The N termini of the inhibitory γ-subunits of phosphodiesterase-6 (PDE6) from rod and cone photoreceptors differentially regulate transducin-mediated PDE6 activation

Phosphodiesterase-6 (PDE6) plays a central role in both rod and cone phototransduction pathways. In the dark, PDE6 activity is suppressed by its inhibitory γ-subunit (Pγ). Rhodopsin-catalyzed activation of the G protein transducin relieves this inhibition and enhances PDE6 catalysis. We hypothesized that amino acid sequence differences between rod- and cone-specific Pγs underlie transducin's ability to more effectively activate cone-specific PDE6 than rod PDE6. To test this, we analyzed rod and cone Pγ sequences from all major vertebrate and cyclostome lineages and found that rod Pγ loci are far more conserved than cone Pγ sequences and that most of the sequence differences are located in the N-terminal region. Next we reconstituted rod PDE6 catalytic dimer (Pαβ) with various rod or cone Pγ variants and analyzed PDE6 activation upon addition of the activated transducin α-subunit (Gtα*-GTPγS). This analysis revealed a rod-specific Pγ motif (amino acids 9–18) that reduces the ability of Gtα*-GTPγS to activate the reconstituted PDE6. In cone Pγ, Asn-13 and Gln-14 significantly enhanced Gtα*-GTPγS activation of cone Pγ truncation variants. Moreover, we observed that the first four amino acids of either rod or cone Pγ contribute to Gtα*-GTPγS–mediated activation of PDE6. We conclude that physiological differences between rod and cone photoreceptor light responsiveness can be partially ascribed to ancient, highly conserved amino acid differences in the N-terminal regions of Pγ isoforms, demonstrating for the first time a functional role for this region of Pγ in the differential activation of rod and cone PDE6 by transducin.

Although the speed and sensitivity of the photoresponses of rod and cone photoreceptors are quite different, the underlying molecular components of the visual excitation pathway are believed to be homologous in both classes of photoreceptors (1,2). In rod photoreceptors, the cGMP signaling cascade is initi-ated when light-activated rhodopsin binds the heterotrimeric G protein transducin (Gt). 2 This leads to guanine nucleotide exchange of GDP to GTP and dissociation of the activated transducin ␣-subunit (Gt ␣ *-GTP). Gt ␣ *-GTP then associates with and activates the rod PDE6 holoenzyme (consisting of two catalytic subunits, ␣ and ␤, and two inhibitory rod-specific ␥-subunits, rP␥). Cone photoreceptors express homologous forms of the visual pigment (cone opsins), cone transducin, and cone PDE6 holoenzyme (comprised of two identical ␣' catalytic subunits and two cone-specific P␥ subunits, cP␥; for a review, see Ref. 3).
One approach to identify the biochemical differences in the rod and cone visual transduction pathways has been to genetically introduce a cone-specific isoform of a phototransduction gene into rod photoreceptor cells. As recently summarized (2), transgenic incorporation of various cone opsin genes, the cone transducin ␣-subunit, or the cone PDE6 ␣'-subunit into mouse rod photoreceptor cells had a very modest or no effect on the light sensitivity of the rod photoresponse. Biochemical and molecular approaches have identified several intrinsic differences between rod and cone opsins (e.g. spontaneous activation, chromophore dissociation, and regeneration), but evidence is lacking that differences in photoactivated rod and cone opsins can account for the different light sensitivity and photoresponse kinetics of rods and cones (4). Characterization of the biochemical properties of rod and cone transducin ␣-subunits have failed to reveal significant functional differences (5)(6)(7). Likewise, the PDE6 catalytic dimers in rods and cones have equivalent hydrolytic activities (8 -10).
The photoreceptor PDE6 is the only member of the 11-member phosphodiesterase superfamily that is known to be regulated by a distinct regulatory protein, the P␥ subunit (3,11). The mechanism of rod PDE6 activation involves binding of Gt ␣ *-GTP to the PDE6 holoenzyme, causing displacement of the C-terminal domain of rP␥ from the entrance of the active site to accelerate hydrolysis of cGMP. The lifetime of Gt ␣ *-GTP and activated rod PDE6 is determined by the GTPase activity of the transducin ␣-subunit, which is regulated by a complex of three GTPase-accelerating proteins (RGS9-1, G␤5L, and R9AP (12)). Importantly, the rP␥ subunit also serves as feedback regulator  Table S1. 1 To whom correspondence should be addressed. Tel.: 603-862-2458; E-mail: rick.cote@unh.edu.
of this GTPase accelerating proteins complex in rods (13). The cone cP␥ sequence is highly homologous to rP␥, and both P␥ isoforms bind to and inhibit the rod PDE6 catalytic dimer with similar affinity (10). However, the cone PDE6 holoenzyme is more easily activated by rod Gt ␣ *-GTP␥S than rod the PDE6 holoenzyme (8). The rod and cone P␥ genes (PDE6G and PDE6H) appear to have evolved with our earliest vertebrate ancestors at the same time when PDE6 catalytic subunit genes arose (14,15) with their unique catalytic and regulatory properties (3). It is noteworthy that the sea lamprey (Petromyzon marinus), a cyclostome that diverged from other vertebrate lineages about 500 million years ago, has a duplex retina in which rod and cone photoreceptors show similar physiological differences to light, as observed with mammalian photoreceptors (16,17). Interestingly, lamprey rods and cones express the same PDE6 catalytic subunit along with distinct rod-and cone-specific P␥ isoforms (18).
To test the hypothesis that differences in rP␥ and cP␥ subunits may contribute to the physiological differences in rod and cone light responsiveness, we examined whether amino acid sequence differences between rP␥ and cP␥ underlie the ability of transducin to more effectively activate cone PDE6 relative to rod PDE6. We first carried out a phylogenetic analysis of available P␥ subunit sequences to identify rod-and cone-specific sequence differences that might underly their different biochemical properties. We then reconstituted the purified rod PDE6 catalytic dimer (P␣␤) with different rP␥ or cP␥ mutants and tested the relative affinity and extent of catalytic activation upon addition of the persistently activated transducin ␣-subunit (Gt ␣ *-GTP␥S). Our results demonstrate that the N-terminal amino acids of rP␥ and cP␥ are responsible for the differential activation of PDE6 by transducin, from which we conclude that the evolution of separate genes for rod and cone P␥ represents one of the regulatory mechanisms distinguishing the rod and cone phototransduction pathways in vertebrate photoreceptors.

