Intrinsically disordered γ-subunit of cGMP phosphodiesterase encodes functionally relevant transient secondary and tertiary structure

Abstract

The retinal phosphodiesterase (PDE6) inhibitory γ-subunit (PDEγ) plays a central role in vertebrate phototransduction through alternate interactions with the catalytic αβ-subunits of PDE6 and the α-subunit of transducin (α_t). Detailed structural analysis of PDEγ has been hampered by its intrinsic disorder. We present here the NMR solution structure of PDEγ, which reveals a loose fold with transient structural features resembling those seen previously in the x-ray structure of PDEγ_46–87 when bound to α_t in the transition-state complex. NMR mapping of the interaction between PDEγ_46–87 and the chimeric PDE5/6 catalytic domain confirmed that C-terminal residues 74–87 of PDEγ are involved in the association and demonstrated that its W70 indole group, which is critical for subsequent binding to α_t, is left free at this stage. These results indicate that the interaction between PDEγ and α_t during the phototransduction cascade involves the selection of preconfigured transient conformations.

Phototransduction, the primary event of vision is critically regulated by the γ-subunit of cGMP phosphodiesterase (PDEγ). In the dark, by binding tightly to the catalytic αβ-subunits of cGMP PDE (PDE6) PDEγ keeps PDE6 inactive; yet upon photoexcitation, interaction of PDEγ with the α-subunit of activated transducin (α_t) relieves its inhibitory constraint on PDE6. For termination of phototransduction, PDEγ interacts with both α_t and RGS9, a regulator of G protein signaling, to accelerate α_t GTPase activity (1). Beyond phototransduction, the importance of PDEγ in photoreceptor cell viability is demonstrated by rapid retinal degeneration in the PDEγ-knockout mouse (2), and growing evidence suggests that PDEγ also interacts with some signaling proteins in nonphototransduction pathways (3).

Biochemical studies have shown that the central polycationic region (residues 24–45) and the C-terminal region of PDEγ constitute two distinct sites of interactions with α_t or PDE6 (4). The crystal structure of a ternary complex representing a partial model for the GTPase-activating protein (GAP) complex, consisting of a fragment of PDEγ (residues 46–87), α_t chimera (α_t/i1), and the catalytic core of RGS9, revealed that the C-terminal region of PDEγ contains three discontinuous helices (5). PDEγ_46–87 interacts with α_t in a GTP-dependent manner, mainly through residues in these helices or their linkers, and the indole side chain of W70 plays the key role in this interaction. Residue V66 of PDEγ interacts with RGS9, which further tightens the association between α_t and RGS9. The C-terminal region of PDEγ also interacts with the catalytic domain of PDE6; however, studies have suggested that this interaction involves the 11 C-terminal residues of PDEγ, a region distinct from the α_t interaction site (6).

It has been demonstrated that PDEγ belongs to the growing family of intrinsically disordered proteins (IDPs) (7, 8). Presumably, the inherent structural plasticity of an IDP, such as PDEγ, confers an advantage in its ability to interact with multiple partners (9). However, the intrinsic disorder raises the question of whether the structure of the interaction site preexists in the IDP or is induced upon binding. A surge of recent evidence has established that both denatured proteins and IDPs can contain residual structure and even adopt compact folds in solution (10–13). The residual structure of an IDP is believed to play a role in initiating the folding process, in protein–protein interactions, or in preventing aggregation (10, 14, 15). Despite the great challenge to structure determination imposed by the dynamic properties of an IDP, recently developed NMR methods, such as paramagnetic relaxation enhancement (PRE) (16) and residual dipolar coupling (RDC) (17), have enabled detailed elucidation of their transient conformations.

In this study we used NMR spectroscopy to determine the major conformational states of full-length PDEγ in solution. These conformations were calculated from a large set of long-range structural constraints generated by introducing a paramagnetic spin-label, 3-maleimido-PROXYL (mPROXYL), into each of 10 single cysteine-containing PDEγ variants. We also used NMR spectroscopy to investigate the interaction between the PDEγ C-terminal fragment (PDEγ_46–87) and the PDE5/6 chimeric catalytic domain, as a mimic of the PDE6 catalytic core (18). Our results demonstrate that the full-length PDEγ contains transiently populated structural elements that provide insight into the regulatory mechanism of PDEγ in the signaling cascade of phototransduction.

