N-terminal Half of the cGMP Phosphodiesterase γ-Subunit Contributes to Stabilization of the GTPase-accelerating Protein Complex*

In the visual signal terminating transition state, the cyclic GMP phosphodiesterase (PDE6) inhibitory γ-subunit (PDEγ) stimulates GTPase activity of the α-subunit of transducin (αt) by enhancing the interaction between αt and its regulator of G protein signaling (RGS9), which is constitutively bound to the type 5 G protein β-subunit (β5). Although it is known from a crystal structure of partial molecules that the PDEγ C terminus contacts with both αt and RGS9, contributions from the intrinsically disordered PDEγ N-terminal half remain unclear. In this study, we were able to investigate this issue using a photolabel transfer strategy that allows for mapping the interface of full-length proteins. We observed label transfer from PDEγ N-terminal positions 50, 30, and 16 to RGS9·β5 in the GTPase-accelerating protein (GAP) complex composed of PDEγ·αt·RGS9·β5. In support of a direct PDEγ N-terminal interaction with RGS9·β5, the PDEγ N-terminal peptide PDEγ(1–61) abolished label transfer to RGS9·β5, and another N-terminal peptide, PDEγ(10–30), disassembled the GAP complex in label transfer and pulldown experiments. Furthermore, we determined that the PDEγ C-terminal interaction with αt was enhanced whereas the N-terminal interaction was weakened upon changing the αt conformation from the signaling state to the transition state. This “rearrangement” of PDEγ domain interactions with αt appears to facilitate the interaction of the PDEγ N-terminal half with RGS9·β5 and hence its contribution to optimal stabilization of the GAP complex.

In vertebrate photoreceptor cells, interactions of the cyclic GMP (cGMP) phosphodiesterase (PDE6) 2 inhibitory ␥-subunit (PDE␥) with its targets switch on and off visual signal transduc-tion (for review, see Refs. [1][2][3][4]. The signaling is turned on when the GTP-bound ␣-subunit of transducin (␣t), which is converted from the GDP-bound conformation by light-excited rhodopsin, activates PDE6 by interacting with PDE␥ and displacing its C terminus from the PDE6 catalytic pocket (signaling state). Lowered cGMP levels then cause hyperpolarization of the plasma membrane by closing cGMP-gated channels, thus leading to signal transmission to the brain. Concomitantly, the GTPase activity of ␣t is accelerated to hydrolyze the bound GTP into GDP, via simultaneous ␣t interactions with PDE␥ and the regulator of G protein signaling (RGS9) constitutively bound with the type 5 G protein ␤-subunit (␤5) (transition state). Once the conformation of ␣t reverts back to the GDPbound inactive form, PDE␥ dissociates from ␣t and reinhibits PDE6. Visual transduction is thus turned off, and the signaling proteins are primed for the next round of the photoresponse.
Throughout this report, the two conformers of ␣t in the signaling state and transition state are represented respectively by ␣t-GTP␥S (5) and ␣t-GDP-AlF 4 Ϫ (6). The four-component (␣t⅐PDE␥⅐RGS9⅐␤5) transition state complex is also referred to as the GAP (GTPase-accelerating protein) complex (2).
Our knowledge regarding the protein functions and interactions within the GAP complex has been learned largely from molecular and biochemical studies, in particular, the crystal structure of a partial GAP complex (7) that includes the GDP-AlF 4 Ϫ -bound ␣t/i1 chimera, the C-terminal half of PDE␥ (residues 46 -87), and the RGS9 catalytic core (RGS9d). RGS9d interacts intimately with the ␣t switch regions thus stabilizing its GTP hydrolysis conformation. The other domains of RGS9 together with ␤5 discriminate between the activated ␣t and the ␣t⅐PDE␥ complex (8,9). PDE␥ enhances the affinity of RGS9⅐␤5 with ␣t by interacting with both simultaneously (10). PDE␥ is a small protein of 87 amino acids containing two major functional domains, the polycationic N-terminal domain (Gly 19 -Gly 49 ) and the hydrophobic C-terminal domain (Thr 62 -Ile 87 ) (3,4). Both domains interact with ␣t as well as with PDE␣␤ (11)(12)(13)(14) and likely in a complementary manner in the signaling state with the C-terminal domain favoring ␣t whereas the central domain binding more tightly with PDE␣␤ (13).
