Complementary Interactions of the Rod PDE6 Inhibitory Subunit with the Catalytic Subunits and Transducin*

Activation of the cyclic GMP phosphodiesterase (PDE6) by transducin is the central event of visual signal transduction. How the PDE6 inhibitory γ-subunit (Pγ) interacts with the catalytic subunits (Pαβ) and the transducin α-subunit (αt) in this process is not entirely clear. Here we have investigated this issue, taking advantage of site-specific label transfer from throughout the full-length Pγ molecule to both αt and Pαβ. The interaction profiling and pull-down experiments revealed that the Pγ C- terminal domain accounted for the major interaction with αt bound with guanosine 5′-3-O-(thio)triphosphate (αtGTPγS) in comparison with the central region, whereas an opposite pattern was observed for the Pγ-Pαβ interaction. This complementary feature was further exhibited when both αtGTPγS and Pαβ were present and competing for Pγ interaction, with the Pγ C-terminal domain favoring αt, whereas the central region demonstrated a preference for Pαβ. Furthermore, αtGTPγS co-immunoprecipitated with PDE6 and vice versa in a Pγ-dependent manner. Either Pαβ or αtGTPγS could be pulled down by the Btn-Pγ molecules on streptavidin beads that were saturated by the other partner, indicating simultaneous binding of these two partners to Pγ. These data together indicate that complementary Pγ interactions with its two targets facilitate the αt·PDE6 “transducisome” formation. Thus, our study provides new insights into the molecular mechanisms of PDE6 activation.

The intricate visual transduction in rod photoreceptor cells provides a paradigm for G protein-coupled signaling. The outstanding visual sensitivity of the rod is largely due to the great signal amplification achieved by the cGMP 2 phosphodiesterase PDE6 (rod photoreceptor cGMP phosphodiesterase), the central effector enzyme (1). Upon absorption of a single photon, light-excited rhodopsin stimulates an exchange of GTP for GDP bound in the transducin ␣ subunit (␣ t ) (2), which in turn relieves PDE6 from the inhibitory constraint exerted by its ␥-subunit (P␥). PDE6 activation causes rapid cGMP breakdown, which closes the cGMP-coupled ion channels, thus relaying visual signals to the brain in a form of electrical pulses (3). PDE6 in the rod is uniquely composed of a large catalytic heterodimer (P␣␤, ϳ100 kDa each subunit) to which bind two small identical P␥ subunits (ϳ10 kDa) keeping the enzyme inactive in the dark (1,4). The PDE6 structure is less well understood compared with the other key players in phototransduction. This is primarily due to the fact that solving the atomic structure of PDE6 has been hindered by the lack of an expression system to produce active P␣␤ heterodimers in large amounts (5). A low resolution electron microscopy image of P␣␤ has revealed a linear alignment of three distinct domains of each subunit: the tandem GAFa and GAFb domains on the N-terminal side that host non-catalytic cGMP binding and the C-terminal catalytic domain that performs cGMP hydrolysis (6). Direct allosteric communication between GAF domains and the catalytic domain has been recently reported (7).
The inhibitory P␥ subunit is an intrinsically disordered protein, yet structural elements important for its function are encoded in the free P␥ molecule (8). The P␥ sequence of 87 amino acids features a polycationic central domain (Gly 19 -Gly 49 ) and a negatively charged C-terminal half that contains a linker region (Phe 50 -Gly 61 ) and a hydrophobic C-terminal domain (Thr 62 -Ile 87 ) (1,9). The last C-terminal dozen or so residues (herein termed the inhibitory region) are involved in the interaction with the P␣␤ catalytic domain (8,10,11). The very recently reported crystal structure of the chimeric PDE5/6 catalytic domain complexed with the P␥(70 -87) inhibitory peptide (5) has confirmed the previous suggestion that the highly hydrophobic C terminus (Y 84 GII 87 ) directly blocks the cGMP entry into the catalytic pocket (12,13). The other important P␣␤-interacting site on P␥ is the central domain, which has been shown to provide most of the binding strength for P␣␤ (14). The central domain of P␥ binds to the P␣␤ GAF domain (15,16) and couples non-catalytic cGMP binding in a positively cooperative manner, thus regulating the PDE-inhibiting function of P␥ (14). Remarkably, the C-terminal domain and the central domain also constitute ␣ t -interacting sites (17)(18)(19).
An overlap of the P␥ C-terminal ␣ t -binding region (Thr 62 -Ile 87 ) and the inhibitory region (Asn 74 -Ile 87 ) forms the structural basis for transducin-mediated PDE6 activation (5,8,20). Various lines of evidence suggest that GTP-bound ␣ t activates PDE6 by physically displacing the inhibitory region of P␥ from the P␣␤ catalytic pocket, thus initiating the signaling state of phototransduction (5, 8, 10 -12, 20). In the ensuing transition state, ␣ t GTP is converted back to the GDP-bound inactive structure, which has lowered affinity with P␥, thus releasing it to reinhibit PDE6 and terminate signaling (3). Fast visual recovery is ensured by great acceleration of the ␣ t GTPase activity, which is achieved by the GTPase-activating protein (GAP) complex composed of ␣ t , P␥, RGS9-1 (the ninth member of the regulators of G-protein signaling in photoreceptors), and its constitutive partner G␤5 as well as the membrane anchoring protein R9AP (3,21). Much of the molecular details of the P␥-␣ t GTP interaction in the signaling state have been learned from the crystal structure of the partial transition state complex, which includes the GDP-AlF 4 Ϫ -bound ␣ t/i1 chimera, the half-P␥ (Gly 46 -Ile 87 ), and the catalytic core of RGS9-1 (20). As visualized by this structure, a stretch of P␥ residues around Trp 70 forms a tight interaction with ␣ t that is further reinforced by additional contacts provided by some residues in the P␥ inhibitory region. Recent NMR (8) and crystallography (5) studies indicated that when the P␥ inhibitory region was associated with the chimeric PDE5/6 catalytic domain, the critical ␣ t -binding residues Trp 70 and Leu 76 , however, were not involved. These studies lend further support to a model of PDE6 activation (5,11); i.e. an engagement of ␣ t GTP with the P␥ residues Trp 70 and Leu 76 triggers a conformational change involving a hingelike rigid body movement of P␥(78 -87) away from the PDE6 catalytic pocket.
