Genetic Analysis Reveals a Hierarchy of Interactions between Polycystin-Encoding Genes and Genes Controlling Cilia Function during Left-Right Determination

During mammalian development, left-right (L-R) asymmetry is established by a cilia-driven leftward fluid flow within a midline embryonic cavity called the node. This ‘nodal flow’ is detected by peripherally-located crown cells that each assemble a primary cilium which contain the putative Ca2+ channel PKD2. The interaction of flow and crown cell cilia promotes left side-specific expression of Nodal in the lateral plate mesoderm (LPM). Whilst the PKD2-interacting protein PKD1L1 has also been implicated in L-R patterning, the underlying mechanism by which flow is detected and the genetic relationship between Polycystin function and asymmetric gene expression remains unknown. Here, we characterize a Pkd1l1 mutant line in which Nodal is activated bilaterally, suggesting that PKD1L1 is not required for LPM Nodal pathway activation per se, but rather to restrict Nodal to the left side downstream of nodal flow. Epistasis analysis shows that Pkd1l1 acts as an upstream genetic repressor of Pkd2. This study therefore provides a genetic pathway for the early stages of L-R determination. Moreover, using a system in which cultured cells are supplied artificial flow, we demonstrate that PKD1L1 is sufficient to mediate a Ca2+ signaling response after flow stimulation. Finally, we show that an extracellular PKD domain within PKD1L1 is crucial for PKD1L1 function; as such, destabilizing the domain causes L-R defects in the mouse. Our demonstration that PKD1L1 protein can mediate a response to flow coheres with a mechanosensation model of flow sensation in which the force of fluid flow drives asymmetric gene expression in the embryo.


Introduction
The internal organs and vasculature of vertebrates are highly asymmetrical between left and right. For example, the heart and stomach become positioned towards the left side while the lungs undergo asymmetric branching, giving rise to different numbers of lobes on the left and right sides. Severe situs defects, such as the loss of concordance in positioning between different organs, called heterotaxia [1], are not usually compatible with survival. Moreover, situs defects frequently manifest as congenital heart malformations [2,3] and therefore represent a major healthcare concern.
During mouse embryogenesis, left-right (L-R) symmetry is broken by leftward fluid flow across the surface of the node, a transient embryonic cavity filled with extracellular fluid [4,5]. Leftward (or nodal) flow is generated by the clockwise motion (when viewed ventrally) of 200-300 motile monocilia, referred to here as nodal cilia. The core logic of this nodal flow model, that symmetry is broken by cilia-driven leftward flow, is now known to apply to other species including the teleost fish, zebrafish and medaka fish, as well as the frog Xenopus laevis [6].
As a result of nodal flow, asymmetries in gene expression emerge around the node in peripherally-located crown cells at the lateral edges of the pit. Cerl2 (Mouse Genome Informatics-Dand5) becomes expressed in crown cells with a right-sided bias (R>L) by the late headfold (LHF) to early somite stages [7,8]. CERL2 then asymmetrically represses Nodal signaling, resulting in an R<L asymmetry in crown cell NODAL activity [9]. This bias is transmitted to the lateral plate mesoderm (LPM) where a cascade consisting of NODAL, its' feedback repressor LEFTY2, and the transcription factor PITX2 (henceforth, the 'Nodal cascade') is established on the left side only [10,11]. This left-restricted Nodal cascade is critical for proper organ L-R asymmetry and is highly conserved throughout vertebrates and some invertebrates [12]. Importantly, expression of Lefty1 at the midline in the prospective floor plate (PFP) inhibits left-sided NODAL signals from spreading and activating the Nodal cascade on the right side of the mouse embryo [13,14], thereby maintaining unilateral left-sided pathway activity.
How nodal flow in the node drives downstream gene expression asymmetries in crown cells has remained unanswered. Previous studies have suggested roles for Polycystin-2 (PC-2, PKD2 or TRPP2) and Polycystin 1-like 1 (PKD1L1) in the response to nodal flow [15][16][17][18][19][20]. PKD2 is a six-pass transmembrane protein that acts as a non-selective cation channel [21], while PKD1L1 is an eleven-pass transmembrane protein which is argued to be sensory. PKD1L1 is expressed within the node and interacts with PKD2 [16,17], leading to the notion that PKD1L1-PKD2 complexes within nodal cilia act as sensors of nodal flow [22]. However, the nature of the asymmetric signal that engages putative PKD1L1-PKD2 sensory complexes is not known. One proposal, the morphogen model, argues that the concentration of a chemical determinant becomes asymmetric in response to flow, thereby initiating a left-sided pathway [5,23]. A second model, historically called the 'two-cilia hypothesis' (here referred to as the mechanosensation model), posits that the force of flow within the node is sensed on the left side, where it is likely to be stronger, thereby initiating events on the left that ultimately activate Nodal [18].
Polycystin-1 family members are known to bind to Polycystin-2 family members (TRPPs) and form receptor-channel complexes in contexts beyond L-R patterning. For example, PKD1 and PKD2 form complexes that are thought to sense urine flow and elicit downstream Ca 2+ signals in the kidney; defects in this process may underlie autosomal dominant polycystic kidney disease (ADPKD) [24][25][26][27]. Moreover, the relative dosage of PKD1 and PKD2 influences the activity of stretch-activated ion channels (SACs) to regulate the sensation of pressure and control the arterial myogenic tone [28]. In this context, PKD1 and PKD2 act in an antagonistic fashion to control downstream events. Thus, diverse roles exist in the sensation of forces by PKD1/PKD2 in epithelial and endothelial cells. Polycystin complexes have also been documented to respond to other kinds of stimuli. PKD1L3/PKD2L1, for example, assemble to form an acid-sensing ion channel complex [29,30] which plays a role in sour taste responses [31]. Thus, pairs of polycystin proteins sense a variety of stimuli via genetic and molecular mechanisms that are not well understood.
Here, we elaborate a genetic cascade acting downstream of nodal flow in L-R patterning that results in initiation of left-sided Nodal. We describe a Pkd1l1 mutant in which the Nodal cascade is bilaterally activated, suggesting that Pkd1l1 is required not for Nodal activation but to restrict Nodal to the left side. Moreover, Pkd1l1 acts genetically upstream of Pkd2 and downstream of flow; this regulates Cerl2 asymmetry in a cilia-dependent fashion. Using artificial flow in a cell culture system, we find that PKD1L1 can mediate a response to fluid flow, initiating a Ca 2+ signaling event upon the onset of flow. Finally, we demonstrate that an extracellular polycystic kidney disease (PKD) domain is critical for PKD1L1 function: a destabilizing mutation in the domain results in a lack of the flow-induced Ca 2+ response in cultured cells and L-R patterning abnormalities in the mouse. This provides evidence that PKD1L1 mediates sensation of fluid flow in L-R patterning.

Results
We have previously analyzed a mouse Pkd1l1 point mutant, called Pkd1l1 rks (Fig 1A) [16]. Pkd1l1 rks/rks homozygotes fail to activate the LPM Nodal cascade and exhibit morphological L-R defects similar to those displayed by both Pkd2 -/null mutants and Pkd2 lrm4/lrm4 point mutants ( Fig 1A) [16,19,32]. However, since Pkd1l1 rks is a point mutant, we set out to address the impact on L-R development of a distinct Pkd1l1 allele, namely the targeted mutation Pkd1l1 tm1Lex (here named Pkd1l1 tm1 ), which is likely a loss-of-function allele (S1 Fig) [33].