Although the partial GAP structure (7) has revealed critical atomic details for the transition state complex, our information regarding the interactions of PDE␥ with its partners is still substantially missing. This is because the current structure contains only the C-terminal half of PDE␥ and less than one-third of the RGS9 sequence with no ␤5 bound. Even though the recent crystal structure of the full-length RGS9⅐␤5 (15) has greatly complemented our knowledge, important questions remain unanswered. One prominent question is in regard to the role of the PDE␥ N-terminal half (hereafter defined as residues 1-50, as opposed to the C-terminal half residues 50 -87 that are resolved in the partial GAP structure (7)). Although the PDE␥ C-terminal peptide PDE␥(63-87) is reported to be fully potent in stimulating the GAP activity (10), the full-length PDE␥ is needed for the high affinity GAP complex formation (8,10). This implicates a necessary role for the PDE␥ N-terminal portion. Additionally, previous studies indicate that practically all of the RGS9 domains as well as ␤5 participate in regulation of the PDE␥-stimulated GAP function (8,9,16). In particular, the noncatalytic constituent RGS9 domains and the N terminus of ␤5 play a decisive role in the RGS substrate selectivity toward the PDE␥ complex with activated ␣t (16). In the partial GAP structure (7), however, only one direct contact between PDE␥ and RGS9d has been discovered. It is therefore important to explore whether the PDE␥ N-terminal half, which is absent in the existing crystal structures (7,15,17), also plays a role in the GAP complex by interacting with RGS9⅐␤5.
In addition, with regard to the PDE␥ interactions with ␣t and RGS9⅐␤5 in the GAP complex, it is not known if PDE␥ interacts with the transition state conformation in the same manner as with the signaling state conformation. Thus far there have been no atomic structures available for the full-length PDE␥ in complex with either ␣t-GDP-AlF 4 Ϫ or ␣t-GTP␥S. Although the crystal structure of the partial transition state complex (7) is often used to infer the signaling state PDE␥⅐␣t interaction, it is not clear whether the PDE␥ interactions with two different ␣t conformers are the same.
Answers to these questions are essential for a clear understanding of the PDE␥ regulation of the photoresponse through dynamic interactions with its partners. However, the intrinsically disordered PDE␥, especially its N-terminal half (18), has been problematic in solving a crystal structure of ␣t and RGS9⅐␤5 bound with the full-length PDE␥. As indicated in our previous reports (13,19,20), the label transfer approach has allowed for systematic mapping of interactions between fulllength molecules, thus circumventing the problems of intrinsic disorder that can impede efforts in solving atomic structures. We have generated PDE␥ photoprobes by derivatization at single-cysteine positions throughout the entire PDE␥ molecule and used them in this study to investigate the interactions of PDE␥ with full-length RGS9⅐␤5 and ␣t in the GAP complex. We have observed an interaction between the N-terminal half of PDE␥ and RGS9⅐␤5, and differential PDE␥ interactions with two ␣t conformers that rationalize the observed PDE␥ N-terminal interaction with RGS9⅐␤5. Our findings afford new insights into the structure and regulation of the GAP complex.
Preparation of Holotransducin, ␣t-GDP, and ␣t-GTP␥S-Using frozen dark-adapted bovine retinas (J. A. & W. L. Lawson Co.), rod outer segment membranes were isolated, from which holotransducin was prepared as described previously (21). ␣t-GDP and ␤␥t were then purified from holotransducin using a Blue-Sepharose CL-6B column. To prepare ␣t-GTP␥S, GTP␥S was added to rod outer segment membranes, ␣t-GTP␥S was thus released and purified on a Blue-Sepharose CL-6B column. The purity of ␣t was determined to be Ͼ95% by SDS-PAGE and Coomassie staining. Correct conformations of ␣t-GTP␥S and ␣t-GDP-AlF 4 Ϫ (prepared using ␣t-GDP) were confirmed by their resistance to trypsin digestion (22-24) (see supplemental Fig. S1).
