INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

The cGMP-gated channels of retinal rods are responsible for light-induced hyperpolarization of the photoreceptor cell: hydrolysis of cGMP upon activation of the light-sensitive cascade of phototransduction leads to closure of the channels and reduction of the cationic current that enters the cell in the dark. Cyclic nucleotide-gated (CNG) channels are directly activated by binding of cyclic nucleotide to a site situated in the cytoplasmic C-terminal region. This site was identified by its high sequence homology with other known cyclic-nucleotide binding proteins: the CRP protein of Escherichia coli and regulatory subunits of the cGMP- and cAMP-activated protein kinases (Kaupp et al., 1989; Shabb and Corbin, 1992). Within this site, a residue, situated near the end of the C helix of the binding site in the rod  subunit (D604), was shown to play an important role in nucleotide specificity: substitution of the charged aspartate residue by uncharged glutamine or asparagine in the rod channel modifies the agonist specificity, and substitution by the non polar methionine residue inverts the specificity, which becomes cAMP > cIMP > cGMP (Varnum et al., 1995). Both the rod and olfactory native channels are heterooligomeric proteins, composed of at least two types of subunits:  (CNG1) and  (CNG4) for the rod channel (Chen et al., 1993; Körschen et al., 1995; Biel et al., 1996), and subunit 1 (CNG2), subunit 2 (CNG5), and a recently discovered CNG4.3 subunit related to the rod  subunit for the olfactory channel (Liman and Buck, 1994; Bradley et al., 1994; Sautter et al., 1998). For the olfactory channel, coexpression of subunit 2 (Liman and Buck, 1994; Bradley et al., 1994) or of CNG4.3 with subunit 1 was found to increase the sensitivity to cAMP, whereas coexpression of the three subunits almost restores native sensitivity (Sautter et al., 1998). Fodor and Zagotta (1996), Gordon et al. (1996), and Shammat and Gordon (1999) also report an increased ratio of cAMP-induced to cGMP-induced currents when the human rod  subunit is coexpressed with the bovine rod  subunit. In the rod  subunit, as in the olfactory subunit 2 and in CNG4.3, the residue corresponding to the acid residue D604 in the rod  subunit (or E581 in the olfactory subunit 1) is an uncharged residue (M in the rat olfactory subunit 2, N in CNG4.3 and in the rod  subunit). Fodor and Zagotta (1996) proposed that the  subunit may be responsible for the increased sensitivity of the native rod channel compared to the expressed  subunit and suggested that the uncharged residue at the position equivalent to D604 (N1201) might explain this effect.

Another characteristic of native rod channels is potentiation of cGMP-induced currents by low concentrations of cAMP (Furman and Tanaka, 1989; Ildefonse et al., 1992); a communication concerning the study of this phenomenon on expressed heteromeric channels has been published (Scott and Tanaka, 1998), but it is not known whether it could be related to a higher sensitivity of the  subunits to cAMP.

We report here a comparative study of the sensitivity to cGMP and cAMP and of cAMP potentiation of cGMP-induced currents of expressed channels consisting of bovine  subunits or of coexpressed bovine  and  subunits. The role of the residue in position 604 in the  subunit and in the corresponding position in the  subunit (1201) is studied by comparing currents from wild-type channels (wt and wt/wt) and from mutated channels (D604N, D604N/wt, and wt/N1201D).