Evolutionary analysis reveals that rod P␥ is far more conserved than cone P␥
To identify differences in the sequences of the rod (PDE6G) and cone (PDE6H) P␥ genes, we sampled the largest phylogenetic diversity possible from public databases. The sequences we obtained include representation from PDE6G and PDE6H sequences from all major lineages of craniates, including cyclostomes (hagfish and lamprey), chondrichthyans (sharks and rays), neopterygian fishes (including teleosts), squamates (lizards and snakes), archosaurs (crocodilians and birds), and mammals. At present, PDE6G and PDE6H sequence data are lacking only for sturgeons, paddlefishes, and bichirs.
Phylogenetic analyses revealed strong support for two clades representing PDE6G and PDE6H for all vertebrate classes; however, because of strong sequence conservation, we found little support for shallow nodes across the tree (Fig. 1). Overall, we found that PDE6G sequences are highly conserved throughout craniates. In contrast, the PDE6H clade has undergone greater amino acid sequence diversification, especially outside of the mammalian class. In particular, PDE6H sequences for neopterygian fishes show radical alterations compared with other PDE6H sequences. In addition, we identified eight N terminus residues that are highly conserved in PDE6G but divergent or missing in PDE6H. We also uncovered two residues that are differentiated, but highly conserved, within PDE6G and PDE6H (Fig. 1).
According to the consensus logo ( Fig. 2), vertebrate PDE6G and PDE6H residues show high sequence similarity except for the N-terminal region of the protein sequence ( Fig. 2A). Within this N-terminal region, PDE6G sequences are significantly more highly conserved than PDE6H sequences, as judged by comparison of the consensus logo for all vertebrates ( Fig. 2A). This pattern of greater conservation of rod versus cone residues in the N-terminal region is even more evident when the analysis is confined to mammalian sequences (Fig. 2B).
Our analyses also reveal four highly conserved rod-cone differences among inhibitory subunit sequences that are virtually invariant since the last common ancestor of PDE6G and PDE6H (Fig. 2, A and B, indicated by vertical lines). With vertebrate PDE6G as a reference, these include residues Val-21, Gln-48, Asn-74, and Tyr-84 ( Fig. 2A). The mutually exclusive nature of these residues is even more strongly pronounced when considering mammalian sequences alone (Fig. 2B), per- Branch colors on the gene tree (left) correspond to the species tree (inset). The tree is rooted between putative rod and cone P␥ subunits. Multiple sequence alignment is plotted on the right. In the N-terminal region, several residues (Rod Conserved, gray boxes) are more than 90% conserved in the PDE6G alignment but divergent or missing in the PDE6H alignment. Two Rod/Cone Differentiated sequences (black boxes) are more than 80% conserved in either PDE6G and PDE6H alignments but differentiated between the two genes.

