|Home | About | Journals | Submit | Contact Us | Français|
Here we report a novel mechanism of PDZ (PSD-95/Dlg/ZO-1) domain regulation that distorts a conserved element of PDZ ligand recognition. The polarity regulator Par-6 assembles a conserved multi-protein complex and is directly modulated by the Rho GTPase Cdc42. Cdc42 binds the adjacent Cdc42/Rac interactive binding (CRIB) and PDZ domains of Par-6, increasing C-terminal ligand binding affinity 10-fold. By solving structures of the isolated PDZ domain and a disulfide-stabilized CRIB-PDZ, we detected a conformational switch that controls affinity by altering the configuration of the conserved ‘GLGF’ loop. As a result, lysine 165 is displaced from the PDZ core by an adjacent hydrophobic residue, disrupting coordination of the PDZ ligand binding cleft. Stabilization of the CRIB:PDZ interface restores K165 to its canonical location in the binding pocket. We conclude that a unique ‘dipeptide switch’ in the Par-6 PDZ transmits a signal for allosteric activation to the ligand binding pocket.
PSD-95/Dlg/ZO-1 (PDZ) domains comprise a large family of interaction domains that typically anchor a modular protein to the C-terminal sequence of a specific binding partner (Dong et al., 1997; Ponting et al., 1997; Songyang et al., 1997). Modular multi-PDZ proteins often serve as mediators of signal transduction by assembling protein complexes at the plasma membrane (Hung and Sheng, 2002; Kim and Sheng, 2004; Zhang and Wang, 2003). Once viewed as simple scaffolds for assembly of other signaling proteins, a subset of PDZ domains participate directly in functional regulation. Par-6(Peterson et al., 2004) and PSD-95 (Petit et al., 2009) contain PDZ domains that are modulated by interactions distant from the peptide binding pocket, indicating the presence of allostery within these domains. Allosteric control of PDZ function as predicted by statistical coupling analysis (Lockless and Ranganathan, 1999) is likely to be an important feature of other scaffold and signaling proteins, but the structural basis for energetic coupling between allosteric sites remains undefined.
Par-6 is the central organizer of an evolutionarily conserved cell polarity complex comprised of Par-6, Par-3, atypical protein kinase C (aPKC), and Cdc42 that regulates apical membrane identity in epithelial cells, leading edge formation and maintenance in leukocyte chemotaxis, and asymmetric cell division in D. melanogaster neuroblasts (Aranda et al., 2006; Aranda et al., 2008; Bose and Wrana, 2006; Gerard et al., 2007; Hurd et al., 2003; Prehoda, 2009; Tsuboi, 2006). The Par-6 protein consists of three known structural units: a Phox/Bem (PB1) domain, a Cdc42/Rac-interactive binding domain (CRIB), and a single PDZ domain (Fig. 1A,)(Penkert et al., 2004). A PB1 domain at the N-terminus associates with the PB1 domain of atypical protein kinase C (aPKC), the kinase responsible for downstream Par complex signaling. Cell polarization requires an input signal delivered by Cdc42, a monomeric GTPase that binds to an unstructured CRIB motif adjacent to the Par-6 PDZ domain. In other Cdc42 effectors, the CRIB sequence supports Rac1 or Cdc42 binding independent of other domains. In contrast, Cdc42-GTP binds Par-6 only when the CRIB and PDZ are linked (Joberty et al., 2000; Lin et al., 2000) because of imperfect conservation in the GTPase binding motif, which has been termed a ‘semi-CRIB’ sequence. Relative to other CRIB-containing proteins, Rac1 binds Par-6 very weakly and is not a viable signaling partner for Par-6 (Garrard et al., 2003). Upon Cdc42-GTP binding, this disordered CRIB region forms a single β-strand (β0) that joins the separate β-sheets of Cdc42 and Par-6 into a single structure (Garrard et al., 2003) (Fig. 1B, PDB entry: 1NF3).
