The ADPKD genes pkd1a/b and pkd2 regulate extracellular matrix formation

Mutations in the polycystin genes PKD1 and PKD2 are responsible for autosomal dominant polycystic kidney disease (ADPKD), the most common heritable human disease (Sutters and Germino, 2003). The proteins encoded by the PKD genes function in mechanosensory ion channel complexes (Sutters and Germino, 2003) that regulate calcium influx via the primary cilium (Nauli et al., 2003) or calcium release from intracellular calcium stores (Koulen et al., 2002). In addition to cystic pathology of the kidney, liver and pancreatic duct (Sutters and Germino, 2003), mutations in polycystins are also linked to abdominal hernia (Morris-Stiff et al., 1997) and intracranial aneurysm (van Dijk et al., 1995); both of these disorders are commonly associated with altered integrity of the extracellular matrix (ECM). Vascular defects are also observed in Pkd1 hypomorphic mouse mutants (Kim et al., 2000; Hassane et al., 2007), where basement membrane thickening is the earliest pathology associated with dissecting aneurysm (Hassane et al., 2007). Other extra renal manifestations of ADPKD that may involve matrix structural defects include gastrointestinal cysts, cardiac valvular defects and pericardial effusion (Hossack et al., 1988; Qian et al., 2007). An altered ECM has long been known to be associated with ADPKD pathology in human tissue (Candiano et al., 1992; Somlo et al., 1993), and in ADPKD animal and cell culture models (Schafer et al., 1994). Matrix genes figure prominently in the lists of upregulated mRNAs in expression profiling studies of human cystic tissue and cystic animal models (Riera et al., 2006). Although matrix defects occur in ADPKD tissue, and mutations in matrix genes have been shown to be sufficient for cyst formation (Shannon et al., 2006), it has been difficult to establish a link between the function of polycystin1 and polycystin2 as regulators of calcium signaling, and to establish a direct role for polycystins in regulating matrix production.

Our group and others have shown that ADPKD can be modeled in zebrafish by mutation or knockdown of zebrafish polycystin2 (pkd2), also known as polycystic kidney disease 2 (Sun et al., 2004; Bisgrove et al., 2005; Obara et al., 2006; Schottenfeld et al., 2007; Fu et al., 2008). pkd2 loss of function results in left-right asymmetry defects, kidney cysts, hydrocephalus, and pronounced dorsal axis curvature. The mechanisms of cyst formation in pkd2 morphants/mutants are distinct from other zebrafish cyst mutants since, as is seen in mouse Pkd2 mutants, cilia function appears normal (Obara et al., 2006; Schottenfeld et al., 2007). Pronephric cyst formation is instead associated with altered morphogenesis of the distal pronephric nephron and absence of a patent pronephric opening (Obara et al., 2006). Cell culture studies and studies in PKD mutant mouse models point clearly to a role for polycystins in regulating intracellular calcium (Koulen et al., 2002; Nauli et al., 2003); however, how altered calcium signaling might lead to ADPKD tissue defects remains poorly understood. Here, we have sought to extend our understanding of the role of the polycystins in the cell biology of ADPKD by cloning and characterizing the function of zebrafish polycystin1 (pkd1). We find that two paralogous pkd1 genes exist in zebrafish and that simultaneous disruption of both pkd1 paralogs results in a multi-organ phenotype that is essentially identical to the pkd2 loss-of-function phenotype. Focusing on the most common and penetrant phenotype, dorsal axis curvature, we provide evidence that both pkd1a/b and pkd2 regulate collagen mRNA and protein expression, and that collagen overexpression can account for dorsal axis curvature defects. Our results suggest that polycystins may participate directly in regulating ECM production and that the matrix defects seen in ADPKD patients may be a central feature of polycystin loss of function.

Polycystin1 is present as two paralogous genes in zebrafish, pkd1a and pkd1b, based on the finding that both zebrafish pkd1 genes are most closely related to mammalian PKD1 genes, and not to mammalian polycystin 1 like (PKD1L) genes or mammalian REJ (receptor for egg jelly) orthologs (Fig. 1A,B). In situ hybridization revealed that pkd1a is expressed broadly in chondrogenic tissues (Fig. 1C–H), whereas pkd1b is expressed primarily in the nervous system (Fig. 1I–N). pkd1a is expressed diffusely throughout the embryonic axis at the 1-somite stage (Fig. 1C), and in cells surrounding Kupffer’s vesicle, an embryonic organ associated with left-right asymmetry (Fig. 1C, inset). By 24 hours post-fertilization (hpf), pkd1a expression becomes concentrated in the brain and ventral head tissue (Fig. 1D), and by 48 hpf, pkd1a expression is strongest in the forming pharyngeal arches (Fig. 1E) and in the notochord (Fig. 1F). At 72 hpf, pkd1a expression in the jaw region is concentrated in perichondrial mesenchyme (Fig. 1G) and is also detectable in cells surrounding cartilage in the fins (Fig. 1H). pkd1b is expressed in the caudal neural plate at the 6-somite stage (Fig. 1I), and later, by the 10-somite stage, it is uniformly expressed throughout the neural plate (Fig. 1J). Uniform expression of pkd1b in the spinal cord at the 18-somite stage (Fig. 1K) is followed by an anterior-to-posterior downregulation of expression, seen at 24 hpf (Fig. 1L). In free-swimming larvae (Fig. 1M,N), pkd1b expression is restricted to ependymal cells of the ventral neural tube (medial floor plate) (Fig. 1N).