Preparation of the PDE␥ Photoprobes-The constructs for expressing the full-length wild-type (WT) PDE␥ (single cysteine at position 68) and the PDE␥ N-terminal peptide (PDE␥(1-61)) were reported previously (21). Single-cysteine PDE␥ mutants were generated using the Strategene QuikChange method (19). These PDE␥ variants were expressed in Escherichia coli and purified by chitin beads followed by reverse-phase HPLC using POROS 20 R2 resin (24). Greater than 95% pure PDE␥ was used for the preparation of PDE␥ photoprobes.
Functional Assay of the PDE␥ Photoprobes-Transducin GTPase activity was determined by using a single-turnover technique as described previously (25). The assays were conducted at room temperature (22-24°C) in a buffer containing 25 mM Tris-HCl (pH 8.0), 140 mM NaCl, and 8 mM MgCl 2 . The urea-treated rod outer segment membranes, lacking endogenous activity of RGS9, were used as a source for the photoexcited rhodopsin required for transducin activation. The reactions were initiated by the addition of 10 l of 0.6 M [ 32 P]GTP (ϳ10 5 dpm/sample) to 20 l of urea-treated rod outer segment membranes (20 M final rhodopsin concentration) reconsti-tuted with transducin heterotrimer (1 M) and recombinant RGS9⅐G␤5 complex (0.5 M). The reactions were performed in either the absence or presence of PDE␥ derivatives (1 M) and terminated by the addition of 100 l of 6% perchloric acid followed by measurement of the 32 P formation. The assays were conducted in the absence of reducing agents due to the presence of the disulfide linkage between the photoreactive group and PDE␥.
The assay of ␣t GTPase stimulation by BBM-PDE␥ photoprobes indicated that the functional activities of the tested photoprobes were similar to WT-PDE␥ (supplemental Fig. S2B). Furthermore, a native gel assay showed that interactions of the PDE␥ photoprobes with ␣t-GDP-AlF 4 Ϫ were not markedly affected (supplemental Fig. S2C).
Photocross-linking/Label Transfer Using PDE␥ Photoprobes-For the strategy of label transfer, please refer to the legend of supplemental Fig. S2A and our previous reports (13,26). Photocross-linking reactions containing PDE␥ photoprobes and target proteins at desired concentrations (for detailed conditions refer to corresponding figure legends) were dark-incubated on ice and then subjected to UV light. The reactions with BBM photoprobes were photolyzed at 5-10°C for 2 ϫ 5 min (with a 5-min dark interval on ice), in an RPR-100 Rayonet Photochemical Reactor equipped with 18 light bulbs of 350 nm (Southern New England Ultraviolet Company). The reactions using [ 125 I]ACTP-PDE␥ photoprobes were exposed to the UV light generated by an AH-6 water-jacketed 1000-watt high pressure mercury lamp for 5 s at a distance of 10 cm (19). Following photolysis, sample buffer was added immediately to the final concentrations of 1% SDS and 50 mM DTT. Proteins were then separated by SDS-PAGE. The gels were either used for Coomassie staining and autoradiography or for electrotransfer to PVDF membranes and subsequent far-Western detection of biotin label transfer. Autoradiography and far-Western blotting were performed as described previously (19,20).
Pulldown Protein Binding Assay-Pulldown by PDE␥ was performed using Ultra-Link Plus immobilized streptavidin gel (Pierce Biotechnology; hereafter termed streptavidin beads), following the previously reported method (13) with minor modifications. Biotinylated full-length PDE␥ (Btn-PDE␥) or the PDE␥ C-terminal half (PDE␥(46 -87)-Btn) were prepared as described previously (13). For detailed experimental conditions please refer to the figure legends. Generally, prior to the experiments, streptavidin beads were equilibrated for 15 min at room temperature with the pulldown buffer indicated in a given experiment, in which a high concentration (0.5-1 g/l) of BSA was included to block possible nonspecific protein-bead interactions. The beads were then washed twice with 500 l of buffer and were ready for use. In the pulldown experiments using Ni-NTA beads, the beads were also preincubated with BSA to eliminate possible protein-bead interactions. Incubation of the beads with the pulldown reactions was performed at 4°C by rotating the microcentrifuge tubes. At the end of the pulldown experiments, proteins on the beads were eluted with the SDS/DTT-containing sample buffer at 80 -90°C for 5 min, resolved on a low cross-link 15% acrylamide gel (27), and then immunodetected by Western blotting using enhanced chemiluminescence.

