Introduction

Top Summary Introduction Materials and Methods Results Discussion References

Cyclic nucleotide-gated (CNG) channels are plasma membrane cation channels directly activated by cytoplasmic cAMP or cGMP (Fesenko et al., 1985; Nakamura and Gold, 1987). They play important roles in visual (Yau and Baylor, 1989) and olfactory (Zufall et al., 1994) signal transduction. For every cloned CNG channel subunit, the deduced amino acid sequence contains a "core" channel domain, followed by a carboxyl-terminal cyclic nucleotide-binding domain. Similar to voltage-gated channels, the core has six putative transmembrane segments and a P region, which constitutes part of the pore (Jan and Jan, 1990). The binding domain is homologous to the cyclic nucleotide-binding sequences conserved from the Escherichia coli catabolite gene-activator protein (CAP), an E. coli transcription regulator, to the regulatory subunits of protein kinase A (PKA) or protein kinase G (Shabb and Corbin, 1992).

The atomic structures of CAP (McKay and Steitz, 1981; Weber and Steitz, 1987) and a type 1 regulatory subunit of bovine PKA (PKA-R1) (Su et al., 1995) have been solved. In both proteins, each binding site consists of three -helices and a distinctive eight-strand, antiparallel -barrel. The availability of these structures allowed the use of homology modeling to construct tertiary structures of binding domains in CNG channels and to predict some important ligand-protein contact points (Kumar and Weber, 1992; Scott et al., 1996).

An early site-directed mutagenesis study identified an alanine/threonine difference that partially underlies ligand discrimination in CNG channels (Altenhofen et al., 1991). However, structural predictions at several other residues have not been tested. At a position that aligns with leucine 61 (Leu61) of CAP, for example, it was predicted that the aromatic residue at this site in CNG channels would form an important contact with bound ligand (Kumar and Weber, 1992; Scott et al., 1996). To test this hypothesis, we have introduced mutations at the predicted site [i.e., tyrosine 512 (Y512) in the  subunit of rat olfactory channel (rOCNC1) (Dhallan et al., 1990) and phenylalanine 533 (F533) in the  subunit of bovine rod photoreceptor channel (Brcng) (Kaupp et al., 1989; Gordon and Zagotta, 1995)]. A strong involvement of the predicted residue in either ligand binding or the conformational changes after binding (gating) would result in significant shifts in the EC₅₀ values (for a review, see Li et al., 1997). Our results, reported herein, revealed shifts in the EC₅₀ values of 2-fold, even with nonaromatic or charged residues. This calls into question the accuracy of the predictions and weakens the hypothesis that the tyrosine/phenylalanine difference at this position between the photoreceptor and the olfactory channels contributes to the fact that both cAMP and cGMP are potent agonists at the olfactory channels, whereas only cGMP can effectively open the photoreceptor channels.

In each of the two cAMP-binding domains of PKA-R1, there is a stacking interaction between an aromatic residue and the adenine ring of cAMP (Su et al., 1995), similar to the interactions predicted for CNG channels. However, the positions in PKA that align to CAP Leu61 are not involved. The residues that are involved, tryptophan 260 and tyrosine 371 (Fig. 1, bold), are situated at the extremes of the amino- and carboxyl-termini of repeat B, and project from outside back into the two binding pockets (Su et al., 1995). It is therefore worth testing whether such aromatic-aromatic interactions also make significant contributions to ligand binding in CNG channels and, if so, which one of the aromatic residues other than rOCNC1-Y512 or Brcng-F535 serves this function.

View larger version (46K): [in this window] [in a new window]   Fig. 1.   Sequence alignment of cyclic nucleotide-binding domains. Shown are rOCNC1, Brcng, the two binding domains (repeat A and repeat B) in PKA R1 (/A) and (/B), and CAP. The six invariant residues are bold and numbered. Also highlighted are Y512 and Y565 in rOCNC1 and the aligning residues in Brcng, F533 and Y586. These two positions were studied in this project. The  helices and  strands defined in the three known structures are underlined. V193 in RI (/A) aligns with Y196 in RII (/A), which can be affinity labeled with cAMP analogs (Bubis and Taylor, 1987), suggesting that the bound ligand is close to this portion of the binding domain.