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Channel expression

The bovine  subunit cDNA (Kaupp et al., 1989) was a gift of Prof. U. B. Kaupp. The  subunit cDNA was amplified by polymerase chain reaction from bovine retinal cDNA, using oligonucleotide primers chosen according to the published sequence (Körschen et al., 1995). Retinal mRNA was prepared from fresh retinas with the Dynabeads mRNA DIRECT kit (DYNAL), and cDNA was synthesized using the First-strand cDNA synthesis kit (Amersham Pharmacia Biotech). The N-terminal domain, which was shown to have no effect on the sensitivity to nucleotides (Körschen et al., 1995), was deleted up to G571, and a methionine residue was introduced before V572 for translation initiation. A Kozak consensus sequence was engineered upstream of the ATG codon, and the truncated cDNA was inserted in the high expression vector pGemHe downstream of the untranslated sequence of the Xenopus -globin gene (Liman et al., 1992). The cDNA sequence of our  subunit from the codon corresponding to V572 is 100% identical to that of CNG4c (Biel et al., 1996), leading to several modifications compared to the sequence published by Körschen et al. (1995): S/A substitution at position 1283, R/A at position 1289, D/E at position 1336, and insertion of A between D1336 and A1337 (all amino acid numbers refer to the sequence of Körschen et al.). Mutations (D604N and N1201D) were created by replacing the GAT (aspartate 604) codon with AAT (asparagine) in the  cDNA, and the AAC (asparagine 1201) codon with GAC (aspartate) in the  cDNA. Mutated sequences were verified by sequencing.

Capped mRNAs were synthesized in vitro from linearized plasmids in the presence of RNA cap structure analogs (New England Biolabs) and injected into Xenopus oocytes (25 ng/oocyte for macroscopic currents or 0.25 ng/oocyte for single channels). For coexpression of  and  subunits,  mRNA and  mRNA were mixed and injected into the oocyte. To reduce the probability of forming homomeric  channels, the : mRNA ratio was 2 for all experiments (except for single-channel analysis, where it was 3). Different ratios were not tested. Oocytes were incubated for 4-10 days in Barth's medium before measurements. As previously reported (Chen et al., 1993; Körschen et al., 1995), the wild-type  subunit did not form functional channels when expressed alone.

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 a RSC100 rapid solution changer (Bio-Logic, Claix, France). Currents induced by voltage steps (500 ms, ±80 mV) were recorded with a RK-400 patch amplifier (Bio-Logic), low-pass filtered at 300 Hz, and digitized at 1 kHz (macroscopic currents, each record averaged three times), or at 10 kHz and digitized at 33 kHz (single channels), using pCLAMP 6.0 (Axon Instruments). For macroscopic currents, the series resistance was compensated for (resulting value < 1 M). Dose-response curves were obtained by plotting the current at +80 mV as a function of nucleotide concentration after subtraction of the leak current.

Probability of channel opening

P_0max(cGMP) (open probability at saturation of cGMP) was estimated by two methods:

1. From the ratio of currents at saturation of cGMP in the absence and in the presence of Ni²⁺ (1 µM). Micromolar concentrations of cytoplasmic Ni²⁺ were previously shown to potentiate cGMP-induced currents (Ildefonse et al., 1992; Gordon and Zagotta, 1995), and the I_max(cGMP)/I_{max(cGMP + Ni)} ratio was shown to be very close to the P_0max(cGMP) value obtained from single-channel measurements for homomeric  (wild type and mutated) channels (Sunderman and Zagotta, 1999). The cGMP-induced currents were measured several times (before and after addition of Ni²⁺) until stabilization; the effect of Ni²⁺ was at maximum after 4-5 min. For these experiments, high-grade KCl or NaCl (containing less than 0.025 ppm transition metals; Merck) was used, and the HEPES concentration was reduced to 5 mM. No EGTA or EDTA was added to the perfusion medium; the solution in the pipette contained 200 µM EDTA and 500 µM niflumic acid.

2. From single-channel records analysis: Amplitude histograms were computed from single-channel records at +80 mV (record duration: 16-38 s for wt, 10-50 s for wt/wt, 18-77 s for D604N), using Bio-Patch software (Bio-Logic). Records were sampled at 33 kHz and numerically filtered (Hanning window) at 4 and 1 kHz. The histograms were fitted with two or three Gaussian curves.

Curve fitting

Fits of dose-response curves were calculated with Microcal Origin software. The error on the value of the parameters calculated by the program is error(i) = (C(i)(i)·²), where C(i)(i) is the covariance matrix for n parameters (i = 1, n).

Hill equation

Monod-Wyman-Changeux model

Statistics

The significance of the difference between two populations of data was analyzed by independent t-tests using Origin software.

