INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

The cyclic GMP-activated rod channels, situated in the plasma membrane of retinal rod outer segments, are responsible for the entry of cations (mainly sodium, but also calcium) in the cell in the dark. Upon light-induced activation of the phototransduction cascade the concentration of cGMP is reduced, which results in closure of the channels and hyperpolarization of the cell. The retinal rod channel belongs to the family of cyclic nucleotide-gated (CNG) channels, involved in sensory transduction, which are activated by direct binding of the cyclic nucleotide to a site situated in the cytoplasmic C-terminal region. The three types of CNG channels (rod, olfactory neuron, and cone channels) are tetramers (for a review see Zagotta and Siegelbaum, 1996) composed of at least two (rod, cone) or three (olfactory neuron) types of homologous subunits (Chen et al., 1993; Körschen et al., 1995; Biel et al., 1996; Sundin et al., 2000; Kohl et al., 2000; Gerstner et al., 2000; Liman and Buck, 1994; Bradley et al., 1994; Sautter et al., 1998; Bönigk et al., 1999). Only the  subunits can form functional channels when expressed alone;  subunits have no channel activity on their own, but when co-expressed with  subunits, modify several characteristics of the channels, among which is the sensitivity to cAMP. The sensitivity to cAMP of heteromeric rod channels expressed in oocytes is higher than that of -only channels (Fodor and Zagotta, 1996; Gordon et al., 1996; Shammat and Gordon, 1999; Pagès et al., 2000), and similar to that of native channels (Tanaka et al., 1989; Gavazzo et al., 1996; Picco et al., 1996). The rod heteromeric channel is also much more sensitive to L-cis-diltiazem (Chen et al., 1993; Körschen et al., 1995) than the homomeric channel. This property can be used to check that the two subunits indeed associate, or to estimate the homogeneity of the channel population obtained when co-injecting  and  mRNAs.

The structure of the nucleotide binding site of the rod  subunit was predicted by homology with the cAMP binding site of the catabolite gene activator protein (CAP) of Escherichia coli (Kumar and Weber, 1992), which is also homologous to that of the regulatory subunits of the cAMP- and cGMP-activated protein kinases (Kaupp et al., 1989; Shabb and Corbin, 1992): it consists of an -helix (A), followed by eight -rolls, and two additional -helices (B and C). In CAP, binding of cAMP modifies the orientation of the DNA binding domain situated downstream of the cAMP binding site (Passner et al., 2000), while in the regulatory subunits of cAMP/cGMP dependent kinases, the effector domain is situated upstream of the nucleotide binding site, revealing some variability in cyclic nucleotide-induced activation. Binding of the nucleotide to the nucleotide binding site in CNG channels induces an allosteric transition that allows the channel to open, and cations to pass through the pore formed by the assembled pore-domains of the four subunits. The pore is homologous to that of voltage-gated K⁺ channels except at the level of the selectivity filter, where it is shorter. Structure predictions of the pore domain of the rod channel, based on the structure of the KcsA K⁺ channel (Doyle et al., 1998; Perozo et al., 1999) have allowed great progress in the approach of the gating mechanism. The pore of the  subunit was shown to undergo a conformation change associated with channel opening (Becchetti et al., 1999; Becchetti and Roncaglia, 2000; Liu and Siegelbaum, 2000). Recently, it was proposed that rotation of the domain situated between the last transmembrane helix (S6, downstream of the pore domain) and the nucleotide-binding site ("C-linker") initiates rotation of S6 and pore opening (Johnson and Zagotta, 2001; Flynn and Zagotta, 2001).