The N terminus of P␥ modulates transducin activation of PDE6
haps indicating purifying selection during the radiation of mammals. (Note that the insertion after Gly-59 of PDE6G ( Fig.  2A) was found in only a single species and is unlikely to have biological significance.) In summary, PDE6H sequences have diversified much more than their rod sisters in the evolutionary history of craniates, and much of this sequence diversity is found within the first ϳ15 N-terminal amino acid residues.

Gt ␣ *-GTP␥S activation of PDE6 is more effective with cone P␥ than rod P␥
It has been reported that the rod and cone PDE6 holoenzymes are very similar in their catalytic properties and their interactions with their inhibitory P␥ subunits but differ in how rod and cone PDE6 are activated by transducin (10). We hypothesized that structural differences in rod and cone P␥ isoforms may be responsible for these different interactions between transducin and rod and cone PDE6.
To test this, we first compared full-length recombinant rP␥ and cP␥ reconstituted with purified rod P␣␤ catalytic dimers and tested the ability of rod Gt ␣ *-GTP␥S to activate PDE6 and accelerate cGMP hydrolysis. Fig. 3A shows that PDE6 containing cP␥ is activated to a much greater extent (72% of the maximum catalytic activity of P␣␤ lacking P␥) than rP␥ (16% maximum activation) when tested under identical experimental conditions. In contrast to the 4.5-fold difference in the extent of activation, the concentration dependence of Gt ␣ *-GTP␥S activation (K1 ⁄ 2 ) did not significantly differ for the two PDE6 enzyme preparations (Table 1).
To evaluate whether this 4.5-fold difference in maximum transducin activation of PDE6 could be accounted for by a lower intrinsic binding affinity of cP␥ for P␣␤, we measured the concentration dependence of rP␥ and cP␥ to inhibit cGMP hydrolysis of P␣␤. As shown in Fig. 3B, no significant difference between IC 50 values of rP␥ and cP␥ was observed. We conclude that the differences in overall binding affinity of P␥ to P␣␤ cannot account for the greater efficacy of cP␥ to promote transducin activation of PDE6.

The N-terminal region of cone P␥ is the locus for the differences in transducin activation of PDE6
To identify the structural basis for the differences in the ability of rP␥ and cP␥ to permit efficient activation of PDE6 by transducin, we first hypothesized that one or more of the four highly conserved amino acid differences identified in the P␥ sequence alignment (vertical lines in Fig. 2B) were responsible. To test this, we generated site-directed mutants of cP␥ to see whether converting the cone residue to its rod counterpart would suppress the extent of transducin activation of reconstituted PDE6. As seen in Fig. 4A, the cP␥ triple mutant (K44Q,S70N,F80Y) was as effective as WT cP␥ in supporting Gt ␣ *-GTP␥S activation of PDE6; a similar result was observed for the cP␥T17V construct (Fig. 4A). From this result, we concluded that the variable N-terminal region of rP␥ and cP␥ was most likely responsible for the observed differences in the ability of transducin to activate PDE6.
To evaluate the role of the N-terminal region of P␥, we created two rod-cone P␥ chimeras consisting of the N-terminal region of one P␥ with the remaining C-terminal portion of the other P␥: rP␥1-18 -cP␥15-83 and cP␥1-14 -rP␥19 -87. We first verified that the two chimeric P␥ constructs inhibited P␣␤ with affinities similar to that of the WT P␥ proteins ( Table 1). As seen in Fig. 4B, the chimera containing the cP␥ N-terminal region behaved essentially the same as WT cP␥, whereas the N-terminal region of rP␥ enhanced Gt ␣ *-GTP␥S activation of PDE6 less than 2-fold (30% maximum activation) compared with WT rP␥ (16% maximum activation). We conclude that differences in the N-terminal region of rP␥ and cP␥, thought previously to lack functional significance, are responsible for regulating the ability of transducin to bind to and efficiently activate catalysis of the PDE6 holoenzyme.

Residues in the N-terminal region of rod P␥ impair the activation efficiency of Gt ␣ *-GTP␥S
Because of sequence dissimilarity in the N-terminal region of rP␥ and cP␥ sequences (Fig. 2), it was difficult to predict  Fig. 1. B, the mammalian consensus logo is based on a multiple sequence alignment of a subset of the entire set of sequences and consisting of 50 mammalian PDE6G sequences and 51 mammalian PDE6H sequences.