Pkd1l1 tm1/tm1 and Pkd1l1 rks/rks Mutants Exhibit Discrete Abnormalities
When assessed on the same genetic background, Pkd1l1 tm1/tm1 and Pkd1l1 rks/rks mutants exhibited distinct phenotypes. First, while the homozygous Pkd1l1 rks mutation is lethal at embryonic day (E) 14.5-15.5 [16], we found that a proportion of Pkd1l1 tm1/tm1 mutants survived until adulthood (S2A and S2B Fig). Surviving mutants either exhibited normal (57%) or reversed (43%) situs based on stomach position in neonates ( Fig 1B). Second, though the laterality of heart and stomach was randomized in both Pkd1l1 tm1/tm1 and Pkd1l1 rks/rks mutants at E13.5 (S2C-S2E Fig), the latter exhibited a significantly higher level of discordance between the two organs, reflecting a heterotaxic phenotype (Fig 1C). Third, whilst all Pkd1l1 rks/rks mutants exhibited right lung isomerism, where four lung lobes develop on both the left and right sides [16], the majority of Pkd1l1 tm1/tm1 embryos displayed the complete opposite phenotype of bilaterally mono-lobed lungs, called left lung isomerism (Fig 1D-1F). This phenotype was not (C) Heart-stomach discordance (H-S Disc.) in Pkd1l1 tm1/tm1 , Dnah11 iv/iv and Pkd1l1 rks/rks mutants scored at E13.5. Normally, the heart apex and stomach are positioned to the left. H-S Disc. is defined as the heart apex and stomach being on opposite sides. ns-not significant; *-p<0.05; **-p<0.001, Fisher's Exact Test applied. (D-F) Lung situs assessed at E13.5 for embryos of the indicated genotypes with the ratio of lung lobes between left and right sides given. The percentage and total numbers of embryos showing each phenotype are indicated in (F). (G-P) Expression patterns of Nodal, Pitx2, and Lefty1/2 in embryos at E8.5 of the indicated genotypes, with the percentage number of embryos exhibiting each phenotype and the total number given. Embryos exhibiting bilateral marker expression are further categorized by whether they show equal or biased expression between the left and right sides. The inset in (M) shows a Pkd1l1 tm1/tm1 embryo with bilateral Pitx2 expression but with a right-sided bias. Arrowheads in (N) and (O) indicate midline Lefty1 expression. t is shorthand for Pkd1l1 tm1 . (Q-R) Sonic hedgehog (Shh) expression in the node (n) and notochord (nc) at E8.5.

Pkd1l1 tm1/tm1 Mutants Bilaterally Activate the Nodal Cascade
To pinpoint the molecular causes of the L-R defects in Pkd1l1 tm1/tm1 mutants, we examined L-R marker gene expression by mRNA whole mount in situ hybridization (WISH). In control embryos, Nodal was nearly always expressed in the left LPM between the 3-7 somite stages (ss) (Fig 1G and 1I), as expected. In contrast, Nodal was most often expressed bilaterally in Pkd1l1 tm1/tm1 mutants (Fig 1H and 1I). Similarly, Pitx2 and Lefty2, both downstream targets of Nodal signaling in the LPM, were most frequently expressed bilaterally in Pkd1l1 tm1/tm1 embryos (Fig 1J, 1K, 1M and 1P). Of those embryos that exhibited bilateral activity of the Nodal cascade, most displayed similar levels and extents of Nodal/Lefty2/Pitx2 expression on both sides, but some embryos exhibited a left or right bias (Fig 1I, 1M and 1P). Whilst the bilateral induction of the Nodal cascade in Pkd1l1 tm1/tm1 embryos explains the high level of left lung isomerism, the fact that some embryos exhibit slight, though randomized, asymmetries in this bilaterality might serve to explain why we find a relatively high incidence of normal, inverted, and other lung lobation patterns (50% in total; Fig 1F) that depart from the left isomerism predicted by truly bilateral Nodal signals. In summary, most Pkd1l1 tm1/tm1 mutant embryos exhibit bilateral activation of the Nodal cascade, representing another striking difference to Pkd1l1 rks/rks mutants that display the opposite phenotypes of loss of LPM Nodal activity [16].
As a corollary of these results, we generated Pkd1l1 rks/tm1 trans-heterozygotes to further understand the differences between the Pkd1l1 rks and Pkd1l1 tm1 alleles. The majority of Pkd1l1 rks/tm1 embryos exhibited right lung isomerism and loss of LPM Pitx2 expression ( Fig  1F, 1L and 1M), suggesting that the Pkd1l1 rks allele is dominant over Pkd1l1 tm1 . We discuss the nature of the Pkd1l1 rks allele further in the Discussion.
Bilateral Nodal cascade activity in L-R mutants can result from midline defects [5,34], including the failure to express Lefty1 in the floor plate which constitutes a molecular 'midline barrier' [13]. The node and notochord were clearly specified in Pkd1l1 tm1/tm1 mutants, as evidenced by normal Sonic hedgehog (Shh) expression (n = 6/6 embryos) (Fig 1Q and 1R). Lefty1 was also expressed at the midline of Pkd1l1 tm1/tm1 mutants; indeed Lefty1 expression appeared more intense than in control embryos, perhaps owing to bilaterally high NODAL activity inducing Lefty1 in both the left and right sides of the floor plate (Fig 1N and 1O). Thus, the Lefty1 midline barrier is formed in Pkd1l1 tm1/tm1 mutants. Since Pkd1l1 tm1 appears to be lossof-function (hypomorphic or null), these data argue that rather than being required for activation of the LPM Nodal cascade, Pkd1l1 is instead needed to restrict Nodal pathway activity to the left side of the embryo in a mechanism that is independent of the midline barrier.