A segment of ~50 residues, situated immediately downstream of the C-helix of the binding site in the rod channel  subunits (I611-K665), is highly conserved among rod, olfactory neuron, and cone channels (Fig. 1). This domain was classified as "PD004548" in the ProDom release 2000.1 (Corpet et al., 2000) when this work started. In the ProDom release 2001.2, PD004548 also includes part of the last helix (C) of the nucleotide binding site. A conserved domain (PD012882) is present in a similar position downstream of the nucleotide binding site of all known (bovine, human, rat) rod  subunits (1202-1282 in the bovine rod  subunit); this domain is not conserved in other  subunits, but the second olfactory  subunit (Sautter et al., 1998: CNG4.3; Bönigk et al., 1999: CNG1b), which associates with the other olfactory OCNC1 and OCNC2 subunits to form the olfactory neuron channel, is identical to the rod  subunit. This suggests that the domains downstream of the cGMP-binding site may be important for channel assembly or function. According to PHD (Profile network prediction HeiDelberg) structure prediction (Rost and Sander, 1993) (Fig. 1), PD004548 ( subunit) downstream of the cGMP-binding site consists for the most part of an -helix (K619-F663), which is predicted by the program COILS (Lupas, 1996) as having a high probability of forming a coiled-coil domain (not shown). The PD012882 domain in the  subunit also presents a predicted -helical structure composed of two helices (P1219-M1232 and R1241-K1270); the second -helix is also susceptible to form a coiled-coil domain, although with a lesser probability than the -helix in PD004548.

View larger version (35K): [in this window] [in a new window]   FIGURE 1   C-terminal domains of the rod channel  and  subunits. A ProDom NCBI-BLASTP2 search was performed for the C-terminal domains downstream of the cGMP-binding site of the bovine rod  (609-690) (CNG1  bovin) and  (1202-1394) (CNG4  bovin) subunits (ProDom server, Corpet et al., 2000. Two conserved domains were found: PD004548 for the  subunit (26 sequences, including rod, olfactory neuron and cone  subunits from mammals, chick rod and cone  subunits, and CNG channel from catfish, Drosophila, and C. elegans), and PD012882 for the  subunit (6 sequences, corresponding to all known rod  subunits). The sequences corresponding to these domains are boxed on the sequence of the C-terminus of the bovine rod  (CNG1) and  (CNG4c, Biel et al., 1996) subunits. The residues that are predicted to form an -helix (The PredictProtein server for protein structure prediction, Rost and Sander, 1993) are in bold letters and boxed. The sites of truncations are indicated by arrows. The nucleotide binding site ends at the C-terminal end of the C helix. The scheme represents the common secondary structure of  and  subunits. Alignments of several sequences corresponding to the part of PD004548 and PD012882 domains situated downstream of the nucleotide binding site are shown below each sequence. The sequence of the bovine rod  subunit in the alignment is that of Körschen et al., 1995.

The pore domains of all CNG channel  subunits are highly homologous, and all lack the YG residues in the GYGD sequence which is characteristic of the K⁺ channel selectivity filter. The pore domain of  subunits from rod and cone channels is homologous to that of  subunits, but has different charged residues. An alignment is shown in Fig. 2.

View larger version (34K): [in this window] [in a new window]   FIGURE 2   Alignment of the pore domains of rod, olfactory neuron (olf), and cone  and  subunits. Conserved charged residues are boxed.

We took advantage of the different properties of homomeric and heteromeric channels concerning the sensitivity to cAMP to investigate the possible role of the conserved C-terminal domains of the  and  subunits, which has not been studied previously. We report here a study of several C-terminal truncations of the  subunit, expressed alone or co-expressed with  subunits, wild type, or truncated in their C-terminal domain. We also studied possible functional differences conferred by the pore of the  subunit by constructing chimeras of the two rod subunits with exchanged pore domains.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

cDNAs and mutations

The bovine rod  subunits were a gift of Prof. U. B. Kaupp. The bovine rod  subunit CNG4c (Biel et al., 1996) starting at V572 downstream of the GARP region previously described (Pagès et al., 2000) was used. Residue numbers in the  subunit correspond to the sequence of the complete  subunit (Körschen et al., 1995). The  and  subunits are inserted in the pGemHe vector (Liman et al., 1992), and the ATG codon is preceded by a Kozak consensus sequence (CCACC) (Kozak, 1984). Truncations were constructed by PCR by introducing a stop codon at the required position.

