Domain Mapping of the Polycystin-2 C-terminal Tail Using de Novo Molecular Modeling and Biophysical Analysis*

In polycystic kidney disease (PKD), polycystin-2 (PC2) is frequently mutated or truncated in the C-terminal cytoplasmic tail (PC2-C). The currently accepted model of PC2-C consists of an EF-hand motif overlapping with a short coiled coil; however, this model fails to explain the mechanisms by which PC2 truncations C-terminal to this region lead to PKD. Moreover, direct PC2 binding to inositol 1,4,5-trisphosphate receptor, KIF3A, and TRPC1 requires residues in PC2-C outside this region. To address these discrepancies and investigate the role of PC2-C in PC2 function, we performed de novo molecular modeling and biophysical analysis. De novo molecular modeling of PC2-C using the ROBETTA server predicts two domains as follows: an EF-hand motif (PC2-EF) connected by a linker to a previously unidentified C-terminal coiled coil (PC2-CC). This model differs substantially from the current model and correlates with limited proteolysis, matrix-assisted laser desorption/ionization mass spectroscopy, N-terminal sequencing, and improved coiled coil prediction algorithms. PC2-C is elongated and oligomerizes through PC2-CC, as measured by analytical ultracentrifugation and size exclusion chromatography, whereas PC2-EF is globular and monomeric. We show that PC2-C and PC2-EF have micromolar affinity for calcium (Ca2+) by isothermal titration calorimetry and undergo Ca2+-induced conformational changes by circular dichroism. Mutation of predicted EF-hand loop residues in PC2 to alanine abolishes Ca2+ binding. Our results suggest that PC2-CC is involved in PC2 oligomerization, and PC2-EF is a Ca2+-sensitive switch. PKD-associated PC2 mutations are located in regions that may disrupt these functions, providing structural insight into how PC2 mutations lead to disease.

Polycystic kidney disease (PKD) 4 is among the most common life-threatening inherited disorders, with clinical consequences characterized by renal and hepatic cysts (1). Most cases of PKD (Ͼ95%) are linked with mutations in the genes Pkd1 or Pkd2, which encode the membrane protein polycystin-1 (PC1) and the calcium (Ca 2ϩ )-permeable channel, polycystin-2 (PC2), respectively (2,3). PC2 belongs to the TRP channel family and is expressed in most tissues (4 -8). It has six transmembrane spans, and both C and N termini are cytoplasmic. PC1 and PC2 interact directly and co-localize to primary cilia where they are hypothesized to be necessary for a mechanosensory or chemosensory response that triggers a rise in intracellular Ca 2ϩ (9). Many aspects of PC2 function are mediated by the cytoplasmic C-terminal tail (PC2-C), including co-assembly with PC1 through the PC1 C-terminal cytoplasmic tail (10 -13). Twothirds of pathogenic mutations in PC1 and Ͼ90% of pathogenic mutations in PC2 result in truncations of their C-terminal cytoplasmic regions and are predicted to abrogate interaction between the proteins. There are no structural data describing PC2, and current descriptions of PC2-C lack functional and biochemical information.
The currently accepted domain model of PC2-C consists of an EF-hand motif overlapping with a short coiled coil (6,12). Helical wheel projection of this purported coiled coil (Glu 772 -Leu 796 ) shows that residues in positions "a" and "d" of the helix (i.e. the coiled coil interface) are hydrophilic or charged and would be unfavorable in a coiled coil interface (Fig. 1D). PKD-associated truncations have been identified outside of these proposed domains in PC2-C (Fig. 1F), suggesting that residues important for PC2 function are located in the truncated regions. Moreover, PC2 binds directly to several proteins, including the inositol 1,4,5trisphosphate receptor (14), TRPC1 (15), tropomyosin-1 (16), Id2 (17), troponin-1 (18), KIF3A (19), and PC1 (9,20). All of these interactions are dependent on residues C-terminal to the currently accepted domain organization of PC2-C. To address these discrepancies and to aid future studies into the molecular basis of PKD pathogenesis and progression, we constructed a validated structural model of PC2-C.