The structural coupling between flexible CRIB and folded PDZ domains creates a molecular switch that regulates PDZ ligand binding. Association of Cdc42-GTP with the CRIB-PDZ130-255 module results in a ~10-fold increase in affinity for a C-terminal PDZ ligand peptide, and GTPase-dependent Par-6 binding is essential for epithelial tight junction formation (Penkert et al., 2004; Peterson et al., 2004). Since the PDZ ligand-binding interface is distant from the Cdc42-binding interface, an allosteric connection is presumed to link the GTPase binding surface and the PDZ ligand binding pocket. Despite being a prototypical example of functional communication between domains (Lee et al., 2008), the structural basis for transmission of the GTPase signal remains undefined.
Formation of the Cdc42 complex with Par-6 creates new intramolecular contacts (CRIB:PDZ) and two intermolecular interfaces (Cdc42:CRIB and Cdc42:PDZ). Comparisons of the previously determined NMR and X-ray crystal structures (Penkert et al., 2004; Peterson et al., 2004) show that intermolecular interactions are formed primarily with the CRIB (residues 130-155). Only 10% are direct contacts between Cdc42 and the PDZ, suggesting that the GTPase functions primarily as a scaffold and the allosteric mechanism is a unique feature of the intramolecular CRIB:PDZ interface. Here we show that a disulfide linking the CRIB and PDZ domains can mimic the structural and functional impact of Cdc42 binding. We measured PDZ ligand binding to the disulfide-stabilized CRIB-PDZ130-255 module and the isolated PDZ (Figure 1C) and solved structures for each new construct. Structural differences between the high and low affinity states demonstrate that Par-6 allostery is encoded in the interdomain interface via a conformational switch that transposes two adjacent PDZ sidechains, L164 and K165. Stabilization of the CRIB:PDZ interface repositions K165 from the PDZ surface to a highly conserved configuration required for C-terminal ligand binding(Doyle et al., 1996; Gee et al., 2000; Harris et al., 2001; Harris et al., 2003; Harris and Lim, 2001; Penkert et al., 2004). We conclude that Par-6 represents a novel example of interdomain communication in which a flexible extra-domain sequence (CRIB) regulates an intra-domain process (PDZ ligand binding). Furthermore, the allosteric control mechanism observed in Par-6 is a striking evolutionary departure from known PDZ domain functionality.
D. melanogaster Par-6 proteins were produced as previously described (Peterson et al., 2004). The CRIB-PDZ130-255 and CRIB-PDZQ144C/L164C constructs encompassed residues 130-255, and the PDZ156-255 contained residues 156-255. Creation of the CRIB-PDZQ144C/L164C variant utilized QuikChange® II Site-Directed Mutagenesis Kit (Stratagene). Mutagenesis was confirmed by DNA sequencing. All proteins were expressed in E. coli strain BL21 (DE3) as a host strain with pBH vectors for hexahistidine fusions. The pET-derived pBH4 vector (Volkman et al., 2002) contains a T7 promoter, an ampicillin resistance gene, and a coding sequence for tobacco etch virus (TEV) protease, allowing removal of the hexahistidine tag during purification. Hexahistidine fusion proteins were purified using Ni-NTA resin followed by incubation with TEV protease to remove the tag. A further purification by Ni-NTA resin captured cleaved the hexahistidine tag and uncleaved fusion proteins, allowing pure protein to be isolated in the initial eluate. Final protein purity was measured as >99% by SDS-PAGE and MALDI-TOF spectroscopy. CRIB-PDZQ144C/L164C purification additionally included guanidine-HCl unfolding at pH 5.0, and rapid dilution into guanidine-free buffer at pH 8.0, allowing simultaneous protein folding and disulfide formation. Subsequent reduction of the inserted disulfide bond for control experiments was achieved via incubation of CRIB-PDZQ144C/L164C in a tenfold molar excess of DTT overnight at room temperature at pH 8.0.