Pkd1l1 Acts Downstream of Nodal Flow
Bilateral establishment of the Nodal cascade in both the right and left LPM can be caused by aberrant or absent nodal flow [35,36]. We therefore imaged the node using differential interference contrast (DIC) microscopy and found that nodal cilia rotated normally, at around 12 Hz, in both Pkd1l1 tm1/tm1 mutants and wild-type controls (S3 Fig). We next assessed flow directly by immersing embryos in fluorescent beads, imaging the node cavity, then performing particle image velocimetry (PIV) analysis [37] to obtain a map of flow movements. This revealed that directional leftward flow was indeed generated both in control embryos and Pkd1l1 tm1/tm1 mutants (Fig 2A, 2B and S4 Fig). Thus, the Nodal cascade is induced bilaterally even in the presence of normal nodal flow in Pkd1l1 tm1/tm1 mutants.
Given that nodal flow is present in Pkd1l1 tm1/tm1 embryos while LPM Nodal expression is predominantly bilaterally activated, we predicted that Pkd1l1 acts downstream of flow. Indeed, this would be expected for molecules required for flow-sensing. In order to determine the genetic control and the order of these events we assessed epistasis between Pkd1l1 rks and Dna-h11 iv ; Dnah11 iv disrupts an axonemal dynein heavy chain, resulting in immotile nodal cilia and loss of nodal flow (Fig 2C and S4 Fig) [35]. At E13.5 (lung situs assessed) and E8.5 (Pitx2 expression assessed), Pkd1l1 rks/rks ;Dnah11 iv/iv double mutants exhibited right lung isomerism and loss of LPM Pitx2, respectively, and therefore phenocopied Pkd1l1 rks/rks mutants rather than Dnah11 iv/iv mutants (Fig 2D and S1 Table). This demonstrates that the Pkd1l1 rks/rks phenotype still manifests in the absence of nodal flow. Similar results were obtained when we assessed epistasis between Dnah11 iv and Pkd2 lrm4 ; in this experiment, Pkd2 lrm4/lrm4 ;Dnah11 iv/iv double mutants phenocopied Pkd2 lrm4/lrm4 and not Dnah11 iv/iv (Fig 2D and S1 Table). Thus, genetically ablating flow, which is otherwise present, does not alter the L-R phenotype of either Pkd1l1 rks/rks or Pkd2 lrm4/lrm4 embryos. This demonstrates that the roles of Pkd1l1 and Pkd2 are Flow was normal in Pkd1l1 tm1/tm1 mutants and wild-type controls but was absent in Dnah11 iv/iv mutants. Black arrowheads denote the direction and speed of flow at that position while the false coloring indicates the direction and magnitude of the flow. Red indicates leftward and blue rightward fluid movements. (D) Lung situs (assessed at E13.5) and Pitx2 expression (assessed at E8.5) for embryos of the indicated genotypes, with the percentage of embryos exhibiting each phenotype and the total number given. (E) Pitx2 expression for Pkd1l1 tm1/tm1 , Dnah11 iv/iv and control embryos for each of the 1-7 somite stages. The onset of Pitx2 expression is delayed in Dnah11 iv/iv mutants but not in Pkd1l1 tm1/tm1 embryos. downstream of nodal flow, supporting the notion that these genes mediate the sensation of flow. We did not generate Pkd1l1 tm1/tm1 ;Dnah11 iv/iv double mutants because the phenotypes of the single mutants are highly similar (S2 Fig), so an analysis of double mutant embryos would be uninformative.
Induction of the Nodal Cascade Is Delayed in Embryos Lacking Flow but Not in Pkd1l1 tm1/tm1 Mutants If the response to flow is mediated through PKD1L1, it might be expected that loss of nodal flow (Dnah11 iv/iv mutants) and targeted mutation of the putative flow sensor (Pkd1l1 tm1/tm1 mutants) would result in the same phenotype. Indeed, at E13.5, the L-R defects of Dnah11 iv/iv and Pkd1l1 tm1/tm1 embryos are highly similar (Fig 1C, 1F and S2E Fig). In wild-type embryos, Nodal is expressed in the left LPM for around 6-8 hours beginning at the 3 ss [38,39]. However, we found that the timing of LPM Nodal cascade induction, assessed by WISH for Pitx2, was subtly different between Dnah11 iv/iv and Pkd1l1 tm1/tm1 mutants. While Pitx2 expression was evident at the 3 ss but not before in control embryos, loss of flow in Dnah11 iv/iv mutants delayed the onset of Pitx2 until the 4 ss ( Fig 2E); a delay of around 2 hours, perhaps resulting from a slower and stochastic activation of the cascade in the absence of the flow signal [35]. No such delay was evident in Pkd1l1 tm1/tm1 mutants; 100% of 3 ss Pkd1l1 tm1/tm1 embryos exhibited LPM Pitx2 expression, and a single embryo displayed expression at the 2 ss ( Fig 2E). Coupled to the above findings, this favors a model wherein nodal flow drives the timely repression of Pkd1l1 on the left side: when Pkd1l1 function is perturbed in Pkd1l1 tm1/tm1 mutants, the Nodal cascade activates bilaterally without delay.

Pkd1l1 Acts Upstream of Cerl2 Asymmetry during Left-Right Development
At the LHF to early somite stages, the action of nodal flow represses Cerl2 in left-sided node crown cells, resulting in the R>L Cerl2 expression bias [8,40,41]. We therefore assessed Cerl2 asymmetry in Pkd1l1 tm1/tm1 mutants by WISH; we quantified in situ staining and then calculated the percentage of total stain that was present in right-sided crown cells. In order to control for observation bias, samples were processed together, scored for L-R expression and genotyped only after analysis. As expected, the percentage of Cerl2 stain was greater on the right side than the left side in control embryos measured at the 1-3 ss when Cerl2 asymmetry was obvious by inspection ( Fig 3A and 3C). In contrast, Cerl2 asymmetry failed to manifest in Pkd1l1 tm1/tm1 mutants, which instead exhibited a symmetrical pattern of Cerl2 expression (Fig 3B and 3C). Consistent with this, most Pkd1l1 tm1/tm1 mutants exhibited symmetrical crown cell Nodal expression whereas Nodal showed the expected subtle R<L bias in control embryos (Fig 3D-3F). Thus, functional Pkd1l1 is not required for crown cell Cerl2 or Nodal expression per se, but it is needed for asymmetries in their expression to manifest downstream of nodal flow.
Close examination of stage-matched Pkd1l1 tm1/tm1 and control embryos stained for Cerl2 under identical experimental conditions, revealed that the bilaterally symmetrical expression in mutants represents bilateral downregulation of Cerl2, rather than the left-side only downregulation observed in wild-type embryos (Fig 3A and 3B). All samples were scored before genotyping, in order to remove observer bias. We furthermore repeated this staining then genotyping procedure multiple times and obtained the same results from different litters, supporting the conclusions (S5 Fig). Thus, genetically, Pkd1l1 normally acts to maintain high levels of Cerl2. In contrast, Cerl2 levels remain bilaterally high (derepressed) in Pkd1l1 rks/rks and Pkd2 lrm4/lrm4 embryos [16]; these differences in crown cell Cerl2 levels between Pkd1l1 tm1/tm1 and Pkd1l1 rks/rks embryos would be expected given the distinct LPM Nodal cascade activities found in these mutants.
The data described so far suggest a genetic pathway linking nodal flow to the emergence of gene asymmetries within crown cells in which flow impacts Pkd1l1 which lies genetically upstream of Pkd2. Pkd2 activity, in turn, influences crown cell Cerl2 and Nodal expression and, ultimately, activation of the LPM Nodal cascade. In depth discussion of this genetic scheme and further interpretation of these results are included in the Discussion.

Pkd1l1 and Pkd2 Require Cilia to Function
Immotile cilia found on crown cells are required for the sensation of nodal flow [42]. Since PKD1L1 and PKD2 localize to nodal cilia [17,18], we asked whether cilia were required for Pkd1l1 and/or Pkd2 function in vivo. Loss of the anterograde intraflagellar transport motor component, KIF3A, results in a failure of ciliogenesis and bilateral induction of the LPM Nodal cascade ( Fig 3H) [34,43]. Both Pkd1l1 rks/rks ;Kif3a -/and Pkd2 lrm4/lrm4 ;Kif3a -/double mutants exhibited bilateral Pitx2 expression, the phenotype of Kif3a -/single mutants, whereas Pkd1l1 rks/rks and Pkd2 lrm4/lrm4 single mutants lacked LPM Pitx2 expression ( Fig 3H and S3  Table). The simplest explanation of this phenotype, that the Pkd1l1 rks/rks and Pkd2 lrm4/lrm4 phenotypes can be suppressed by loss of cilia, suggests that Pkd1l1 and Pkd2 require cilia to function in vivo. This genetic evidence is backed up by the findings that PKD2 and PKD1L1 localize to cilia [16][17][18].