Thus, P␥ plays a pivotal role, not only for turning on but also for turning off phototransduction and keeping the signaling system inactive in the dark (9). Despite a wealth of information regarding phototransduction mechanisms, dynamic interactions of P␥ with ␣ t and P␣␤, as well as RGS9-1, are not well understood. There has been controversy as to whether P␥ completely dissociates from P␣␤ in the process of PDE6 activation. It is possible that whereas ␣ t sequesters the P␥ C-terminal region from the P␣␤ catalytic domain, the central domain of P␥ stays bound to the P␣␤ GAF domain until the binding is allosterically reduced by the dissociation of cGMP from the GAF domain (1). This scenario of simultaneous P␥ interactions with both ␣ t and P␣␤ is consistent with the proposition of an intermediate ␣ t ⅐PDE6 complex during PDE6 activation (17,(22)(23)(24)(25)(26). Earlier studies suggested that direct ␣ t -P␣␤ contacts may be a driving force in forming the intermediate complex in the presence of disc membranes (24, 27). However, it has not been determined whether the P␥ interactions with ␣ t and P␣␤ contribute important elements to the intermediate PDE6 activation complex.
As presented in this study, the label transfer approach, which has proven to be powerful for systematically detecting interactions of full-length molecules (16,28,29), offered us an opportunity to investigate this issue from a unique perspective. The data obtained through label transfer, immunoprecipitation, and pull-down suggest that complementary interactions, in which the P␥ C-terminal domain forms a strong interaction with ␣ t while the central region binds tightly with P␣␤, assist the transducin⅐PDE6 complex formation, which elicits PDE6 activation.

EXPERIMENTAL PROCEDURES
The chemicals and reagents used in this study were from the sources described previously (16,28) unless otherwise stated. The C-terminal P␥ peptide (P␥(62-87)) was custom-synthesized at the Peptide Synthesis Facility of the Biotechnology Center, University of Wisconsin (Madison, WI).
Transducin Preparation-Using frozen dark-adapted bovine retinas (J. A. & W. L. Lawson Co.), rod outer segment (ROS) membranes were isolated, from which holotransducin was prepared as described previously (29,30). ␣ 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 ROS membranes, and ␣ t GTP␥S was thus released and purified on the blue Sepharose CL-6B column. The purity of ␣ t was determined to be Ͼ95% by SDS-PAGE and Coomassie staining. The purified proteins were stored at Ϫ80°C.
Preparation of PDE6-The samples of bovine PDE6 were kindly provided by Dr. Nikolai O. Artemyev at the University of Iowa and prepared according to established methods (4). Briefly, holo-PDE6 was extracted from bleached ROS membranes, and P␣␤ was then obtained by removing P␥ through mild tryptic proteolysis of holo-PDE6. More vigorous tryptic treatment generated the P␣␤ heterodimer with a nick at Lys 146 / Lys 147 on P␤. It has been reported that nicked P␣␤ has unaltered functional properties (12,16). Unless otherwise stated, "P␣␤" refers to nicked P␣␤ throughout this paper. The P␣␤ preparations were purified to Ͼ95% by a Mono-Q column (Amersham Biosciences), as judged from Coomassie-stained SDS gels.
Preparation of P␥ Photoprobes-The constructs for expressing the full-length wild type P␥ with the single cysteine at position 68 (29), and the single cysteine mutants were generated as described previously (28). They were expressed in E. coli and purified by chitin beads, followed by reversed-phase HPLC using the POROS 20 R2 resin (31). The truncated P␥ variants (29) with and without a His 6 tag at the N terminus (HisP␥(1-61) and P␥(1-61), respectively) were prepared using the same protocol. Full-length P␥ (Ͼ95% pure) was used for preparation of P␥ photoprobes. The radioactive [ 125 I]ACTP-P␥ and nonradioactive [ 127 I]ACTP-P␥ photoprobes were prepared as described earlier (28).
The maleimido benzophenone (mBP)-P␥ photoprobes were prepared as described previously (16). Briefly, P␥ was derivatized with mBP in 10 -20-fold molar excess, and mBP-P␥ was then separated from unreacted P␥ and free mBP through reversed phase HPLC. Correct molecular masses of the [ 127 I]ACTP-P␥ and mBP-P␥ photoprobes have been confirmed by electrospray ionization mass spectrometry conducted at the Chemistry Department Mass Spectrometry Facility of the University of Wisconsin (Madison, WI).
Functional Assay of the P␥ Photoprobes-The transducin GTPase activity assay was kindly conducted by Dr. Kirill A. Martemyanov (now at the University of Minnesota) and Dr. Vadim Y. Arshavsky (now at Duke University), using a single turnover technique as described previously (32). The assay was 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 .

Transducin-PDE6 Interaction
The urea-treated ROS membranes, lacking endogenous activity of RGS9-1, 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 ROS membranes (20 M final rhodopsin concentration) reconstituted with transducin heterotrimer (1 M) and recombinant RGS9-1⅐G␤5 complex (0.5 M). The reactions were performed in either the absence or presence of P␥ derivatives (1 M). The reaction was stopped by the addition of 100 l of 6% perchloric acid. The 32 P formation was measured with activated charcoal. All assays were conducted in the absence of reducing agent due to the presence of the disulfide linkage between the photoreactive group and P␥.
Photocross-linking/Label Transfer Using P␥ Photoprobes-A scheme is presented in supplemental Fig. S1A to explain the label transfer strategy. Unless otherwise described, photocrosslinking reactions were performed in the HEPES buffer (10 mM HEPES, pH 7.5, 120 mM NaCl, 5 mM MgCl 2 ). Samples were contained in ultraclear polypropylene microcentrifuge tubes (Axygen). The reactions using [ 125 I]ACTP-P␥ photoprobes were exposed to the UV light generated by an AH-6 waterjacketed 1000-watt high pressure mercury lamp for 5 s at a distance of 10 cm (28). The reactions with mBP-P␥ were photolyzed at 5-10°C for 2 ϫ 15 min with a 5-min dark interval on ice in an RPR-100 Rayonet photochemical reactor equipped with 18 bulbs of 350 nm (Southern New England Ultraviolet Company). Immediately after photolysis, sample buffer was added to the reactions to final concentrations of 1% SDS and 50 mM DTT. The proteins were separated by SDS-PAGE and then subjected to Coomassie Blue staining and autoradiography. Autoradiography and protein quantitation were performed as described previously (28).