To assess the function of polycystin1 paralogs, we designed antisense morpholino oligonucleotides targeting the translation start site and the exon 8 splice donor site of pkd1a, and the exon 44 and exon 45 splice donor sites of pkd1b. Reverse transcription (RT)-PCR on single morphant embryos revealed that pkd1a exon 8 morpholino (exon8MO) embryos (Fig. 2A) display a smaller pkd1a RT-PCR product, reflecting a partial deletion of exon 8 and out-of-frame fusion with exon 9, which generates a premature stop codon in the mRNA encoding the N-terminal extracellular domain of Pkd1a (Fig. 2B). Disruption of pkd1b mRNA processing resulted in in-frame deletions of the predicted Pkd1b transmembrane domains tm10 (exon44dMO; supplementary material Fig. S1) or tm11 (exon45dMO; Fig. 2C,D). Both of the pkd1b internal deletions are predicted to alter the transmembrane topology of Pkd1b and force the C-terminal tail, a domain that is essential for polycystin1 function, into the extracellular space. pkd1a exon8MO and pkd1b exon45MO splice disruption completely eliminated wild-type transcripts (Fig. 2A,C) and were used as primary morpholinos for subsequent analyses.

Knockdown of pkd1a alone using either ATG or exon 8 morpholinos resulted in hydrocephalus, jaw defects and kidney cysts (Fig. 2E). Hydrocephalus (Fig. 2E,F) was observed in all morphant embryos, whereas kidney cysts (Fig. 2F) were present in only 10–20% of morphant embryos (three out of 31 embryos in a representative experiment) despite the complete absence of wild-type pkd1a mRNA (Fig. 2A). pkd1a knockdown also resulted in short jaw defects (Fig. 2H) compared with control embryos at five days of age (Fig. 2G). Surprisingly, disrupting the expression of pkd1b with morpholinos targeted to either exon44 or exon45 did not result in a visible phenotype (data not shown). Control inverted sense and randomized sequence morpholinos for both pkd1a and pkd1b did not cause any developmental abnormalities, indicating that the observed phenotypes were specific to polycystin loss of function.

The existence of two pkd1 paralogs in zebrafish and the apparent lack of effect of pkd1b knockdown suggested that zebrafish pkd1 genes may have partially overlapping or redundant functions. To reveal the full pkd1 loss-of-function phenotype, we knocked down both pkd1a and pkd1b (hereafter referred to as pkd1a/b morphants) simultaneously by injecting a combination of the pkd1a exon 8 splice donor MO and the pkd1b exon 45 splice donor MO. Double pkd1a/b morphants exhibited a strong and fully penetrant dorsal axis curvature (Fig. 2I), in contrast to the variable axis curvature seen with knockdowns of pkd1a alone (Fig. 2E). As seen with pkd1a knockdown alone, pkd1a/b morphants developed hydrocephalus (Fig. 2I), which was confirmed in histological sections (data not shown). Both hydrocephalus and axis curvature phenotypes were strikingly similar to phenotypes observed in pkd2 morphant embryos (Fig. 2J). Jaw development was assessed in col1a1:EGFP (enhanced green fluorescent protein) transgenic embryos. Compared with wild-type embryos (Fig. 2K), pkd1a/b double morphants exhibited defects in jaw morphogenesis (Fig. 2L) that were similar to pkd1a morphants (Fig. 2H). Chondrocytes of Meckel’s cartilage persisted in a rounded immature morphology and appeared disorganized (Fig. 2L) when compared with the ordered stacking of the flattened differentiated chondrocytes in the jaws of control embryos (Fig. 2K). Calcification of jaw bones was assessed by calcein staining and also appeared delayed or inhibited in pkd1a/b morphants (Fig. 2N) compared with controls (Fig. 2M). The frequency of kidney cysts was slightly increased in pkd1a/b double morphants but never exceeded 20% of the injected embryos. Approximately 40% (in one experiment 18 of 45 injected embryos) of the pkd1a/b morphants showed severe edema at 5 days post fertilization (dpf), which is a sign of kidney failure.