Construction of the [-pore] chimera

Construction of the [-pore] chimera

Channel expression

Capped mRNAs were synthesized in vitro from linearized plasmids in the presence of RNA cap structure analogs (New England Biolabs, Beverly, MA), and injected into Xenopus oocytes (25 ng/oocyte for macroscopic currents or 0.25 ng/oocyte for single-channel records). For coexpression of  and  subunits,  mRNA and  mRNA were mixed and injected into the oocyte. The / ratio was between 2 and 3. Oocytes were incubated in Barth's medium for 2-10 days at 19°C for macroscopic currents (or at 4°C after 18-24 h at 19°C for single-channel currents) before measurements.

Patch-clamp recording of excised inside-out patches

The solution in the pipette and in the perfusion medium was 100 mM KCl, 10 mM EGTA/KOH, 10 mM Hepes/KOH, pH 7.2. The cytoplasmic face of the patch was superfused by solutions containing variable nucleotide concentrations using an RSC100 rapid solution changer (Bio-Logic, Claix, France). The saturating cGMP and cAMP concentrations were, respectively, 500 µM and 20 mM. Currents were recorded with an RK-400 patch amplifier (Bio-Logic).

P_omax (open probability at saturating concentrations of cGMP and cAMP) was obtained from single-channel record analysis. Single-channel records at +80 mV (20-30-s duration) sampled at 33 kHz and numerically filtered (Hanning window) at 1 kHz were used to compute amplitude histograms using the Bio-Patch software (Bio-Logic). The same current interval (4 pA) for all records was divided into 50 classes for building the histograms. The maximum number of events/pA was at least 10⁶. The histograms were fitted with two or three Gaussian curves. Similar P_omax values were obtained from records filtered at 4 kHz, as previously described (Pagès et al., 2000).

Macroscopic currents induced by voltage steps (500 ms, ±80 mV) were low-pass filtered at 300 Hz and digitized at 1 kHz using pCLAMP 6.0 (Axon Instruments, Union City, CA). The series resistance was compensated for (resulting value <1 M). Each record was averaged three times. Dose-response curves were obtained by plotting the current at +80 mV as a function of nucleotide concentration after subtraction of the leak current.

Curve fitting

Data were fitted with the Hill equation or the Monod-Wyman-Changeux (MWC) model (Monod et al., 1965) using the Microcal Origin software.

Hill equation

MWC model

Chemicals

L-cis-Diltiazem was a gift of Synthelabo Recherche (Bagneux, France).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Role of the C-terminal domains

To determine whether the C-terminal domains downstream of the nucleotide binding site may be involved in channel gating, we studied the effect of C-terminal truncations on the efficacy of cAMP and cGMP. The sites of truncation were chosen according to the predicted structure of the C-terminal domains (Fig. 1). The  subunit was truncated after K665 (the last residue of the conserved PD004548 domain), R656 (deleting part of PD004548), and D608 (deleting the whole C-terminal domain downstream of the nucleotide binding site, including the last two residues of the C-helix). The  subunit was truncated after S1271 (at the end of the predicted helical domain of PD012882), L1221 (the third residue of the first predicted  helix of PD012882 downstream of the nucleotide binding site), and N1202 (deleting the whole C-terminal domain downstream of the nucleotide binding site). Like the wt subunit, the truncated  subunits did not give rise to channel activity when expressed alone. The selectivity for cGMP versus cAMP was determined from single-channel records (from the open probability at saturating cGMP and cAMP concentrations) and from macroscopic currents (I_max(cAMP)/I_max(cGMP) ratios) to obtain sufficient data for statistical analysis. When measuring macroscopic currents, the extent of inhibition by L-cis-diltiazem was used as a tool to test the homogeneity of the channel population: since the effect of diltiazem on mutated channels is unknown, this was first determined from single-channel records. The EC₅₀ of several constructions for cGMP and cAMP were determined from macroscopic currents. All measurements were performed on excised inside-out patches.

