctions, revealed that no major differences could be observed in the choroid plexus epithelia between Pkd1 wild-type and knock-out brains (Figures 7I-L).

Discussion
In this report, we present the generation and characterization of two novel knock-in mouse models carrying either HA or Myctagged endogenous Polycystin-1. We have shown that the manipulations required to generate these alleles do not alter the function of PC-1, since both tagged alleles were able to rescue the embryonic lethality caused by inactivation of the Pkd1 gene, and adult heterozygous mice carrying the tagged-allele (Pkd1 HA/2 or Pkd1 Myc/2 ) have an identical incidence of renal, pancreatic, and liver cystic lesions to that observed in heterozygous animals carrying the wild-type allele (Pkd1 +/2 ). In addition, we have performed a series of biochemical studies to compare the tagged versus wild-type alleles and shown that both HA and Myc-PC-1 have an identical pattern of expression, cleavage, association with the NTF, and heterodimerisation with PC-2 as compared to wildtype protein. All of this evidence leads us to conclude that the endogenous tagged Polycystin-1 behaves identically to endogenous wild-type protein. In addition, we have shown that inclusion of either tag allows for much more efficient detection of PC-1, as we were able to visualize the endogenous protein in total lysates in the absence of an immunoprecipitation enrichment step, as well as by immunofluorescence. Furthermore, using this system, we were able to provide conclusive evidence of the interaction between endogenous PC-1 and PC-2. Thus, we expect our model system to enormously facilitate future studies on the biology of PC-1.
Notably, we have found that PC-1 is developmentally regulated with a dramatic downregulation of its expression in most tissues during the perinatal life, as was reported previously. In particular, expression in the kidneys is dramatically downregulated during a narrow window of time, between days P4 and P10. In addition, we have shown that expression of PC-1 is ubiquitous but found at very low levels in most adult tissues, including the kidney, liver, pancreas, muscle (not shown), heart, and brain. Dissection in the different areas of the brain revealed that PC-1 is expressed in all Figure 7. Cilia morphology and apico-basal polarity in the brain from knock-out embryos at E16.5. (A to H) Centrosomes and cilia in the choroid plexus from E16.5 brain were stained with anti-pericentrin (red) and anti-acetylated alpha-tubulin antibodies (green), respectively. Multicilia, monocilia, and bundles from lateral ventricles in knock-out (B, D, and F, respectively) and in wild-type (A, C, E) do not show differences in shape or number. Cilia from choroid plexus of the third ventricle of wild-type (G) and knock-out (H) animals do not show any differences. Staining of b-catenin of the choroid plexus in the third ventricle (I and J) and Aquaporin-1 in the choroid plexus of the lateral ventricle (K, L) did not reveal major differences. Transverse sections of the cilium of an ependymal cell from the third ventricle of wild-type (M) or mutant (N) animals carried out by electronic microscopy revealed the normal structure of the cilium. Scale bar A to F 5 mm; G to L 10 mm; M and N 100 nm. doi:10.1371/journal.pone.0007137.g007 areas with slightly higher expression in the olfactory bulbs and cerebellum. Surprisingly, the organ showing the highest expression levels in the adult is the lung and we have attributed this to the absence of postnatal downregulation of PC-1 in the lung. Of great interest, a recent study has reported a bronchiectasis defect in ADPKD patients [44], suggesting that PC-1 function in lungs might be essential in humans.
In this study, we have focused our attention on characterization of PC-1 role in brain, where we were able to track the protein thanks to the novel system developed. This is not a compartment known to be affected in ADPKD, despite the reasonably high expression levels of PC-1. This is not surprising, but rather provides further support to the two-hit hypothesis, which predicts that renal cystogenesis generates as a consequence of loss of the normal copy of the PKD1 gene. Thus, our data support the notion that loss of function of both copies of the PKD1 gene are necessary in order for its function to be inactivated. Thus, based on our evidences it appears that either the PKD1 gene does not undergo second-hits in the ependyma and choroid plexus, or else that its inactivation in a single cell in this compartment does not provide a growth advantage necessary for the phenotype to manifest.
However, our data also highlight the broad role of the PC-1 receptor in mammalian development, as previously suggested based on the severe phenotype observed in Pkd1 mutant mice [2][3][4][5]. Therefore, although the brain is not an organ affected in ADPKD, we believe that studying PC-1 function in this compartment will help shedding light on fundamental features of this receptor, including its subcellular distribution and function. In this study we show that PC-1 can be detected in ependyma and Choroid Plexus (CP), where we found a strong and specific staining in all three types of cilia observed in this region of the brain, i.e., multiciliated and monociliated CP cells and multiciliated ependymal cells. We believe that the distribution observed in these cilia is very informative, as PC-1 appears to localize in a slightly different manner in the different type of cilia. In fact, PC-1 is present at the base of the cilium in the kidney and in tight bundles of multiciliated choroid plexus cells, whereas it distributes along