The Pkd1l1 rks Mutation Destabilizes an Extracellular PKD Domain
We next addressed how the Pkd1l1 rks mutation impacts PKD1L1 protein function. Our genetic experiments comparing the targeted allele Pkd1l1 tm1 to Pkd1l1 rks had suggested that Pkd1l1 rks is a non-null allele (Fig 1F and 1M). The Pkd1l1 rks point mutation itself resides in the second of two polycystic kidney disease (PKD) domains within the extracellular N-terminal portion of the protein (Fig 1A) [16]. An NMR structure of the first PKD domain from human PKD1, a close homolog of PKD1L1, revealed the domain to be a ß-sandwich fold consisting of seven ßstrands ( Fig 4A) [44]. By extending our previous fold recognition modeling approaches [16], we have predicted the second PKD domain of mouse PKD1L1 to have a similar fold topology (Fig 4B). To validate our model, we performed structural studies on PKD1L1 by synchrotron radiation circular dichroism (SRCD) spectroscopy [45]. Conformational studies in solution of the purified second PKD domain from mouse PKD1L1 showed the typical appearance of a ßstrand protein (Fig 4D) consisting of 44% ß-strand and 56% disordered (linkers and loops between strands) by Raussens algorithm [46], in agreement with our molecular model.
We then assessed the impact of the Pkd1l1 rks point mutation upon the structure of PKD1L1's second PKD domain. Solution studies by SRCD of rks-mutated purified PKD domains shifted the SRCD spectra, which now exhibited characteristics of increased disorder (Fig 4D), with structural content changed to being 31% ß-strand and 69% disordered. Thermal denaturation studies on both wild-type and rks-mutated PKD domains showed the wild-type domain to have a higher melting temperature (Tm) of 68.6°C, compared to 56.4°C for the rksmutated domain (Fig 4E). Both of these lines of evidence support the notion that the Pkd1l1 rks point mutation destabilizes the extracellular PKD domain. Indeed, such a destabilization was also observed in our molecular models which predicted an increase in disorder reflected by a loss of secondary structure in the C' ß-strand in particular, marked by asterisks in Fig 4C (see [44] for PKD domain ß-strand nomenclature).
In Pkd1l1 rks/rks mutants, it is possible that the point mutation causes a drastic inability of the protein to properly fold or be localized to the correct cellular compartment. Whilst we cannot fully rule out this possibility, it seems unlikely since the phenotype of Pkd1l1 rks/rks mutants is very different to Pkd1l1 tm1/tm1 . Moreover, Pkd1l1 rks is dominant over Pkd1l1 tm1 whilst neither Pkd1l1 rks nor Pkd1l1 tm1 exhibit phenotypes in the heterozygous state. Thus, we favour a model in which the PKD domain structural destabilization we have observed results in abrogated functionality of that domain within the protein, thereby implicating the PKD domain in PKD1L1 function in L-R patterning. It is worth noting that similar point mutations in PKD domains of PKD1 also impact the stability and mechanical properties of PKD domains [47].

PKD1L1 Mediates a Response to Artificial Fluid Flow
After establishing that Pkd1l1 is necessary for L-R patterning, we then sought to test sufficiency of PKD1L1 in the response to flow in a context in which we can control fluid flow and assay the cells Ca 2+ signal response. We therefore addressed this issue by asking the question: can PKD1L1 function in the cellular response to shear stress induced by fluid flow?
In vascular endothelial cells and kidney epithelial cells, flow-induced shear stress elicits a Ca 2+ signaling response. This signaling event depends on primary cilia as well as the Polycystin protein PKD1; cells isolated from Pkd1 -/mice do not undergo flow-induced Ca 2+ signaling (FICS) [48,49]. We utilized this experimentally tractable system to ask how expression of PKD1L1 impacted FICS. We loaded ciliated endothelial cells with the Ca 2+ -binding dye Fur-a2-AM, applied fluid flow at 7.2 dyne/cm 2 , then recorded Ca 2+ response by ratiometric fluorescence imaging. In Fig 5A and 5B we show representative images from selected time-points, whilst in Fig 5C and 5D we show averaged traces from multiple fields of view for the changes in intracellular Ca 2+ after the onset of flow (arrows). Finally, the results of three independent transfections, with quantitation of 50 GFP+ and 50 GFP-cells per transfection, are collated in Fig 5E (see S5 Table for numerical results); statistical comparison among groups was performed using ANOVA followed by Tukey's posttest, with p<0.05 taken as statistically significant differences. In wild-type cells, a transient Ca 2+ response upon the onset of fluid flow was observed (Fig 5A, 5C and 5E). As anticipated, loss of PKD1 in cells generated from Pkd1 -/- mice abolished the FICS response (Fig 5B, 5D and 5E). Expression of a PKD1L1-GFP construct was able to restore FICS in Pkd1 -/cells and to increase the magnitude of the response in Pkd1 +/+ cells (Fig 5A-5E). Importantly, we did not see a restoration in either untransfected cells (GFP-negative) within the same sample or within separate samples that were transfected with GFP alone (Fig 5A-5E). These experiments demonstrate that PKD1L1 can mediate a Ca 2+ response to fluid flow. In contrast to wild-type PKD1L1-GFP, expression of PKD1L1 rks -GFP protein did not rescue FICS in Pkd1 -/cells (Fig 5A-5E). Thus, in our cell line system, PKD1L1 restores FICS in the absence of PKD1. In the node, Pkd1l1 acts downstream of flow and upstream of Pkd2 and Cerl2/Nodal asymmetries in the sensation pathway.

Proper Localization of PKD2 to Nodal Cilia Requires PKD1L1
Finally, we turned our attention to the localization of PKD2 protein in the node. In contexts where PKD1 and PKD2 function together, PKD2 localization to cilia has been shown to depend on the presence of PKD1 [48]; this perhaps explains why PKD2 is not functional in the absence of PKD1 [50]. By contrast, we find bilateral Nodal expression in Pkd1l1 tm1/tm1 mutants, implying bilateral PKD2 activity in this Pkd1l1 mutant background. Two possibilities that could explain this apparent contradiction are: (1) PKD2 localization to cilia is not required for its role in L-R patterning; or (2) PKD2 can still localize to cilia and signal in the Pkd1l1 tm1/tm1 mutant. We tested these by assessing PKD2 subcellular localization within the node of wildtype and mutant embryos at 8.0 dpc.
In wild-type control embryos, PKD2 localized to all nodal cilia (Fig 6A and 6B), as previously reported [18]. Most cilia showed continuous PKD2 staining along the ciliary axoneme, marked by acetylated α-tubulin, or greater than five PKD2 puncta per cilium (Fig 6A and 6B). By contrast, PKD2 was absent from the vast majority of cilia in Pkd2 lrm4/lrm4 mutants (Fig 6A  and 6E), in agreement with the finding that the lrm4 mutation prevents a PKD2-Venus fusion from localizing to nodal cilia in transient transgenic mouse embryos [51]. The lrm4 point mutation changes a residue in the first extracellular loop of PKD2 (Fig 1A) [32], but does not affect its channel activity [51]. It is currently not known why PKD2 lrm4 protein fails to properly localize to nodal cilia. Nevertheless, given the striking L-R defects of Pkd2 lrm4/lrm4 mutants, these data suggest a requirement for PKD2 protein localization to nodal cilia; without PKD2 in cilia, the downstream Nodal cascade is not activated on either the left or right sides, resulting in right isomerism in mutants.
In Pkd1l1 tm1/tm1 mutants, we observed reduced PKD2 ciliary levels compared to controls (Fig 6A and 6C), showing that proper levels of functional PKD1L1 are needed for the efficient import or retention of PKD2 in nodal cilia. Pkd1l1 +/tm1 heterozygous embryos exhibited reduced ciliary PKD2 (Fig 6A), suggesting that PKD2 levels within nodal cilia are controlled, at least in part, by the dosage of PKD1L1. Thus, whilst we do observe reduced PKD2 in the nodal cilia of Pkd1l1 tm1/tm1 mutants, this is not a complete loss of PKD2 as has been observed in distinct cell types when Pkd1 is ablated [48]. An assessment of PKD1L1 localization and levels in these mutant backgrounds will require functional PKD1L1 antibodies, which are currently unavailable. Finally, we also assessed the PKD2 localization to nodal cilia in Pkd1l1 rks/rks mutants and found that it was also reduced (albeit to a lesser extent than in Pkd1l1 tm1/tm1 ) in Pkd1l1 rks/rks mutants (Fig 6A and 6D). Together, these data suggest that PKD2 localization to cilia does not depend entirely on functional PKD1L1.