There are typically six to eight aromatic residues in a CNG-channel binding domain of ~120 amino acids. To examine their functional importance, we changed them to alanine, one at a time, in the binding domain of rOCNC1. In this series of experiments, all but one aromatic residue was mutated; we expressed the mutant channels in Xenopus laevis oocytes and measured the dose-response relations. We also mutated some of these residues to leucine, tryptophan, glutamate, or serine. Some of the alanine mutations rendered the channel nonfunctional, whereas all but one of the functional mutants had only minor shifts in sensitivity. The interesting exception, Y565A in rOCNC1, showed 10-fold greater sensitivity to both cAMP and cGMP. The effect is caused, at least in part, by facilitated gating transitions, as indicated by the prominent spontaneous activities. Replacement of Y565 by other residues revealed that for the residue at this position, the function is related more to the size of its side chain than to such properties as the aromaticity or charge.

	Materials and Methods

Top Summary Introduction Materials and Methods Results Discussion References

Mutagenesis. The cDNA for rOCNC1, kindly provided by Dr. K.-W. Yau (Howard Hughes Medical Institute, Johns Hopkins University School of Medicine, Baltimore, MD), was subcloned into pGEMHE (Liman et al., 1992) at the EcoRI and HindIII sites, between the 5'- and 3'-untranslated sequences of the X. laevis major -globin gene for enhanced expression in oocytes. The cDNA for Brcng, already in pGEMHE, was kindly provided by Dr. William Zagotta (University of Washington, Seattle, WA).

Expression. cRNA was synthesized in vitro (Ambion T7 mMESSAGE mMACHINE kit; Ambion, InC., Austin, TX) using plasmid linearized with PstI as template. Stage V and VI X. laevis oocytes were each injected with 50 nl of cRNA, with concentrations ranging from 20 to 300 ng/µl. The oocytes were incubated in ND96 solution, containing 96 mM NaCl, 2 mM KCl, 1 mM MgCl₂, 1.8 mM CaCl₂, and 5 mM HEPES, pH 7.4. Recordings were made at room temperature from 24 to 120 h after injection. To improve the viability of oocytes, horse serum (HyClone Laboratories, Logan, UT) was added at 5% to the incubation solution (Quick et al., 1992).

Electrophysiological Recording and Analysis. All recordings were performed at room temperature from inside-out patches in symmetrical solutions containing 140 mM NaCl, 5 mM HEPES, and 0.2 mM EDTA, pH 7.4. The oocytes were stripped of the vitelline membrane as described previously (Quick et al., 1992), and membrane seals were formed in ND96. The patch was then excised by withdrawing the pipette, and excision was signaled by the flow of current through the endogenous Ca²⁺-activated Cl channels. The perfusion solutions containing various concentrations of cAMP or cGMP were applied to the patch using an RSC100 rapid solution changer (Molecular Kinetics, Pullman, WA). Upon perfusion of divalent cation-free solution containing no cyclic nucleotide, the endogenous Ca²⁺-activated Cl current disappeared, leaving a patch with a resistance of 1 to 4 G. cAMP and cGMP were both obtained from Sigma Chemical Co. (St. Louis, MO).

3dB

	Results

Top Summary Introduction Materials and Methods Results Discussion References

The sequence alignment shown in Fig. 1 includes the binding domains of Brcng, rOCNC1, the two tandem binding domains (repeat A and repeat B) in the splice variant  of PKA-R1 [denoted R1 (/A) and R1 (/B)], and CAP. In the latter three sequences, the secondary structural motifs identified in the atomic structures are underlined. These include three -helices (A, B, and C), and eight -strands (1 through 8). In PKA-R1, there is also an additional  helix (B') between 6 and 7. Highlighted in bold are the six residues absolutely conserved in the complete alignment of the more than 40 CAP-related cyclic nucleotide-binding proteins. They are at positions aligning with CAP G33, G45, G71, E72, R82, and A84.