Results

PDEγ Is Intrinsically Disordered but Exhibits Transient Secondary Structure in Solution.

Amino acid analysis and biochemical studies of PDEγ had led to the conclusion that PDEγ is an IDP (7, 8). We found that the peaks in the 2D [¹H-¹⁵N]-HSQC NMR spectrum of PDEγ exhibited narrow linewidths with poorly dispersed amide proton chemical shifts [see supporting information (SI) Fig. 6a], consistent with structural disorder of PDEγ in solution. NMR spectra collected at pH values between 2.7 and 7.0 indicated that PDEγ remains disordered over a wide pH range (data not shown). In addition, as expected for a protein with low backbone rigidity, the [¹H-¹⁵N] heteronuclear NOEs observed for PDEγ were consistently <0.4 (Fig. 1a). Although the NOE values were low, they varied along the sequence as might be explained by partial residual secondary structure. The first 21 residues, with NOE values for every residue below −0.3, constituted the least structured region. By contrast, residues 64–81, with NOEs ranging from 0.1 to 0.38 and with higher ¹⁵N R₁ and R₂ relaxation rates than for residues elsewhere in the sequence (SI Fig. 7), constituted the most structured region of PDEγ.

Fig. 1.

Evidence that PDEγ contains regions of transient secondary structure in solution. (a) [¹H-¹⁵N] NOE of A68-PDEγ plotted as a function of residue number. (b) The ¹H^α secondary shifts and combined ¹³C^α and ¹³C^β secondary shifts (20) observed for PDEγ. The locations of the three helices of PDEγ_46–87 in the partial GAP complex are indicated at the top.

To evaluate the hydrodynamic property of PDEγ, we measured the translational diffusion coefficient (Ds) of PDEγ at various concentrations by using a water-suppressed longitudinal encode-decode (sLED) pulse-field-gradient experiment (19). The results showed that the Ds of PDEγ is concentration dependent (SI Fig. 6b). For example, when the concentration of PDEγ was lowered from 350 to 90 μM, the Ds of PDEγ increased from 0.92 × 10⁻⁶ cm²·s⁻¹ to 1.19 × 10⁻⁶ cm²·s⁻¹. The measured Ds did not change when the concentration of PDEγ was lowered from 90 to 45 μM. These results suggest that PDEγ exists as an oligomer at higher concentrations but is monomeric at <90 μM. Furthermore, comparison of the diffusion coefficient for monomeric PDEγ (10 kDa, Ds = 1.2 × 10⁻⁶ cm²·s⁻¹) with that for monomeric lysozyme (14 kDa, Ds = 1.11 × 10⁻⁶ cm²·s⁻¹) (21), suggested that the structure of PDEγ in solution is relatively compact, despite its dynamic state.

¹H^α, ¹³C^α, and ¹³C^β chemical shifts depend on local backbone geometry. Their deviations from random coil values, termed secondary shifts, provide sensitive probes to detect the secondary structural propensities of both folded and IDPs (22, 23). The proton (Δδ¹H^α) and combined carbon (Δδ¹³C^α–Δδ¹³C^β) secondary shifts of PDEγ (Fig. 1b) show a continuous stretch of negative proton secondary shifts (−0.2 to −0.1 ppm) and positive combined carbon secondary shifts (1–1.5 ppm) for C-terminal residues 68–84. Because a rigid helix typically yields proton secondary shifts in the range of −0.39 ppm (23) and combined carbon secondary shifts in the range of 2 ppm (24), this pattern suggests that residues 68–84 are ≈50% helical. The presence of residual structure in this region is further supported by the observation of helix-type sequential NOEs [e.g., d_αN(i, i + 3) and d_αN(i, i + 4)] (SI Fig. 8), [¹H-¹⁵N] NOEs (Fig. 1a), and ¹⁵N R₁, R₂ relaxation parameters (SI Fig. 7). Residues 51–67 exhibited positive combined carbon secondary shifts but proton secondary shifts that varied randomly between positive and negative; thus the results provide only marginal evidence for helix in this region. By contrast, the proton and combined carbon secondary shifts of residues 1–50 were much smaller and more evenly distributed between positive and negative values as expected for a random coil.