Pull-down Assays of P␥ Interactions with ␣ t GTP␥S and P␣␤ Using Affinity Beads-To immobilize the full-length P␥ to the streptavidin beads, biotinylated P␥ was prepared by covalently attaching maleimide-PEO 2 -biotin (Pierce Biotechnology) to the single cysteine at position 3 of the P␥ mutant, L3C (28). The derivatization reaction and purification of the Btn-L3C derivative were performed following the protocol of mBP-P␥ preparation (16). To prepare Btn-P␥(46 -87) , the P␥ 87C mutant was first derivatized with maleimide-PEO 2 -biotin and then trypsinized, and Btn-P␥(46 -87) was purified by reversed phase HPLC using a C4 column (31).
For each pull-down reaction, 0.4 l of Ultra-Link Plus immobilized streptavidin gel (Pierce) was first equilibrated with the pull-down buffer, which contains 20 mM HEPES, pH 7.5, 120 mM NaCl, 5 mM MgCl 2 , 1 mM DTT, 0.1% n-dodecanoylsucrose (Calbiochem), and 50 g/ml BSA, and then incubated with 4 M Btn-P␥ by rotating the microcentrifuge tube for 10 min at room temperature. A high concentration (1 g/l) of BSA or soybean trypsin inhibitor was added at this step to block possible nonspecific protein-bead interactions. After this incubation, Btn-P␥ was found completely bound to the streptavidin beads (data not shown). To test if the P␥ peptides disrupt the P␥-␣ t GTP␥S or P␥-P␣␤ interaction, P␥(62-87) or P␥(1-61) in excess over Btn-P␥ was first incubated with ␣ t GTP␥S or P␣␤ in the pull-down buffer for 1 h on ice and then added to the Btn-P␥-streptavidin beads. After rotating the reactions at 4°C for 1-2 h, the beads were washed twice with 400 l of ice-cold pull-down buffer. Proteins were then eluted from the beads with SDS/DTT-containing sample buffer, run on a low crosslink 15% acrylamide gel (33), and visualized by staining with Coomassie Blue R-250, or SilverSNAP Stain Kit II (Pierce) when lower amounts of proteins were used.
To study the interactions of the P␥ central region with P␣␤ and ␣ t , the P␥ construct HisP␥(1-61) was used. For each reaction, 0.5 l of His-Select High-Flow nickel beads (Sigma) were first washed with 500 l of H 2 O and then with 300 l of pulldown buffer. HisP␥(1-61) of 5 M was immobilized to the beads by incubation in the pull-down buffer (supplemented with 1 g/l trypsin inhibitor at this step) at room temperature for 10 min on rotating. Twenty mM imidazole was included in the pull-down buffer throughout the experimental procedures to prevent possible nonspecific binding of proteins to nickel beads. After the beads were washed with 2 ϫ 500 l of pulldown buffer to remove unbound HisP␥(1-61), P␣␤ or/and ␣ t GTP␥S were added and incubated with the beads for 1-2 h at 4°C. The beads were then washed twice with 200 l of pulldown buffer. The proteins on the beads were eluted with the sample buffer and resolved by SDS-PAGE using a low cross-link 15% gel (33), which was then silver-stained using SilverSNAP Stain Kit II (Pierce).
Immunoprecipitation Assay of the ␣ t GTP␥S-PDE6 Interaction-Co-immunoprecipitation of ␣ t GTP␥S with holo-PDE6 was carried out using nProtein A Sepharose Fast-Flow beads (Amersham Biosciences) and the antibody against bovine rod P␣ (Affinity Bioreagents). For each reaction, 0.5 l of Protein A beads were first equilibrated with the HEPES buffer (10 mM HEPES, pH 7.5, 120 mM NaCl, 5 mM MgCl 2 ) and then incubated with 0.5 g of the anti-P␣ antibody by rotating for 1 h at 4°C. One g/l soybean trypsin inhibitor was included to block possible nonspecific protein binding sites on Protein A beads. The beads were washed three times with 300 l of HEPES buffer prior to the immunoprecipitation reaction. Meanwhile, 0.5 g of holo-PDE6 was incubated for 1-2 h on ice with 0.5 g of ␣ t GTP␥S in the HEPES buffer containing 50 g/ml trypsin inhibitor and 1 mM DTT. The reaction was then added to the washed Protein A beads with anti-P␣ bound and incubated for 1 h by rotating at 4°C. P␥ peptide P␥(1-61) or P␥(62-87) in a 200-fold molar excess over holo-PDE6 was added as a competitor to disrupt P␥ interactions. After washing the beads three times with 300 l of HEPES buffer (containing trypsin inhibitor and DTT), the immunoprecipitate was eluted with the sample buffer and subjected to SDS-PAGE and then detected by Western blotting using the anti-␣ t antibody (K-20, Santa Cruz Biotechnology, Inc. (Santa Cruz, CA)).
Co-immunoprecipitation of P␣␤ with ␣ t GTP␥S was performed similarly but in a reverse manner. Briefly, 1.5 g of anti-␣ t was immobilized onto 0.75 l of Protein-A beads. In separate tubes, 0.95 g of P␣␤ was incubated with 0.1 g of P␥ on ice for 0.5 h with or without a competitor (P␥(1-61) or P␥(62-87) in a 200-fold molar excess over P␣␤), and 0.2 g of ␣ t GTP␥S was then added. Following incubation on ice for 1 h, the reaction was mixed with washed anti-␣ t /Protein A beads and rotated for 1 h at 4°C. The P␣␤ immunoprecipitate was Transducin-PDE6 Interaction MAY 14, 2010 • VOLUME 285 • NUMBER 20 then eluted off of the washed beads and detected by Western blotting using the anti-P␣ antibody.