The N terminus of P␥ modulates transducin activation of PDE6
potential functional sites responsible for suppressing the ability of transducin to activate P␣␤ reconstituted with rP␥. Instead, we generated several N-terminal truncation mutants of rP␥ to test whether removing a portion of the N terminus would enhance Gt ␣ *-GTP␥S activation of PDE6. As shown in Fig. 5A, removal of the first four or eight amino acids from the rP␥ N terminus did not alter Gt ␣ *-GTP␥S activation efficiency, whereas removal of the first 18 amino acids resulted in less than a 2-fold increase in transducinactivated PDE catalysis.
We next constructed several internal deletion mutants of rP␥ to search for possible "inhibitory" elements within the N-terminal region that could suppress the ability of transducin to activate rod PDE6. Deletion of amino acid residues 5-8 (i.e. rP␥1-4_9 -87) resulted in a modest increase in the maximum extent of Gt ␣ *-GTP␥S activation of PDE6 (25% of P␣␤ activity; Fig. 5B and Table 1). Further enhancement of Gt ␣ *-GTP␥S activation of PDE6 was observed when amino acids 5-18 were deleted (55% of maximum P␣␤ activity; Table 1). A positive role of the first four amino acids of the rP␥ sequence in promoting transducin activation can be inferred by comparing the rP␥19 -87 truncation mutant (28% of maximum P␣␤ activity) with the internal deletion mutant (rP␥1-4_19 -87; 55% of maximum activity). We further localized the region of rP␥ that suppresses transducin activation of rod PDE6 by testing another internal deletion mutant of rP␥ lacking only amino acids 9 -18; in this instance, P␣␤ reconstituted with rP␥1-8_19 -87 achieved close to the same level of transducin activation (64%) as we observed for WT cP␥ (72%; Fig. 5B and Table 1). These results demonstrate that amino acids 9 -18 of mammalian rP␥ represent a highly conserved sequence (E(I/ F)RSATRV*G) that is absent in mammalian cP␥ and that significantly reduces the ability of transducin to maximally activate PDE6. These results also indicate that the first eight amino acids of rP␥ can enhance the ability of Gt ␣ *-GTP␥S to activate PDE6, although this effect is only observed when residues 9 -18 are deleted.
Having identified a rP␥-specific element that decreased the ability of transducin to fully activate PDE6, we sought to determine whether substituting the nearby highly conserved rodcone P␥ difference (Val-21 in rP␥ and Thr-17 in cP␥) would enhance transducin activation when the rod inhibitory region was deleted. Fig. 5C shows that the rP␥ construct (P␥1-8_19 -87V21T) achieved a slightly greater maximum extent of activation (80%) compared with WT cP␥ (72%); a similar enhancing effect of this V21T substitution was also observed when the N-terminal 18 amino acids were deleted (rP␥18 - Figure 3. PDE6 reconstituted with cP␥ is more effectively activated by Gt ␣ *-GTP␥S than PDE6 containing rP␥. A, Gt ␣ *-GTP␥S activation assay. 1 nM P␣␤ was preincubated with a 10-fold molar excess of WT rP␥ or cP␥ to reconstitute the PDE6 holoenzyme. Gt ␣ *-GTP␥S was then added at the indicated concentrations and incubated for 1 h prior to measuring PDE activity with 2 mM cGMP as substrate. PDE activity is reported relative to the activity of P␣␤ in the absence of P␥. The data are the mean (Ϯ S.D.) for 19 or 20 separate determinations of PDE6 reconstituted with WT rP␥ or cP␥, respectively. The data were fit by nonlinear regression analysis using a three-parameter hyperbolic equation and are reported in Table 1; basal activities for PDE6 reconstituted with rP␥ and cP␥ were 2% Ϯ 0.8% and 8% Ϯ 3%, respectively. B, P␥ inhibition assay. P␣␤ (0.2 nM) was incubated with the indicated concentrations of WT rP␥ or cP␥ for 10 min, and then the catalytic activity was measured using 2 mM cGMP as substrate. The data are the mean (Ϯ S.D.) of eight experiments with rP␥ and cP␥. A threeparameter logistic equation was used to estimate the IC 50 : rP␥, IC 50 ϭ 0.56 nM; cP␥, IC 50 ϭ 0.58 nM.

Table 1
Effectiveness of P␥ constructs in promoting Gt ␣ *-GTP␥S activation of PDE6 P␣␤ (1 nM) was preincubated with one of the rP␥ or cP␥ constructs (10 nM) to reconstitute the PDE6 holoenzyme. Various concentrations of Gt ␣ *-GTP␥S were then added and incubated for 1 h at room temperature. The PDE activity was measured using 2 mM cGMP as substrate and expressed as the percent activity referenced to fully activated P␣␤. The K1 ⁄ 2 and maximum percent activation were calculated by fitting the data to a three-parameter hyperbola. The IC 50 of all P␥ constructs was determined using 0.2 nM P␣␤ and 2 mM cGMP as substrates. The values represent the mean Ϯ S.D. for n individual experiments.