Chemicals

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

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Coexpression of the wt subunit with the wt subunit increases the sensitivity to cAMP; role of residues D604 in  subunit and N1201 in  subunit

Plots of currents at saturating cGMP and cAMP concentrations obtained from the same patch in one experiment with homomeric wt channels and in one experiment with heteromeric channels expressed in oocytes coinjected with the wt and wt subunit mRNAs are shown in Fig. 1 A. The presence of the  subunit clearly increases the current at saturating cAMP concentration compared to the current at saturating cGMP concentration. The effect is observed independently of the voltage but is more evident at positive voltage because of the larger amplitude of cAMP-induced currents. To test whether the increased sensitivity conferred by coexpression of the  subunit is due to the uncharged residue N1201, as suggested by the results of Varnum et al. (1995) and Fodor and Zagotta (1996), we constructed a mutated  subunit in which D604 is replaced by the neutral asparagine residue present in the  subunit at the corresponding place (D604N) and the symmetric mutated  subunit N1201D.

View larger version (20K): [in this window] [in a new window]   FIGURE 1   Comparison of cGMP- and cAMP-induced currents for different channel compositions. (A) Examples of current recordings at saturating cGMP (0.5 mM) and cAMP (20 mM) concentrations from oocytes injected with wt or wt + wt mRNAs. The voltage step protocol is shown above. When accumulation/depletion was observed, the value for the current was taken as the average between the initial value and the value at 500 ms; this approximation gives values close to those obtained with the correction proposed by Zimmerman et al. (1988). (B) Mean ratios of the current at saturation of cAMP and cGMP for different channel compositions. Saturating concentrations were 20 mM for cAMP and 0.5 mM (wt, wt + wt, wt + N1201) or 5 mM (D604N, D604N + wt) for cGMP. For each patch, cAMP- and cGMP-induced currents were measured several times, as closely as possible, with control measurement of the leak current before and after. Mean values of Imax(cAMP)/Imax(cGMP) (± SE) at +80 mV and 80 mV are listed in Table 1. Imax(cGMP) was between 1098 pA and 4953 pA (mean value ± SE:2733 pA ± 378 pA) at + 80 mV and between 2474 and 5127 (mean value 4009 pA ± 371 pA) at 80 mV for wt; between 1188 pA and 3969 pA (mean value ± SE: 2505 pA ± 242 pA) at +80 mV and between 1440 pA and 4778 pA (mean value ± SE: 2968 pA ± 373 pA) at 80 mV for wt + wt; between 660 pA and 3500 pA (mean value ± SE: 2425 pA ± 300 pA) at +80 mV and between 2032 pA and 4749 pA (mean value ± SE: 3102 pA ± 277 pA) at 80 mV for wt + N1201D; between 455 pA and 3653 pA (mean value ± SE: 1733 pA ± 272 pA) at + 80 mV and between 132 pA and 2200 pA (mean value ± SE: 938 pA ± 168 pA) at 80 mV for D604N; between 431 pA and 2623 pA (mean value ± SE: 1353 pA ± 184 pA) at + 80 mV and between 124 pA and 2200 pA (mean value ± SE: 713 pA ± 131 pA) at - 80 mV for D604N + wt.

Mean I_max(cAMP)/I_max(cGMP) ratios at +80 mV and 80 mV from several experiments with different channel compositions are plotted in Fig. 1 B, and values are listed in Table 1. Coexpression of the wt subunit with the wt subunit increases the I_max(cAMP)/I_max(cGMP) ratio, although to a lesser extent than does the D604N mutation in the  subunit, which is as expected if the effect is due to the uncharged N1201 or N604 residues. However, upon coexpression of the mutated N1201D subunit with the wt subunit, the I_max(cAMP)/I_max(cGMP) ratio remains intermediate between those of the wt and wt/wt channels, and coexpression of wt with D604N further increases the I_max(cAMP)/I_max(cGMP) ratio compared to D604N channels. These effects are more clearly observed at positive voltage.