the axoneme with an IFTlike appearance in the monociliated CP cells and it shows more diffuse axonemal staining in the multiciliated ependymal cells. These various distributions might represent differences in the functions of PC-1 or else in the speed of intraflagellar transport in the different cilia, with slower movement in the monociliated CP cells and a longer residence time at the base in the renal monocilia and in the tight multiciliated bundles of the CP. Whatever the explanation for these different staining patterns, our data are in agreement with a previous report showing that inactivation of Polaris, an IFT protein, in the Tg737 mouse results in accumulation of PC-1 at the tip of monocilia, consistent with its possible transportation along the axoneme [34]. Furthermore, studies in the biflagellate unicellular alga Chlamydomonas reinhardtii have shown that the Pkd2 orthologue fused to GFP is able to move along the axoneme, further suggesting that the Polycystins can indeed be transported via intraflagellar transport along the ciliary axoneme [45].
Finally, we not only showed that PC-1 is localized to ependymal and choroid plexus cilia, we also showed a novel phenotype of Pkd1 mice consistent with a function for PC-1 in these structures. In fact, we have observed a mild hydrocephalus both in Pkd1 mutant embryos and in newborn conditional mice carrying a Nestin-Cre transgene. Our data suggest that the brain ventricular system is impaired in the absence of a functional Pkd1 gene, most likely due to a developmental defect. During the last phases of embryonic development starting at E14.5 in the mouse, the ventricles, appearing as large cavities, start narrowing to become very compact in the newborn mouse [31]. This process appears to be impaired in the Pkd1 mutant brain at E16.5, with a major defect in the third ventricle and a much milder effect in the lateral ventricles. These data are particularly interesting when placed in the context of ependymal cilia development. In fact, it was prteviously reported that ependymal cilia are not visible in the lateral ventricles of E16.5 brains, while they are already detectable in the third ventricle [46]. At later time points during development, ependymal cilia become progressively visible in the lateral ventricles as well. Our data strongly suggest that the hydrocephalus phenotype in Pkd1 mutant mice originates from the third ventricle, the first compartment to develop ciliated ependyma, and subsequently expands to the lateral ventricles. In agreement with the phenotype observed in ubiquitously mutant embryos, we found that the third ventricle was massively enlarged at P8 in newborn mice carrying conditional inactivation of the Pkd1 gene in brain. In this case, however, the lateral ventricles were also subsequently enlarged at P10, further suggesting a progressive dilatation of all the ventricles initiating from the third ventricle. Unfortunately, the Nestin-Cre model employed in this study is also known to be active in the kidney [33,42]. The Pkd1 Flox/2 ;Nestin-Cre mice develop massive renal cystogenesis after birth, eventually causing death of the mice due to renal failure by day P12 (not shown), thus preventing us from studying the effect on hydrocephalus at later time points. However, we would predict that the phenotype becomes more severe with time. Future studies using a faithful brain-specific Cre that also targets the ependyma and choroid plexus will be necessary to clarify this aspect.
Interestingly, we found no major structural ciliary defects in the brains of these mice. This situation is different from all previously reported mice showing hydrocephalus due to a defective ciliary protein [33][34][35][36][37]. These data are in agreement with the previously proposed role of PC-1 in cilia, which was shown to be dispensable for a correct ciliogenesis, but necessary for ciliary function at least in the kidneys [27,28]. In addition, our data showing no massive defects in the apico-basal polarity of the epithelium in the choroid plexus are also in agreement with the model previously proposed for the role of cilia in CSF formation [34,37]. Previous studies on the role of Polaris (IFT88) in preventing hydrocephalus showed similar results. In that case, differences in the secretory properties of the choroid plexus cells were shown in addition to defective ependymal cilia motility, demonstrating that defective ciliary beating is not the sole explanation for hydrocephalus due to defective ciliary function [34,37]. These data were further supported by the fact that hydrocephalus in the Tg737 mouse develops prior to the initiation of ependymal cilia beating (postnatal life) [34]. Identical conclusions could be drawn from our results.
The function of cilia in most tissues is still a mystery. However, several studies have proposed a role for cilia in the regulation of different signalling pathways [38]. Therefore, in the ventricular system of the brain, it is conceivable that cilia might be important signalling platforms to allow precise control and coordinate the rate of secretion, movement, and re-absorption of the CSF. PC-1 is a large receptor able to regulate different signalling pathways. Thus, its function in the brain cilia might be to integrate this information in order to achieve the correct regulation of fluid balance.
The anti-PC-1 antibodies used in this study were previously described [21]. They are both polyclonal antibodies generated in chicken (anti-CC) or rabbit (anti-LRR) by immunization using recombinant fragments of PC-1 corresponding to the entire C-tail or to the N-terminal LRR repeats, respectively [21].