The N terminus of P␥ modulates transducin activation of PDE6
87V21T; Fig. 5D). When tested as a single site substitution, the rP␥V21T construct caused a 10% increase in Gt ␣ *-GTP␥S activation of PDE6 (Table 1). These results indicate that the valine at position 21 of the rP␥ sequence contributes to the decreased ability of Gt ␣ *-GTP␥S to bind to and effectively activate PDE6.

The cone P␥ residues Asn-13 and Gln-14 contribute to efficient Gt ␣ *-GTP␥S activation of PDE6
To determine which N-terminal residues of cP␥ modulate the ability of transducin to effectively activate PDE6, several cP␥ N-terminal mutants were created. Fig. 6A shows that  Table 1.

Figure 5. Residues in the N-terminal region of rP␥ affect the Gt ␣ *-GTP␥S activation efficiency.
A, three N-terminal truncated rP␥ mutants (rP␥5-87, rP␥9 -87, and rP␥19 -87) were reconstituted with P␣␤ to test Gt ␣ *-GTP␥S activation efficiency. B, three rP␥ mutants with internal deletions were tested to evaluate the role of the N terminus in promoting PDE6 activation by Gt ␣ *-GTP␥S. C and D, Gt ␣ *-GTP␥S activation assays of PDE6, in which rP␥ mutants were tested with and without the site-directed mutant V21T. Gt ␣ *-GTP␥S activation assays were performed and analyzed as described in Fig. 3A, and the K1 ⁄2 and percent of maximum activation are summarized in Table 1.

The N terminus of P␥ modulates transducin activation of PDE6
removal of the first four (cP␥5-83) or eight (cP␥9 -83) residues of cP␥ has an insignificant effect on the ability of Gt ␣ *-GTP␥S to activate PDE6 containing these mutant cP␥ constructs. Although truncation of the first 14 amino acids from cP␥ (cP␥15-83) significantly reduced the maximum extent of transducin activation, inclusion of amino acids Asn-13 and Gln-14 (cP␥13-83) was sufficient to restore transducin activation to levels observed for WT cP␥ (Fig. 6B and Table 1).
Because the first four or eight N-terminal amino acids of rP␥ were able to enhance the ability of Gt ␣ *-GTP␥S to bind to and activate PDE6 when the rP␥ inhibitory region was deleted (Fig.  5B), we also investigated whether the N-terminal amino acids of cP␥ contributed to transducin activation efficacy. Comparison of cP␥9 -83 (62% of P␣␤ activity; Fig. 6A and Table 1) with cP␥1-4_9 -83 (80% maximal activation; Fig. 6C and Table 1) revealed a significant enhancing effect of the first four amino acids of cP␥. More significantly, the impaired activation efficiency of cP␥15-83 (Fig. 6B) could be completely reversed by inclusion of the first four or eight N-terminal amino acids (Fig.  6C), rising from 45% for cP␥15-83 to 83% of the P␣␤ activity for either cP␥1-4_15-83 or cP␥1-8_15-83 (Table 1). We conclude from Fig. 6 that either residues Asn-13 and Asn-14 or the first several N-terminal amino acids of cP␥ are sufficient for Gt ␣ *-GTP␥S to maximally activate PDE6 containing these cP␥ constructs.

Discussion
Using a phylogenetic analysis to identify functionally important differences between the rod and cone isoforms of P␥ in conjunction with biochemical assays of transducin activation of PDE6 reconstituted with different rP␥ and cP␥ mutants, we determined that some of the well-established physiological differences in light responsiveness of rod and cone photoreceptors (1,2) can be ascribed to structural differences in the N-terminal region of rod and cone P␥. This work also identifies, for the first time, a regulatory role of the N terminus of P␥ in modulating the ability of transducin to bind to the PDE6 holoenzyme and relieve the inhibitory constraint on cGMP hydrolysis imposed by the last 10 C-terminal residues of P␥ (19).
Current evidence suggests that the physiological differences in the response to light exhibited by rods and cones had already evolved prior to the last common ancestor of living vertebrates (16,18,20). Our analyses of P␥ sequence evolution provide additional context to this suggestion, as we demonstrate that these physiological differences are explained in part by ancient divergences in the N terminus region of P␥. A recent paper by Lagman et al. (15) analyzed P␥ sequences but did not attempt a phylogenetic analysis because of high sequence conservation and a proposed lack of phylogenetic signal. However, our phylogenetic analyses of P␥ demonstrate two well-supported Figure 6. Identification of residues in the N-terminal region of cP␥ that enhance Gt ␣ *-GTP␥S activation efficiency. A and B, four N-terminal truncated cP␥ mutants (cP␥5-83, cP␥9 -83, cP␥13-83, and cP␥15-83) were reconstituted with P␣␤ to test the Gt ␣ *-GTP␥S activation efficiency of PDE6. C, three internal deletion mutants of cP␥ (cP␥1-4_9 -83, cP␥1-4_15-83, and cP␥ 1-8_15-83) were also tested under the same experimental conditions. The Gt ␣ *-GTP␥S activation assay was carried out and analyzed as described in the legend for Fig. 3A, and the results are summarized in Table 1.