Discussion
Addressing how nodal flow is sensed by the embryo to elicit downstream asymmetries in gene expression is critical if we are to understand the early phases of L-R patterning. We, and others, have previously implicated the Polycystin proteins PKD1L1 and PKD2 in the sensation of nodal flow. Here, we reveal that Pkd1l1 acts in a genetic pathway at the level of flow sensation, placing Pkd1l1 function between flow and Pkd2. Moreover, we demonstrate that PKD1L1 protein is sufficient to elicit a Ca 2+ signal in response to artificial fluid flow.

A Genetic Repression Model of Nodal Flow Sensation
Our genetic experiments suggest a pathway in which flow acts upstream of Pkd1l1 which, via Pkd2 (that lies downstream on the pathway), signals to affect crown cell Cerl2 and Nodal levels. The result of this pathway, initiated by nodal flow, is to increase left-sided crown cell NODAL signalling and thereby activate the Nodal pathway in the left LPM only.
We can further extend our genetic results to interpret the nature of the interactions between adjacent genes on the pathway (summarized in Fig 7). Since the Pkd1l1 tm1/tm1 is likely to be hypomorphic or null (S1 Fig) i.e. loss-of-function, and Pkd1l1 tm1/tm1 mutants exhibit bilateral LPM Nodal cascade activity, we suggest that Pkd1l1 normally represses Nodal cascade activation and that this repression must be relieved by nodal flow on the left side only. Moreover, since Pkd1l1 tm1/tm1 and Pkd2 lrm4/lrm4 mutants have opposite crown cell Cerl2 and Nodal phenotypes as well as distinct LPM Nodal phenotypes, and that the Pkd2 lrm4/lrm4 phenotype manifests in double mutants, we suggest that, genetically, Pkd1l1 acts as an upstream repressor of Pkd2. Finally, Pkd2 is known to act as a genetic repressor of Cerl2 which itself is an antagonist of the Nodal pathway. This model, described pictorially in Fig 7A and 7B, coheres with the genetic tests we have performed in this study. It is worth noting that this interpretation of the data relies on Pkd1l1 tm1 being a hypomorphic or null allele, something which is very likely (S1 Fig), but owing to the absence of functional PKD1L1 antibodies, not something we have formally tested at the protein level. It is formally possible that a cryptic initiation codon downstream of the deleted exons in Pkd1l1 tm1 results in a portion of PKD1L1 being expressed in mutants, and our repression model must be considered with this caveat in mind. Nevertheless, the model predicts bilateral Nodal cascade activity in the absence of Pkd1l1 (Fig 7C), something we do see in Pkd1l1 tm1/tm1 mutants, as well as absent Nodal cascade activity in loss-of-function Pkd2 mutants like Pkd2 lrm4/lrm4 (Fig 7E), again in agreement with experiment. Moreover, it is worth emphasizing that this is a genetic model and that our study leaves open several possibilities regarding the nature of the molecular interactions between adjacent members of this genetic cascade. The genetic relationships uncovered here could manifest via direct molecular repression of PKD2 by PKD1L1 or more indirectly by the control of PKD1L1 and/or PKD2 localization.
Our multi-repression-based genetic model ascribes two functions to Pkd1l1 in L-R patterning: (1) to mediate a flow-based signal (or, genetically, to receive a signal from Dnahc11, mutants of which we use in our genetic experiments to eliminate flow) and (2) to genetically repress Pkd2 activity. In the absence of flow, Pkd1l1 represses Pkd2 in the node. The onset of flow (Dnahc11 function) relieves this repression, causing Pkd2 to activate and thereby initiates a cascade that results in left-sided Nodal signals. In the absence of Pkd1l1, repression of Pkd2 is relieved and signals are activated bilaterally independently of flow (since the Dnahc11 iv mutation does not manifest in the presence of Pkd1l1 rks ). By contrast, since Pkd2 is required for Cerl2 repression and Nodal activation, loss of Pkd2 function results in bilateral absence of the Nodal cascade. The repression model provides an explanation for the previously published Pkd1l1 rks mutant phenotype since the Pkd1l1 rks mutation seems to uncouple the two functions of Pkd1l1. The Pkd1l1 rks mutation structurally disrupts an extracellular PKD domain, resulting in PKD1L1 rks protein being unable to elicit Ca 2+ signals downstream of artificial flow. Moreover, NODAL activity is bilaterally repressed in Pkd1l1 rks/rks mutants, implying that Pkd2 is inactive regardless of the presence of normal nodal flow in these mutants. This is consistent with the idea that the Pkd1l1 rks mutation is loss-of-function with respect to role (1), flow sensation, but, owing to its insensitivity to flow, it appears gain-of-function with respect to role (2), Pkd2 inhibition (i.e. Pkd2 remains inhibited regardless of the presence of the upstream activating signal, nodal flow) (Fig 7D).
Furthermore, the repression model coheres with the subtle phenotypic differences between Pkd1l1 tm1/tm1 and Dnah11 iv/iv mutants. Though these two mutants exhibit overtly similar phenotypes, with high levels of left lung isomerism and bilateral Nodal cascade activity, it is noteworthy that these outcomes result from distinct mechanisms. In Dnah11 iv/iv mutants, loss of flow results in stochastic (unbiased) Nodal pathway activation; crucially, initiation of the Nodal cascade in the LPM occurs with a delay of around 2 hours in the absence of flow. This is consistent with the repression model, since the pathway to activate Nodal robustly requires repression of Pkd1l1 by flow; this would not occur in a timely fashion in the absence of flow. In stark contrast, the Pkd1l1 tm1 allele leads to bilateral LPM Nodal pathway activity without delay, interpreted in our model as loss of Pkd1l1 leading to the fast derepression of Nodal. Why the Nodal cascade activates at all in the absence of flow is an intriguing question and suggests that repression of Pkd1l1 by flow is not absolutely required for Nodal activation but, rather, the repression is required for biased expression to manifest at the correct time. If this pathway is not activated in this way owing to lack of flow, stochastic changes in fluid movement, or other variations, followed by positive feedback loops operating at the node and LPM [14,40] likely result in randomized or bilateral Nodal activity.