Profiling of the ␣ t GTP␥S Interaction with the Full-length P␥-
A full spectrum profiling of P␥-␣ t interaction has not been reported. Previous peptide mapping studies showed that C-terminal P␥ peptides had slightly lower affinity for ␣ t GTP␥S than peptides from the central region (17,18). The affinity of these peptides with activated ␣ t (ϳ1 M), however, is nearly 100-fold lower than the full-length P␥ (10 -12 nM (18,34)). This conspicuous affinity difference indicates that the full-length P␥ molecule is required to assume an optimal conformation for binding with ␣ t . Obviously, peptide mapping is not an optimal approach to measure relative P␥ domain contributions in the full-length P␥-␣ t GTP␥S interaction. Our label transfer experiments, in which the full-length P␥ could be used, however, offered a better means to address this issue.
Eleven photoprobes were prepared, with [ 125 I]ACTP sitespecifically attached through mixed disulfide to the single cysteines placed at various positions throughout the P␥ molecule. Functional properties of these ACTP-P␥ probes have been carefully characterized when previously used to map the P␥-P␣␤ interaction interface, and no major change due to the ACTP modification was observed for the PDE6 inhibition potency of P␥ (28). A possible impact of ACTP on the P␥ function of ␣ t GTPase stimulation was further assessed in this study, which showed that the functional activities of these P␥ photoprobes were similar to that of the unmodified native P␥ (supplemental Fig. S1B).
We therefore used these probes in the photocross-linking/ label transfer experiments to profile the P␥-␣ t GTP␥S interaction. Upon UV illumination, the azide group of ACTP is photoactivated into a nitrene (35), which then inserts into the nearby P␥-interacting site(s) on ␣ t , forming a covalent bond with the ␣ t backbone. After DTT reversal of the S-S link between ACTP and the cysteine on P␥, the 125 I radiolabel is transferred from P␥ to ␣ t (see the diagram in supplemental Fig. S1A), which can be detected by autoradiography (Fig. 1A). Thus, the label transfer efficiency reflects the interaction intensity between ␣ t and a given P␥ position where [ 125 I]ACTP is attached.
The [ 125 I]ACTP-P␥ photoprobes, which were previously proven to transfer radiolabel to P␣␤ specifically (28), were shown here to also specifically transfer radiolabel to ␣ t GTP␥S. Specificity of the observed label transfer to ␣ t was manifested not only by the absence of radiolabel on BSA (Fig. 1A), which was included as an internal control, but also by the P␥ position dependence of label transfer (Fig. 1B).
The label transfer yield from each P␥ position was quantified by normalizing the intensity of radiolabel on ␣ t with the specific radioactivity of the corresponding [ 125 I]ACTP-P␥ photoprobe  Table S1 or Ref 28). Statistical analyses were performed by t test (Microsoft Excel). *, p ϭ 0.01-0.05 (significant); **, p ϭ 0.001-0.01 (very significant); ***, p Ͻ 0.001 (extremely significant); ns, p Ͼ 0.05 (not significant). As shown in the figure, there is a very significant difference between label transfer from the P␥ Phe 30 -Cys 68 region and that from the Trp 70 -Leu 76 region. Position 87 is not included in the latter group for statistical analysis because label transfer from the hydrophilic ACTP probe substituting the hydrophobic Ile 87 residue could not accurately reflect the interaction between this position and ␣ t GTP␥S (see the discussion of Figs. 1 and 2 under "Results"). The difference between each position and position 76 is also analyzed: positions 16 and 21 (***); 40, 50, and 68 (**); 30, 60, and 87 (*); 70 and 73 (not significant).
(supplemental Table S1) (28). Interestingly, the resultant profile of label transfer to ␣ t showed a pattern in which ␣ t GTP␥S was highly labeled by [ 125 I]ACTP from the P␥ C-terminal positions around Trp 70 (positions 70, 73, and 76) (Fig. 1B). This region is known to have intimate contacts with ␣ t (20). Consistently, experiments performed under the same conditions but using ␣ t GDP-AlF 4 Ϫ , which shares a high similarity with ␣ t GTP␥S in their three-dimensional structures and functional properties (20,36), showed a similar profiling pattern of label transfer (data not shown).
To further confirm the observed P␥-␣ t GTP␥S interaction pattern (Fig. 1B), label transfer profiling was also carried out using P␥ photoprobes containing mBP, a photoreactive group different from ACTP ( Fig. 2A). Under UV light (350 -365 nm), the ketone in benzophenone is activated into a diradical by disproportionation and reacts with neighboring C-H bonds (35) in ␣ t to form a C-C link. Because the maleimide group forms a C-S bond with the cysteine on P␥, which cannot be cleaved by DTT, the cross-linked P␥-␣ t complex stays covalently linked and migrates as a band higher than the ␣ t band on the gel after SDS/DTT treatment ( Fig. 2A). The mBP-P␥ photoprobes, which have been previously characterized, showed no significant functional changes in PDE6 inhibition due to mBP modifications on P␥ (16). Moreover, in a previous study from our laboratory (29), another benzophenone probe, benzoyl-L-phenylalanine, which is very similar to mBP, caused only minor changes in the P␥ stimulation of ␣ t GTPase when incorporated into a P␥ C-terminal position at 66, 73, 76 or 86.
In a good agreement with the above results of [ 125 I]ACTP label transfer, the cross-linking experiments using mBP-P␥ resulted in a similar profiling pattern (Fig. 2B). Thus, the fact that profiling with two different photophores (ACTP and mBP) led to similar patterns eliminates concerns regarding possible chemical selectivities of the photoprobes.