RESULTS
Label Transfer from PDE␥ to RGS9⅐␤5 in the Transition State GAP Complex-To map the entire PDE␥⅐RGS9⅐␤5 interaction interface in the GAP complex that is composed of the fulllength molecules, photocross-linking experiments were performed using 13 PDE␥ photoprobes prepared with BBM at positions throughout the PDE␥ molecule. The BBM photolabel transfer to RGS9⅐␤5 was then detected by far-Western blotting after DTT reversal. As shown in Fig. 1, A and B, prominent biotin label transfer onto RGS9 can be observed from PDE␥ position 68, and interestingly, also from the PDE␥ positions N-terminal to 68, such as 60, 50, 30, and 16. In contrast, there was no obvious label transfer observed from the positions C-terminal to 68. It is known that Val 66 , which is close to Cys 68 , makes direct contact with Trp 362 in the RGS domain, as exhibited by the partial GAP structure that includes only the C-terminal half of PDE␥ (7). Remarkably, from PDE␥ N-terminal positions 30 and 16, label transfer was found on ␤5 as well (Fig.  1, A and B). Although the protein amount in each lane is approximately equal (Fig. 1A, lower panel), the labeling intensity of RGS9⅐␤5 is PDE␥ position-dependent, manifesting the specificity of the observed label transfer.
Supporting the GAP-specific nature of the observed label transfer from PDE␥ to RGS9⅐␤5 (Fig. 1, A and B), prominent label transfer from positions 68, 60, 50, and 30 could be detected in the presence of ␣t-GDP-AlF 4 Ϫ , but not in the GDPbound inactive conformation (Fig. 1C) or the GTP␥S-bound form (Fig. 1D).
To answer the question of whether the label transfer to RGS9⅐␤5 resulted from protein-protein interaction or was just due to close proximity of the PDE␥ probe with RGS9⅐␤5, label transfer experiments were performed using PDE␥ peptides as competitors of the PDE␥ probes ( PDE␥ N-terminal Peptide PDE␥ (10 -30) Disrupts the RGS9 Pulldown from the Transition State GAP Complex-To confirm further a direct PDE␥ N-terminal interaction with RGS9⅐␤5 in the GAP complex that was determined from the label transfer approach, we performed pulldown experiments using Btn-PDE␥ and streptavidin beads (Fig. 3). The GAP complex was first reconstituted by incubating Btn-PDE␥ with RGS9⅐␤5 and ␣t-GDP-AlF 4 Ϫ and then captured on the streptavidin beads followed by Western blotting analyses of the co-precipitated proteins. As can be seen in . This indicates that PDE␥ (10 -30), which is part of the PDE␥ N-terminal sequence, competed with the N-terminal interaction of the full-length Btn-PDE␥ with RGS9⅐␤5, further supporting direct interactions between RGS9⅐␤5 and the PDE␥ N-terminal positions that were observed from the label transfer ( Fig. 1). In contrast to PDE␥(10 -30), the peptide PDE␥ (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26), which has similar positive charge properties, however, did not reduce the level of RGS9 pulldown at all (lane 4). This result argues against the possibility that PDE␥(10 -30) eliminated RGS9 pulldown (lane 3) nonspecifically by a charge effect, and it also supports the sequence specificity of the PDE␥(10 -30) result. A consistent result was also obtained in the label transfer experiments (Fig. 2, lanes 7-9). Because PDE␥(10 -30) contains two of the positions (16 and 30) on PDE␥ that were shown to interact with RGS9 as well as ␤5 (Fig. 1, A and B), it is likely to be much more potent than PDE␥ (15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26) in competing with the full-length PDE␥ interaction.