We have determined a de novo molecular model of PC2-C using the ROBETTA server (21)(22)(23) and validated characteris-* This work was supported, in whole or in part, by National Institutes of Health Grants DK57328 (to B. E. E.), CA009085 (training grant to E. T. P.), RR022200 (UltraScan software development), and CA054174 (to the Center for Analytical Ultracentrifugation of Macromolecular Assemblies facility). This work was also supported in part by a grant from the PKD Foundation, UltraScan software development was also supported by the Robert J. Kleberg Jr., and Helen C. Kleberg Foundation, and AUC calculations were supported by National Science Foundation Grant TGMCB070038 (to B. D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Figs. 1-5 and coordinates for the ROBETTA models and the theoretical model PC2-C. 1 Both authors contributed equally to this work. 2 To whom correspondence may be addressed. tics of this model using biophysical and biochemical analysis. We show that PC2-C contains two domains, a single EF-hand motif (PC2-EF) connected by a linker to a coiled coil domain (PC2-CC). We propose that PC2-CC is the actual coiled coil domain that has been assumed in the literature. Our results suggest that PC2-CC is involved in PC2 oligomerization and that PC2-EF acts as a Ca 2ϩ -sensitive switch. PKD-associated PC2 truncation mutations are located in regions that would disrupt these functions, providing insight into how PC2 mutations may lead to disease.
Limited Proteolysis Domain Mapping-PC2-C was treated with trypsin (1:100 w/w) or proteinase K (1:10,000 w/w) in 25 mM Tris, 250 mM NaCl, pH 8.0. Reactions were quenched using proteinase inhibitor mixture (Roche Applied Science). Products were resolved on SDS-PAGE, transferred to polyvinylidene difluoride, and bands sent for N-terminal sequencing (Tufts University Core Facility). Identical reaction products were sent for MALDI-MS (Tufts University Core Facility). MALDI-MS results were analyzed using FindPept (25).
Analytical Ultracentrifugation-Sedimentation equilibrium experiments were performed with a Beckman Optima XL-I at the Center for Analytical Ultracentrifugation of Macromolecular Assemblies (University of Texas Health Science Center, San Antonio, Dept. of Biochemistry). AUC experiments were analyzed with UltraScan version 9.5 (29). Hydrodynamic corrections were made according Laue et al. (30), as implemented in UltraScan. The partial specific volumes of PC2-EF, PC2-CC, and PC2-C were estimated according to Cohn and Edsall (31), as implemented in UltraScan yielding 0.722, 0.732, and 0.717 cm 3 /g, respectively. Equilibrium data were fit to multiple models, and the best chosen based on fitting statistics and visual inspection of the residual run patterns. 2DSA and Monte Carlo analyses were calculated on a Linux Beowulf cluster at the Bioinformatics Core Facility, Department of Biochemistry, University of Texas, Health Science Center, San Antonio, and on Lonestar/Teragrid (Texas Advanced Computing Center, University of Texas, Austin). Samples were analyzed in a buffer containing 300 mM NaCl. Sedimentation velocity experiments were performed at 20°C and 60,000 rpm. Absorbance samples were spun in two-channel Epon/charcoal centerpieces in the AN-60-TI rotor. Loading concentrations of 1.0 OD (optical density at 280 nm) for PC2-CC, 0.25 OD for PC2-EF, and 1.0 OD for PC2-C were scanned at 280 nm with 0.001-cm step size setting and no averaging, corresponding to 120 M for PC2-CC (⑀ ϭ 6970 OD 280 mol Ϫ1 cm Ϫ1 ), 163 M for PC2-EF (⑀ ϭ 1280 OD 280 mol Ϫ1 cm Ϫ1 ), and 87 M for PC2-C (⑀ ϭ 9530 OD 280 mol Ϫ1 cm Ϫ1 ). Extinction coefficients were estimated according to Gill and von Hippel (32). Velocity data were analyzed with 2DSA (33)

Molecular Modeling of the C-terminal Tail of Polycystin-2 Predicts Two Domains Connected by a Flexible Linker-De novo
structural models of PC2-C (Ile 704 -Val 968 ) and PC2-EF (Lys 719 -Met 800 ) were obtained as output from the ROBETTA server (coordinate files are available as supplemental material).