Fluorescence polarization (FP) binding assays were performed on a Photon Technology International spectrofluorometer. Solutions were prepared with increasing amount of protein and constant rhodamine-labeled VKESLV (120 nM), and FP values were measured using excitation and emission wavelengths of 560 nm and 585 nm, respectively. All solutions were prepared in a buffer containing 25mM NaHPO4 and 50mM NaCl at pH 5.5. Binding curves were analyzed by nonlinear fitting to an equation describing 1:1 ligand binding as previously described (Tyler et al., 2010). The reported uncertainties in Kd values represent the error estimated by the nonlinear regression algorithm for a representative binding curve, and are consistent with the Kd variations observed in replicate titration experiments.
NMR experiments were carried out at 25 °C on a Bruker Avance II 600 MHz or Bruker Avance III 500 MHz spectrometer equipped with a triple resonance z-axis gradient CryoProbe®. All NMR samples contained 0.75-1.25 mM 15N- or 13C/15N-labeled protein and were prepared in 90% H2O/10% D2O containing 20 mM NaHPO4, 50 mM NaCl and 0.05% sodium azide (pH 5.5). Resonance assignments and distance constraints of PDZ156-255 and CRIB-PDZQ144C/L164C were obtained from the following experiments: 15N-HSQC, HNCO, HNCA, HNCACB, HN(CO)CA, HN(CA)CO, HN(CO)CACB, C(CO)NH, HCCH-TOCSY, HBHA(CO)NH, HC(CO)NH, 15N NOESY-HSQC, 13C NOESY-HSQC (one each for aromatic and aliphatic side-chain 13C regions). All NOESY mixing times were 80 ms. Heteronuclear NOEs were measured from an interleaved pair of 2D 15N-1H gradient-sensitivity enhanced correlation spectra. NOEs were calculated as the ratios of peak heights in spectra recorded with and without a 5s proton saturation period. NMR data were processed with the NMRPipe software package. The XEASY program was used for resonance assignments and analysis of NOE spectra. Using the program TALOS, 139 PDZ domain-mediated target binding Φ and ψ dihedral angle constraints were generated from 1H, 13C, and 15N secondary shifts (Cornilescu et al., 1999). Initial structures were generated using the NOEASSIGN module of the torsion angle dynamics program CYANA (Herrmann et al., 2002). Structures were refined by iterative rounds of manual refinement to eliminate constraint violations. Final refinement with explicit solvent with experimental constraints and non-bonded energy terms was performed in XPLOR-NIH (Schwieters et al., 2003) using a protocol that improves the quality of NMR structures in terms of validation criteria like Ramachandran statistics and Z-scores (Linge et al., 2003).
The CRIB domain of Par-6 (residues 130-155) is unstructured when Cdc42-GTP is not present, but it becomes ordered when bound to Cdc42-GTP (Peterson et al., 2004). We hypothesized that specific CRIB:PDZ contacts are formed in the absence of the GTPase because residues 159-163 (the PDZ β1-strand) are severely broadened in 1H-15N HSQC spectra of the CRIB-PDZ130-255 module. Removal of CRIB residues induced small chemical shift perturbations and restored PDZ β1 peak intensities to a uniform level equivalent to other peaks in the spectrum (Figure 2A). Highlighting broadened residues (from CRIB-PDZ130-255) on the PDZ surface reveals a region equivalent to the CRIB:PDZ interaction surface in Cdc42:CRIB-PDZ (Figure 2B-D). This suggests that transient association of the CRIB and PDZ domain is a weak but specific interaction that may correspond to the folded conformation observed in the Cdc42:CRIB-PDZ crystal structure.