Single-channel records

View larger version (41K): [in this window] [in a new window]   FIGURE 3   Examples of unitary cGMP and cAMP-induced currents for control wild-type channels and heteromeric channels with truncated  subunits (K665stop, R656stop, and D608stop) co-expressed with wt or  L1221stop. cGMP: 500 µM; cAMP: 20 mM. Pomax values are obtained from the histograms computed from the whole length of the records shown. wt: 0.93 (cGMP), 0.03 (cAMP); wt + wt: 0.98 (cGMP), 0.12 (cAMP); K665stop + wt: 0.83 (cGMP), 0.11 (cAMP); K665stop + L1221stop: 0.94 (cGMP), 0.17 (cAMP); R656stop + wt: 0.99 (cGMP), 0.32 (cAMP); R656stop + L1221stop: 0.84 (cGMP), 0.45 (cAMP); D608stop + L1221stop: 0.93 (cGMP), 0.86 (cAMP). A two-channel record is shown for D608stop + wt (Pomax cGMP = 0.83, Pomax cAMP = 0.68): the high Pomax cAMP, with numerous simultaneous openings of two channels, unambiguously indicates that both channels are heteromers. Inhibition of the cGMP-induced current (500 µM cGMP) by L-cis-diltiazem (50 µM) is close to 100% for all channels, except wt-only channels (11%), D608stop/wt (60%), and D608stop/L1221stop (46%). For calculation of Pomax in the patch with 2 channels (D608stop/wt), the following formula was used: Pomax = 1  Sq. root (1  Po(measured)). Mean values of several experiments are given in Table 1. The activity of heteromeric channels with K665stop and R656stop subunits is more flickering than that of wt channels, as previously described for wt/wt (Körschen et al., 1995). In the examples shown, the activity of heteromers containing the D608stop subunit seems less flickering, and the amplitude of the cGMP- and cAMP-induced unitary currents appears smaller than that of wild-type channels (~1 pA compared to 1.35 pA).

View this table: [in this window] [in a new window]   TABLE 1   Characteristics of unitary currents from wild-type rod channels (wt, wt/wt), and heteromeric channels composed of truncated  subunits (K665stop, R656stop, or D608stop) and wt or L1221stop

Macroscopic current measurements

View this table: [in this window] [in a new window]   TABLE 2   Characteristics of macroscopic currents from rod homomeric  channels (wt, K665stop, R656stop, or D608stop), and from heteromeric channels composed of  (wt or truncated) and  (wt or L1221stop) subunits

View larger version (13K): [in this window] [in a new window]   FIGURE 4   Examples of macroscopic cGMP and cAMP-induced currents illustrating the effect of C-terminal truncation of the  subunit on the cAMP sensitivity of heteromeric channels containing the L1221stop subunit. cGMP: 500 µM; cAMP: 20 mM. The activity of the four channel types can be attributed for the most part to heteromeric channels by comparison of their sensitivity to L-cis-diltiazem and Imax(cAMP)/Imax(cGMP) ratio to those of single channels. Mean values of several experiments are indicated in Table 2. The left part of each record is the current recorded at 80 mV, and the right part the current at +80 mV (see Methods).