An alternative explanation for the PDE␥(10 -30) effect is that in the GAP complex PDE␥(10 -30) may have competed with the interaction between PDE␥ and ␣t and thus indirectly removed the RGS9 that was bound to ␣t. The data in Fig. 3E, however, shows only a slight decrease of ␣t-GDP-AlF 4 Ϫ in the presence of 1500-fold of PDE␥(10 -30), whereas there was nearly complete loss of RGS9 pulldown (Fig. 3D). Considering that Btn-PDE␥ was in a molar excess (25 mol of PDE␥ versus 1 mol of RGS9⅐␤5 dimer) and ␣t was also in excess (2.5-fold over the RGS9⅐␤5 heterodimer), only a portion of ␣t was bound in the GAP complex whereas the majority of ␣t molecules were In the experiments of A, B, and C, to focus on profiling label transfer from PDE␥ to RGS9⅐␤5, Ni-NTA beads were used to "fish out" RGS9⅐␤5 from the photocross-linked reaction mixture. This strategy allowed us to retain His-tagged RGS9⅐␤5 on the beads and detect the biotin label transfer by far-Western blotting with a relatively clean background. During vigorous washing of the Ni-NTA beads, some ␣t may have been lost, making quantification of label transfer to ␣t and "within experiment" comparison with the label transfer to RGS9⅐␤5 inappropriate. For this reason, the ␣t band is not shown in A, B, and C. The experiments in D were performed without using Ni-NTA beads, and biotin label was shown to be transferred to both RGS9 and ␣t from PDE␥ position 68 (lane 2, left panel). Please note, however, that the ␣t band does not necessarily represent only the label transfer to the ␣t population that was in the GAP complex. There is a possibility that the label transfer to the ␣t population that was not in the GAP complex also mixed into the ␣t band on the blot, considering that the interaction of PDE␥ with ␣t-GDP-AlF 4 Ϫ alone (without RGS9⅐␤5) is strong (33). A, far-Western blot showing the biotin label transfer from BBM-PDE␥ photoprobes to RGS9⅐␤5 (upper panel) in the presence of ␣t-GDP-AlF 4 Ϫ . The same blot was stained with Amido Black to reveal equal amounts of RGS9⅐␤5 proteins used in each reaction (lower panel). Alignment of the upper and lower panels indicates that the far-Western signal below the RGS9 band at position 30 was from ␤5, rather than from the asterisk-marked band, which is a commonly observed C-terminally truncated RGS9 species due to proteolysis (49).  bound to Btn-PDE␥ with no RGS9⅐␤5 bound. Therefore, the observed reduction of ␣t pulldown in the presence of PDE␥(10 -30) (Fig. 3E, lane 3) was most likely due to the depletion of RGS9⅐␤5 to which ␣t was bound (Fig. 3D, lane 3). This further supports the conclusion that PDE␥(10 -30) competed with the Btn-PDE␥ N-terminal interaction with RGS9⅐␤5 rather than that with ␣t-GDP-AlF 4 Ϫ . Indeed, in the absence of RGS9⅐␤5, PDE␥(10 -30) in 1500-fold molar excess did not reduce ␣t pulldown (Fig. 3F). This result is also consistent with a previous mutational study indicating that the three lysine residues (Lys 41 , Lys 44 , and Lys 45 ) outside the Ile 10 -Phe 30 region were mainly responsible for the PDE␥ N-terminal interaction with activated ␣t (28). Figs. 1-3), it is important to explore further whether the PDE␥⅐␣t interaction is dynamically adjusted upon a change from the signaling state with PDE6 being a partner (13,14), to the transition state when RGS9⅐␤5 becomes a target for PDE␥ (8,10).