Pkd1l1, Pkd2, and Cilia
Our experiments, in which we generated Pkd1l1 rks/rks ;Kif3a -/and Pkd2 lrm4/lrm4 ;Kif3a -/double mutants and found that they exhibited the same LPM Nodal pathway activity as Kif3a -/mutants, show that cilia (or, strictly, Kif3a) are required for the manifestation of the Pkd1l1 and Pkd2 mutant phenotypes. A simple explanation of this finding is that PKD1L1 and PKD2 require cilia to function, in agreement with their sub-cellular localization to cilia [16-18, 42, 52]. However, this relatively complicated genetic interaction warrants further discussion. In Kif3a -/mutants, the vast majority of nodal cilia are lost [34,43] and the Nodal cascade is bilaterally activated. This suggests that cilia provide an inhibitory signal preventing LPM Nodal cascade activation. In contrast, loss of Pkd2 results in the absence of LPM Nodal, implying that Pkd2 acts to overcome this inhibition. In Pkd2 lrm4/lrm4 ;Kif3a -/double mutants, the loss of ciliamediated inhibition results in Nodal being activated (now bilaterally) independently of Pkd2. Therefore, loss of that cilia-based inhibition (in Kif3a -/-) suppresses loss of Pkd2. However, since Pkd1l1 tm1 does not suppress Pkd2 lrm4 , it seems likely that the cilia-based inhibitory signal is not provided by Pkd1l1. These data speak to an apparent paradox in the field; since PKD2 acts to activate LPM Nodal and is thought to act in cilia, how can loss of cilia result in bilateral Nodal activity? One possibility is that Hedgehog signaling, which acts in the LPM bilaterally upstream of Nodal activation [53], could be augmented sufficiently to affect Nodal expression upon loss of cilia. Moreover, loss of Kif3a impacts both flow generation and sensation in the node [5,18] and also affects midline structures resulting in barrier defects. Thus, the inhibitory signal (in terms of Nodal activation) provided by cilia independently of Pkd1l1/Pkd2 function could be acting at the node, the LPM, the midline, or in a combination of these locations. However, the most important contribution likely comes from crown cell cilia since embryos expressing Kif3a solely in crown cells express Nodal exclusively in the left LPM following application of leftward flow [42]. This result coheres with our interpretation of the Pkd1l1 rks/rks ;Kif3a -/and Pkd2 lrm4/lrm4 ;Kif3a -/double mutant phenotypes because the ciliabased repressive signal, that cannot be overcome by mutation of Pkd1l1 or Pkd2, appears to be acting most dominantly within the sensory cilia of the crown cells.
The finding that loss of cilia masks the Pkd2 lrm4 mutant phenotype reinforces the idea that PKD2 must localize to nodal cilia to activate downstream Nodal signals [51], (and this manuscript). Others have demonstrated a requirement for PKD1 or PKD1L1 in the localization of PKD2 to cilia [16,17,48]. If PKD2 localization to cilia were entirely dependent on wild-type PKD1L1, we might expect no PKD2-mediated Nodal cascade activation in Pkd1l1 tm1/tm1 mutants, the opposite of the bilateral Nodal activity that we do see. However, our results show this clearly not to be the case; some PKD2 still localizes to nodal cilia in Pkd1l1 tm1/tm1 mutants. Although there is less PKD2 in the nodal cilia of Pkd1l1 tm1/tm1 mutants compared to wild-type, interpreting this result in light of our genetic model suggests this lower amount of PKD2 to be active in the absence of PKD1L1. These findings further suggest that PKD2 cannot respond to nodal flow in the Pkd1l1 tm1/tm1 mutant since in these mutants nodal flow is normal and some PKD2 localizes to cilia; however, we see no evidence for consistent lateralized downstream Nodal cascade activity, implying that no flow sensation is occurring.

PKD1L1 and the Response to Nodal Flow
Whether PKD1L1 is part of a pathway mediating sensation of an asymmetrically distributed chemical determinant positioned by nodal flow (chemosensation), or the force of flow itself (mechanosensation), remains a topic of debate [16,22,42,54]. Here, we demonstrate that, in a tissue culture system where the onset of flow and the presence/absence of PKD1L1 is controlled, PKD1L1 cell autonomously mediates Ca 2+ signaling downstream of fluid flow. Thus, our work favours a mechanosensation model of nodal flow sensation.
Left-biased Ca 2+ signals have been observed around the node and thus correlate with leftsided Nodal activation [18,36,55]; these signals depend on both flow and PKD2. Thus, nonfunctional Pkd2 alleles, such as Pkd2 lrm4 , should abolish this Ca 2+ signal and lead to loss of Nodal induction, a result we find in Pkd2 lrm4/lrm4 mutants. In Pkd1l1 tm1/tm1 mutants, we observe bilateral Nodal, suggesting bilateral Ca 2+ signaling. Extrapolation of our genetic model, in which Pkd1l1 tm1/tm1 represents loss of Pkd1l1, into a simple molecular model can explain this: on the left side of the node, PKD2 repression by PKD1L1 is relieved by flow, resulting in the opening of PKD2 channels and activation of the Nodal cascade. In our tissue culture system, we do not see Ca 2+ signals upon loss of the cells endogenous flow sensor, PKD1. Importantly, upon expression of PKD1L1 in this system, the cells again become competent to respond to flow, demonstrating that PKD1L1 can mediate flow sensation. PKD2 activity in Pkd1l1 tm1/tm1 mutant embryos could not be recapitulated in the cell line we used because PKD1, not PKD1L1, is the endogenous flow sensor. Thus, the two systems provide complementary information, but are not directly analogous. Of note is that either PKD1 or PKD1L1 can function independently to mediate flow sensation in our tissue culture flow chamber system. The embryonic system provides us with a genetic model that can explain the mouse mutant phenotypes, whereas the tissue culture experiments reveal PKD1L1 protein to be able to mediate a cellular response to flow sensation.
With this in mind, though our data supports the mechanosensation model, this study does not rule out the importance of asymmetrically distributed molecules in the node cavity. Modeling studies demonstrate the viability of the concept of L-R asymmetric morphogen gradients within the node [56]. Nodal vesicular parcels (NVPs), large membrane-encased vesicles, have been described transiting the node and opening on the left side [23] and these might asymmetrically deliver a left-side determinant. Indeed, recently CERL2 protein has been demonstrated to transition from right-biased to left-biased in response to nodal flow; the timing of this event is notable in that left-biased localization is not detected until several hours after the establishment of L-R asymmetry at the node, when it functions to shut down NODAL activity and thereby lock in asymmetry [57]. Future study will determine whether flow-driven asymmetric distribution of molecules functions solely in this later event.
Nevertheless, other considerations favour the mechanosensation hypothesis. Firstly, the fact that the Pkd1l1 rks point mutation impacts one of two extracellular PKD domains implicates this domain in PKD1L1 function during L-R patterning. Interestingly, PKD domains are required in other contexts where mechanosensing occurs. For example, PKD1, which harbors fifteen PKD domains, is argued to act as a mechanosensor of fluid flow in the kidney [58]. However, this alone does not formally rule out a role for the PKD domain, or PKD1L1 in general, as a binding site for an unknown chemical determinant. Indeed, the Polycystin-1 family member PKD1L3 acts as a chemical receptor needed for sour taste perception [59], though it is noteworthy that PKD1L3 does not contain PKD domains.
Secondly, PKD domains exhibit remarkable mechanical strength along the N-to-C terminal axis [60,61]. Some mutations in human PKD1 that cause ADPKD have been found within the PKD domains [47,62]. An assessment of domain strength within these disease variants by atomic force microscopy found that many human disease associated point mutations caused a weakening of the PKD domains, suggesting that domain strength is critical for function [47]. Here, we found by molecular modeling, SRCD spectroscopy, and melting temperature analysis, that the Pkd1l1 rks point mutation structurally destabilizes the second PKD domain of PKD1L1. We thus hypothesize that such a destabilization leads to reduced domain strength and, therefore, the inability of PKD1L1 rks to transduce the force of nodal flow. This is corroborated by our finding that unlike PKD1L1, PKD1L1 rks protein does not mediate mechanosensation in response to fluid flow.
PKD1L1-PKD2L1 complexes in cilia of neural derived cells are known to facilitate a constitutive ciliary Ca 2+ flux [52,63], while a pressure clamp failed to induce signaling through PKD1L1-PKD2L1 complexes at physiological force levels. Whilst this somewhat contrasts with our data, it is worth noting that different cell types and a different PKD protein pair are examined in each study. Moreover, the nature of flow/force sensation by cilia will not necessarily translate to simple pressure upon the channel; cilia become deformed by motion within a flow leading to stresses along as well as perpendicular to the membrane. It is unclear to what extent all such forces will be modeled by pressure clamping.
The mechanistic details of how PKD1L1 mediates flow sensation in the node will require a multi-disciplinary approach combining structural biology, computational approaches, and genetics. Our finding that PKD1L1 and PKD2 function during L-R patterning requires primary cilia is intriguing; it has been argued that the bending of the ciliary membrane as a result of flow might be critical for mechanosensation by Polycystin proteins [48]. One possibility is that the N-terminus of PKD1L1 is tethered to the ciliary membrane so that PKD domains lie parallel to the membrane. Thus, when the membrane is bent by flow, the PKD domains would be placed under mechanical force along their N-to C-terminal axis, where they exhibit mechanical strength; this might aid in force transduction.
In summary, we reveal a genetic pathway for the early phases of L-R patterning which encompasses the generation of nodal flow, its' sensation, and subsequent downstream changes in gene expression. Moreover, in a tissue culture system we demonstrate that PKD1L1 can mediate the sensation of fluid flow to elicit Ca 2+ signals. Finally, we implicate a PKD domain of PKD1L1 in the response to flow and L-R patterning.