View larger version (20K): [in this window] [in a new window]   FIGURE 5   Dose-response curves for cGMP and cAMP-induced macroscopic currents from oocytes expressing heteromeric wt/L1221stop and D608stop/L1221stop channels. The symbols represent the mean of different experiments, ±SE. cGMP-induced currents are normalized to 1, and cAMP-induced currents to the Imax(cAMP)/Imax(cGMP) ratio from Table 2. The fits of all the points to the Hill equation are shown on the graph. For wt/L1221stop: EC50 cGMP = 27 ± 1 µM, nH = 2.3 ± 0.1 (7 patches), and EC50 cAMP = 1353 ± 18 µM, nH = 1.5 ± 0.1 (6 patches). For D608stop/L1221stop: EC50 cGMP = 22 ± 2 µM, nH = 2.2 ± 0.2 (8 patches), and EC50 cAMP = 224 ± 6 µM, nH = 2.2 ± 0.1 (8 patches). Hill fits of the dose-response curves for wt/wt (from Pagès et al., 2000) are indicated on both graphs with thin lines (EC50 cGMP: 35 ± 0.6 µM, nH = 2 ± 0.1; EC50 cAMP: 1607 ± 31 µM, nH = 1.45 ± 0.04). Note that the Hill number of the cAMP dose-response curve is lower than that of the cGMP dose-response curve for wt/wt and wt/L1221stop channels, while the two nH values are identical for the double mutant, consistent with the predictions of the MWC model according to which nH increases with the gating efficacy of the ligand (Rubin and Changeux, 1966; see also Pagès et al., 2000). The sets of data for EC50 cGMP, EC50 cAMP, and Imax(cAMP)/Imax(cGMP) measured from each experiment with wt/L1221stop are significantly different from the corresponding sets of data measured with wt/wt, with p = 5.104, 0.006, and 0.028, respectively (independent t-test on two populations). The set of data for EC50 cGMP measured with D608stop/L1221stop is significantly different from that measured with wt/wt with p = 7.105.

View larger version (31K): [in this window] [in a new window]   FIGURE 6   Summary of the effect of C-terminal truncations of  and  subunits on EC50 cGMP, EC50 cAMP, and Imax(cAMP)/Imax(cGMP). Statistical box charts for EC50 (obtained by fitting each individual dose-response curve with the Hill equation) and Imax(cAMP)/Imax(cGMP) ratios (obtained from Imax cAMP and Imax cGMP measured on the same patch), from macroscopic current measurements. Dose-response curves for cAMP were measured on patches that had very large cGMP-induced currents in the case of channel constructs with a low cAMP sensitivity (usually different patches were used for the two ligands). Data for wt and wt/wt are from Pagès et al. (2000). The vertical lines in the box denote the 25th, 50th, and 75th percentile values. The error bars denote the 5th and 95th percentile values. The crosses are the extreme values, and the square in the box is the mean of the data. EC50 cGMP: 29.9 ± 08 µM, 8 patches (wt); 33 ± 1 µM, 7 patches (D608stop); 35 ± 0.6 µM, 11 patches (wt + wt); 27 ± 1 µM, 7 patches (wt + L1221stop); 22 ± 2 µM, 8 patches (D608stop + L1221stop). EC50 cAMP: 2077 ± 172, 8 patches (wt); 1607 ± 31 µM, 9 patches (wt + wt); 1353 ± 18 µM, 6 patches (wt + L1221stop); 224 ± 6 µM, 8 patches (D608stop + L1221stop). For mean values of Imax(cAMP)/Imax(cGMP) ratios, see Table 2.

Role of the pore domains

In an attempt to identify other domains that may be responsible for cAMP sensitivity, we constructed two chimeras of  and  subunits: a  subunit with the pore domain of the  subunit ([-pore]) and the symmetric construction of the  subunit with the pore domain of the  subunit ([-pore]).