The PC2-C model predicts an ␣-helical, two-domain elongated structure connected by a linker containing a known PC2 phosphorylation site (Ser 812 ) (37). Domain 1 (ϳIle 704 -Ser 794 ) consists of a globular ␣-helical bundle (PC2-EF), whereas domain 2 (ϳGly 828 -Val 968 ) contains a striking ϳ40-residue-long central ␣-helix (Tyr 836 -Lys 876 ) characteristic of coiled coil-containing proteins (PC2-CC) (Fig. 1A). Unexpectedly, domain 1 was not  OCTOBER 17, 2008 • VOLUME 283 • NUMBER 42 modeled as an EF-hand motif in the top theoretical structure received as output from the ROBETTA server. Because residues believed to participate in Ca 2ϩ binding were improperly positioned for Ca 2ϩ coordination in this model, and the de novo protocol utilized in ROBETTA is optimized for single domain proteins, we submitted the sequence for the predicted EF-hand domain alone (Lys 719 -Met 800 ) for de novo modeling to the ROBETTA server. Modeling the smaller PC2-EF fragment resulted in a canonical single EF-hand motif with a well defined Ca 2ϩ -binding site (Fig. 1C). The ROBETTA protocol does not attempt to predict or model Ca 2ϩ -binding sites; thus the presence of a credible Ca 2ϩ coordination loop in our model is independent evidence in support of the presence of a Ca 2ϩ -binding EF-hand motif in PC2-C. Interestingly, a structural domain with similarities to a coiled coil (domain 2) was found C-terminal to the position of the purported coiled coil domain in the currently accepted model of PC2-C. To investigate the possibility that a previously undiscovered coiled coil motif is present in the C-terminal cytoplasmic tail of PC2, the sequence of PC2-C was analyzed for coiled coil propensity using the program MARCOIL (38) which contains an improved, more stringent prediction algorithm. The resulting output from MAR-COIL predicts that residues Phe 839 -Asp 919 have a high probability to form a coiled coil (supplemental Fig. 3). Heptad repeat residues (a-d) were mapped and found to correspond to a hydrophobic stripe on the long ␣-helix in the PC2-CC model (Fig. 1, B and E and supplemental Fig. 1). Typically, a-d residues are hydrophobic and make up the coiled coil interface (39).

Domain Mapping of the Polycystin-2 C-terminal Tail
The results of our de novo modeling for the cytoplasmic tail of PC2 suggest that PC2-CC is the actual coiled coil domain assumed in the literature. Residues within PC2-CC are necessary for direct binding to PC2-interacting proteins (PC1, TRPC1, and KIF3A) and PC2-PC2 oligomerization (9, 14 -20). Furthermore, PC2-CC includes the most C-terminal pathogenic PKD-associated truncation variant (R872X) (40). This truncation product is unable to interact with the C terminus of PC1 (12), leading to a plausible hypothesis for PKD pathogenesis. These modeling studies identify a previously unreported coiled coil domain within the C-terminal cytoplasmic tail of PC2, which may serve as an oligomerization interface ablated by PKD-associated truncations.