Ethics Statement
All experiments were performed under the guidelines and approval of the Home Office, UK and the MRC Harwell, UK Ethics Committee; Euthanasia was by cervical dislocation of adults and decapitation of embryos.

Mice
The mouse lines used in this study are: Pkd1l1 rks ; Pkd1l1 tm1Lex (Pkd1l1 tm1 ); Pkd2 lrm4 ; Dnah11 iv ; Kif3a tm1Gsn (Kif3a -); Pkd1 tm1Jzh (Pkd1 -). All experiments were performed under the guidelines and approval of the Home Office, UK; Euthanasia was by cervical dislocation of adults and decapitation of embryos. Pkd1l1 rks , Pkd2 lrm4 , and Dnah11 iv lines were congenic on C3H/HeH, while Pkd1l1 tm1 and Kif3awere only used for experiments after at least four and two backcrosses to C3H/HeH, respectively.

In Situ Hybridization
Whole mount in situ hybridization (WISH) was performed using digoxygenin-labeled riboprobes using standard procedures. Quantitation of crown cell gene expression was performed using ImageJ, NIH.

Analysis of Embryonic Nodes
To visualize nodal cilia rotation, embryos were mounted on slides with the node facing up. Differential interference contrast (DIC) microscopy was then performed using a Leica DM2500 compound microscope with a monochrome high-speed Hamamatsu camera attached. Quantitation was performed by counting cilia rotations from movies taken at approximately 100 frames per second for at least 10 cilia per embryo and at least 3 embryos per genotype. For detection of nodal flow, fluorescent beads (0.2 μm; Invitrogen F-8848) were diluted 1:10 with dissection medium supplemented with 2% fetal bovine serum and then placed over mounted embryos. The node was imaged using a Zeiss Axio Observer Z1 microscope with a VivaTome extension to allow for high-speed optical sectioning. Particle Image Velocimetry (PIV) analysis was performed as previously described [37].

Quantitative Reverse Transcription PCR (qRT-PCR)
RNA was extracted from single embryos using a Microplus kit (Qiagen) according to manufacturers' instructions. 500 ng of RNA was used for cDNA synthesis using a SuperScriptIII First-Strand Synthesis SuperMix for qRT-PCR kit (Invitrogen). qPCR was performed in triplicate on a 7900 Fast Machine (Life Technologies) using Fast SYBR Green Mastermix (Life Technologies) with 20 ng cDNA and 500 nM forward and reverse primers in a final 20 μl reaction volume (see S4 Table for primer sequences). Quantitation was relative to Hprt and fold changes were calculated using the ΔΔC T method (7500 Software v2.0.6., Life Technologies).

Protein Expression and Purification
Both wild-type and rks-mutated versions of PKD domain 2 from mouse PKD1L1 were tagged at the N-terminus with His, expressed in E. coli, purified using a His column, refolded at decreasing urea concentration by affinity chromatography then vacufuged and dialysed. Protein concentration was calculated based on the measured ultraviolet (UV) absorbance (at 1 mm path length) at 280 nm on an Implen NanoPhotometer.

In Vitro Flow-Induced Ca 2+ Signaling (FICS) Assessment
Vascular endothelial cells were extracted from wild-type or Pkd1 -/embryos as previously described [49]. Cells were cultured in permissive conditions (in the presence of interferon-γ at 33°C) to induce proliferation then transfected with 1 μg/mL pEGFP-N1 plasmid (Clontech Laboratories) containing EGFP only, PKD1L1-EGFP, or PKD1L1 rks -EGFP. To promote cell differentiation and cilia growth, cells were grown under non-permissive conditions (in the absence of interferon-γ at 39°C) to induce differentiation.
For FICS experiments we used previously described protocols [64]. Briefly, cells were loaded with 5 μM Fura2-AM (Invitrogen) for 30 minutes at 39°C. Basal fluorescence was measured for one minute before the onset of fluid flow using a Nikon TE2000 microscope with Metafluor software. A shear stress of 7.2 dyne/cm 2 was achieved in an FCS2 chamber with electrical enclosure heater (Biotechs, Inc.) at a flow rate of 550 μL/sec. Flow was then applied and fluorescence was monitored every 4 seconds. At the end of the experiment, minimum fluorescence measurements were obtained by treating cells with 2 mM EGTA and 10 μM ionomycin. After achieving the minimum signal, the maximum fluorescence was obtained by treating cells with excess calcium (10 mM). All fluorescence measurements were corrected for auto-fluorescence.
All quantifiable experimental values are expressed as mean +/-standard error of the mean (SEM), with values of p<0.05 being considered statistically significant. Data analysis was performed using SigmaPlot software Version 11 and comparisons among groups were done using ANOVA followed with Tukey's posttest.