Like wt subunits, [-pore] alone had no channel activity. In contrast, when [-pore] mRNA was co-injected with wt mRNA, cGMP-induced currents that could be inhibited by L-cis-diltiazem (67% ± 2%) were observed (Table 3), indicating that the channels were at least for a large part heteromeric channels. However, the cAMP sensitivity of these channels was reduced to that of -only channels, or even lower (I_max(cAMP)/I_max(cGMP) = 0.024 ± 0.002). Because the sensitivity to cAMP of heteromeric channels containing truncated  and  subunits is markedly increased (see above), we also studied the cAMP sensitivity of heteromeric D608stop/[-pore]L1221stop and R656/[-pore]L1221stop channels to determine which of the two modifications (pore exchange or C-terminal truncation) predominates. For these two types of channels (Table 3), cAMP sensitivity was inferior or equal to that of -only channels, while L-cis-diltiazem sensitivity was unambiguously different from that of -only channels, and close to that of D608stop/L1221stop or R656/L1221stop channels, respectively (Table 1), indicating that the channel population was mainly composed of heteromers. The I_max(cAMP)/I_max(cGMP) measured for heteromeric channels with chimeric  subunits having an -pore domain are compared in Fig. 7 to those measured with the corresponding channels containing nonchimeric  subunits. These results show that the pore domain of the  subunit plays a key role concerning the gating efficacy of cAMP, and that C-terminal truncation of the  subunit has no effect if the pore of the  subunit is exchanged for that of the  subunit.

View this table: [in this window] [in a new window]   TABLE 3   Characteristics of macroscopic currents from heteromeric rod channels with chimeric  subunits having an -pore domain

View larger version (18K): [in this window] [in a new window]   FIGURE 7   Effect of exchanging the pore domain of the  subunit by that of the  subunit on the sensitivity to cAMP of heteromeric channels. Statistical box charts for Imax(cAMP)/Imax(cGMP) ratios (from macroscopic Imax(cAMP) and Imax(cGMP) measured on the same patch) obtained for heteromeric channels containing wt or L1221stop subunits (right part, same data as in Fig. 6), or the corresponding chimeric  subunits in which the pore domain was exchanged for that of the  subunit (left part). All the channel populations mainly consisted of heteromeric channels, as proved by the extent of inhibition by diltiazem (Tables 2 and 3). The Imax(cAMP)/Imax(cGMP) ratio of wt channels is shown for comparison. Mean values are given in Tables 2 and 3. cGMP: 500 µM; cAMP: 20 mM.

No channel activity was detected with the [-pore] subunit mRNA, either when injected alone or when co-injected with wt mRNA, wt mRNA, or with [-pore] mRNA.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

The results presented here demonstrate the role of two domains of the retinal cGMP-activated channel, distinct from the nucleotide-binding site, concerning the sensitivity to cAMP: a conserved C-terminal domain situated downstream of the cGMP-binding site in the  subunit (PD004548), and the pore domain of the  subunit. Exchanging the pore domains or removing the C-terminal domains produces much larger changes in sensitivity than any previously reported mutation.

The pore domain of the  subunit

The possible influence of the pore of the  subunit on the activity of heteromeric channels has not been studied previously. The striking finding presented here (Fig. 7) is that replacing the pore of the  subunit by that of the  subunit completely abolishes the effect of the  subunit on the sensitivity to cAMP of heteromeric channels. Moreover, truncation of the C-terminal domain PD004548 in the  subunit, which increases cAMP sensitivity, has no effect if the pore of the  subunit is exchanged for that of the  subunit: this suggests that high cAMP sensitivity is in fact due to the -pore domain but is partially masked in the presence of wt, and that the effect of deleting PD004548 in the  subunit is to relieve this inhibition of cAMP sensitivity rather than to directly increase cAMP sensitivity.