Compared with [ 125 I]ACTP-P␥, the mBP-P␥ probes showed a more profound preference for the P␥ C-terminal positions in cross-linking with ␣ t GTP␥S. This difference very likely stemmed from the fact that mBP is more hydrophobic (than ACTP) and thus more similar to the C-terminal hydrophobic residues that were replaced by the photoprobe. An interesting example is Ile 87 , the prominent hydrophobicity of which is known to play an important role in the function of P␥ (12,13). Accordingly, the hydrophobic mBP at position 87 yielded the highest P␥-␣ t GTP␥S cross-link efficiency (Fig. 2B), whereas [ 125 I]ACTP, a relatively hydrophilic probe due to the presence of a carboxyl group (Fig. 1A), resulted in less cross-link efficiency at position 87 (Fig. 1B). In this regard, the mBP crosslinking profile (Fig. 2B) may better represent P␥ domain contributions to the P␥-␣ t GTP␥S interaction.
Notably, both [ 125 I]ACTP and mBP yielded a substantial cross-link at P␥ position 70 (Figs. 1B and 2B), which is critical for the P␥-␣ t interaction (20). This observation agrees with our previous study in which benzoyl-L-phenylalanine replacement of Trp 70 also yielded a relatively high cross-link efficiency (29). The simplest explanation is that, due to a similarity to tryptophan, the photoprobe placed at position 70 could remain in close proximity to the Trp 70 -interacting residues of ␣ t GTP␥S. This proposition is also supported by the native gel assay in which the ACTP-P␥⅐␣ t complexes still formed substantially, although the ACTP derivatization at position 70, 76, or 87 affected the P␥⅐␣ t interaction to some extent (supplemental Fig. S1C). It is thus reasonable to conclude that the ␣ t GTP␥S interaction with Trp 70 would be actually stronger than observed here with a photoprobe substitution at this position. Nevertheless, our profiling experiments using full-length purified proteins revealed a clear pattern of the P␥-␣ t GTP␥S inter-FIGURE 2. Profiling of the P␥-␣ t GTP␥S interaction using mBP photoprobes. A, photocross-linked mBP-P␥-␣ t GTP␥S is shown as a higher band above the ␣ t band on the Coomassie-stained SDS gel. The photocross-linking reactions were performed as described under "Experimental Procedures." Each reaction included 1.6 g of ␣ t GTP␥S and an equal molar amount of mBP-P␥ in the HEPES buffer (10 mM HEPES, pH 7.5, 120 mM NaCl, 5 mM MgCl 2 ). DTT was added at a concentration of 2 mM to prevent possible nonspecific cross-link. B, the photocross-link efficiency of mBP-P␥ and ␣ t GTP␥S at each P␥ position was quantified as a percentage ratio of the protein amount in the cross-link band versus the sum in both the ␣ t band and the cross-link band.  MAY 14, 2010 • VOLUME 285 • NUMBER 20 action; the C-terminal domain of P␥ contributed the major strength for the interaction with ␣ t GTP␥S compared with the remainder of the P␥ molecule (Figs. 1B and 2B).

The C-terminal Domain of P␥ Provides Major Binding Strength for the P␥ Interaction with ␣ t GTP␥S, as Does the Central Domain for the Interaction with P␣␤-In order to further assess the role of the P␥ C-terminal domain in the interaction
with ␣ t GTP␥S, a different approach (pull-down) was applied, using Btn-P␥ bound to streptavidin beads. As shown in Fig. 3A, ␣ t GTP␥S was specifically pulled down by Btn-P␥ (lane 2), because no ␣ t was detected in the control with no Btn-P␥ (lane 1). Interestingly, the P␥ C-terminal peptide P␥(62-87) effectively abolished ␣ t GTP␥S pull-down (lane 3), but the N-terminal peptide P␥(1-61) of the same concentration did not (lane 4). This result demonstrates a dominant role of the P␥ C-terminal domain in the P␥-␣ t GTP␥S interaction. This conclusion is also supported by the observation that the C-terminal half of P␥ (Btn-P␥(46 -87)) efficiently pulled down ␣ t GTP␥S (Fig. 3D,  lane 3), but the N-terminal HisP␥(1-61) peptide did not (Fig.  3C, lane 4). Importantly, the observation that the P␥ C-terminal domain provided the major P␥-␣ t GTP␥S interaction strength ( Figs. 1-3) is in accord with the crystal structure of the partial GAP complex (20), in which the P␥ C-terminal domain is engaged in a hydrophobic interlock between the ␣ t/i1 switch II and ␣ 3 helix.
It is noteworthy that the full-length P␥ in a 20-fold excess completely abolished the ␣ t GTP␥S pull-down (Fig. 3A, lane 6), and the full-length P␥ pulled down ␣ t GTP␥S much more efficiently than the C-terminal half-peptide (compare lane 1 with lane 3 in Fig. 3D). This is consistent with the previous assessment that although either the central or the C-terminal peptides could interact with ␣ t GTP␥S separately, the full-length P␥ interacted with ␣ t with a much higher affinity (17,18). Here we further assert that in the full-length P␥-␣ t interaction, although the C-terminal domain contributes more than the central domain, both domains are required to forge a strong P␥-␣ t GTP␥S interaction.
Interestingly, in contrast to the major role of the P␥ C-terminal domain in the P␥-␣ t interaction, it is P␥(1-61) (Fig. 3C, lane  2) rather than the C-terminal half (Fig. 3D, lane 6) that efficiently pulled down P␣␤. The fact that P␥(1-61) (lane 4 in B) but not P␥(62-87) (lane 5 in B) abrogated the P␥-P␣␤ interaction also indicates a dominant role of the P␥ N-terminal side for interacting with P␣␤. This conclusion is consistent with the previous observations that the P␥ central domain binds P␣␤ much more strongly than the C-terminal domain (14). It is noteworthy that P␥(1-61) that was 20,000-fold (lane 4 in B) but not 1000-fold (lane 2 in B) in molar excess could disrupt the P␥-P␣␤ interaction. This result reflects an exceptionally tight full-length P␥-P␣␤ interaction, the optimum K d of which is in the subpicomolar range (1,14). Therefore, the stark contrast of the binding of P␣␤ and ␣ t GTP␥S to the same P␥ domain raised an important question as to whether P␥ interacts with ␣ t GTP␥S and P␣␤ differentially.