We chose to focus on the dorsal axis curvature phenotype since (1) it was the most penetrant and consistent pkd1a/b-deficiency phenotype in zebrafish and (2) it was a quantitative trait, with embryos showing dose-dependent severity of curvature. Surprisingly, kidney cyst formation was poorly penetrant (10–20% of knockdown embryos), suggesting that this phenotype has a high threshold and may only be manifest in the most severely affected embryos. Axis curvature in pkd1a/b morphants was strikingly similar to the phenotype seen in pkd2 mutants and morphants (Fig. 2J) (Obara et al., 2006; Schottenfeld et al., 2007), presenting an opportunity to assess whether pkd1a/b and pkd2 function in a common pathway affecting axial morphogenesis. Previous studies have shown that polycystin 1 and polycystin 2 physically interact (Qian et al., 1997; Tsiokas et al., 1997) and that this heterodimeric assembly produces a non-selective cation channel (Hanaoka et al., 2000). To test for pkd1a/b: pkd2 genetic interactions in zebrafish, we generated partial loss-of-function embryos for both pkd1a/b and pkd2 using threshold doses of morpholinos and then assessed whether the combined low doses of pkd1a/b: pkd2 morpholinos produced a phenotype stronger than either alone. Embryos were sorted as having a curvature of less than 90 degrees (mild), more than 90 degrees (moderate), and full curvature with the tail crossing the body axis (severe) (Fig. 3). Reducing the dose of injected pkd1a/b and pkd2 morpholinos to a tenth of the full knockdown dose resulted in the majority of embryos having mild axis curvature and very few having severe dorsal axis curvature (20% of pkd1a/b morphants and 10% of pkd2 morphants) (Fig. 3). Knockdown of pkd1a/b and pkd2 together using sub-threshold doses of morpholinos revealed a striking interaction in pkd1a/b and pkd2 activity; combined knockdown resulted in a synergistic increase in the fraction of embryos displaying a severe axis curvature (56%) (Fig. 3), indicating that pkd1a/b and pkd2 function in the same molecular pathway regulating axis formation.

Axial elongation in chordate embryos is dependent on the formation of the notochord and two additional rigid structures: the floor plate and the hypochord, which are dorsal and ventral to the notochord, respectively (reviewed in Scott and Stemple, 2005). pkd1a/b-deficient embryos often displayed an undulating notochord phenotype at 24 hpf (Fig. 4B) and 48 hpf (Fig. 4C) compared with control embryos (Fig. 4A), suggesting that altered notochord structure could be a primary defect in pkd1a/b morphants. Notochord structure is dependent on the formation of a surrounding basement membrane or sheath consisting of multiple ECM proteins including laminin 5 and collagen types II, VIII and IX (Scott and Stemple, 2005). Expression of the most abundant notochord collagen protein, collagen type II alpha 1 (Col2a1), was significantly increased in the notochord sheath by knockdown of pkd1a/b (Fig. 4E) or pkd2 (Fig. 4F) compared with control embryos (Fig. 4D). These results raised the possibility that dorsal axis curvature in polycystin morphants could be the result of collagen overexpression in the notochord sheath.

To assay directly whether misregulation of notochord collagen gene expression occurred owing to pkd1a/b or pkd2 deficiency, we performed in situ hybridization using probes for col2a1, col9a2 and col27a1. The col2a1 gene is strongly expressed early during notochord formation at 20 hpf and is progressively downregulated as the notochord matures, such that, by 48 hpf, only cells of the hypochord and floor plate show col2a1 expression and, by 96 hpf, axial col2a1 expression in the notochord is minimal or absent (supplementary material Fig. S2). A comparison of notochord col2a1 expression in control injected embryos with pkd1a/b or pkd2 morphants revealed that polycystin knockdown results in abnormally persistent col2a1 mRNA expression in the notochord (Fig. 4G–I). Expression of the col9a2 and col27a1 mRNAs was also abnormally upregulated in pkd1a/b and pkd2 morphants (Fig. 4J–O), indicating that polycystin knockdown had pleiotropic effects on the expression of multiple collagen genes. Interestingly, expression of col27a1 in the distal end of the pronephric nephron (red arrowheads in Fig. 4M–O) was enhanced in pkd2 morphants but diminished in pkd1a/b morphants, suggesting that although both pkd1a/b and pkd2 affect collagen expression and can be shown to interact, they may have unique effects on collagen gene regulation in different tissues. Abnormally persistent col2a1 expression was also observed in the pkd2 retroviral insertional mutant pkd2^hi4166Tg (Fig. 4P,Q), indicating that our results using polycystin morpholino knockdowns were not the result of non-specific developmental delay. To further exclude the potential effects of non-specific developmental delay, we examined the expression of sox9a, a transcription factor that is responsible for the early activation of collagen expression in the notochord. sox9a was strongly expressed in the notochord at 24 hpf, and subsequently decayed to baseline levels by 72 hpf in wild-type embryos (supplementary material Fig. S3A–C), pkd1a/b morphants (supplementary material Fig. S3D–F) and pkd2 morphants (supplementary material Fig. S3G–I). The normal rate of sox9a downregulation in polycystin morphants excludes general developmental delay as an explanation for the observed phenotypes, and argues for a specific role for polycystins in collagen gene expression and maturation of the notochord.