The exchanged domains only comprise 16 residues: the -pore segment V₃₄₈YSLYWSTLTLTTIGET₃₆₄ replacing I₉₄₇RCYYWAVKTLITIGGL₉₆₃ in the  subunit. The main difference between the two pore domains (Fig. 2) is the presence of distinct charged residues. E363 near the selectivity filter of the rod  subunit was shown to be responsible for calcium block of homomeric  channels (Root and MacKinnon, 1993; Eismann et al., 1994). Mutating the corresponding residue in the olfactory neuron  subunit is also responsible for calcium blockage, and decreases the gating efficacy of both cGMP and cAMP (Gavazzo et al., 2000). The residue at the corresponding place in the  subunit is uncharged (G962), and an acid residue is situated three residues downstream (D965) in place of P366 in the  subunit. In addition, there are two basic residues (R948 and K955) in the pore of the  subunit (corresponding to the hydrophobic residues Y349 and L356 in the  subunit). Point mutations of the corresponding residues in the pore of the  subunit will be needed to determine the residue(s) responsible for the effect. It is interesting that the pore of the third olfactory channel subunit (CNG4.3 or CNG1b, identical to the rod  subunit: Sautter et al., 1998; Bönigk et al., 1999) and that of the  subunit of the cone channel (human and mouse: Sundin et al., 2000; Kohl et al., 2000; Gerstner et al., 2000), which are also responsible for increased cAMP sensitivity of these channels, contain two basic residues and an acid residue at the same positions as the rod  subunit (Fig. 2). However, although the second olfactory channel subunit (OCNC2) also confers increased cAMP sensitivity to the  subunit (Liman and Buck, 1994; Bradley et al., 1994) its pore does not contain the three charged residues as that of the other  subunits; in fact, for its pore domain, this subunit is homologous to an  subunit.

The fact that the symmetric construct [-pore], expressed alone or with wt subunits, has no channel activity suggests that four -pore domains are unable to form a functional cationic channel. However, this is not the only reason for the absence of activity of  subunits alone, because [-pore]s have no channel activity either when expressed alone. The fact that [-pore] has no channel activity when expressed with wt or [-pore] moreover suggests that the -pore domain may be unable to function when inserted in the  subunit. Alternatively, the absence of detected activity could be due to rapid inactivation, as previously observed for several mutants of the residue E363 in the pore of the rod  subunit (Bucossi et al., 1996); given our experimental conditions, however, we estimate that inactivation should be complete in <10 ms after the start of the voltage pulse to remain undetected.

The C-terminal domains

The other striking finding is the role of the C-terminal domain downstream of the cGMP-binding site of the  subunit in heteromeric channel function. When the  subunit truncated downstream of D608, which corresponds to the deletion of the whole C-terminal domain downstream of the cGMP-binding site, is co-expressed with truncated  subunits, a drastic effect on cAMP sensitivity is observed (P_omax cAMP measured from single-channel records and I_max(cAMP)/I_max(cGMP) ratio measured from macroscopic currents: Table 1 and 2). Truncation of the  subunit C-terminus domain downstream of K665 has little effect, while truncation downstream of R656 has a partial but well-marked effect, suggesting that the presence of the entire predicted -helix of the PD004548 domain downstream of the cGMP-binding site plays a major role for cAMP sensitivity of wild-type heteromeric channels.

The discrepancies between the values obtained from single-channel records and macroscopic current measurements might be explained by the irregular occurrence of long closed periods in cAMP-, and (although to a lesser extent) in cGMP-induced currents (see Fig. 3), which are fully taken into account in macroscopic currents, but poorly in single-channel analysis. The errors on P_omax (cAMP) and P_omax (cGMP) add up in the P_omax (cAMP)/P_omax (cGMP) ratio. However, a proportion of -only channels can be present in the patches from oocytes injected with  and  mRNAs used for macroscopic current measurements; comparing the values for the extent of inhibition by diltiazem in Table 1 (single-channel) and Table 2 (macroscopic currents) suggests that part of the macroscopic currents in patches from oocytes co-injected with  and  mRNAs is due to homomeric channels.