P␥ Interacts with ␣ t GTP␥S and P␣␤ in a Complementary
Manner-Label transfer profiling with the full-length, photoprobe-derivatized P␥ constructs allowed us to compare the pattern of P␥-␣ t GTP␥S interaction (Figs. 1B and 2B) with the previously observed P␥-P␣␤ interaction profile (28). A "complementary" feature of the two separately determined profiles was thus revealed, in which the P␥ central region provides most of the strength for binding with P␣␤, whereas the C-terminal region accounts for the major interaction with ␣ t GTP␥S. These data prompted us to investigate the competition between ␣ t GTP␥S and P␣␤ for interacting with P␥, by comparing the [ 125 I]ACTP label transfer to ␣ t GTP␥S and to P␣␤ from various P␥ positions in a systematic manner. For this purpose, label transfer to ␣ t GTP␥S and P␣␤ from a certain P␥ position in a photocross-linking reaction with these two partners present could be directly compared in the same lane on the SDS gel (Fig.  4A). These experiments were intended to mimic the signaling state when ␣ t GTP interacts with and displaces P␥ from PDE6 to activate the enzyme. P␥ was utilized substoichiometrically in comparison with ␣ t GTP␥S and P␣␤ (0.7 P␥, 1 ␣ t , 1 subunit of P␣␤) so that ␣ t GTP␥S and P␣␤ could effectively "compete" in their interactions with P␥ and therefore provide a greater opportunity for revealing a preferential labeling on ␣ t GTP␥S or P␣␤ from a given P␥ position. As shown in the autoradiogram (Fig. 4A) and the data that are summarized as the labeling ratios

Transducin-PDE6 Interaction
for the two targets (Fig. 4B), a distinct preference of photolabel transfer for ␣ t GTP␥S over P␣␤ occurred from the P␥ C-terminal positions, in particular from position 76. Accordingly, the recent crystal structure (5) showed that whereas the P␥ C terminus bound to the chimeric PDE5/6 catalytic domain, Leu 76 pointed away, positioning itself in a direction ready for interaction with ␣ t . In clear contrast to the C-terminal positions, however, more labeling from the P␥ central positions (Val 21 -Leu 60 ) was observed on P␣␤ than on ␣ t GTP␥S (Fig. 4B). These experiments of label transfer competition further confirm the complementary nature of the P␥ interactions with ␣ t GTP␥S and P␣␤, which was first exhibited by the different interaction profiles.
Complementary P␥ Interactions with ␣ t and P␣␤ Constitute the ␣ t ⅐PDE6 Complex-Given the above evidence that P␥ differentially interacted with its two targets, with the C-terminal domain favoring ␣ t GTP␥S and the central region preferring P␣␤, we sought to test whether the complementary interactions play an important role in the intermediate transducin⅐PDE6 complex, using co-immunoprecipitation approaches (Fig. 5, A and B). If the ␣ t GTP␥S⅐PDE6 complex is indeed primarily held together by the P␥ C-terminal interaction with ␣ t GTP␥S and the central region interaction with P␣␤ (diagramed in Fig. 5C), disrupting either of these interactions using a P␥ peptide should dissociate the ␣ t ⅐PDE6 complex.
We first observed that ␣ t GTP␥S was co-immunoprecipitated specifically with the holo-PDE6 that was immobilized to Protein A beads via the anti-P␣ antibody (Fig. 5A, lane 2), as compared with the control with no holo-PDE6 added (lane 1). The immunoprecipitation of ␣ t proved to be GTP-dependent because only a background level of ␣ t GDP was co-immunopre-   Fig. 1 were used. Photocross-linking reactions were conducted with the [ 125 I]ACTP-P␥ photoprobes in the presence of both ␣ t GTP␥S (4 M) and P␣␤ (without a nick) at a molar ratio of 0.7 P␥, 1.0 ␣ t , 1.0 P␣␤ subunit. The reactions were then subjected to SDS-PAGE (bottom) and autoradiography (top). BSA was used as an internal control to demonstrate the specificity of label transfer. B, labeling ratio of ␣ t GTP␥S versus P␣␤. Each bar value represents a mean Ϯ S.D. (error bars) (of 3-5 experiments). The amount of label transfer to ␣ t GTP␥S or P␣␤ from each P␥ position was normalized for the ␣ t (or P␣␤) protein amount and the specific activity of the corresponding [ 125 I]ACTP-P␥ probe (supplemental Table S1 and Ref. 28).  To further explore the P␥-dependent nature of the ␣ t ⅐PDE6 coimmunoprecipitation, experiments were performed in a reverse manner, detecting co-immunoprecipitation of P␣␤ with ␣ t GTP␥S, which was immobilized on Protein A beads through the anti-␣ t antibody (Fig. 5B). Similar to the results shown in Fig. 5A, P␣␤ was co-immunoprecipitated with ␣ t GTP␥S in the presence of P␥ (lane 3). Obviously, P␣␤ was precipitated by the P␥⅐␣ t GTP␥S complex, because no P␣␤ precipitation was observed when P␥ was absent (lane 4). P␣␤ precipitation was diminished in the presence of P␥(1-61) (lane 1) or P␥(62-87) (lane 2). A significant portion of P␣␤ was still precipitated with excess P␥(1-61) present (lane 1), reflecting tight binding of the P␥ N-terminal half with P␣␤ (see Fig. 3, B and C). These data further confirm the P␥-mediated complementary interactions.
With the data in Fig. 5 showing P␥-dependent co-immunoprecipitation of ␣ t GTP␥S with PDE6 as well as co-immunoprecipitation of PDE6 with ␣ t GTP␥S, we obtained additional evidence (Fig. 6, A-C) supporting the conclusion that complementary and simultaneous binding of P␥ to its two partners constituted the ␣ t ⅐PDE6 complex.