PRE Effects Detect Long-Range Transient Interactions in PDEγ.

Unlike [¹H-¹H] NOEs, which are limited to detecting pairs of protons located closer than 5 Å and are quenched by internal dynamics, PRE effects, which arise from proton relaxation by unpaired electrons, are observable at distances up to 20 Å and are not quenched by internal dynamics (16). With proteins like PDEγ that do not contain a paramagnetic center, the common strategy is to introduce a spin label by attaching it covalently to a free cysteine side chain. Wild-type PDEγ contains a single cysteine at position 68, which we labeled with mPROXYL. To collect additional PRE constraints, we mutated C68 to alanine and constructed a series of nine PDEγ variants by mutating selected hydrophobic residues to cysteine. These cysteine-containing PDEγ variants were then labeled with mPROXYL groups. Comparison of [¹H-¹⁵N]-HSQC NMR spectra of each of the reduced mPROXYL-labeled PDEγ variants with that of wild-type PDEγ indicated that introduction of the mPROXYL-cysteine in place of the original hydrophobic residue did not result in an appreciable conformational change (data not shown). To rule out possible intermolecular interactions, the concentrations of the mPROXYL-labeled PDEγ variants were limited to 50–90 μM. These concentrations appeared to be low enough because equivalent PRE effects were determined for two of the variants (V21C-mPROXYL and F50C-mPROXYL) at concentrations in this range (70–80 μM) and at a higher concentration (120 μM) (SI Fig. 9).

The experimental PREs, observed for all 10 PDEγ variants (Fig. 2) extended beyond those estimated for random coil-like structures from theoretical calculations. However, the PRE effects differed among the 10 spin-labeled sites. For example, in I10C-mPROXYL, no appreciable PRE effects were observed except for residues 68–71, which exhibited slight signal reductions. In V21C-mPROXYL and F30C-mPROXYL, modest interactions were observed between the spin-labeled sites and residues 50–87 and 48–74, respectively. By contrast, all mPROXYL derivatives located between residue 38 and the C terminus exhibited prominent PRE effects for almost all residues C-terminal to V21 but only slight PRE effects for residues located in the N-terminal tail. Because the PRE effects arising from each site were mutually consistent, there was no evidence for appreciable shifting of the conformational population as the result of introduction of a spin label at any of these sites. In accordance with the measured dynamic parameters, these results suggest that residues between 38 and the C terminus form a relatively compact tertiary core undergoing slow dynamics, which further forms a transient tertiary structure with residues 21–30, whereas N-terminal residues 1–20 are highly flexible and do not interact appreciably with the rest of the sequence.

Fig. 2.

PRE results for amide protons in spin-labeled PDEγ. Ten Cys-substituted and spin-labeled variants of PDEγ were analyzed. Dashed lines indicate paramagnetic effects expected for a random coil polypeptide (14).

Structure Calculations Reveal that PDEγ Assumes a Loose Tertiary Fold.

Input to the structure calculations consisted of 29 ϕ and ψ dihedral angle constraints, 651 PRE constraints, and 494 intraresidue, sequential, or medium-range NOE constraints (SI Table 1). No long-range (residues i − j >5) NOEs were observed. Structure calculations were performed on an extended conformation by using a simulated annealing protocol and torsion angle dynamics. The PRE effects reflect ensemble-averaged dipole–dipole interactions between two sites, whose proximity may be transient, and the structure calculations are biased by a r⁻⁶ weighting function. As a consequence, such calculations may generate structural features that do not coexist in a given molecule at a single point in time. To accommodate the conformational flexibility and take into consideration the large errors in the PRE distances, we introduced 5-Å upper and/or lower bounds to each PRE restraint enforced by a soft-square potential. This conformational freedom, combined with the randomization step of the molecular dynamics calculation, allows the generation of a large variety of structures whose distances for a given electron–proton pair fall into a statistical distribution within the confined space. In the generated structural pool, each individual structure may not fit the experimental PRE restraints well, but the representation of a large number of structures can reasonably back-calculate the PRE data. This loose interpretation of the restraints takes into account the intrinsic flexibility of the protein and the attached mPROXYL group and avoids the generation of overly compact structures. A similar semiquantitative treatment of restraints was used in structure calculations for α-synuclein, another IDP, and the procedure yielded a family of conformers that were in agreement with those derived by the independent approach of ensemble-based, restrained molecular dynamics simulation (10, 15).