Differential PDE␥ Domain Interactions with the ␣t Conformers of the Transition State and Signaling State-Based on the new observation of a PDE␥ N-terminal interaction with RGS9⅐␤5 (
We asked the question whether PDE␥ interacts with two conformers of ␣t, ␣t-GTP␥S and ␣t-GDP-AlF 4 Ϫ differentially. We compared label transfer from the same PDE␥ position to the two conformers of ␣t side by side on the same gel under exactly the same conditions using [ 125 I]ACTP-PDE␥ photoprobes (19) (Fig. 4). With this side-by-side comparison on the same gel we were able to eliminate variations caused by different gels, etc.
As indicated in our previous study (13), these [ 125 I]ACTP-PDE␥ photoprobes did not have a major impact on the PDE␥⅐␣t interaction and photolabeled ␣t in a specific manner. Consistently, in this study, although radiolabel was transferred from PDE␥ to ␣t, both of the internal controls BSA and ␤t were not labeled (Fig. 4A), indicating the specificity of the label transfer.
Interestingly, as indicated by the quantitative results presented in Fig. 4B, compared with the photolabel transfer to ␣t-GTP␥S, label transfer to ␣t-GDP-AlF 4 Ϫ from the PDE␥ N-terminal positions 21, 30, 50, and 60 decreased, whereas label transfer from the C-terminal positions 70 -87 increased. Although the difference in label transfer to the two ␣t conformers appeared less remarkable from position 40 or position 70, statistical analysis indicated a very significant overall difference between the PDE␥ N-terminal 21-60 region and the C-terminal 70 -87 region in terms of a change of label transfer to the two ␣t conformers. In other words, upon a change from the signaling state conformation to the transition state conformation of ␣t, the PDE␥ C-terminal interaction with ␣t was enhanced whereas the N-terminal interaction with ␣t was weakened. The same pattern of change was also observed using the BBM photoprobes (data not shown).
This observation was further confirmed using pulldown as a secondary approach (Fig. 5). Because the N-terminal peptide PDE␥(1-61) could not readily pull down ␣t-GTP␥S due to the weak interaction (13), we used the biotin-tagged PDE␥ C-terminal half (PDE␥(46 -87)-Btn), which was previously shown to pull down ␣t-GTP␥S (13). As indicated by the data in Fig. 5B, PDE␥(46 -87)-Btn pulled down ␣t-GDP-AlF 4 Ϫ ϳ2-fold more efficiently than it did ␣t-GTP␥S at the stoichiometry. Statistical analysis of pulldown of ␣t by PDE␥(46 -87)-Btn indicated a significant difference between the two ␣t conformers (Fig. 5C). Considering that the PDE␥ C-terminal domain (residues 62-87) contributes the major strength to binding with ␣t (13), these data are consistent with the label transfer results indicating an enhanced PDE␥ C-terminal interaction with ␣t-GDP-AlF 4 Ϫ compared with that with ␣t-GTP␥S (Fig. 4B).

PDE␥ N-terminal Interaction with RGS9⅐␤5 Contributes to PDE␥ Function in Stabilizing the GAP Complex-A PDE␥
N-terminal interaction with RGS9␤5 has not been reported previously. In fact, the PDE␥ N-terminal half has been missing in the relevant atomic structures (7,15,17). Using a label transfer method, we detected interactions of PDE␥ N-terminal positions with RGS9⅐␤5 in the GAP complex (Figs. 1 and 2). This result was then confirmed by the effective competition of PDE␥(10 -30) with the full-length Btn-PDE␥ when pulling down RGS9⅐␤5 from the GAP complex (Fig. 3). Importantly, our data indicate that the PDE␥ N-terminal interaction with RGS9⅐␤5 contributes to the role of PDE␥ as an enhancer of the RGS9⅐␤5⅐␣t interaction in the transition state complex.
The label transfer approach applied here is validated by the existing crystal structure of the partial GAP complex that includes the C-terminal half of PDE␥ (7). In this structure, Val 66 on PDE␥, which is near position 68, makes contact with Trp 362 of RGS9d. This hydrophobic interaction is complemented by an electrostatic RGS9 Arg 360 interaction with PDE␥ Glu 52 ,  which is close to Phe 50 . Thus, the photoprobe placed at position 68 or 50 on PDE␥ was able to reach (as measured by PyMOL) and cross-link with RGS9, resulting in the label transfer to RGS9 that we have observed (Fig. 1, A and B).