Inspection of the Par-6:Cdc42 crystal structure shows that direct Cdc42-to-PDZ contacts are a small portion (~10%) of the total Par-6:GTPase interface, and CRIB residues mediate the vast majority of new PDZ interactions. Consequently, we hypothesized that stabilization of the CRIB:PDZ interface could shift the PDZ to a high-affinity state in the absence of Cdc42-GTP. The sidechains corresponding to Q144 and L164 pack against each other at the CRIB:PDZ interface in the complex with Cdc42 (Garrard et al., 2003; Peterson et al., 2004), so we substituted each residue with cysteine (CRIB-PDZQ144C/L164C) to permit formation of an interdomain disulfide bond. The 1H-15N-HSQC spectrum showed that CRIB-PDZQ144C/L164C is folded (Figure 3A), and, while specific differences are evident, the overall pattern of shifts resembles those of the Cdc42:CRIB-PDZ complex and the free CRIB-PDZ130-255 module (Figure S1) (Peterson et al., 2004). The 13Cβ chemical shifts for C144 and C164, 39.0 and 42.8 ppm, respectively, are diagnostic for the oxidized sulfur, indicating that the disulfide bond was formed. A comparison of heteronuclear 1H-15N NOEs showed that, relative to the unstructured residues observed in CRIB-PDZ130-255, the disulfide in CRIB-PDZQ144C/L164C stabilizes a portion of the CRIB domain (Figure 3B)(Peterson et al., 2004). Residues 130–143 remain disordered, but beginning with C144 the NOE values for CRIB-PDZQ144C/L164C match the pattern observed for Cdc42-bound CRIB-PDZ. As expected, reduction of the disulfide bond in CRIB-PDZQ144C/L164C with DTT restores the HSQC spectrum to the pattern observed for CRIB-PDZ130-255 (Figure S2A). Loss of the disulfide also diminishes the heteronuclear NOE values of CRIB residues consistent with a loss of folded structure (Figure S2B).
To assess the impact that CRIB residues have on PDZ ligand binding, we compared PDZ156-255, CRIB-PDZ130-255, CRIB-PDZQ144C/L164C and Cdc42:CRIB-PDZ binding to rhodamine-labeled VKESLV, a C-terminal PDZ ligand previously shown to bind the CRIB-PDZ130-255 module in a Cdc42-dependent manner (Penkert et al., 2004; Peterson et al., 2004). Fluorescence polarization measurements and nonlinear fitting revealed that the isolated PDZ domain (PDZ130-255) binds the C-terminal PDZ ligand VKESLV with an equilibrium dissociation constant (Kd) of 72 ± 5 μM for PDZ156-255, whereas the intact CRIB-PDZ130-255 binds with a Kd of 54 ± 5 μM (Figure 4). CRIB-PDZQ144C/L164C bound the VKESLV peptide with a KD of 13 ± 1.2 μM, a significant increase relative to CRIB-PDZ130-255 that approaches the affinity of the Cdc42:CRIB-PDZ complex (Kd = 6 ± 2 μM). Reduction of the CRIB-PDZQ144C/L164C disulfide shifts the Kd to 51 ± 8 μM (Figure S2C & D), and the Cdc42:CRIB-PDZQ144C/L164C complex binds with a Kd of 6.2 ± 2.2 μM (Figure S2E). Taken together, these results suggest that stable association of the predominantly disordered CRIB motif increases the affinity of the PDZ domain for a C-terminal ligand peptide even when Cdc42-GTP is not present. Furthermore, by stabilizing a high-affinity state of Par-6 in the absence of Cdc42, we enabled NMR structural studies on the high-affinity state so that a specific allosteric mechanism could be elucidated.
To determine whether disulfide stabilization of the CRIB had a discernable effect on the PDZ conformation, we solved the structure of CRIB-PDZQ144C/L164C (Figure 5A). Refinement statistics (Table 1) and the NMR ensemble show that a portion of the CRIB motif is positioned in a well-defined conformation adjacent to the β1 strand of the PDZ domain (Figure 5B). A series of cross-strand NOEs confirms that three residues (V145-S146-A147) make contacts equivalent to the β0 strand observed in the Cdc42:CRIB-PDZ complex (Figure 5C). CRIB:PDZ contact surfaces in CRIB-PDZQ144C/L164C and Cdc42:CRIB-PDZ are quite similar when mapped on each structure (Figure 5D), effectively recapitulating the configuration formed upon GTPase binding. In total, the CRIB:PDZ interface comprises 536.1 Å2 in the complex with Cdc42 and 509.4 Å2 in CRIB-PDZQ144C/L164C. This 26.7 Å2 difference corresponds to a loss of 5% of the CRIB:PDZ contact area. Interestingly, the Gibb’s free energy of VKESLV peptide binding to Cdc42-bound Par-6 (ΔGCdc42:CRIB-PDZ = −7.1 kcal/mol) and to CRIB-PDZQ144C/L164C (ΔGCRIB-PDZQ144C/L164C = −6.7 kcal/mol) differ by 0.4 kcal/mol, a ~5% difference in binding energies between the two constructs. We speculate that the absence of marginal CRIB:PDZ contacts accounts for the difference in ligand binding affinity for Cdc42:CRIB-PDZ and CRIB-PDZQ144C/L164C. These data demonstrate that CRIB-PDZQ144C/L164C closely resembles the Cdc42:CRIB-PDZ crystal structure, consistent with the increase in ligand binding affinity.