Of 200 calculated conformers, the 100 structures with lowest energies were chosen for analysis. Upon superposition, 86% of these conformers fell into two major families, one represented by 56 conformers and the other by 30 conformers (Fig. 3). Both clusters contain a loose tertiary fold formed by the central polycationic domain (residues 24–45) and the C-terminal domain, surrounded by the highly flexible N-terminal tail. The two families differ primarily by the relative orientation of the two partially structured domains. In the more abundant (56 conformers) fold (Fig. 3a), PDEγ adopts a right-hand serpentine coil topology where the C-terminal domain spirals on top of the polycationic domain. In the less abundant (30 conformers) fold (Fig. 3b), PDEγ appears as a twisted antiparallel hairpin, with the C terminus further curled toward the turn portion of the protein (residues 42–49). In both clusters, the long-range order is driven by two kinds of interactions: electrostatic attraction between the central polycationic region and the acidic region near the C terminus (residues 52–80), and hydrophobic interaction within the C-terminal domain (residues 50–87), primarily mediated by residues F50, F73, L76, and I87. The mean Stokes' radii (Rs) for the two clusters as estimated by the HYDROPRO program (25) were 22.1 ± 0.7 Å (56 conformers) and 21.9 ± 0.5 Å (30 conformers). These values more closely match the experimental Rs of 22.2 Å determined by analytical ultracentrifugation (26) than the Rs of 28.2 Å expected for an 87-residue random coil (27).

Fig. 3.

Conformational ensembles representing the solution structure of native PDEγ. Shown in stereoview are the two most populated clusters represented by 56 conformers (a) and 30 conformers (b) from the 100 analyzed structures. The five conformers with lowest energy from each cluster are shown; they are colored from blue at the N terminus to red at the C terminus. Residues R24, K45, D52, and E81 are labeled to delineate the boundaries of two oppositely charged segments.

Comparison of the Structure with RDC Measurements.

RDC values provide important information on local and long-range orientations (28). The observation that the polycationic region and the C terminus of PDEγ superimpose as two autonomous domains was also reflected by our N-H RDC measurement (SI Fig. 10). In contrast to the overall bell-like distribution of RDCs expected for a random coil polypeptide chain (17, 29), the RDCs from PDEγ exhibited two distinct distributions: those from the N-terminal region (residues 1–45) showed consistently negative couplings, with a minimum at residues 32–33, whereas those from the C-terminal domain (residues 46–84) showed smaller couplings of mixed sign. Our RDC observations can be explained by assuming that the magnetic alignment tensor of PDEγ coincides with the long axis of the chain, as observed for many denatured proteins (29). The extended backbones of residues 1–45 statistically favor the β and polyproline II (PPII) conformations (30), whose backbone N-H vectors are oriented nearly perpendicular to the long axis. Thus, their RDCs are consistently negative (29). By contrast, the smaller N-H RDCs observed for the residues in the C-terminal domain are consistent with our model (SI Fig. 10), which indicates significant population of α-helix and chain compaction in this region (29).

Residues 74–87 of PDEγ_46–87 Interact Directly with the PDE5/6 Catalytic Domain.