rOCNC1-Y512F and Brcng-F533Y. Brcng F533 and rOCNC1 Y512 are also highlighted in bold in Fig. 1. They align with each other, and according to the model of Scott/Tanaka (Scott et al., 1996), a tyrosine (Y) at this position would form a strong interaction with either cAMP or cGMP, whereas a phenylalanine (F) would form only a weak interaction. In addition to this prediction, there are other reasons to study aromatic residues at this position: 1) A tyrosine in repeat A of bovine PKA-RII, Y196 (sequence not shown), can be affinity-labeled by cyclic nucleotide analogues (Bubis and Taylor, 1987). Y196 in PKA-RII (/A) aligns with V193 of R1 (/A) (the latter is shown in Fig. 1) and is near Y512 of rOCNC1. This strongly suggests a close contact between this portion of the binding domain and the ligand molecule. 2) The position aligning with Y512 in rOCNC1 is Y in most olfactory channels (the catfish olfactory channel is the only exception, with F) and F in most of the rod or cone photoreceptor channels (Fig. 2). The olfactory channels can be potently activated by both cGMP and cAMP, whereas for photoreceptor channels, only cGMP is an effective ligand. This correlation between Y/F identity and agonist specificity raised the possibility that an F at this position in photoreceptor channels contributes to selectivity to cGMP over cAMP. In olfactory channels, if the extra hydroxyl group in tyrosine makes a hydrogen bond with cAMP, but not with cGMP, it could provide stronger affinity or greater efficacy to cAMP, offsetting, if only partially, the selectivity against cAMP displayed by photoreceptor channels. A typical hydrogen bond brings 3 to 7 kcal/mol of free energy difference; at room temperature, 1.36 kcal/mol produces a 10-fold change in an equilibrium constant. Therefore if Y-to-F or F-to-Y mutations add or eliminate one hydrogen bond in either ligand binding or the subsequent conformational changes, a dramatic shift in EC₅₀ value is expected.

View larger version (37K): [in this window] [in a new window]   Fig. 2.   Sequence alignment of a portion of the cyclic nucleotide-binding domain. Highlighted are the residues at the position homologous to Y512 in rOCNC1.

View larger version (17K): [in this window] [in a new window]   Fig. 3.   Dose-response comparison between wild-type rOCNC1 and its mutant Y512F. The current responses at 80 mV were fitted to the Hill equation with Imax, EC50, and nH as free parameters. The current values were then normalized to Imax and refitted using only EC50 and nH as free parameters. The horizontal axis is on a logarithmic scale.

View larger version (13K): [in this window] [in a new window]   Fig. 4.   Dose-response comparison between wild-type Brcng and its mutant F533Y at 80 mV. The fitting and normalization were as described in Fig. 3. The horizontal axis is on a logarithmic scale.

Mutations of Other Aromatic Residues. To examine the functional importance of other aromatic residues in the binding domain, we mutated six additional aromatic residues in the binding-site region of rOCNC1. These were F477, Y482, Y494, Y547, F551, and Y565. Along with Y512, seven of the eight aromatic residues in the rOCNC1 binding domain have been examined. We were unable to generate a mutation for the eighth aromatic residue, F521. The annealed fragments failed to amplify during the "bridge" PCR, and this position was not studied further.

Y565A, a Hypersensitive Mutant. The last mutation to be described, Y565A, increased the sensitivity for both cAMP and cGMP (Fig. 5). The dose-response relations shown in Fig. 6 and the parameters in Table 1 indicate approximately 10-fold reductions in the EC₅₀ value. The major properties of all of the mutations studied were summarized in Fig. 7. Y565A had the same conductance as the wild-type channel (~44 pS at 60 mV, more details below), yet the expression level was reduced by 10- to 30-fold. Single-channel recordings showed 5 to 30% spontaneous opening probabilities (Fig. 8A), which fluctuated in time (Fig. 8B) and were variable among oocytes both within and across batches. The prominence of spontaneous activities provides a strong indication for facilitated gating transitions in the mutated channels. The open- and closed-time histograms of the spontaneous openings contained multiple components (Fig. 9), indicating complex underlying kinetic states. In the presence of cyclic nucleotides, the maximum open probability was well over 50% and approached 100% in several patches. With increasing concentrations of cAMP, the open times became longer; the closed times became correspondingly briefer (Fig. 10). There was no bursting behavior. In the presence of cyclic nucleotide, there were also fluctuations in P_open, which often led to higher sensitivities of the channel. After "sensitization", the channel maintained a high level of activity even after prolonged washing with the ligand-free solution. The parallel mutation in Brcng, Y586A, failed to express. Y586F and Y586W, more conservative mutations, increased cGMP sensitivity by 2-fold and 30%, respectively (Table 1). Apparently, in Brcng, this position is much less tolerant to changes than in rOCNC1.