The experiments were designed based on the following idea. If P␣␤ binds P␥ simultaneously along with ␣ t GTP␥S (diagramed in Fig. 6D, 1), P␣␤ should be pulled down by the Btn-P␥ molecules on streptavidin beads that are saturated by excess ␣ t GTP␥S, and vice versa. As shown in Fig. 6A, P␣␤ was readily pulled down by Btn-P␥, which was bound to the beads and preincubated with ␣ t GTP␥S at a 5-fold molar excess (lanes 2 and 4), supporting the notion that co-binding of ␣ t GTP␥S and P␣␤ to Btn-P␥ occurred as depicted in Fig. 6D, 1. We cautioned that the observed co-pull-down of P␣␤ and ␣ t might be subject to such alternate explanations as depicted in Fig. 6D, 2-4. However, these concerns can be ruled out by detailed analysis of our data. First, P␣␤ was FIGURE 6. Simultaneous P␥ binding with ␣ t GTP␥S and P␣␤ evidenced by pull-down experiments. Pull-down was performed as described under "Experimental Procedures." Trypsin inhibitor was used to block nonspecific protein-bead interactions. 4 M Btn-P␥ was used in A, and the gel was Coomassie-stained; 0.5 M Btn-P␥ was used in B and C, and the gels were silver-stained. Each gel shown in the figure is a representative of at least three similar experiments. A, ␣ t GTP␥S in a 5-fold molar excess versus Btn-P␥ (4 M) was added to the Btn-P␥ streptavidin beads to saturate the ␣ t -binding sites on Btn-P␥ by incubation for 1 h at 4°C. After washing the beads, P␣␤ (0.2 M dimer) was added and incubated for another 1 h. In the reaction of lane 3, P␥ in 5-fold molar excess was added to compete with the Btn-P␥-P␣␤ interaction. In lane 4, ␣ t GTP␥S in a 5-fold molar excess was added again (marked as ␣ t *) after washing the ␣ t -saturated Btn-P␥ beads to resaturate the P␥ molecules that could have become free due to extensive washing. B, in order to determine whether ␣ t could replace P␣␤ that is prebound with Btn-P␥, the P␣␤-binding sites of Btn-P␥ were first saturated by incubation with 1 M P␣␤, the beads were washed, and 1 M ␣ t GTP␥S (lane 2) or equal volume of pull-down buffer (lane 1) was then added. Similarly, to address whether P␣␤ could replace ␣ t that is bound on the Btn-P␥ beads, the ␣ t -binding sites were first not pulled down by a population of Btn-P␥ that could have become free due to extensive washing of the ␣ t -saturated beads (see Fig. 6D, 2), because the same amount of P␣␤ was pulled down in the presence of excess ␣ t that was added back to the ␣ t -saturated beads (Fig. 6A, lane 4). Second, the addition of P␣␤ to ␣ t -saturated beads did not reduce ␣ t pull-down (Fig. 6B, compare lane 4 with lane 3), indicating that the P␣␤ pull-down did not result from replacement of a portion of ␣ t by the P␣␤ binding to Btn-P␥ (see D3). In this case, replaced ␣ t should have been washed off and would have led to a lowered intensity of the ␣ t band. This was also true for the ␣ t pull-down on P␣␤-saturated beads (compare lane 2 with lane 1 in B). Third, P␣␤ was not pulled down by directly interacting with ␣ t (see D4), because ␣ t was readily pulled down by Btn-P␥(46 -87) but P␣␤ was not pulled down together with ␣ t (Fig. 3D, lane 5). Similarly, ␣ t was not pulled down together with P␣␤, which was pelleted with His-P␥(1-61) on nickel beads (Fig. 3C, lane 3).
Finally, P␥(1-61) in a 1000-fold excess could not compete with the strong full-length P␥-P␣␤ interaction (Fig. 3B, lane 2, and Fig. 6C, lane 2) but could do so in the presence of ␣ t GTP␥S (Fig. 6C, lane 3). The simplest explanation is that simultaneous binding of ␣ t GTP␥ to Btn-P␥ along with P␣␤ weakened the P␥-P␣␤ interaction by sequestering the P␥ C-terminal domain and thus kept P␣␤ from a high affinity binding with the fulllength P␥. Taken together, these lines of evidence suggest that P␥ facilitated the formation of the ␣ t GTP␥S⅐PDE6 complex by binding to both P␣␤ and ␣ t GTP␥S simultaneously.

DISCUSSION
PDE6 activation mediated by ␣ t is the central step in the visual transduction cascade. How P␥ interacts with its two targets (␣ t and P␣␤) is the key to understanding the molecular mechanism of PDE6 activation, yet it has been difficult to capture a snapshot of this dynamic process through structural biology. We have obtained evidence here supporting the conclusion that complementary interactions (a strong P␥ C-terminal interaction with ␣ t GTP and a tight binding of the central region to P␣␤) occur in favor of PDE6 activation and at least partly account for the binding force in the intermediate ␣ t ⅐PDE6 complex.
It is known that the P␥ central region and the C-terminal domain are the two primary binding sites not only for ␣ t GTP␥S but also for P␣␤ (8, 10, 16 -18, 29, 37, 38), but it is not yet clear how the two P␥ domains differentiate their interactions with ␣ t and P␣␤ (9) when both targets are involved during PDE6 activation.
The observation of complementary P␥ domain interactions with ␣ t GTP␥S and P␣␤ was based on comparison of the interaction profiles with the two targets, both obtained from systematic mapping of the entire interaction interfaces. A stronger ␣ t GTP␥S interaction with the P␥ C-terminal domain than with the P␥ central region (Figs. 1B and 2B) was in interesting contrast to a stronger P␣␤ interaction with the P␥ central region than with the P␥ C-terminal domain (28), as also indicated by the pull-down experiments (Fig. 3). More significantly, this complementary feature was also observed when both of the targets were present competing for P␥ interaction (Fig. 4).
The complementary P␥ interactions with its two targets may have important implications for the molecular mechanism of PDE6 activation. The significance is 2-fold. First, because the ␣ t -interacting C-terminal domain (Thr 62 -Ile 87 ) includes the PDE6-inhibiting region, a strong P␥ C-terminal interaction with ␣ t GTP is essential to compete with the P␥-P␣␤ interaction in order to displace the P␥ C terminus from the P␣␤ catalytic site (5,8,20). Accordingly, a weak interaction between the P␣␤ catalytic domain and the P␥ C-terminal domain should therefore provide ␣ t GTP a stronger competitive edge over P␣␤, ensuring an efficient PDE6 activation. Moreover, tight binding of the P␥ C-terminal domain to ␣ t GTP helps keep P␥ from reinhibiting PDE6 before the visual signal is adequately amplified (20).