To understand the molecular mechanism underlying the functional diversity of PDEγ and the activation kinetics of PDE6, we carried out an NMR study to determine whether PDEγ undergoes a conformational change upon interacting with the chimeric PDE5/6 catalytic domain. We have been unsuccessful in producing recombinant active PDE6 catalytic core. However, the chimeric PDE5/6 catalytic domain contains most of the PDE6 residues that interact with the C-terminal domain of PDEγ. In addition, PDEγ_63–87 inhibits PDE5/6 as potently as full-length PDEγ with a K_i of 1.7 μM, comparable with that for PDE6 from bovine retina (18, 31). Thus the PDE5/6 construct recapitulates the contacts between PDE6 and the PDEγ C-terminal domain and offers a model for the native PDEγ/PDE6 interaction. To simplify the spectral analysis, we used a C-terminal fragment of PDEγ (A68-PDEγ_46–87) to map its interaction with the PDE5/6 catalytic domain. Those residues of A68-PDEγ_46–87 making contacts with PDE5/6 are expected to exhibit the largest NMR signal broadenings and chemical shift perturbations. The NMR cross-peaks of free A68-PDEγ_46–87 are well resolved, with narrow linewidths, characteristic of a small flexible peptide (Fig. 4). Upon binding to the PDE5/6 catalytic domain, the peaks are collectively broadened because of the slower molecular tumbling rate. The most severe signal attenuation was detected for residues 74–87. These residues did not exhibit observable signals in the NMR spectrum, suggesting that they directly contact PDE5/6. Only moderate signal attenuation and slightly perturbed chemical shifts were observed for residues 70–73, which likely arose from a secondary or indirect binding effect. Although the backbone amide peak of W70 was broadened, the indole group exhibited relatively high signal intensity with a negligible chemical shift change. NMR signals from the remainder of sequence (residues 46–69) remained reasonably strong, and no appreciable chemical shift changes were observed, which indicates that these residues do not interact with PDE5/6.

Fig. 4.

Overlaid 2D [¹H-¹⁵N]-HSQC spectra highlight the spectral changes of A68-PDEγ_46–87 upon binding to the PDE5/6 catalytic domain. Assigned peaks from free A68-PDEγ_46–87 are labeled in red, and those for the complex form are labeled in black. The side-chain imino group of W70 and side-chain amide group of Q83 are labeled with the respective residue name followed by a superscript ε.

Discussion

The crystal structure of PDEγ_46–87 in the partial GAP complex (5) revealed three discontinuous helical elements: α1 (residues 62–65), α2 (residues 69–73), and α3 (residues 78–83). In comparison with our chemical-shift analysis, these helices, in particular α2 and α3, match well with the helical propensities of full-length PDEγ alone (Fig. 1b). It has been observed that many IDPs exhibit regions of residual structure, which are believed to function as primary interaction sites (14). In support of this hypothesis, our study indicates that two of the functional secondary structural elements of PDEγ are partially formed in the uncomplexed protein.

The central polycationic domain and the C-terminal domain of PDEγ constitute two discrete sites of interaction with α_t or PDEαβ catalytic heterodimer (32). When bound to PDEαβ, the polycationic domain interacts with the PDEαβ regulatory GAF domains, whereas the C-terminal domain blocks the catalytic sites (33–35). The whole PDEγ molecule likely assumes an extended conformation (33, 36, 37). Thus, the transient interaction between the polycationic site and the C-terminal region in the native PDEγ may provide an internal control switch that protects the two domains intramolecularly in the absence of intermolecular protein–protein interactions. In addition, such intramolecular electrostatic communication within PDEγ may facilitate the complete sequestering of PDEγ from PDEαβ in the late stage of visual signaling (38).

A similar interaction between two oppositely charged regions was detected in the structure of α-synuclein (10, 15). In this case the interaction is believed to help prevent α-synuclein from intermolecular aggregation (10, 15).

In the crystal structure of the partial GAP complex (5) (Fig. 5a) PDEγ_46–87 is observed to form a bent structure with residues 50 and 87 close to each other and the three discontinuous helices arrayed perpendicular to each other. In the solution structure of PDEγ described here, this region is also highly defined (Fig. 3). Using an 8-Å cutoff, the conformations of the corresponding region in 96% of the analyzed structures can be grouped into one cluster. Comparison of the equivalent regions between native PDEγ (Fig. 5b) and PDEγ_46–87 in the partial GAP complex further revealed a significant conformational resemblance. First, the helical structures observed for the transition-state complexed PDEγ_46–87 match well with the transient helical structures observed for native PDEγ. Second, their backbones are also well superimposed. In particular, the average backbone rmsd for residues 62–84 is 4.60 Å between the native and complexed PDEγ, comparable with the pairwise rmsd of 4.66 Å within the structural cluster of the native PDEγ.