The N terminus of P␥ modulates transducin activation of PDE6
clades consisting of PDE6G (rod) and PDE6H (cone) sequences. We do observe generally low support for internal nodes by either bootstrapping or approximate likelihood ratio indices, but the delineation between the major clades is clear. Lagman et al. (15) also proposed a new clade of P␥ called PDE6I. In our analyses, PDE6I sequences are found within the PDE6H clade. This finding does not dispute the existence of PDE6I genes, but our phylogenetic results indicate that they are best interpreted as duplicate PDE6H genes. We chose to root our phylogeny in the most evolutionarily parsimonious manner, which places P␥ sequences from the hagfish Eptatretus stoutii at the base of both the rod and cone clades. However, we note that electrophysiological evidence for cyclostome photoreceptor light responsiveness exists only for the lamprey (17,21,22), and other rootings for the P␥ phylogenetic tree could change the affinity of the hagfish P␥ sequences. On the amino acid sequence level, we find that PDE6G sequences are generally far more conserved than PDE6H sequences ( Figs. 1 and 2). Because rod phototransduction shows single-photon sensitivity and has been considered a paragon of the efficacy of natural selection (23,24), this strong conservation among PDE6G sequences could be interpreted as evidence for purifying selection. Alternatively, sequence divergence among PDE6H sequences might be an indication of diversifying selection, as the kinetics of cone phototransduction and its role in color vision might facilitate selection for adaptive but alternative regulatory properties. These two scenarios are not mutually exclusive, and functional studies from the cyclostomes will be needed to provide further resolution.
In addition to highly reliable single photon detection, rod photoreceptors differ from cone photoreceptors in having a lower level of continuous "dark noise" in their dark-adapted state (25,26) as well as a longer latency period for the initial "rising" phase of the response to flash stimuli (27). Recently, it has been reported that rod PDE6 activation by transducin requires binding of two activated transducins to the rod PDE6 holoenzyme (28); furthermore, computer simulations of this dimeric activation model for rod PDE6 (29) support the idea that this dimeric transducin activation mechanism serves to reduce rod PDE6 spontaneous activation (i.e. reduced dark noise) and accounts for the longer delay in the rising phase of visual excitation compared with the cone photoresponse in vivo. Because cone PDE6 consists of a catalytic homodimer, Lamb et al. (29) postulated that the symmetrical cone catalytic dimer may permit each cone catalytic subunit to be activated by transducin binding in an independent, noncooperative manner. Our results support an alternative hypothesis in which differences between rod and cone P␥ isoforms in the N-terminal region of the protein and not differences between rod and cone PDE6 catalytic subunits determine whether transducin can activate each catalytic subunit independently (in the presence of cP␥) or, alternatively, require binding of two transducins to the rod PDE6 holoenzyme (containing rP␥) for maximal activation. However, it should be pointed out that our in vitro biochemical data were obtained with protein concentrations several orders of magnitude lower than in the photoreceptor outer segment, and it will be important to evaluate the importance of the N-terminal region of P␥ under physiological conditions. Previous work delineating the sites of interaction of the intrinsically disordered rP␥ subunit (11) with the rod PDE6 catalytic dimer have identified several functionally important regions (Fig. 7): the approximately 10 C-terminal amino acid residues of rP␥ are known to directly interact with the catalytic domain to regulate access of substrate to the active site and directly control catalytic activity (19,30); a region in the N-terminal half of rP␥ (amino acids 10 -30) enhances the affinity of cGMP to noncatalytic binding sites in the regulatory domain of the catalytic dimer (31), most likely by direct binding of this region of rP␥ to the cGMP binding pocket (32,33); and multiple binding sites of rP␥ with P␣␤ have been identified along practically the entire length of the rP␥ sequence, with its central region making a major contribution to the high overall affinity of rP␥ for P␣␤ (31). In contrast to these functionally important regions of rP␥, the N-terminal region of rP␥ (specifically its first 17 residues) was not observed to contribute to the ability of P␥ to inhibit cGMP hydrolysis or to stimulate cGMP binding (31). Because of the difficulty of isolating biochemical quantities of purified cone PDE6 holoenzyme, comparable characterizations of the interactions of cP␥ with cone catalytic homodimer are lacking.
Upon rod PDE6 activation by transducin, Gt ␣ *-GTP interacts with multiple sites within the C-terminal half of rP␥ (Fig. 7, black boxes) to displace rP␥ from the catalytic domain, allowing diffusion of cGMP into the active site (34 -36). Studies of the interactions of rP␥ with the activated transducin ␣-subunit have also identified a second region of interaction in the polycationic central region of rP␥ (37,38) whose function is uncertain. This work is the first report showing that the N-terminal region of P␥ (preceding its polycationic region) plays an important role in Gt ␣ *-GTP activation of PDE6. For both rP␥ and cP␥, Figure 7. Functionally important sites of the rP␥ and cP␥ inhibitory subunits. Blue-shaded residues represent highly conserved rod-cone P␥ differences. Solid green boxes represent regions that enhance Gt ␣ * activation. The solid red boxes represent the regions that impair Gt ␣ * activation. The red letters identify two amino acids in the bovine P␥ sequences that differ from the mammalian consensus sequences: M17V (rP␥) and T11A (cP␥). Also shown are previously identified regions that enhance rP␥ inhibition potency (white boxes), enhance noncatalytic cGMP binding to the GAFa domain (gray box), or are critical for activation of PDE6 by activated transducin (black boxes) (31).