Previous reports have suggested the existence of a negative feedback signaling mechanism that monitors the formation of the notochord collagen sheath (Anderson et al., 2007; Gansner et al., 2007). Inhibition of collagen crosslinking by copper chelation, direct inhibition of lysyl oxidase, or knockdown of lysyl oxidase mRNAs results in a characteristic kinked notochord and abnormally persistent expression of col2a1, similar to what we observe in polycystin morphants (Anderson et al., 2007; Gansner et al., 2007). To better define the nature of the feedback signals regulating notochord collagen expression, we knocked down col2a1 and assayed the expression of the remaining notochord collagens, col9a2 and col27a1. col27a1 (Fig. 4R,S), as well as col9a2 (supplementary material Fig. S4), were expressed ectopically in col2a1 morphants. Reciprocal experiments knocking down col9a2 and assaying the expression of col2a1 mRNA resulted in a similar ectopic expression of col2a1 (data not shown), indicating that notochord cells regulate collagen genes as a cohort and downregulate collagen expression only after a complete matrix has formed.

Overexpression of the col2a1 mRNA and protein in polycystin-deficient notochords suggested that excessive collagen deposition could underlie notochord axial defects. We therefore tested whether knocking down col2a1 expression would rescue polycystin axis defects. A morpholino oligonucleotide targeting the col2a1 exon 1 splice donor effectively eliminated normal col2a1 mRNA and generated two mis-spliced products that were predicted to truncate the Col2a1 protein immediately after the exon 1 coding sequence (supplementary material Fig. S5). As shown above, morpholino knockdown of either pkd1a/b (Fig. 5B) or pkd2 (Fig. 5C) resulted in severe dorsal axis curvature compared with control embryos (Fig. 5A). Knockdown of col2a1 alone caused a mild ventral body axis curvature (Fig. 5D). Strikingly, combined knockdown of col2a1 and pkd1a/b (Fig. 5E) or pkd2 (Fig. 5F) completely rescued the dorsal axis curvature defect caused by polycystin knockdown alone (Fig. 5B,C). The notochord sheath collagen structure can also be modified by inhibition of lysyl oxidase, an enzyme responsible for collagen crosslinking. Prevention of collagen crosslinking using the small molecule 3-aminopropionitrile (βAPN) also prevented the dorsal axis curvature in pkd1a/b (Fig. 5H) and pkd2 morphants (Fig. 5I). The results demonstrate that the polycystin axis curvature phenotype is dependent on the abnormal persistence of embryonic collagen gene expression, and further suggests that polycystins could play a direct role in regulating axial matrix assembly or deposition.

The upregulation of multiple collagen genes in polycystin morphants suggested that polycystins could function as part of a negative feedback system that detects completely assembled, crosslinked ECM and that generates signals that repress collagen gene expression. To examine the relationship between polycystin function and matrix synthesis and assembly, we assayed whether regulation of collagen gene expression was synergistically affected by disruption of collagen crosslinking (lysyl oxidase inhibition) and polycystin knockdown. MCP (2-mercaptopyridine N-oxide) is a potent inhibitor of lysyl oxidase, an enzyme that is required for collagen crosslinking and notochord matrix assembly (Anderson et al., 2007). MCP-treated embryos fail to downregulate col2a1 mRNA expression by 48 hpf and exhibit an undulating notochord phenotype (Anderson et al., 2007). As expected, embryos treated with MCP alone showed overexpression of col2a1 at 48 hpf (data not shown) (Anderson et al., 2007). However, by 4 days (96 hpf), col2a1 expression had returned to baseline (Fig. 5J), despite the fact that the notochord remained structurally distorted (Fig. 5M). By contrast, pkd2 (Fig. 5K,N) and pkd1a/b (Fig. 5L,O) morphants treated with the same dose of MCP failed to suppress col2a1 expression by 96 hpf. The results suggest that polycystins are required for negative feedback control of embryonic collagen gene expression and could play a role in signaling the completion of morphogenesis.

To identify signaling pathways that might control collagen synthesis in the context of polycystin function, we screened chemical inhibitors that had previously been implicated in both polycystin signaling and collagen gene regulation. Polycystins function in ion channel complexes that regulate calcium release from intracellular calcium stores (Koulen et al., 2002; Nauli et al., 2003; Sutters and Germino, 2003), and maintain calcium store levels in the lumen of the endoplasmic reticulum (ER). Non-toxic doses (50 μM) of the SERCA [sarco(endo)plasmic reticulum Ca²⁺ ATPase] calcium pump inhibitor cyclopiazonic acid phenocopied polycystin knockdown with respect to col2a1 overexpression (Fig. 6C). Epidermal growth factor receptor (EGFR), JAK/STAT and AKT/PI3K have also been implicated as kinase signaling pathways that mediate polycystin function. Although inhibition of EGFR or JAK/STAT had little effect on col2a1 expression, treating embryos from 14 hpf to 44 hpf with the phosphoinositide 3-kinase (PI3K) inhibitors wortmannin and LY294002 resulted in a dose-dependent enhancement of notochord col2a1 expression (Fig. 6D–I), similar to what we observed in pkd1a/b and pkd2 morphants. Despite the upregulation of collagen gene expression, PI3K inhibition did not cause dorsal axis curvature, suggesting that these inhibitors may have more pleiotropic effects on embryogenesis. Our results indicate that maintenance of intracellular calcium stores and PI3K signaling are required for the developmental regulation of col2a1 expression in the notochord.