Moreover, previous studies using an evolutionary trace method combined with mutational analysis predicted that residues 314, 353, 357-360, and 362-364 in the catalytic core of RGS9 form part of the effector-GAP interface (29,30). Yet only one direct contact, that is between Trp 362 and Val 66 (on PDE␥), was resolved in the partial GAP structure (7). This raises the real possibility that the remainder of the predicted interface between RGS9 and PDE␥, i.e. the PDE␥ N-terminal half that is missing in this crystal structure, was not revealed due to the PDE␥ truncation.
Essentially, all the domains within the RGS9⅐␤5 complex contribute to the PDE␥-regulated GAP function. In particular, the noncatalytic RGS9 domains together with ␤5 that are missing in the partial GAP structure (7), play an important role in discriminating between the activated ␣t alone and the PDE␥⅐␣t complex (8,9). This evidence also supports the likelihood that PDE␥ N-terminal interactions with RGS9 and ␤5 occur but are missing in the partial GAP structure.
More importantly, the fact that the N-terminal PDE␥ peptides could effectively disassemble the GAP complex (Figs. 2 and 3) indicates that the N-terminal half of PDE␥ contributed considerably to enhancing the affinity between RGS9⅐␤5 and ␣t. The greater potency of PDE␥(1-61) than PDE␥(62-87) in preventing label transfer from the PDE␥ photoprobes to RGS9 (Fig. 2) suggests that the N-terminal half of PDE␥ plays an important role for the function of PDE␥ as an affinity enhancer to tighten the binding of ␣t with RGS9⅐␤5 in the GAP complex. As shown in previous studies, whereas the PDE␥ C-terminal peptide PDE␥(63-87) was equally potent in stimulating the GAP activity, the full-length PDE␥ was needed to maintain a high affinity GAP complex (10). We propose that whereas the PDE␥ C-terminal interaction with RGS9 (in particular, PDE␥ Val 66 with Trp 362 in the RGS domain) is essential to secure a correct conformation of ␣t for efficient GTP hydrolysis (7), the PDE␥ N-terminal interactions with RGS9␤5 complement the C-terminal interaction to ensure sufficient affinity between ␣t and RGS9⅐␤5.
Differential PDE␥ Interactions with the Two ␣t Conformers Suggest a "Rearrangement" That Primes the PDE␥ N-terminal Half for Interaction with RGS9⅐␤5-In this study, we also identified differential PDE␥ domain interactions with ␣t in two different signal transduction states, the signaling state conformation preserved in ␣t-GTP␥S (5) and the transition state conformation mimicked by ␣t-GDP-AlF 4 Ϫ (6). The interaction of the PDE␥ N-terminal domain with ␣t-GDP-AlF 4 Ϫ was determined to be weaker compared with that with ␣t-GTP␥S, whereas the PDE␥ C-terminal domain bound to ␣t-GDP-AlF 4 Ϫ stronger than with ␣t-GTP␥S (Figs. 4 and 5). Differential PDE␥ domain interactions with the two ␣t conformers may not be readily detectable using methods other than the label transfer approach applied here. Given the high similarities between the two ␣t conformers in their crystal structures (5,6) and their affinities with the full-length PDE␥ (31)(32)(33)(34), this issue appears to have been previously overlooked.