Despite a dramatic difference in ligand binding affinity, structural comparisons between CRIB-PDZQ144C/L164C and the previously solved CRIB-PDZ130-255 NMR structure(Peterson et al., 2004) revealed no structural differences that would explain the disulfide-induced shift in affinity. Further inspection of the CRIB-PDZ130-255 structure revealed a lack of distance constraints in the β1-2 loop of the PDZ domain due to exchange broadening. We speculated that interconversion between distinct high- and low-affinity states in CRIB-PDZ130-255 might obscure their important structural differences. Strikingly, superimposition of a number of 15N HSQC peaks from the ligand-free PDZ domain in four different contexts (PDZ156-255, CRIB-PDZ130-255, CRIB-PDZQ144C/L164C and Cdc42:CRIB-PDZ) revealed the characteristic pattern of a two-state, fast exchange conformational equilibrium (Volkman et al., 2001) (Figure 6A, Figure S1), in which the position of an 1H-15N HSQC peak along a linear progression reports on the ratio of high- and low-affinity states. Relative to the CRIB-PDZ130-255 spectrum, these peaks shift in one direction when the CRIB domain is removed and the opposite direction when the CRIB is stabilized by a disulfide link or by GTPase binding. Moreover, their positions in the absence of ligand can be correlated with Gibbs’ free energies calculated from the Kd values for VKESLV binding (Figure 6B). These data suggest that CRIB-PDZ130-255 favors the low-affinity binding state based on its similarity to PDZ156-255 in terms of binding energy and HSQC peak positions. Because PDZ156-255 appears to be dominated by the lowest affinity state and exhibits little or no conformational exchange, we speculated that its three-dimensional structure could reveal specific elements associated with allosteric PDZ activation.
To characterize the stable low-affinity Par-6 conformation, we solved the NMR structure of PDZ156-255, which includes only the PDZ domain and no CRIB residues (Figure 7A). Refinement statistics are summarized in Table 1. The domain fold of PDZ156-255 generally matches the NMR and X-ray crystal structures of CRIB-PDZ130-255 solved in the absence and presence of Cdc42, respectively. However, the precision of the NMR ensemble is higher than the CRIB-PDZ130-255 NMR structure, based on a comparison of backbone atomic r.m.s.d. values (0.73 vs. 1.27 Å, respectively, over residues 156-253)(Peterson et al., 2004). Heteronuclear NOE experiments show that the profile of picosecond-nanosecond backbone dynamics for PDZ156-255 is similar to CRIB-PDZ130-255 (Figure 7B). These trends, in addition to the reduction of exchange broadening in PDZ residues relative to CRIB-PDZ (Figure 2A), suggest that removal of the unstructured CRIB produced a slight decrease in backbone flexibility for the PDZ domain.
Careful examination of the PDZ156-255 ensemble revealed that the β1-2 loop conformation is altered in comparison to the crystal structure of Cdc42:CRIB-PDZ. K165 is the most striking difference, as its solvent-exposed sidechain projects away from the PDZ domain in PDZ156-255 (Figure 8A), whereas in structures of CRIB-PDZQ144C/L164C and the Cdc42:CRIB-PDZ complex this conserved basic residue extends through the domain to form one end of the ligand-binding pocket (Figure 8B)(Doyle et al., 1996; Lemmers et al., 2004; Penkert et al., 2004). In PDZ156-255, this rearrangement is compensated by insertion of an adjacent hydrophobic residue (L164) in the location normally occupied by K165. Since the configuration of the L164-K165 dipeptide appears to be inverted in PDZ156-255 relative to the structures representing the high-affinity conformation (Figure 8B), we sought to verify the positions of both side chains in each construct.