Fig. 5.

The conformation of PDEγ favorable for binding α_t is populated in solution. (a) The structure of PDEγ (blue ribbon) in complex with α_t/i1 and RGS9 (surface representation) (5). (b) Superposition of the solution structure of native PDEγ with that in the partial GAP complex.

An IDP generally undergoes a disorder-to-order transition upon interaction with its functional partner (39). The detailed mechanism underlying this structural transition remains poorly understood. Several models have been proposed to describe the process, including binding coupled folding/induced fit (40) and preexisting equilibrium/conformational selection (41). The binding coupled folding/induced fit model emphasizes that the folding of an unstructured protein is induced by the molecular interaction, whereas the preexisting equilibrium/conformational selection model suggests that in the native state of a protein the functional (folded) conformation preexists in a dynamic equilibrium with other closely related conformations, and the binding shifts the equilibrium toward the functional conformation. In this study, the structural similarity between transition-state complexed PDEγ_46–87 and the equivalent region in native PDEγ indicates that the α_t binding conformation of this region is already encoded in the native structure of PDEγ. Indeed, chemical-shift analysis provides independent evidence that the structural propensity of residues 51–73 in free PDEγ is stabilized in the PDEγ:PDE6 complex. This finding suggests that the conformational selection mechanism holds both for the interaction between free PDEγ and α_t and between PDE6-bound PDEγ and α_t. In other words, the conformational states of PDEγ (residues 50–87) suitable for binding α_t and RGS9 are dynamically populated in native PDEγ, providing a molecule pool for binding the α_t/i1:RGS9 complex directly. Because a single structure is observed in the complex, it appears likely that induced fitting of the PDEγ tertiary structure is involved in stabilizing the GAP complex. Future structural investigation on the interactions between the α_t/i1:RGS9 complex and different regions of PDEγ will be needed to fully reveal the functional significance of this dynamic conformation.

This study is consistent with previous biochemical studies (6, 18, 31, 36) suggesting that residues 75–87 of PDEγ are responsible for its interaction with the PDE6 catalytic core. In particular, our result that the side chain of W70 is not involved in forming the A68-PDEγ_46–87:PDE5/6 complex is consistent with the finding that mutation of W70 did not affect its affinity for the PDE6 catalytic core, although it completely abolished PDEγ interaction with α_t (42). In addition, the minimal chemical shift perturbations observed for residues 46–73 of PDEγ in the A68-PDEγ_46–87:PDE5/6 complex suggest that their structure is preserved in the complex, which may further provide ready availability of these residues for binding α_t during PDE6 activation and subsequent GAP complex formation.

In conclusion, our study has shown that transient secondary and tertiary structural elements present in free PDEγ provide a mechanism for the functional specificity and controlled internal switching of this protein. In addition, these structural features are partly preserved in the PDE6:PDEγ complex in which the exposure of W70 facilitates efficient recognition by α_t in the ensuing signaling cascade of photoactivation. Our observation of transient population in PDEγ of secondary and tertiary structural elements required for formation of the ternary complex with α_t and RGS9 clearly demonstrates that this IDP uses a conformational selection mechanism in adapting to this complex.

Methods

Protein Production.