Domain Mapping Using Limited Proteolysis Validates the Presence and Relative Positions of Two Domains within the C-terminal Tail of Polycystin-2-
To test the validity of our de novo structural model, we mapped the domain structure of PC2-C using limited proteolysis experiments coupled with MALDI-MS and N-terminal sequencing. PC2-C (Ile 704 -Val 968 ) was treated with trypsin or proteinase K to identify proteolytically resistant fragments under single hit kinetics (Fig. 2). Two proteolytically resistant bands from each reaction were analyzed by N-terminal sequencing and correspond to fragments that begin with the sequences: "ALVKL" (Ala 711 -Leu 715 ) and "NTVDD" (Asn 720 -Asp 724 ) from trypsin digestion and "SFPRS" (Ser 804 -Ser 808 ) and "SSRRR" (Ser 823 -Arg 827 ) from proteinase K digestion. Notably, both proteinase K cleavage sites occur within the predicted linker region of our PC2-C model. Given the broad substrate specificity of proteinase K, these results support the presence of an interdomain linker between residues Arg 803 and His 822 . Identical proteolysis reaction products were analyzed by MALDI-MS and matched with fragments of the PC2-C sequence using the N-terminal sequencing results as a constraint, corresponding to Ala 711 -Pro 797 and Gly 828 -His 927 . These results were obtained before modeling was complete because full chain automated structure prediction using the ROBETTA server can take several months. The residues corresponding to these PC2-C fragments were mapped onto our de novo model of PC2-C and overlap with the positions of the predicted domains, validating our two domain model and the position of the linker region (Fig. 1D). Using these results we generated PC2-EF (Asn 720 -Pro 797 ) and PC2-CC (Gly 828 -His 927 ) for biophysical analysis.

Circular Dichroism Shows That PC2-C Has a Highly ␣-Helical Fold and Undergoes Ca 2ϩ -induced Conformational
Changes-The de novo model of PC2-C predicts an all ␣-helical protein. To test this, we compared the secondary structure content of the PC2-C model with values obtained by CD spectroscopy. CD spectra were recorded for PC2-C, PC2-EF, and PC2-CC and show each to contain high ␣-helical content in agreement with our modeling results (Fig. 3). Secondary structure content, calculated using DICHROWEB (27), was compared with percentage ␣-helicity from our structural model of PC2-C for the entire model and for individual domains. Because the model of PC2-C suggested a Ca 2ϩ binding domain, we also investigated the Ca 2ϩ sensitivity of the conformation of PC2-C. PC2-C and PC2-EF spectra were recorded in the absence and presence of Ca 2ϩ . PC2-EF shows an increase from ϳ27 to ϳ45% ␣-helicity after Ca 2ϩ addition, and PC2-C increases from ϳ34 to ϳ50% ␣-helicity with Ca 2ϩ . These values for Ca 2ϩ -bound PC2-C (ϳ50%) and PC2-EF (ϳ45%) correspond with our structural models of PC2-C (62%) and PC2-EF (51%). Secondary structure content calculated from CD spectra for PC2-CC (ϳ70% ␣-helix) is also in agreement with our structural model of the coiled coil domain (74%). CD thus validates the ␣-helical fold predicted by our model of PC2-C and suggests that Ca 2ϩ binding induces a conformational change within the EF-hand domain.

PC2-C Contains a Ca 2ϩ -binding EF-hand Domain by Isothermal Titration Calorimetry That May Be Regulated by Residues Outside of the EF-hand Fold-
The presence of an EFhand motif within the C-terminal cytoplasmic tail of PC2 has been reported (41), but not experimentally confirmed. Our structural model and CD analysis suggest that this EF-hand domain binds Ca 2ϩ . To verify and quantify this Ca 2ϩ binding, we conducted isothermal titration calorimetry (ITC) (Fig. 4). Fitting ITC isotherms with a onesite binding model indicates that PC2-C binds Ca 2ϩ with a K d of ϳ12 M and a stoichiometry of 1:0.4, whereas PC2-EF binds Ca 2ϩ with lower affinity (K d ϳ 214 M) and 1:1 stoichiometry. To verify that residues predicted by our ROBETTA model of the EF-hand domain of PC2 are directly involved in Ca 2ϩ binding, we created PC2-EF-X-Z (T771A/E774A) by site-directed mutagenesis. Thr 771 and Glu 774 are predicted to be directly involved in Ca 2ϩ coordination according to our model (supplemental Fig. 4), and the resulting ITC isotherms following Ca 2ϩ titration indicate a complete loss of affinity for Ca 2ϩ (Fig. 4B).