Dose-response curves at +80 mV for each channel composition are shown in Fig. 2. The values of EC₅₀ and n_H for cGMP and cAMP indicated in Table 1 are the parameters of the fits to the Hill equation of all of the data (normalized to the current at saturation of nucleotide) from all experiments. Inversely to the variation of the I_max(cAMP)/I_max(cGMP) ratio, the EC₅₀ for cAMP varies in the order wt > wt/N121D > wt/wt > D604N > D604N/wt. The errors calculated by the program suggest that the EC₅₀ for cAMP of the five channel types are significantly different, although this assumption should be made with caution because of the large variations between experiments. The larger effect is observed upon coexpression of wt with D604N, compared to all other channel types. The population of EC₅₀ obtained from the fit of each experiment for D604N/wt channels is also significantly different from that of all other channel types, including D604N (p < 10²); therefore, coassembly of the wt subunit with D604N seems to further reduce the EC₅₀ for cAMP compared to D604N channels, suggesting that the effect is not only due to the charge of residues N1201 or N604.

View larger version (27K): [in this window] [in a new window]   FIGURE 2   Dose-response curves for cGMP- and cAMP-induced currents from oocytes injected with mRNAs for wt (a), wt + wt (b), wt + N1201D (c), D604N (d), and D604N + wt (e). , cGMP; , cAMP. Several dose-response curves for cGMP and cAMP were measured for each channel composition (different patch for each experiment). For each experiment, currents were normalized to the current at saturating cGMP or cAMP concentration (calculated from the fit of the raw data to the Hill equation). Hill fits shown on the graphs were obtained using all of the normalized data points from all experiments for each channel composition. The cAMP dose-response curves were then multiplied by the Imax(cAMP)/Imax(cGMP) ratios from Table 1. For clarity, only the mean of all data points for each nucleotide concentration (± SE) is shown. Parameters that give the best fits and the number of experiments for each channel composition are listed in Table 1. Hill fits of cGMP and cAMP dose-response curves for wt channels (a) are shown for comparison (·····) in b, c, d, and e.

Note also that whereas a 15-fold increase of EC₅₀ for cGMP is observed for D604N channels compared to wt channels, only a limited (if significant) increase is observed when wt is coexpressed with wt.

Estimates for the open probability of homomeric (wt, D604N) and heteromeric (wt/wt) channels

The P_0max(cGMP) of wt, wt/wt, and D604N channels was estimated by two different methods (see Materials and Methods): from the I_max(cGMP)/I_max(cGMP+Ni ratio (Gordon and Zagotta, 1995; Varnum et al., 1995; Varnum and Zagotta, 1996; Sunderman and Zagotta, 1999) and from single-channel recordings (Fig. 3). Values obtained with the two methods are indicated in Table 2. They are similar for wt and wt/wt channels, but the agreement is less satisfying for D604N channels (see below). Nevertheless, whatever the method used, P_0max(cGMP) decreases in the order P_0max(cGMP) (wt) > P_0max(cGMP) (wt/wt) > P_0max(cGMP) (D604N). P_0max(cAMP) can be deduced from the value of P_0max(cGMP) and the I_max(cAMP)/I_max(cGMP) ratio (Table 1) and varies in the inverse order P_0max(cAMP) (wt) < P_0max(cAMP) (wt/wt) < P_0max(cAMP) (D604N).

View larger version (48K): [in this window] [in a new window]   FIGURE 3   Single-channel records of expressed wt/wt (a), wt (b), and D604N (c) channels. Single-channel recording and analysis were performed as described in Materials and Methods. The records were filtered at 1 kHz for the figure. Amplitude histograms corresponding to the full records filtered at 4 kHz (·····) or 1 kHz () are shown. Histograms at 4 kHz were used to calculate P0max. Values of P0max for the examples shown are 0.88 and 0.01 (wt/wt in the absence or presence of 50 µM L-cis-diltiazem); 0.98 and 0.86 (wt in the absence or presence of 50 µM L-cis-diltiazem); and 0.06 and 0.27 (D604N after 20 s and 3 min in the presence of cGMP). For D604N, NEM (2 mM) was added to the bath at the end of the experiment to check the number of channels in the patch (a single channel in the example shown).