Although the overall differences between the crystal structures of ␣t-GTP␥S and ␣t-GDP-AlF 4 Ϫ are not remarkable (5, 6), a considerable difference occurs in the ␣t Switch II region, which is flexible and contains primary binding sites for both RGS9d and the PDE␥ C-terminal domain (7,33,35). Interestingly, a single mutation in the ␣t Switch II region (E203A) turned ␣t constitutively active for PDE6 activation even in the GDP-bound form (36). Another Switch II mutant, W207F, underwent the same type of conformational change of Switch II as did the wild-type ␣t upon activation, yet had ϳ100-fold lower affinity with PDE␥ (37). These examples together with a systematic mutational analysis of the Switch II region (35) strongly argue for the notion that even a minor structural difference in ␣t Switch II is sufficient to mediate significant changes in the ␣t⅐PDE␥ interactions. Moreover, the RGS domain binds to the two similar ␣t conformers with drastically different affinities with a preference for the transition state conformation (35, 38 -40). Further, PDE␥ together with RGS9 allosterically changes the ␣t Switch II region into a stable conformation optimal for GTP hydrolysis (7). Allosteric interplay between the RGS and effector binding sites via the Switch II region appears possible for all G␣ subunits that bind RGS proteins. Thus, subtle changes in the conformation of G␣ Switch II region could have profound effects on the affinity of effectors and GAPs (41).
Accordingly, the observed changes in the binding strength of the PDE␥ N-terminal and C-terminal domains with ␣t-GDP-AlF 4 Ϫ relative to that with ␣t-GTP␥S (Figs. 4 and 5) may be rationalized by the function of PDE␥ in the transition state complex. Because PDE␥ acts by physically enhancing the interaction between ␣t and RGS9⅐␤5 (8,10), the interaction of the PDE␥ C terminus with ␣t is an important binding force for stabilizing the GAP complex. An enhanced binding of the PDE␥ C terminus with ␣t in the transition state complex is therefore induced (Figs. 4 and 5) after its removal from PDE␣␤ in the signaling state (7,17). On the other hand, the PDE␥ N-terminal interaction with ␣t in the transition state is reduced compared with that in the signaling state (Fig. 4). Thus, in the transition state the most reasonable role for the PDE␥ N-terminal regions is to bind RGS9⅐␤5, providing additional strength to stabilize the GAP complex.
Therefore, we propose that upon a change of the ␣t conformation from the signaling state to the transition state, a "rearrangement" of the PDE␥ domain interactions with ␣t occurs, even though the overall PDE␥ affinity with ␣t is not significantly altered (31)(32)(33)(34). In the signaling state, a "transducisome" is formed by the complementary PDE␥ C-terminal interaction with ␣t-GTP and the N-terminal interaction with PDE␣␤ (13). Evidenced by a high sensitivity to trypsinization (42,43) and an extended structure in the PDE␣␤-bound state (20), the PDE␥ N-terminal half is likely exposed on the surface of PDE␣␤ GAF domains (44), providing an interface to interact with ␣t simultaneously in the transducisome. In the ensuing transition state, RGS9⅐␤5 recognizes and binds to the PDE␥ C-terminal complex with ␣t. This may allosterically change ␣t into a conformation that tightens its binding with the PDE␥ C-terminal domain while loosening its interaction with the PDE␥ N-terminal half (Fig. 4), which now becomes available to bind to RGS9⅐␤5 (Figs.  1-3). With the C-terminal domain tightly bound to ␣t and the N-terminal regions (and Val 66 ) interacting with RGS9⅐␤5, PDE␥ likely serves as a molecular "clasp" to stabilize the ␣t interaction with RGS9. Indirect evidence suggests that PDE␥ stays associated with PDE␣␤ in the transition state (45), perhaps binding to both PDE␣␤ and RGS9⅐␤5 via alternate N-terminal residues in different orientations. Although the PDE␥ C-terminal domain together with the RGS domain induce an optimal ␣t Switch II structure for GTP hydrolysis (7,15), the role of the PDE␥ N-terminal domain is most reasonably to be an interaction with RGS9⅐␤5 to secure a stable GAP complex, which is critical for efficient visual signal termination.
As an intrinsically disordered small protein (18), PDE␥ plays important physiological roles in the photoreceptor neurons (46 -48) via interactions with PDE6, transducin, as well as RGS9⅐␤5 (3,4). In the context of activation and ensuing termination of phototransduction, it remains enigmatic as to how PDE␥ alternates/adapts to different targets to regulate the amplitude and duration of the photoresponse. Future studies on comparison of PDE␥ interactions in the signaling state and transition state are warranted.