Wild-type PDEγ contains a single cysteine at position 68. To introduce a spin label to various other sites of PDEγ, single-cysteine mutations were constructed at sites 10, 21, 30, 38, 50, 60, 73, 76, and 87 on the A68-PDEγ background by using the Strategene QuikChange method (37). [U-¹⁵N], [U-¹³C,¹⁵N], or [U-²H,¹³C,¹⁵N]-labeled PDEγ variants were expressed in Escherichia coli and purified by using a chitin column followed by RP-HPLC, as described (7). Approximately 99% pure [U-¹⁵N]PDEγ was obtained as estimated by SDS/PAGE. [U-¹³C,¹⁵N]A68-PDEγ_46–87 and [U-²H,¹³C,¹⁵N]A68-PDEγ_46–87 were prepared through trypsin digestion of the full-length labeled A68-PDEγ molecules. Typically, 1 mg of PDEγ was digested with 12 μg trypsin (Promega) for 10–12 h at room temperature, in 20 mM Tris buffer (pH 8.0) containing 100 mM NaCl. A68-PDEγ_46–87 was then purified by RP-HPLC.

A typical spin-labeling reaction included 40 mM Na₂HPO₄ of pH 6.8, 200 mM NaCl, 50 μM PDEγ, and ≈1 mM mPROXYL. The reaction was allowed to proceed at room temperature for 30–60 min under argon, and then quenched by addition of 0.1% trifluoroacetic acid (TFA). The reactions were >90% complete, as judged from HPLC chromatograms. The unreacted PDEγ was then removed by RP-HPLC by using a POROS column. A gradient of 0.7% acetonitrile per minute in the presence of 0.1% TFA was applied. The purity and correct masses of various mPROXYL-PDEγ derivatives were confirmed by electrospray ionization MS.

The NMR samples for structural determination of PDEγ contained 50–90 μM [U-¹⁵N]/[U-¹³C,¹⁵N]-labeled A68-PDEγ or mPROXYL-PDEγ derivatives, 90% H₂O/10% D₂O at pH 4.0. RDCs were measured for wild-type PDEγ aligned in 14 mg/ml of bacteriophage Pf1 (Alsa), 20 mM MOPS, and 200 mM NaCl, pH 6.5. To determine the interaction between PDEγ and PDE5/6 catalytic domain, unlabeled PDE5/6 was prepared as described (18), mixed with [U-²H,¹³C,¹⁵N]-labeled PDEγ_46–87, and concentrated by using a YM-3 Centricon filter (Millipore). The final NMR sample contained 0.7 mM A68-PDEγ_46–87 and the PDE5/6 catalytic domain in a molar ratio of 1:1, dissolved in 20 mM MOPS, 200 mM d₂-glycine, 50 mM NaCl, 5 mM MgSO_4, 1.0 mM DTT, and 90% H₂O/10% D₂O, pH 7.0. No appreciable protein aggregation was observed for the complex under this condition. In addition, 0.1–0.2 mM [U-¹³C,¹⁵N]-labeled PDEγ_46–87, dissolved in 20 mM MOPS, 50 mM NaCl, and 90% H₂O/10% D₂O at pH 7.0, was also prepared for chemical-shift assignment of PDEγ_46–87.

NMR Spectroscopy.

All NMR spectra were recorded at the National Magnetic Resonance Facility at Madison, WI (NMRFAM) by using 600-MHz Varian Inova and 750-MHz Bruker spectrometers equipped with ¹H, ¹⁵N, and ¹³C triple-resonance cryogenic probes. The sample temperature was controlled at 25°C. A suite of 3D multinuclear NMR experiments, including HNCACB, CBCA(CO)NH, HNCO, HBHA(CO)NH, C(CO)NH, HC(CO)NH, and HCCH-TOCSY were collected for A68-PDEγ for sequential backbone and nonaromatic side-chain assignments. 3D ¹⁵N-edited NOESY-HSQC (τ_mix = 100 ms) and 3D ¹³C-edited NOESY-HSQC (τ_mix = 100 ms) spectra were collected for NOE observations. In addition, a 3D HNHA spectrum was collected to measure the ³J_HNHA scalar couplings. 2D [¹H-¹⁵N]-HSQC spectra were acquired to measure the PRE effects. Data were first collected from each sample containing a nitroxide free radical. Subsequently, 0.5 mM ascorbic acid was added at room temperature to reduce the paramagnetic spin label, and after a period of 5–6 h, spectral data were collected for the reduced sample. To ensure complete reduction of the spin label and accurate measurement of the NMR signals, an additional HSQC spectrum was then acquired for signal comparison. Steady-state [¹H-¹⁵N] NOE and ¹⁵N relaxation (R₁, R₂) data were collected by standard pulse sequences (43). The relaxation rates were calculated by least-squares fitting of peak heights versus relaxation delay to a single exponential decay. Steady-state [¹H-¹⁵N] NOE values were calculated from the ratio of peak heights in a pair of NMR spectra acquired with and without proton saturation. The signal-to-noise ratio in each spectrum was used to estimate the experimental uncertainty. One-bond N-H RDCs were determined by using the in-phase anti-phase [¹H-¹⁵N]-HSQC sequence (44). The translational diffusion coefficient of PDEγ in aqueous solution was determined by a water-sLED pulse-field-gradient experiment (19). The gradient was calibrated by reference to data for 10 mg/ml lysozyme, which has a diffusion coefficient of 1.11 × 10⁻⁶ cm²·s (21). In studying the interaction between A68-PDEγ_46–87 and the PDE5/6 catalytic domain, 3D HNCACB and CBCACONH spectra were collected for free A68-PDEγ_46–87, and a 3D HNCA spectrum was acquired for A68-PDEγ_46–87 in the complex. The spectra were processed and analyzed, respectively, with the NMRPipe (45) and Sparky (www.cgl.ucsf.edu/home/sparky) software packages.