Ethical approval of all animal work
All animal care and experimental protocols, including the generation of chimeras, were conducted in accordance with the guidelines provided by the Italian Ministry of Health, upon approval of a specific protocol (IACUC-303) by the institutional care and use ethical committee (I.A.C.U.C.) at the San Raffaele Scientific Institute. Personnel from our own laboratory carried out all aspects of the mouse work under strict guidelines to insure careful, consistent and ethical handling of the mice.

Generation of knock in/out mouse line
Transgenic ES cell lines were established using standard methods with the targeting vectors HALM and MLHA (Supporting Information) and injected in C57BL/6 blastocysts to generate chimeras. F1 knock-in Pkd1 HAneo/+ and Pkd1 Mycneo/+ mice (C57BL/ 6129Sv) were crossed with Flp-expressing mice [47] to excise the neomycin cassette. To get a pure genomic background, the knockin lines were crossed with wild-type C57BL6 mice for seven further generations. Knock-out animals were generated by breeding knock-in lines with ubiquitous ATCB-Cre expressing mice [48] or by crossing knock in/out lines with Nestin-Cre mice [42].

Genotyping and MEF cells isolation and culture
Genomic DNA was isolated from tails of mice and embryos as well as MEF cells. Knock-in mouse genotyping was performed by PCR using a three primer strategy (Supporting Information) to distinguish between wild-type, heterozygous, and homozygous knock-in and knock-out mice, respectively. PCR products were analysed on agarose gels. Mouse embryonic fibroblasts (MEF) were isolated from 11.5 day-old embryos as previously described [13], and cultured in DMEM, 10% FCS, 100 U/ml penicillin G sodium, and 100 mg/ml streptomycin sulphate.
For histological analysis, sections were stained with 10% Harris Hematoxylin (Sigma Aldrich) and 1% Eosin-G solution (Bio Optica). For immunofluorescence assays, sections were permeabilised (0.1% Triton X-100) and incubated in blocking solution (3% BSA/5% goat serum) and subsequently with primary Abs overnight at 4uC and secondary Abs for 1 hour at room temperature. After staining with DAPI (1:10,000), sections were mounted in Mowiol/anti-fade solution (Invitrogen). Images were acquired and processed by Olympus IX70 DeltaVision RT Deconvolution System.

Immunoprecipitation and immunoblotting
For immunoprecipitation studies tissues and cells were harvested and homogenised in lysis buffer (150 mM NaCl, 10% Glycerol, 1 mM glycerophosphate, 1 mM sodium orthovanadate, 1 mM sodium fluoride). After centrifugation, lysates were incubated for 1 hour at room temperature or overnight at 4uC with immobilised a-HA (agarose beads) or a-Myc Abs followed by incubation with G sepharose beads. After washing the beads with lysis buffer, precipitated proteins were resolved in sample buffer, boiled, separated by 4%, 3-8% or 4-12% gradient NuPAGE pre-cast gels (Invitrogen), transferred onto PVDF membranes, and analysed with the indicated primary and secondary Abs.