Interestingly, our attempts to prepare PC2-C in a Ca 2ϩ -free state by treatment of proteins with EDTA resulted in a binding stoichiometry of 1:0.4, indicating that a significant fraction of PC2-C remained in the Ca 2ϩ -bound state. This result contrasts with PC2-EF, where identical treatment seems to have removed all bound Ca 2ϩ . The reduced effectiveness of EDTA treatment correlates with the 17-fold lower affinity of the isolated EFhand domain. The stoichiometry values for Ca 2ϩ binding do not affect the final K d values obtained from fitting procedures. The differential affinity for Ca 2ϩ of PC2-EF alone versus PC2-C implies that regions of PC2-C outside of PC2-EF are involved in stabilization or regulation of the Ca 2ϩ -binding site and support the presence of a physiologically relevant Ca 2ϩ -binding site within PC2.

Oligomerization of the C-terminal Tail of Polycystin-2 Is Mediated by an Extended Coiled Coil Domain as Shown by Size
Exclusion Chromatography-Our structural model and MAR-COIL analysis predict that PC2-CC contains a coiled coil domain with a hydrophobic stripe. We investigated whether this region of PC2-C is important for oligomerization. Analytical SEC was used to analyze the conformation and oligomerization states of PC2-C, PC2-EF, and PC2-CC (Fig. 5). PC2-C (32 kDa) elutes at ϳ670 kDa, suggesting both oligomerization and an elongated molecular conformation. Stripping Ca 2ϩ from PC2-C reduces the apparent molecular weight by 50%, implying a change in oligomerization state or molecular conformation driven by Ca 2ϩ . PC2-EF (12 kDa) elutes at ϳ19 kDa, consistent with the calculated molecular weight of a monomer, whereas PC2-CC (12 kDa) elutes at ϳ75 kDa, indicating both oligomerization and an elongated conformation. To test the possibility that residues corresponding to positions a and d in the PC2-CC coiled coil interface are directly mediating PC2  oligomerization, we created PC2-CC variants with substitutions of hydrophobic a or d residues to either glutamate, lysine, or proline. (PC2-CC L842P, V846E, M849K, I853P, I856K,  V846E/I856K, or M849K/V863E). All point mutations tested resulted in complete insolubility of recombinant PC2-CC suggesting the importance of the integrity of the coiled coil interface for proper folding and possibly oligomerization. (supplemental Fig. 5) It is possible that the extended hydrophobic interface created by these a and d residues was disrupted by the insertion of charged residues or putative helix breaking proline residues resulting in nonspecific aggregation and misfolding. Interestingly, although mutations resulting in truncation of the entire coiled coil segment have been identified in PKD patients within our proposed coiled coil domain, no point mutations resulting in amino acid substitutions have been found. These findings support our structural model of PC2-C and further show that PC2-C is an extended oligomeric macromolecule with PC2-CC responsible for this oligomerization.

Analytical Ultracentrifugation Shows That the C-terminal Tail of Polycystin-2 Has an Extended Oligomeric Structure with
Oligomerization Mediated by the Coiled Coil Domain-To further examine the oligomerization state and molecular conformation of PC2-C, PC2-EF, and PC2-CC, we performed high speed sedimentation velocity ultracentrifugation experiments. In a velocity experiment, sample composition can be determined and multiple components resolved. The van Holde-Weischet analysis provides a model-independent view of the sedimentation coefficient distribution (42), whereas 2DSA resolves heterogeneity in the solution according to both shape and molecular weight. Combining the 2DSA with a Monte Carlo analysis reduces the effects of stochastic noise on the appearance of false-positive peaks and enhances the inherent signal in the data (34). We used a color gradient indicating the relative concentration of solutes to create pseudo-three-dimensional plots (Fig. 6). Van Holde-Weischet analysis showed that all samples were mostly free of aggregates and degradation products, and 2DSA analysis showed one predominant species with narrow confidence bands, and several minor species with little signal. Velocity sedimentation results suggest that PC2-CC is mostly present as an oligomer (trimer or tetramer) with an elongated shape, and that PC2-EF is monomeric with globular shape (Fig. 6, A  and B). PC2-C appears to be present in a monomer-dimer equilibrium in velocity sedimentation experiments, which at the concentration K AV for PC2-C (ϪCa 2ϩ ) was measured following treatment with EDTA. PC2-C elutes at ϳ670 kDa, but this shifts to ϳ330 kDa following removal of Ca 2ϩ . PC2-EF elutes at ϳ19 kDa consistent with a monomeric state. PC2-CC elutes at ϳ75 kDa consistent with an extended conformation and oligomerization. . f/f 0 Ͼ1.0 indicate an extended molecular shape. The relative concentration of solutes is shown using a color gradient (light blue to black with black centered peaks corresponding to the highest concentration). A, PC2-EF exists as a monomer with globular shape. B, PC2-CC is present as an elongated oligomeric species (trimer or tetramer). C, PC2-C appears to be elongated and present in a monomer-dimer equilibrium. D, van Holde-Weischet differential sedimentation distributions plotted as molecular weight distributions. All proteins appear narrowly distributed in the sedimentation coefficient.