View larger version (23K): [in this window] [in a new window]   Fig. 5.   Representative current-voltage families during voltage ramps from 80 mV to 120 mV of 800-ms duration. Left, rOCNC1; right, rOCNC1-Y565A. A, cAMP. B, cGMP. The basal current recorded in the absence of cyclic nucleotide was subtracted. For wild-type rOCNC1, the basal current comprised almost entirely the leak current at the seal, which was usually 30 to 60 pA at +80 mV. For rOCNC1-Y565A, the basal current was 5 to 30% of maximal current (not shown). This percentage agrees with the single-channel results, where the spontaneous open probability fell into this range (see Fig. 8).

View larger version (16K): [in this window] [in a new window]   Fig. 6.   Dose-response comparison between wild-type rOCNC1 and Y565A at 80 mV. The fitting and normalization were as described in Fig. 3. The horizontal axis is on a logarithmic scale.

View larger version (32K): [in this window] [in a new window]   Fig. 7.   Summary of mutants made and their major properties.

View larger version (26K): [in this window] [in a new window]   Fig. 8.   A, consecutive current traces of a single Y565A channel in the absence of any cyclic nucleotide, recorded at 60 mV. The channel is opening downward. B, the fluctuations of Popen during 90 s of spontaneous activity of this channel. The Popen values were calculated within 1-s time windows, and the mean ± standard deviation in the total of 90 windows is 21.3 ± 15.9%. Other channels had different Popen values, which ranged from 5 to 30%, yet after exposure to cAMP or cGMP, the channels often become sensitized, yielding spontaneous Popen values as high as 80%.

View larger version (18K): [in this window] [in a new window]   Fig. 9.   The open-time (A) and closed-time (B) histograms of the spontaneous openings of the channel shown in Fig. 8. The smooth lines are the fits by sums of exponential functions.

View larger version (22K): [in this window] [in a new window]   Fig. 10.   The open-time (left) and closed-time (right) histograms at two different cAMP concentrations for a Y565A channel other than the one shown in Figs. 8 and 9. A, 5 µM. B, 10 µM. Note that with increasing cAMP, open times were lengthened, whereas the closed times were shortened. The smooth curves were the sum of exponential functions used to fit the observed distributions. The fitting parameters, expressed as time constant (fractional area), are as follows: A, open times, 1 ms (18%) and 18 ms (82%); closed times, 2.1 ms (21%) and 179 ms (79%). B, open times, 4.3 ms (3%), 37 ms (87%), and 128 ms (10%); closed times, 3.8 ms (5%) and 63 ms (95%).

	Discussion

Top Summary Introduction Materials and Methods Results Discussion References

The homology-based models of binding domains in CNG channel depend heavily on the template structure of CAP and are sensitive to 1) the uncertainties inherent in defining energy terms and 2) the assumptions regarding the intricate and extensive couplings between neighboring side chains. It is therefore important to test these models, particularly the explicit predictions of prominent interactions. In one such experimental test, Thr560 in Brcng was investigated through site-directed mutagenesis and expression (Altenhofen et al., 1991). It was found that Thr560 determined the selectivity of cGMP over cAMP, a finding consistent with the interpretation of an earlier structure model of cGMP-dependent protein kinase (Weber et al., 1989). In another study, Tibbs et al. found that the highly conserved Arg559, one residue upstream of Thr560, formed a favorable ionic bond with cGMP (Tibbs et al., 1998), again agreeing with previous models. The largest shifts caused by amino acid substitutions amounted to 150-fold and 1000-fold differences in EC₅₀ values for Thr560 and Arg559, respectively, reflecting free energy differences on the level of 3 to 4 kcal/mol, typical of a hydrogen bond or an ionic pair.