Second, a greater binding strength of P␣␤ for the P␥ central region keeps it sequestered by P␣␤. Because it is the complex formed by ␣ t GTP and the full-length P␥ that can be readily recognized by RGS9-1⅐G␤5 to maximally fulfill the GAP function (32), sequestration of the P␥ N-terminal half by P␣␤ may prevent visual signaling from being terminated too early. An extremely high rod visual sensitivity is thus achieved not only because of an exceptionally high efficiency of the activated PDE6 but also for its extended lifetime (1).
Furthermore, the complementarity of the P␥-␣ t interaction and the P␥-P␣␤ interaction also reveals some insights with regard to the molecular topology of the ␣ t ⅐PDE6 complex. ␣ t GTP␥S was co-immunoprecipitated by holo-PDE6, and P␣␤ was co-immunoprecipitated by ␣ t GTP␥S, all in a P␥-dependent fashion (Fig. 5). This indicates a molecular organization Our previous studies revealed that the P␥ Phe 30 region preferred binding to P␣, whereas the Ser 40 region favored binding to P␤, suggesting simultaneous P␥ interactions with P␣ and P␤ (16,28). Because PDE6 can only be efficiently activated 50% by transducin (26,46), it is highly likely that only one P␥ is displaced by ␣ t GTP during PDE6 activation in the mammalian retina, whereas the other P␥ (on the opposite side) stays tightly bound to P␣␤ (1). A movement of the P␥(78 -87) segment from P␣␤ to ␣ t GTP was suggested by Barren et al. (5), based on the crystal structure of the P␥(70 -87)⅐PDE5/6 catalytic domain complex. MAY 14, 2010 • VOLUME 285 • NUMBER 20

Transducin-PDE6 Interaction
such that the C-terminal domain of P␥ binds tightly with ␣ t GTP␥S while the central region forms a strong interaction with P␣␤, thus "gluing" ␣ t GTP␥S and P␣␤ into the PDE6 activation complex, or "transducisome." This proposed organization is also supported by the data from co-pull-down of ␣ t GTP␥S and P␣␤ by Btn-P␥ (Fig. 6). Thus, the complementary P␥ interactions may explain a long held puzzle; although the P␣␤-binding regions and ␣ t -binding regions overlap on P␥ (1,9), an intermediate transducin⅐PDE6 complex could still occur during visual transduction (22)(23)(24). Our data do not exclude the possibility that in the PDE6 activation complex, the P␥ central region may also be involved in binding with ␣ t GTP, albeit probably through weak interactions. A mutagenesis study showed that Lys 41 , Lys 44 , and Lys 45 on the C-terminal side of the P␥ polycationic region were involved in the interaction with ␣ t but not with P␣␤ (39), raising the possibility of simultaneous non-competitive ␣ t GTP␥S and P␣␤ binding to the P␥ central region.
Early studies using rod disc membranes suggested that direct ␣ t -P␣␤ interaction accounted for an important binding force in the ␣ t ⅐PDE6 complex (24, 27). In the current study, however, ␣ t GTP␥S and P␣␤ were not co-immunoprecipitated with each other in the absence of P␥ (Fig. 5). Moreover, when either ␣ t GTP␥S or P␣␤ was bound to Btn-P␥ (46 -87) or HisP␥(1-61) on affinity beads, the other was not co-pulled down (Fig. 3). These results indicate a lack of direct ␣ t -P␣␤ interaction in the ␣ t ⅐PDE6 complex under our experimental conditions. Because no disc membranes were involved in our experiments, we suggest that the disc membranes used in the early studies may have played a role in organizing the proteins in such a way that ␣ t and P␣␤ make direct contacts that further tighten the ␣ t ⅐PDE6 complex. In support of this proposition, increasing evidence indicates that disc membranes enhance protein functions in phototransduction (40 -42). Nevertheless, our data indicate that the complementary P␥ interactions with its two targets, at least in part, account for the binding force in the ␣ t ⅐PDE6 complex.
Based on the data presented herein and evidence from previous studies, a possible scenario of protein-protein interactions during PDE6 activation is depicted in Fig. 7. GTP-bound ␣ t may initially engage PDE6 by making contacts with the P␥ residues Trp 70 and Leu 76 , which are not involved in intimate interactions with the P␣␤ catalytic site (5,8,11), and may also interact via part of the polycationic region (19,39) (Figs. 1B and  2B). These initial ␣ t contacts with P␥ may trigger a conformational change that results in a rigid body movement of the P␥ C terminus away from the P␣␤ catalytic site with Leu 76 serving as a "hinge." The P␥ C-terminal domain can now bind tightly with ␣ t , and PDE6 is deinhibited (5). At this stage, an intermediate complex containing ␣ t GTP⅐P␥⅐P␣␤, or transducisome, probably exists due to the complementary binding of P␥ to ␣ t GTP and P␣␤ (Figs. 4 -6), with the P␥ C-terminal domain tightly bound to ␣ t GTP and the central region bound to the P␣␤ GAF domain with high affinity. The P␥ central region could stay bound to the P␣␤ GAF domain until the binding is weakened by lowered cGMP levels, through a mechanism of positive cooperativity of P␥ and non-catalytic cGMP in binding to the GAF domain (1, 14, 26, 43, 44).
Thus, the duration of the ␣ t GTP⅐P␥⅐P␣␤ transducisome is probably subject to regulation by the cGMP occupancy in the P␣␤ GAF domain. However, the activity of the GAP complex may have a more significant effect on the lifetime of the transducisome, because GTP hydrolysis in ␣ t accelerated by RGS9-1 is the rate-limiting step of the rod photoresponse (45). Interesting questions hereby arise as to how P␥ dynamically and differentially interacts with RGS9-1/G␤5, ␣ t , and P␣␤ upon a transition from its role in the PDE6 activation complex to that in the GAP complex and how the cGMP binding in the GAF domain regulates this process. Because disruption of P␥ interactions with its partners in phototransduction causes impaired visual functions (9), systematic investigations of these interactions will advance our understanding of the molecular mechanisms of the related retinal diseases.