The identification of the PKD1 and PKD2 genes as the cause of ADPKD provided important leads in determining the cellular basis of cystic disease (Hughes et al., 1995; Mochizuki et al., 1996). However, the cellular mechanisms maintaining normal tissue architecture that are controlled by polycystins remain unclear. Our finding that pkd1a/b knockdown results in pleiotropic defects in the kidney, cartilage and brain are consistent with knockout models of polycystin1 in the mouse. In addition to cystic kidneys, we observe cartilage defects similar to the undulating spinal chord and short jaw phenotypes that have been reported in mouse Pkd1 knockouts (Boulter et al., 2001; Kolpakova-Hart et al., 2008). Expression of Pkd1 in the mouse is observed in the notochord and floor plate, as well as in other chondrogenic tissues (Boulter et al., 2001), similar to what we observe in zebrafish. Expression of zebrafish pkd1 in perichondrial mesenchyme in the head parallels the expression seen in long bone perichondrial mesenchyme in the mouse (Lu et al., 2001). The delayed ossification of the skull and spinal chord in Pkd1-deficient mice (Lu et al., 2001) is consistent with defects in ossification (calcein staining) that we observe in the forming skull and jaw bones in zebrafish pkd1a/b knockdowns. We also present genetic evidence that pkd1 and pkd2 interact in zebrafish as they do in mammalian kidneys (Wu et al., 2002). These results argue that the zebrafish is a useful model to further pursue the cellular mechanisms of polycystin function.

Somewhat surprisingly, pronephric kidney cyst formation was not highly penetrant in pkd1a/b morphants. Although we could demonstrate complete knockdown of pkd1a and pkd1b mRNA, kidney cysts formed in only 20% of injected embryos. Our results may be most relevant to conditional knockout of Pkd1 in juvenile mouse models (after postnatal day 14) where cyst formation is also poorly penetrant and slow to develop (Piontek et al., 2007; Takakura et al., 2008). In this context, cyst formation may require additional kidney insults beyond PKD1 loss of function (Takakura et al., 2008). Loss of pkd1a and pkd1b may sensitize embryos to other morphogenesis defects or injury, but by itself may not be sufficient to cause kidney cysts at high frequency. To analyze the cellular basis of pkd1 function, we focused instead on the role of polycystins in chondrogenesis and fully penetrant axial morphogenesis defects. The requirement for double pkd1a/b knockdown for full expression of axis curvature (compared with single gene knockdown in pkd2 morphants) probably reflects the cell type-specific expression of pkd1a in the notochord and pkd1b in the floor plate, whereas pkd2 is expressed ubiquitously (Obara et al., 2006). In this view, pkd1a/b axis curvature is a compound phenotype involving defects in both the notochord and floor plate. Electron microscopic examination of the notochord sheath did not reveal strong qualitative differences in collagen structure between pkd1a/b morphants and controls; collagen fibril periodicity and diameter did not appear different (data not shown). It is therefore unlikely that pkd1a/b deficiency caused collagen crosslinking defects in the notochord like those observed with βAPN. The notochord undulations that were seen during early development in pkd1a/b morphants probably arise by a different mechanism involving overproduction of collagen.

In mouse models of Pkd1 deficiency, raising or lowering the amount of polycystin1 protein expression is sufficient to cause polycystic kidney disease (PKD) kidney phenotypes (Pritchard et al., 2000; Lantinga-van Leeuwen et al., 2004), indicating that the Pkd1 gene dosage is crucial for normal tissue architecture. For both zebrafish pkd1a/b and pkd2, we found that the axial curvature phenotype is highly sensitive to gene dosage and provides a quantitative readout of polycystin function. The overt similarity and genetic interaction that we demonstrate in pkd1a/b and pkd2 axis curvature defects substantiates the idea that pkd1 and pkd2 interact in an underlying cellular process or signaling pathway that ensures proper morphogenesis (Wu et al., 2002).