In the present study, however, the changes in EC₅₀ value we found with mutations at Y512 of rOCNC1 and F533 of Brcng were much less than expected (Scott et al., 1996). These residues were predicted to interact specifically with cyclic nucleotides and to govern the selectivity of cGMP over cAMP in the photoreceptor channels. Contrary to this prediction, our results indicated that this side chain is unlikely to form a hydrogen bond or direct hydrophobic contact with the ligand molecule in the olfactory channel. Not only is the Y/F difference not important for the higher efficacy of cGMP at the photoreceptor channels, but even the replacement by alanine or glutamate caused no changes appropriate to the free energy for a hydrogen bond or an ion pair. The surprisingly mild effect of introducing a negative charge with glutamate suggested that there is a charged or polar environment for this portion of the binding domain. This result, however, does not rule out hydrophobic interactions mediated by other regions of the binding site.

A recent study showed that the photoaffinity analog of cGMP, 8-(p-azidophenacylthio)-[³²P]cGMP specifically labels Brcng at V524, V525, and A526, residues that align with 4 in CAP (Brown et al., 1995). This region of the binding site was not noted by the models to interact with the ligand.

The validity of the structural models were further confounded by the uncertainties about the conformation of cAMP and cGMP molecules. These ligands can bind in either the syn- or the anti- conformation; the two variations lead to profoundly different energetic outcomes. A more definitive examination of related hypotheses would require testing individual combinations of mutant receptors and cyclic nucleotide analogs that have relatively defined preferences for anti- or syn- conformation.

In the regulatory subunit of PKA, the aromatic-aromatic interactions play an important role in ligand binding. PKA-R1 is one of the many examples of the importance of aromatic residues in ligand recognition, protein folding, and protein conformational changes in general (Hunter, 1994). In a review of 34 high-resolution protein structures, Burley and Petsko (1985) concluded that an average of 60% of aromatic side chains in proteins are involved in aromatic pairs, 80% of which form networks of three or more interacting aromatic side chains. In most common cases, the aromatic rings prefer an edge-to-face configuration and are separated by 4.5 to 7 Å. Nonbonded potential energy calculations indicate that a typical pairwise interaction has a stabilizing energy of 1 to 2 kcal/mol. The conservation of aromatic-aromatic interactions in related proteins is striking, indicating that these interactions play important roles in structure or function. A potassium channel contains crucial edge-to-edge aromatic contacts (Doyle et al., 1998).

The aromatic-aromatic interactions in PRA-R1 are not mediated by residues that align with Leu61 of CAP; instead, the relevant aromatic residues reside in other, unexpected positions (Su et al., 1995). We therefore attempted to examine all of the aromatic residues in the binding domain of rOCNC1 through mutagenesis. At some of these positions, alanine mutations resulted in nonfunctional channels, but tryptophan mutations resulted in responses indistinguishable from those of the wild-type channels. Perhaps channels with reduced function would result from side chains with properties intermediate between alanine and tryptophan. Most of the aromatic-to-alanine mutations that are functional did not result in significant changes in channel properties. However, one mutation, Y565A of rOCNC1, increased agonist sensitivity by 10-fold, indicating that this position is important for channel functioning. Nonetheless, this residue probably does not interact directly with the ligand molecule, for two reasons: 1) the increased spontaneous activity strongly suggests gating changes, and the reduced expression level may result from the desensitization of the constitutively active channels; and 2) assuming that the overall folding pattern is conserved between CAP, PKA-R1,and rOCNC1, this residue would be located between B and C, outside the binding pocket. It is also formally possible that Y565 is a site of tyrosine phosphorylation in the wild-type channel (Molokanova et al., 1997).

The C helix is likely to be mobile during channel activation. In fact, rotation of this helix, either induced or stabilized by the bound ligand molecule, is probably one of the conformational changes that link ligand binding and its ultimate effect, the opening of the pore (Varnum et al., 1995). Y565A is therefore positioned at the "hinge" of the motion. Our data suggest that larger residues at this "pivotal" position favor the closed state of the channel. For instance, after replacement with alanine, we see increased spontaneous activities and an corresponding increase in ligand sensitivity. The fact that Y565A affects cAMP and cGMP to equal extents also argues for a general functional role that is indifferent to the fine structure of the ligand. Residues that interact directly with the ligand probably include D604 in the C helix of Brcng (Varnum et al., 1995).

The distances between the binding pocket, the pore, and Y565 are likely to span a considerable portion of the entire protein, demonstrating the remarkable range of allosteric coupling. The effect of Y565 mutations, particularly the correlation between channel properties and the size of the side chain at this site, place important constraints on future structural models.