tested (87 M) appears to be mostly shifted to the dimer state (Fig. 6C). These results are also in agreement with our SEC analysis, and support a model of PC2-C as an oligomeric elongated macromolecule whose oligomerization is mediated by the C-terminal coiled coil domain, PC2-CC, identified in this study.
Conclusion-In PKD, PC2 is frequently truncated in PC2-C. The currently accepted domain model of PC2-C fails to explain the mechanisms by which truncations in PC2-C lead to PKD. Moreover, direct PC2 binding to proteins involved in PKD progression requires residues outside this region of PC2-C. To address these discrepancies and provide a framework to investigate the role of PC2-C in PC2 function and PKD progression, we performed de novo molecular modeling and validated this model through biochemical and biophysical analysis.
Structural modeling using the ROBETTA server predicts an extended ␣-helix C-terminal to the currently accepted coiled coil domain of PC2. The correlation of our structural model with limited proteolysis and MARCOIL results suggests that there is a previously unidentified coiled coil domain C-terminal to the EF-hand motif, and that this is the actual coiled coil domain assumed in the literature. Using AUC and SEC, we found the EF-hand motif domain to be monomeric, whereas the coiled coil domain forms oligomers. Mutagenesis of a and d heptad repeat residues predicted to form the PC2 coiled coil interface result in complete insolubility of recombinant PC2-CC, suggesting their importance in folding and possibly oligomerization. This is in agreement with findings for TRP channels where a C-terminal coiled coil region mediates channel oligomerization (43), but it remains to be tested on active fulllength PC2 ion channels.
Ca 2ϩ binding to the EF-hand motif domain was investigated using ITC, CD, and site-directed mutagenesis. This is the first experimental data supporting the presence of a Ca 2ϩ -binding site within PC2. EF-hand motif proteins bind Ca 2ϩ with affinities ranging from nanomolar to millimolar (44), and our ITC experiments for the EF-hand domain of PC2 suggest one Ca 2ϩbinding site with micromolar affinity. Given the relatively high concentration of Ca 2ϩ predicted at the mouth of the PC2 channel pore and locally because of opening of associated ryanodine or inositol 1,4,5-trisphosphate receptor channels at the endoplasmic reticulum surface, a micromolar affinity for Ca 2ϩ may allow PC2-EF to sense local changes in Ca 2ϩ concentration. This result, coupled with the conformational changes observed upon Ca 2ϩ binding, suggests a functional role for Ca 2ϩ in PC2 activation.
We report here the first molecular model of the C-terminal cytoplasmic tail of PC2. This model provides a framework to guide experiments probing normal PC2 function and the molecular basis of PKD pathogenesis and progression. Our results validate the presence of a functional Ca 2ϩ -binding site and identify a previously unreported coiled coil domain in PC2. These analyses further suggest that the coiled coil domain may serve as a PC2 oligomerization interface. Additionally, the presence of a coiled coil encompassing ϳTyr 836 -Lys 876 may provide a structural explanation for previous reports that show the importance of this region for PC2 binding to PC1 and other proteins. Significantly, all known PKD-asso-ciated PC2 truncations disrupt this new PC2 coiled coil domain. Our results thus provide a plausible molecular mechanism for PKD pathogenesis.