Multiple NOEs in the CRIB-PDZQ144C/L164C spectra constrain K165 to the buried configuration seen in Cdc42:CRIB-PDZ (Figure 8C) and throughout the PDZ family (Doyle et al., 1996; Gee et al., 2000; Harris et al., 2001). In contrast, no medium- or long-range contacts were detected for K165 in the NOESY spectra of PDZ156-255, while numerous long-range NOEs position the L164 sidechain in the position normally occupied by K165 (Figure 8D). Taken together, these results show that the lowest affinity state of Par-6, the isolated PDZ domain, displays interchanged L164 and K165 sidechain positions, marking a divergence from conserved PDZ domain structure (Fanning and Anderson, 1996). We speculated that association of the CRIB with the PDZ domain might govern Par-6 C-terminal PDZ ligand binding affinity by flipping this ‘L/K switch’.
The configuration of the dipeptide switch is poorly defined in the NMR ensemble of CRIB-PDZ130-255 due to line broadening in the β1-2 loop (Peterson et al., 2004). To determine whether K165 is in the high affinity ‘in’ or low affinity ‘out’ configuration, we used the chemical shift profiles of CRIB-PDZQ144C/L164C and PDZ156-255 as fingerprints for the two states, respectively. Correlation plots comparing CRIB-PDZ130-255 chemical shifts with both PDZ156-255 and CRIB-PDZQ144C/L164C indicate that the L/K switch in CRIB-PDZ130-255 favors the low affinity configuration (L164 ‘in’ and K165 ‘out’, Figure 8E). Additionally, the largest shift deviations in the domain highlighted two regions of the PDZ: the CRIB:PDZ130-255 interface, and the pocket that surrounds K165 in the high-affinity conformation (Figure 8F). High chemical shift deviations often indicate structural rearrangements, and large shift deviations around the K165 pocket suggest that the low-affinity state of the PDZ is governed by the position of K165.
Amino acid substitutions of this conserved lysine or arginine in other PDZ domains decreased their binding affinity for C-terminal ligands (Gee et al., 2000; Harris et al., 2003), consistent with its specific contribution to the carboxylate binding pocket. Because either L164 or K165 can occupy this location in the Par-6 conformational equilibrium, we substituted each residue in the CRIB-PDZ130-255 construct to more precisely define their contributions. VKESLV binding to the K165L mutant was substantially weaker (KdK165L = 219 ± 64 μM, Figure 9A), while substitution of K165 with an isosteric methionine decreased the affinity slightly (KdK165M = 60.4 ± 9.3 μM, Figure 9A). CRIB-PDZK165M peptide binding is enhanced in the presence of Cdc42 (KdCdc42:K165M = 15.9 ± 2.2 μM, Figure 9C), though somewhat less than in CRIB-PDZ130-255 (Kd = 6 ± 2 μM, Figure 4). The difference in VKESLV binding affinity of Cdc42:CRIB-PDZK165M and Cdc42:CRIB-PDZ130-255 corresponds to a change in binding energy (ΔΔGVKESLV) of ~590 cal/mol. Taken together, these results demonstrate that when the carboxylate binding pocket contains a leucine in place of the conserved lysine, binding to a C-terminal ligand is sharply reduced, while a methionine is largely compatible with PDZ function and the Par-6 dipeptide switch.
Additionally, we mutated L164 to both a lysine (L164K) and a glutamic acid (L164E). The L164E substitution reduced the binding affinity significantly (KdL164E = 165 ± 64.3, Figure 9B) but also made the protein more susceptible to aggregation and precipitation. In contrast, the binding affinity of L164K was enhanced relative to CRIB-PDZ130-255 (KdL164K = 35 ± 4.5 μM, Figure 9B). Thus, replacement of the leucine at position 164 with a lysine partially mimics the high affinity conformer.