Structural Constraints.

NOE constraints were derived from a 3D ¹⁵N-edited NOESY-HSQC spectrum and a 3D ¹³C-edited NOESY-HSQC spectrum. Backbone ϕ and ψ angles were derived from TALOS-based analysis of backbone chemical shifts (46) and analysis of the 3D HNHA spectrum. The PRE restraints were derived as described from the intensity (heights) ratios between two HSQC spectra acquired in the presence and absence of the nitroxide radical (47). The paramagnetic contribution to the transverse relaxation rate (R₂^sp) was calculated from:

where I_ox and I_red are the peak intensities (heights) of oxidized and reduced resonances, respectively. t is the total INEPT (insensitive nuclei enhanced by polarization transfer) evolution time of the HSQC (≈10 ms). R₂, the intrinsic transverse relaxation rate, was estimated from the peak line-width in the reduced spectra. Then the distance between the spin label and the amide group, r, was calculated from:

where K is 1.23 × 10⁻³² cm⁶·s⁻² for the interaction between a free electron and a proton. ω_h is the Larmor frequency of a proton. τ_c, the correlation time for the electron-nuclear dipole-dipole interaction, was set to 3.8 ns (17). To avoid overly constrained structure, an intensity ratio of 0.85 was used in classifying the PRE restraints. All peaks with an intensity ratio >0.85 were assigned only a lower bound; this bound corresponded to that calculated from intensity ratio = 0.85 minus 5 Å. Peaks detected in oxidized spectra with intensity ratios <0.85 were assigned upper bounds equal to the target distance plus 5 Å and lower bounds equal to the target distance minus 5 Å. Peaks not detectable in oxidized spectra because of severe relaxation enhancement were assigned an upper bound only; this bound corresponded to the target distance calculated from the noise level in the oxidized spectrum plus 5 Å.

Structure Calculation.

Structure calculations were performed with the X-PLOR NIH software package (48). The coordinates of mPROXYL-cysteine was incorporated into the structural model to allow for accurate representation of the PRE distance. The topology and energy parameters for the mPROXYL-cysteine group were generated with the PRODRG program (49) and then modified to be compatible with the existing libraries. The calculation started with 1,000 cycles of torsion angle molecular dynamics (TAD) at 50,000 K, followed by 1,000 cycles of TAD with increasing values of interatomic repulsion while cooling to 1,000 K, and subsequent 2,000 cycles of molecular dynamics in Cartesian space while cooling to 300 K and 1,000 cycles of final energy minimization. A soft square energy potential was used for distance and dihedral restraints. Of 200 calculated properly geometrized structures, the 100 with the lowest energies were analyzed. The final structures were clustered based on residues 10–87 by using the “cluster_struc” module in the HADDOCK program (50) with 10-Å cutoff distance.