Synchrotron Radiation Circular Dichroism (SRCD) Spectroscopy
Far UV SRCD experiments were performed using Module B of Beamline B23 at the Diamond Light Source (Didcot, Oxfordshire UK). Measurements of protein solutions were carried out in 20 mM Tris and 50 mM NaCl, pH8.5, at a concentration of 0.145 mg/ml (wild-type PKD domain) or 0.104 mg/ml (rks-mutated PKD domain). Four scans were taken at 20°C (1 nm increment, 1 s integration time and a scan rate of 38 nm/min) and averaged. Spectra of buffer alone were subtracted from the sample spectra. Secondary structure composition was calculated from the experimental spectra using Raussens algorithm [46].
Supporting Information S1 Fig. Characterisation of the Pkd1l1 tm1 allele. The Pkd1l1 tm1 allele (Pkd1l1 tm1Lex ; [33]) comprises a beta-galactosidase-neomycin fusion gene (beta-geo) inserted in place of exons 3, 4 and 5 (labelled 'LacZ' in the schematic given in D). Importantly, this insertion contains a stop codon and polyA signal at its 3' end. The data in this figure demonstrate that in Pkd1l1 tm1/tm1 mutants, the gene is disrupted either by splicing onto the beta-geo cassette or by splicing around the cassette in a fashion that introduces a premature stop codon which truncates the protein very early, suggesting that Pkd1l1 tm1 is a null or strong hypomorph. (AA'-AG') Expression of Pkd1l1 was assessed by LacZ staining, revealing that expression from the beta-geo reporter locus mirrors the endogenous expression pattern [16]. Pkd1l1 +/tm1 embryos that were phenotypically normal; 7.5 dpc (AA'-AB'), 8.5 dpc (AC'-AE'), and 9.5 dpc (AF'-AG') embryos were assessed. (B) WISH analysis of Pkd1l1 expression in Pkd1l1 +/tm1 and Pkd1l1 tm1/tm1 embryos at 8.5 dpc. If all transcripts splice into beta-geo, then we would predict there to be no mRNA present for 3' portions of Pkd1l1. However, WISH revealed equivalent expression patterns in both wild-type and mutant embryos (assessed with a probe covering exons 20-24); expression in the Pkd1l1 tm1/tm1 embryos appears slightly reduced. (C) A proportion of Pkd1l1 tm1 transcripts splice from exon 2 to exon 6, skipping the beta-geo insertion. It is documented that a proportion of gene trap alleles produce novel splice products that 'jump over' the gene trap. We therefore investigated whether the message detected by WISH (B) might result from such a splicing event around the targeted insertion. cDNA was prepared from wild-type, Pkd1l1 +/tm1 and Pkd1l1 tm1/tm1 8.5 dpc embryos. PCR primers in exons 1 (5'-TTGGCAGGTGCAACTACTGT-3') and 6 (5'-CCCATGTTCTT-CACTGGGGG-3') were used to amplify the intervening region. This resulted in a band of the predicted size (~800bp) in wild-type and Pkd1l1 +/tm1 samples and a smaller band (~350 bp) in Pkd1l1 +/tm1 and Pkd1l1 tm1/tm1 samples (gel on right; t refers to tm1). The wild-type band was missing from the Pkd1l1 tm1/tm1 samples. The smaller band is of the size predicted for a splicing event between exons 2 and 6, as confirmed by Sanger sequencing. The resulting message in Pkd1l1 tm1/tm1 is out of frame such that any resulting protein would truncate within 18 amino acids of the exon 2-exon 6 splicing event. This would lead to a very small product, lacking all characterised protein domains including having no transmembrane domains. Thus, in Pkd1l1 tm1/tm1 embryos, the second exon of Pkd1l1 either splices onto the beta-geo cassette (which harbours a polyA and stop codon) or splices from exon 2-exon 6 producing an out of frame transcript which contains a stop codon after 18 amino acids. (D) Quantitative analysis of Pkd1l1 transcripts reveals that the level of transcript detected in 3' potions of the locus in Pkd1l1 tm1/tm1 is equivalent to the level of transcript that splices exons 2-6 in Pkd1l1 tm1/tm1 embryos. Only single long Pkd1l1 Havana-curated transcripts exist in both mouse and humans. To test whether additional start sites might exist we utilised quantitative reverse transcription PCR (qRT-PCR) to measure the expression levels of different regions of the transcript. The following assays were used: • Exon 1-2 assesses expression from the known start site of the locus. Surprisingly, this revealed a 2-fold upregulation of Pkd1l1 in Pkd1l1 tm1/tm1 embryos, suggesting that a negative feedback loop controls Pkd1l1 expression.
• Exon 2-3 and Exon 5-6 assess wild-type expression. As exon 3, 4 and 5 are absent from the Pkd1l1 tm1/tm1 allele, as expected, no expression in this region of the transcript is evident in Pkd1l1 tm1/tm1 mutants.
• Exon 2-LacZ assesses the splicing from exon 2 into the beta-geo insertion; the predicted splice product. As predicted, this product is present in both Pkd1l1 +/tm1 and Pkd1l1 tm1/tm1 and is approximately twice as highly expressed in the mutant relative to the heterozygous state.
• Exon 2-6 assesses splicing around the beta-geo insertion, which is predominant in Pkd1l1 tm1/tm1 and Pkd1l1 +/tm1 samples. We also observed a very small amount of exon 2-6 splicing in wild-type samples in this assay. However, additional analysis suggests this low level of expression to be an artefact of the qPCR assay and not a biologically meaningful splice variant.
• Exon 21-22 assesses expression of a 3' region of the locus. This product overlaps the WISH probe used above (B).
(E) The relative levels of the exon 2-6 versus exon 2-LacZ were calculated, revealing that exon 2-6 splicing occurs at~35% of the level for exon 2-LacZ splicing in both the Pkd1l1 +/tm1 and Pkd1l1 tm1/tm1 embryos. In combination with the doubled level of transcription of the Pkd1l1 locus in Pkd1l1 tm1/tm1 embryos, this explains the level of transcript that we detect by exon 21-22 qRT-PCR. For all experiments, error bars show the RQ min and RQ max when confidence levels are set at 95%. (DOCX) S2 Fig. Pkd1l1 tm1/tm1 mutants exhibit variable times of death and gross heart and stomach situs defects that are similar to Pkd1l1 rks/rks and Dnah11 iv/iv mutants. (A-B) Charts showing the observed (Obs) and expected (Exp) frequencies of Pkd1l1 genotype for embryos dissected at E13.5 (A) or recovered as surviving adults (B). There is a statistically significant loss of Pkd1l1 tm1/tm1 mutants at these time points (chi-square test applied). When dissected at E13.5, 32% of Pkd1l1 tm1/tm1 (n = 13/41) had already arrested in utero (at various times between E9.5-E12.5). Approximately 35% of the expected number of homozygotes survived until adulthood. (C-D) Examples of reversed heart (H) and stomach (S) laterality in Pkd1l1 tm1/tm1 embryos (D) compared to a control (C) at E13.5. Normally, the heart apex and stomach are positioned to the left of the body cavity, but this is reversed in a proportion of Pkd1l1 tm1/tm1 mutants. R-L refers to right-left. (E) Heart and stomach laterality scored at E13.5 for Pkd1l1 tm1/tm1 , Dnah11 iv/iv and Pkd1l1 rks/rks mutants. The percentage of embryos showing each phenotype and the total number of embryos examined is given. Cerl2 expression at the node of Pkd1l1 tm1/tm1 mutants and control embryos at the 1-3 somite stage from three separate litters. In each case, expression is more symmetrical and expression levels are lower in Pkd1l1 tm1/tm1 embryos than in controls. Embryos from the same litter were treated identically throughout the procedure and scoring of expression was performed prior to genotyping. (DOCX) S1