As the central organizer of a conserved protein complex that controls cell polarity, Par-6 is activated upon Cdc42-GTP binding to its CRIB domain. The GTPase-induced change in binding affinity to a C-terminal PDZ ligand is required for Par-6 to function in epithelial tight junction formation (Peterson et al., 2004). Unlike other CRIB sequences, the Par-6 CRIB fails to bind Cdc42-GTP unless it is linked to the PDZ domain, due to the absence of two conserved histidine residues required for high-affinity GTPase binding (Garrard et al., 2003). Another difference between Par-6 and other CRIB-containing proteins is the mechanism by which Cdc42-GTP alters the activity of its effector. In other CRIB:Cdc42 interactions, an autoinhibited complex becomes activated when Cdc42-GTP disrupts a folded CRIB conformation, releasing a constitutively active functional domain. For example, the GTPase binding domain (GBD) of N-WASP is a small helical domain that sequesters both the CRIB motif and cofilin homology domain in an autoinhibited state. Cdc42-GTP binding disrupts the GBD allowing the unstructured cofilin motif to interact with the Arp2/3 complex and catalyze actin polymerization (Abdul-Manan et al., 1999; Kim et al., 2000; Morreale et al., 2000; Prehoda et al., 2000). Cdc42-GTP binding likewise disrupts a folded state of the CRIB in PAK, ACK, and WASP (Abdul-Manan et al., 1999; Morreale et al., 2000; Mott et al., 1999). In our model for Par-6 activation, Cdc42-GTP instead binds the CRIB cooperatively with the PDZ domain (Figure 8A), shifting a conformational equilibrium and promoting new interactions that alter Par complex activity via its ten-fold enhancement of C-terminal ligand-binding affinity (Peterson et al., 2004). Our results indicate that allosteric regulation of Par-6 ligand binding is encoded in interdomain contacts that trigger a novel conformational switch in the PDZ domain. Furthermore, Cdc42 activates the switch indirectly by stabilizing a weak CRIB:PDZ interaction. Thus, Cdc42 activation of Par-6 illustrates that GTPase binding to the conserved CRIB motif can promote downstream signaling by either disrupting or stabilizing protein structure.
K165 in Par-6 occupies a conserved position at the start of the β1-2 loop (Figure 10)(Doyle et al., 1996; Fanning and Anderson, 1996; Gee et al., 2000; Harris et al., 2001; Harris et al., 2003; Harris and Lim, 2001), and the PDZ domain family can be divided into two subgroups based on the identity of this amino acid. The larger group possesses either a lysine or arginine residue at this position, and a terminal amino group from its sidechain helps to stabilize a structural water molecule that aids in C-terminal ligand binding (Doyle et al., 1996; Gee et al., 2000; Harris et al., 2001). However, in the PDZ155-256 structure the K165 sidechain is reoriented outward from the PDZ ligand binding pocket and the sidechain of L164 instead occupies that space within the PDZ core. In the Cdc42:CRIB-PDZ crystal structure, β0-β1 hydrogen bonds linking L164 with Val 145 and K165 with Arg 143 (D. melanogaster Par-6 numbering) immobilize this segment of the β0-β1 backbone and redirect K165 to the internal position observed throughout the PDZ family. Additionally, folded CRIB residues occupy the region in which K165 sits in the low affinity configuration, preventing this arrangement upon Cdc42 binding (Figure 10). In the CRIB-PDZQ144C/L164C NMR structure, the disulfide enforces the outward orientation of the sidechain at the 164 position permitting K165 to occupy its normal position in the ligand binding pocket. Thus, the position of K165 is strongly correlated with ligand binding affinity in Par-6. In the absence of a stable CRIB:PDZ interface, peptide binding is weak and K165 is predominantly in the ‘out’ orientation. We conclude that the β0-β1 interface is incompatible with the ‘out’ conformation of K165, and that Cdc42-GTP binding or disulfide stabilization switches the L/K dipeptide to its high affinity configuration.