The N terminus of P␥ modulates transducin activation of PDE6
the first four amino acids significantly contribute to enhancing the maximum extent of transducin activation under conditions where neighboring amino acids have been deleted (⌬5-8; Figs. 5 and 6). Of greater significance are the inhibitory (rP␥ residues 9 -18 and Val-21; Fig. 5) and stimulatory loci (cP␥ Asn-13 and Gln-14; Figs. 4 and 6) we discovered that modulate the efficacy with which transducin can bind to P␥ and activate PDE6 catalysis.
The fact that these sites near the N terminus of P␥ overlap with the region of rP␥ shown previously to enhance cGMP binding affinity to the GAFa domain (Fig. 7) is consistent with the following allosteric mechanism. We hypothesize that rP␥ residues 9 -21 may be allosterically coupled to the central polycationic region of rP␥ (residues ϳ22-45) and reduce the affinity of transducin binding to this region of rP␥, which, in turn, reduces the ability of transducin to bind to and displace the C-terminal residues of rP␥ required for disinhibition of PDE6 catalysis. Conversely, residues 13-17 in cP␥ may serve to allosterically enhance transducin binding to the central region of cP␥, facilitating full activation of PDE6 catalysis. The lower affinity of cGMP for the GAFa domains of cone PDE6 than for rod PDE6 (8,39) and the well-established reciprocal regulation of rP␥ and cGMP binding affinity to the PDE6 catalytic dimer (reviewed in Ref. 3) also likely contributes to differences in the allosteric regulation of rP␥ and cP␥ interactions with activated transducin.
In conclusion, this work demonstrates that the N-terminal region of the rod and cone inhibitory P␥ subunits allosterically regulate the efficiency with which activated transducin is able to displace P␥ from the active site of the PDE6 holoenzyme. This supports the hypothesis that the activation mechanism of PDE6 differs in rods and cones and can account for at least some of the physiological differences in rod and cone light responsiveness. These results will also contribute to a better understanding of the molecular etiology of disease-causing mutations in the catalytic and inhibitory subunits of photoreceptor PDE6, particularly those that occur in the regulatory domains of these proteins (Ref. 40 and references cited therein).

Materials
Bovine retinas were purchased from W. L. Lawson, Inc. Synthetic peptides (rP␥19 -87 and cP␥15-83) were purchased from New England Peptide. All synthetic DNAs of rP␥ and cP␥ mutants were purchased from Thermo Fisher-Invitrogen. Sitedirected mutants of P␥ were generated using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies). Filtration membranes and chemicals were from Millipore-Sigma.

Sequence alignments and phylogenetic analyses
PDE6G and PDE6H sequences from a diversity of vertebrate species were obtained from Uniprot or from the literature (14,18). Additional sequences were obtained from the gene models of 56 additional craniate species with whole-genome sequences using a custom phylogenomics pipeline. This pipeline uses BLAST (e value ϭ 0.0001) to search individual datasets using PDE6G and PDE6H query sequences (CNRG_HUMAN, P18545.1; CNCG_BOVIN, P22571.1) (41). Redundant sequences were then removed using cdhit (Ϫc ϭ 1.0 (42)), and the resulting dataset was aligned using the progressive algorithm implemented in PASTA with default parameters (43). Following the initial PASTA alignment, we removed four sequences that were either shorter than 50 amino acid residues or had internal deletions longer than 30 amino acids. Sequences were then realigned using PASTA under default parameters; we confirmed that additional iterations of progressive alignment did not produce improvements. The resulting P␥ amino acid sequences from different vertebrate species are listed in Table  S1.
The phylogeny of PDE6G and PDE6H was then estimated using IQ-tree under the best fit model (44); see Fig. S1 for the resulting tree with species labels. Node support was ascertained using both bootstrap and approximate likelihood ratio scores (45). The final tree with tip label and approximate likelihood ratio scores is available as a .tre file in the Github repository associated with this paper. All scripts and command lines used in sequence analyses are located at https://github.com/plachetzki/PDE6_GAMMA. 3