Our results suggest that both pkd1a/b and pkd2 are required for notochord cells to transition to a mature differentiated phenotype. In the absence of polycystins, notochord cells persistently express embryonic collagen genes at high levels, resulting in overproduction of the ECM and distortion of the embryonic axis. These results support the idea that polycystin1 may act as a sensor of differentiation to repress embryonic programs of gene expression. Similar evidence for a role for polycystin1 as a sensor of differentiation is found in the persistent overexpression of embryonic matrix genes and receptors (collagens type I, III and IV, laminins, heparan sulfate proteoglycan, β4 integrin), cell surface proteins (erb-B2, β2 Na/K ATPase subunit), and secreted proteins (periostin) in ADPKD tissue (Ebihara et al., 1988; Haverty and Neilson, 1988; Ebihara et al., 1995; Joly et al., 2003; Wallace et al., 2008). These results have implications for understanding the fibrotic defects that are often associated with PKD (Cuppage et al., 1980; Wilson et al., 1992), as well as understanding other Pkd1 matrix-associated phenotypes, including intracranial aneurysm (Brooke et al., 2003; Ruigrok et al., 2005); vascular fragility and maintenance of the aortic wall ECM (Kim et al., 2000; Hassane et al., 2007); overlapping connective tissue disorders (Somlo et al., 1993; Kaplan et al., 1997); abdominal wall hernia (Morris-Stiff et al., 1997); and increased pericardial compliance associated with pericardial effusion in ADPKD patients (Qian et al., 2007). Our finding that a reduction in matrix gene expression by col2a1 knockdown can rescue polycystin axis curvature defects suggests that abnormalities in matrix composition or integrity are likely to be developmental defects linked directly to polycystin function, as opposed to secondary consequences of tissue damage or deformity. This view of polycystins as sensors of mature ECM may be analogous to ‘outside-in’ integrin signaling (Arnaout et al., 2005). A role for polycystin1 as a matrix sensor (Fig. 7) is supported by findings that (1) polycystin1 can bind to multiple matrix molecules, including collagen types I, II and IV (Weston et al., 2001); (2) polycystin1 is localized, in part, to focal complexes in both muscle and epithelial cells (Wilson and Burrow, 1999; Qian et al., 2003a); (3) polycystin1 interacts with focal adhesion proteins (Wilson et al., 1999); and (4) polycystin1 can recruit focal adhesion signaling molecules to alter cell adhesiveness and migration (Wilson et al., 1999; Joly et al., 2006).

Cells sense their local environment by both ligation-induced signaling, which depends on cell-ECM biochemical interactions, and by traction-induced signaling, which depends on cellular force generation between points of cell-ECM contact (Arnaout et al., 2005; Giannone and Sheetz, 2006). In our experiments with collagen cross-linker inhibitors and polycystin knockdowns, we find no evidence for the absence of a matrix ligand; indeed, we observe that more type II collagen protein is synthesized from increased expression of col2a1 mRNA. Nonetheless, the embryos appear to lack a negative feedback signal that represses collagen gene expression. This would argue that ligation-induced signaling is unlikely to initiate negative feedback signaling. Inhibition of collagen crosslinking produces a non-rigid ECM that would interfere with the ability of the cell to generate traction forces, cluster adhesion receptors, and initiate focal adhesion kinase cascade signaling (Wozniak et al., 2004; Giannone and Sheetz, 2006). The idea that polycystin1 participates in traction-induced signaling remains to be tested.

Whatever the nature of the initiating signal, our studies suggest that negative feedback regulation of collagen gene expression involves activation of PI3K and calcium signaling. Consistent with our results, cultured mammalian epithelial cells overexpressing polycystin1 activate PI3K and the downstream kinase Akt (Boca et al., 2007). G-protein inhibition with pertussis toxin was sufficient to downregulate PI3K activity in Pkd1-overexpressing cells, indicating that polycystin1 may activate Gi as part of its signaling cascade (Calvet and Grantham, 2001; Boca et al., 2007). Alternatively, PI3K is also known to be activated by focal adhesion kinase (FAK) (Hanks et al., 2003). Interestingly, a small C-terminal fragment of polycystin1 is sufficient to recruit FAK (Joly et al., 2006), and FAK recruitment to focal complexes containing polycystin1 does not occur in ADPKD (polycystin mutant) cells (Wilson et al., 1999). Together, these findings suggest that, in both fish and mammalian cells, polycystin1 may signal the state of their environment by activating PI3K.

Immunolocalization studies have revealed that polycystin1 can function in several different subcellular domains including the apical cilium, focal adhesions, desmosomes, or intracellular ER membranes (Wilson et al., 1999; Scheffers et al., 2000; Xu et al., 2001; Yoder et al., 2002; Grimm et al., 2003; Qian et al., 2003a). It is currently not known which of these locations is involved in the regulation of axial morphogenesis in zebrafish. Expression of a mutant form of polycystin2 that is retained in the ER has been reported to partially rescue zebrafish pkd2 morphant body axis curvature, suggesting that polycystin2 regulation of ER calcium stores or calcium release may be an important site of polycystin2 signaling in axis morphogenesis (Fu et al., 2008). Our finding that treating embryos with cyclopiazonic acid, an inhibitor of the ER calcium ATPase, phenocopies polycystin loss of function in terms of the notochord defects and the persistence of collagen gene expression is consistent with this idea. These results suggest an alternative model of polycystin function in the ER to ensure proper collagen processing in the secretory pathway (Fig. 7). Both polycystin1 and polycystin2 are known to be required to maintain proper ER lumenal calcium levels (Koulen et al., 2002; Grimm et al., 2003; Nauli et al., 2003; Qian et al., 2003b; Qian et al., 2003a; Hooper et al., 2005; Anyatonwu et al., 2007; Xu et al., 2007; Weber et al., 2008). It is possible that, in addition to a role as a matrix sensor, polycystin maintenance of ER calcium may also be required for proper collagen crosslinking and assembly.