While our results provide a structural explanation for GTPase regulated C-terminal ligand binding, the Par-6 PDZ domain also binds an internal sequence of the Stardust/Pals1 protein. In the crystal structure of a Par-6 PDZ/Pals1 complex (Penkert et al., 2004) an aspartic acid sidechain in the Pals1 sequence occupies the site normally filled by the backbone carboxylate of a C-terminal ligand. As in the crystal structure of the complex with VKESLV, K165 is in the high affinity ‘in’ position suggesting that the dipeptide switch also rearranges to accommodate binding of this internal sequence ligand. Strikingly, Cdc42 binding has a negligible impact on the affinity of internal ligand binding, in stark contrast to C-terminal ligand binding (Peterson et al., 2004). While the functional consequence of Cdc42-GTP activation may be to shift the association of Par-6 with an internal ligand (e.g. Pals1) to a C-terminal ligand like the Crumbs receptor based on the change in relative binding affinities, our present results do not yet explain the ligand-selective nature of CRIB-PDZ allostery.
Par-6 has a unique capacity for allosteric regulation that derives partly from the CRIB:PDZ linkage, but PDZ sequence features that permit inversion of the L164-K165 dipeptide switch remain uncertain. One possible factor is the length and flexibility of the β1-2 loop of the PDZ. According to an alignment of 286 PDZ domain sequences on the SMART server (http://smart.embl-heidelberg.de)(Letunic et al., 2009), this loop in Par-6 contains a ~2 residue insertion between the conserved K165 and ‘GLGF’ motif (PLGF in Par-6). Residues 166–168 appear flexible in each of the NMR structural ensembles because the NOESY data contains few medium and long-range NOEs. In addition, 1H-15N heteronuclear NOE data also indicates increased flexibility in the β1-2 loop in of the Par-6 PDZ constructs (Figure 3B, ,7B),7B), whereas other PDZ domains exhibit a relatively rigid β1-2 loop (Li et al., 2006; Terrien et al., 2009; Tochio et al., 2000; Walma et al., 2002). Thus, increased mobility in this region may enable the switching motion required to reversibly replace K165 with L164 in the PDZ core.
PDZ affinity switching observed in Par-6 may also require a specific amino acid type adjacent to the conserved lysine. The presence of leucine in Par-6 is anomalous, since this position is typically occupied by a polar residue in the PDZ family (Letunic et al., 2009; SMART, 2011). Our results for the low-affinity state represented by the isolated PDZ domain indicate that insertion of L164 into the hydrophobic core is energetically favored over the conserved lysine. In contrast to other PDZ domains, stabilization of the CRIB:PDZ interface provided by Cdc42 binding or a disulfide crosslink is required to maintain the ‘K165 in’ high-affinity configuration. Replacement of L164 with a negatively charged glutamate residue significantly diminished binding (Kd = 240 μM) while the L164K substitution enhanced VKESLV binding (Kd = 38 μM). Since L164 normally occupies the ‘in’ configuration in the absence of Cdc42, we conclude that the L164K substitution ensures that a lysine sidechain occupies the carboxylate binding pocket and partially mimics the high-affinity state. In conclusion, while the ‘dipeptide switch’ employed by Par-6 to transmit the GTPase signal to the ligand binding pocket is unprecedented, the combination of a flexible β1-2 loop and an apolar residue adjacent to the conserved lysine might introduce similar conformational rearrangements and allosteric regulation of binding affinity in other members of the PDZ family.
This work was supported by a grant from the National Institutes of Health (AI508072).
Accession Numbers The structures were deposited in the Protein Data Bank (PDB) and the Biological Magnetic Resonance Bank (BMRB). PDB accession numbers for CRIB-PDZQ144C/L164C and PDZ156-255 are 2LC6 and 2LC7, respectively. BMRB accession numbers for CRIB-PDZQ144C/L164C and PDZ156-255 are 17599 and 17600, respectively.
The authors declare no conflicts of interest. D.S.W. and F.C.P. performed experiments, analyzed data, and wrote the manuscript. B.F.V. designed research, analyzed data, and wrote the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.