Construction and purification of P␥ mutants
Codon-optimized synthetic DNA fragments coding for various bovine rP␥ and cP␥ constructs used in this study (Fig. S1) were inserted into the NdeI and BamHI sites of the pET11a vector, followed by transformation into the Escherichia coli BL21(DE3) strain. The sequence of all P␥ mutants was confirmed by DNA sequencing (Functional Biosciences).
Following expression of recombinant P␥ mutants in E. coli BL21(DE3), the bacterial extract was purified by HiTrap SP FF column from GE Healthcare. The P␥ mutants were further purified by C18 reverse-phase HPLC following standard procedures (46). The purity of these proteins was determined to be Ͼ90%, as judged by SDS-PAGE. Protein concentrations were determined by the bicinchoninic acid protein assay using bovine ␥-globulin as a standard.

PDE6 and P␣␤ purification and activity assays
The bovine rod PDE6 holoenzyme was purified from bovine retinas as described previously (47). P␣␤ catalytic dimers were prepared by limited trypsin proteolysis and repurified by Mono Q anion exchange chromatography prior to use (47). PDE6 catalytic activity was measured in 20 mM HEPES, 10 mM MgCl 2 , and 0.5 mg/ml BSA using a colorimetric assay (48). The PDE6 concentration was estimated based on the rate of cGMP hydrolysis of trypsin-activated PDE6 and knowledge of the k cat of the enzyme (5600 mol cGMP hydrolyzed per 1 mol P␣␤ per second (39). The IC 50 values of all P␥ constructs reported in Table 1 were determined using 0.2 nM P␣␤ and 2 mM cGMP as substrates (39,49); in all cases, the IC 50 values did not differ significantly from those observed for WT rP␥ and cP␥ (Fig. 3B). Note that the difficulty in obtaining biochemical quantities of purified bovine cone photoreceptor PDE6 holoenzyme and bovine cone transducin precluded carrying out complementary experiments with these cone isoforms.

Purification of Gt ␣ *-GTP␥S and measurements of transducin activation of PDE6
Transducin ␣-subunits were extracted from the PDE6-depleted rod outer segment membranes by addition of 50 M GTP␥S. The extracted Gt ␣ *-GTP␥S was purified on a Blue Sepharose column as described previously (50,51), followed by gel filtration chromatography to remove residual PDE6. The concentration of Gt ␣ *-GTP␥S was determined by a colorimetric protein assay. Purified Gt ␣ *-GTP␥S (in 50% glycerol) was stored at Ϫ20°C.
To measure transducin activation of PDE6, 1 nM P␣␤ was preincubated with 10 nM P␥ mutant for 10 min prior to addition of the indicated concentrations of Gt ␣ *-GTP␥S for 1 h at room temperature. The PDE6 activity was then assayed at a final concentration of 0.2 nM using 2 mM cGMP as substrate. Note that these assay conditions were chosen to maximize differences between P␣␤ reconstituted with rP␥ and cP␥; however, even with a 10-fold higher concentration of purified rod PDE6 holoenzyme (2 nM) and a 1000-fold excess of Gt ␣ *-GTP␥S, we failed to observe more than 50% maximal activation by Gt ␣ *-GTP␥S (similar to literature values; see Refs. 8, 10, 31).

Data analysis
All experiments were repeated at least three times. The transducin concentration dependence of PDE6 activation (Figs. 3A and 4 -6) was analyzed by nonlinear regression (Sigmaplot v.12.5, SPSS, Inc.) using a three-parameter hyperbolic equation: y ϭ y 0 ϩ a ϫ x/(b ϩ x), where a is the maximum percent activation, b is the K1 ⁄ 2 , and y 0 is the basal PDE6 activity in the absence of Gt ␣ *-GTP␥S. P␥ inhibition potency (Fig. 3B) was determined using a three-parameter logistic equation: y ϭ a/(1 ϩ (x/x 0 ) b ), where a is the amplitude, b is the slope, and x 0 is the IC 50 value.