The EC₅₀ for cAMP measured from macroscopic currents of oocytes expressing D608stop/L1221stop channels is also very much reduced (224 ± 6 µM), as is the EC₅₀ for cGMP (22 ± 2 µM) compared to those of wt/wt channels (1607 ± 31 µM and 35 ± 0.6 µM, respectively) (Fig. 5). After normalizing the dose-response curves to the P_omax values obtained from single-channel records, the dose-response curves for cAMP and cGMP of the D608stop/L1221stop channel were fitted with the MWC allosteric model (Monod et al., 1965) (Fig. 8), which can be used as an approximation of the CNG channel mechanism (Zagotta and Siegelbaum, 1996). According to this model, stabilization of the open state is achieved independently of the ligand by decreasing the parameter L (L = [T]/[R], equilibrium constant between closed and open states in the absence of ligand), which is a characteristic of the protein. The equilibrium constant between fully liganded closed and open states being equal to Lc⁴, the open state can be stabilized in the presence of ligand by decreasing L and/or by decreasing c = K_R/K_T (ratio of the affinity of the ligand for the closed state to the affinity for the open state). Fig. 8 shows that it is possible to fit the data by only reducing c and K_R for both ligands (without reducing L), but that decreasing L only cannot account for the data: when L is reduced, whatever its value, it is necessary to also decrease c (increase the gating efficacy) and K_R (increase the affinity of the open state) for cAMP. Thus, interpreting the data with the MWC model suggests that, in D608stop/L1221stop channels, the conformation of nucleotide binding sites is different from that of the wild-type channel; i.e., that the C-terminal domains constrain the conformation of the nucleotide binding sites, resulting in lowering the efficacy of cGMP and more particularly, cAMP (increase "c" and "K_R"). The fit does not exclude that the truncation may in addition facilitate the transition to the open state (reduce "L"). It should be noted that although high cAMP sensitivity is conferred by the  subunit (as discussed above), most of the constraint is relieved by truncating the C-terminus of the  subunit (compare dose-response curves for wt/L1221stop and for D608stop/L1221stop, Fig. 5), suggesting that intersubunit interactions are involved in the conformation of the oligomer.

View larger version (17K): [in this window] [in a new window]   FIGURE 8   Fits of dose-response curves for D608stop/L1221stop channels according to the MWC allosteric model (Monod et al., 1965). The data, ±SE, are the same as in Fig. 5, but are expressed in open probabilities; currents are normalized to the value of Pomax cGMP (0.93) and Pomax cAMP (0.86) from Table 1. The dose-response curves cannot be fitted with a reasonable error with the MWC model without fixing any of the parameters (L, c, KR). We have fitted the data for varying values of L between L = 7999 (measured for wt, Tibbs et al., 1997) and L = 100 (chosen as the lowest value compatible with the absence of detectable spontaneous openings), corresponding to a probability of spontaneous openings Psp = 1/(L + 1) ~ 0.01. The fits corresponding to the extreme values are shown in A: the solid line corresponds to L = 7999 (KRcGMP = 1.47 ± 0.17 µM, c = 0.047 ± 0.007; KRcAMP = 17.6 ± 1.6 µM, c = 0.037 ± 0.005), and the dashed line corresponds to L = 100 (KRcGMP = 7.4 ± 1.5 µM, c = 0.094 ± 0.04; KRcAMP = 80 ± 8 µM, c = 0.077 ± 0.017). Note that these are only global values, since the simple MWC model does not take into account the existence of two types of sites. The values of c (for cGMP and cAMP) and of KR (for cGMP and cAMP) that give the best fit for each value of L are plotted in B and C as a function of L; error bars are shown for several points. The values calculated for wt/wt channels with L = 7999 (Pagès et al., 2000) are indicated on the two plots with solid (cGMP) or open (cAMP) circles: c (cGMP) = 0.065 ± 0.001, c (cAMP) = 0.166 ± 0.001, and KRcGMP = 1.89 ± 0.04 µM, KRcAMP = 79 ± 0.4 µM. The graph in A shows that the data can be reasonably fitted with any value of L between 7999 and 100, although the fit with L = 100 appears less satisfying at low nucleotide concentrations. B and C show that it is not possible to fit the cAMP dose-response curve by reducing L only: reducing both KR(cAMP) and c are also necessary.

Finally, our results indicate that although L-cis-diltiazem sensitivity is conferred to the rod channel by the  subunit, the presence of part of the PD004548 domain in the  subunit (D608-R656) is necessary for full inhibition of heteromeric channels. Complete inhibition of heteromeric channels containing truncated L1221stop subunits (K665stop/L1221stop and R656stop/L1221stop, Table