Although our group and others do not detect any structural or motility defects in pkd1 or pkd2 morphant cilia (data not shown) (Obara et al., 2006; Sullivan-Brown et al., 2008), the primary cilium could function as an alternate site of polycystin-initiated signaling. If this were the case, it might be expected that both polycystin and primary cilium loss-of-function mutants would show similar phenotypes. Instead, we find that zebrafish intraflagellar transport protein (IFT) mutants that lack functional cilia (ift172, double bubble; ift88, oval) do not exhibit dorsal axis curvature or persistent collagen gene expression in the notochord. In addition, the ventral axis curvature that is exhibited by IFT mutants is not rescued by col2a1 knockdown (Y.L. and I.A.D., unpublished). In a similar vein, kidney-specific knockout of Pkd1 in mice generates a more severe and early cystic phenotype (Shibazaki et al., 2008) than a similar (using the same ksp-Cre recombinase line) kidney-specific knockout of Kif3a, a gene essential for ciliogenesis (Lin et al., 2003). The increased severity of kidney phenotypes in mutants lacking polycystin1 versus those specifically lacking cilia suggests that, in addition to its function in the cilium, polycystin1 is likely to function at additional sub-cellular locations.

In summary, our results suggest a role for polycystin1 and polycystin2 in a negative feedback signaling system that detects mature crosslinked ECM formation and that regulates the expression of multiple collagen genes accordingly. Although our work has focused on the zebrafish notochord, studies with humans and mouse models of cystic kidney disease suggest that the sensor malfunction model for ADPKD may also be applicable to epithelial cells in cystic kidneys (Bissler and Dixon, 2005).

Wild-type TüAB or WIK zebrafish were maintained and raised, as described previously (Westerfield, 1995). Dechorionated embryos were kept at 28.5°C in E3 solution with or without 0.003% PTU (1-phenyl-2-thiourea; Sigma) to suppress pigmentation and were staged according to somite number (som) or hpf (Westerfield, 1995). The retroviral insertion line pkd2^hi4166Tg was a gift from Zhaoxia Sun (Yale University) and the col1a1:EGFP transgenic line was a gift from Shannon Fisher.

Zebrafish pkd1a cDNA was amplified from total RNA at 48 hpf by RT-PCR using a primer design based on TBLASTN searches of the Sanger Center zebrafish genomic sequence using the mammalian PKD1 protein sequence. Overlapping 3–4-kb RT-PCR fragments were sequenced directly and the sequence was assembled using Phred/Phrap (Ewing et al., 1998). pkd1b was isolated as a partial cDNA clone from a 24-hpf whole-embryo zebrafish cDNA library using low-stringency hybridization conditions and the human PKD1 C-terminus as a probe. The predicted protein sequences for each gene were compared with human polycystins using ClustalW, and the output was plotted using Drawtree (PHYLIP 3.65) as an unrooted phylogenetic tree.

In situ hybridization was performed using standard techniques, as described previously (Drummond et al., 1998). The accession numbers for the cDNAs used are: col2a1a (NM_131292), col9a2 (NM_212579) and cb221(col27a1) (BQ169308). The cb221 expressed sequence tag (est) represents the 3′-UTR of the predicted transcribed locus hmm358864 (predicted protein XP_001343932), which encodes the zebrafish homolog of collagen type XXVII, alpha 1 (col27a1). Whole-mount immunocytochemistry was performed on embryos fixed in methanol:DMSO (80:20) (anti-chicken collagen II alpha 1 monoclonal antibody II-II6B3, Developmental Biology Hybridoma Bank). Embryos were blocked in 10% normal goat serum (NGS) and then incubated with primary antibody in PBS with 1% DMSO, 2% NGS and 0.1% Tween-20 overnight at 4°C. After washing in incubation medium, secondary antibodies (Alexa 548 anti-mouse; Molecular probes) were used at 1:1000 and the nuclei were counterstained with DAPI (1:3000, Molecular Probes). Immunostaining of notochord anti-collagen II alpha 1 was performed on 100-μm vibratome sections (Leica VT 1000S vibratome) of agarose-embedded embryos. Immunostaining was preceded by treatment in 6 M urea, 100 mM glycine, pH 3.5, for 1 hour at 4°C. Washed tissue sections were cleared in mounting media (53% benzyl alcohol, 45% glycerol, 2% N-propyl gallate) and photographed on a Nikon E800 fluorescence microscope or on a Zeiss LSM5 Pascal confocal fluorescent microscope. The col1a1:EGFP transgenic and calcein-stained embryos (Du et al., 2001) were imaged on a Zeiss LSM5 Pascal confocal microscope as image stacks, and then projected using Zeiss software to make the final image. For the sections, stained embryos were dehydrated and embedded in glycolmethacrylate (JB-4) resin (Polyscience Inc.) following the manufacturer’s instructions and cut at a thickness of 1–4 μm.

Wild-type embryos (TüAB or WIK) at the 1- to 2-cell stage were microinjected with 4.6 nl of a 0.05–0.5 mM antisense morpholino oligonucleotide solution (Gene Tools LLC) in 200 mM KCL with 0.1% Phenol Red using a nanoject2000 microinjector (World Precision Instruments). The final antisense morpholino oligo concentrations that were used for the full loss-of-function experiments were sufficient to eliminate wild-type mRNA as judged by RT-PCR. The final amounts of morpholino ranged from approximately 1 to 10 ng/embryo. Randomized and inverted sequence morpholinos for pkd1a and pkd1b were used as negative controls and showed no effect on development; the sequence of the Gene Tools standard negative control MO that was used is: 5′-CCTCTTACCTCAGTTACAATTTATA-3′. The sequence of the translation-blocking oligonucleotide for pkd1a was pkd1a MO ATG: 5′-GTCTGTTCCTGAGACAGTACCGG-3′, and for cyclops (ndr2) was cyc MO: 5′-GCGACTCCGAGCGTGTGCATGATG-3′ (Karlen and Rebagliati, 2001). The splice donor-blocking oligonucleotide sequences were pkd1a MO ex8: 5′-GATCTGAGGACTCACTGTGTGATTT-3′; pkd1b MO ex45: 5′-ACATGATATTTGTACCTCTTTGGTT-3′; pkd1b MO ex44: 5′-TAAAAATACTGTACCATCATGCCTA-3′ and col2a1 MO ex1: 5′-TGAAAAACTCCAACTTACGGTCATC-3′. Morpholinos targeting pkd2 have been previously described (Obara et al., 2006). Nested RT-PCR primers were designed in flanking exon sequences to confirm the morpholino oligonucleotide efficacy and to characterize altered mRNA splicing products: for pkd1a MO ex8: outside F: 5′-GGCGGAGCTTTCTCTGGTCA-3′, outside R: 5′-GAACGTGGCGTGTGAACTGG-3′, inside F: 5′-GGTCAACTGGGGTGGAGTGG-3′, inside R: 5′-TTTGGCGGTGCAGGAGTGTA-3′; for pkd1b MO ex44: outside F: 5′-TGGCTTTGGCAGCTGCTGTTCTA-3′, outside R: 5′-AGCAGCACATAGGCGTCCCAGTA-3′, inside F: 5′-TCTTAGGCGATGGTTGGTCATGG-3′, inside R: 5′-ACCCTGTGCGACCCCTTAACAGA-3′; for pkd1b MO ex45: outside F: 5′-TGGCTTTGGCAGCTGCTGTTCTA-3′, outside R: 5′-GGGCTTAGTGGGGTC CAGCAGTT-3′, inside F: 5′-TCTTAGGCGATGGTTGGTCATGG-3′, inside R: 5′-CCTGAGAGGATCTGGAGGGTGGA-3′; for col2a1 MO ex1: outside F: 5′-GGCGACTTTCACCCCTTAG-3′, outside R: 5′-GGACTTCCCTTCTCACCCTTA-3′, inside F: 5′-CACCCCTTAGGACCTGCAT-3′, inside R: 5′-CCTCTTTCTCCACGTGTTCC-3′.

Collagen crosslinking was inhibited by treating embryos in E3 with the lysyl oxidase inhibitors MCP (25-200 nM, as indicated) or βAPN (10 mM) from post-gastrulation (6–8 hpf) to approximately 48 hpf (as indicated). The PI3K inhibitors wortmannin and LY294002 were dissolved in DMSO as 10 mM stock solutions and used at the indicated concentrations in E3 egg water containing 1% DMSO. Embryos were treated with PI3K inhibitors from 14 hpf to 44 hpf, fixed in 4% formaldehyde/0.1 M phosphate buffer, and then processed for in situ hybridization, as described previously.

For histology analysis, embryos were fixed in 1% paraformaldehyde, 1.5% glutaraldehyde, 70 mM NaPO₄, pH 7.2, 3% sucrose, then embedded in JB-4 resin (Polyscience, Inc.) and sectioned at a thickness of 4 μm. For Alcian Blue staining, embryos were fixed at 5.5 dpf in 4% formaldehyde overnight at 4°C, followed by four 20-minute washes in PBST and digestion with proteinase K for 30 minutes. Embryos were postfixed for 1 hour in 4% PFA, washed four times for 20 minute in PBST, and then stained in 0.1% Alcian Blue dissolved in 70% ethanol/30% glacial acidic acid overnight at room temperature. Embryos were destained using 1% KOH/3% H₂O₂ and photographed on a Leitz MZ12 dissecting scope with a Spot Insight digital camera (Diagnostic Instruments).

This work was supported by NIH grants DK53093 and DK54711 to I.A.D.; PKD Foundation fellowship grants to S. Mangos and A.Z.; and a PKD Foundation Grant in aid to I.A.D. We also acknowledge the assistance of our colleagues David Beier for sequencing a pkd1b bac, Richard Sandford for help in cloning pkd1b, Shannon Fisher for sharing her col1a1:GFP transgenic, Nathan Hellman and Alexandra Petrova for editing the manuscript, and Amin Arnaout for helpful suggestions. Deposited in PMC for release after 12 months.