When wt and wt mRNAs were coinjected, the nature of the expressed single channel was checked by the addition of 50 µM L-cis-diltiazem (Fig. 3, a and b); for 12 of 13 records, diltiazem reduced P₀ to almost 0, whereas it was only reduced by less than 20% for the other, as well as for patches from oocytes injected with  mRNA only (9-18%, four patches). This suggests that the population of channels is mainly composed of heteromeric channels when the two mRNAs are coinjected, consistent with the results of Shammat and Gordon (1999), who suggest that channel assembly may be biased toward the inclusion of  subunits. In our experiments, the unitary current was similar for all single wt/wt channel records (1.38 ± 0.05 pA at +80 mV, nine patches), suggesting that the population of heteromeric channels is also homogeneous, in contrast to the results of Torre et al. (1997), who describe three distinct heteromeric channel types. The unitary current is also similar to that of wt channels (1.35 ± 0.04 pA, four patches) and D604N channels (1.37 ± 0.07 pA, nine patches). The low value compared to published data can be related to the fact that the channel conductance is smaller when the permeating cation is K⁺ than when it is Na⁺ (Nizzari et al., 1993; G_Na/G_K = 1.24 at +140 mV).

As the P_0max(cGMP) for D604N channels measured with the Ni²⁺ method (0.35 ± 0.04, Table 2) was considerably higher than that measured by Sunderman and Zagotta (1999) (0.08 with the same method), we asked whether this could be related to the nature of the permeating ion (which is K⁺ instead of Na⁺ in our experiments); when K⁺ was replaced by Na⁺, P_0max(cGMP) was indeed reduced to 0.20 ± 0.03 (eight patches), still remaining higher, however, than the value reported by Sunderman and Zagotta (1999). In our experiments, the onset of the cGMP-induced current is slow, reaching a steady state in 1-2 min, and single-channel records of D604N channels reveal a heterogeneity in the channel activity that was not previously reported. The P_0max(cGMP) value obtained from single-channel analysis for D604N channels varies with time for a given patch, usually starting with a low value (0.08 ± 0.03, six different patches) within the first minute in the presence of cGMP, and then increasing to higher values (0.41 ± 0.09, eight patches), as in the example shown in Fig. 3 c. The maximum P_0max(cGMP) value obtained was also variable for different patches (between 0.15 and 0.83, eight patches). For three of these patches, long records (4-6 min) were analyzed, showing that after reaching a higher value, the activity did not clearly stabilize, but varied with periods of tens of seconds at any level between the two extreme values, and with long closed periods of several seconds or tens of seconds. The mean value indicated in Table 2 (0.25 ± 0.04) includes all of the 29 records (18-77 s, total length 1302 s) from the eight patches. The number of channels in the patch was measured at the end of the experiment (Fig. 3 c) by the addition of N-ethylmaleimide (2 mM) (Serre et al., 1995; Gordon et al., 1997), which increases P_0max(cGMP) to almost 1; only single-channel records were retained. P_0max(cGMP) in the presence of NEM was estimated to be 0.92 ± 0.03 from single-channel analysis. This allows a third independent determination of P_0max(cGMP) from the ratio of macroscopic currents in the absence and in the presence of NEM: a value of 0.39 ± 0.02 (seven patches) was obtained for I_max(cGMP)/I_{max(cGMP+NEM)}, corresponding to a P_0max(cGMP) of 0.36, close to the estimate obtained from Ni²⁺ potentiation. Thus, although there is a large variation between the different estimates (which is probably due to the fact that the estimate from single-channel analysis includes the low initial values, whereas estimates from macroscopic currents are obtained after stabilization of the currents), the P_0max(cGMP) of D604N channels is higher than reported by Sunderman and Zagotta (1999), whatever the method used.

In conclusion, coexpression of the wt subunit with the wt subunit reduces the gating efficacy of cGMP and increases that of cAMP. Similar though larger effects are produced by the D604N mutation in the  subunit.

Potentiation of cGMP-induced currents by low concentration of cAMP for homomeric (wt) and heteromeric (wt/wt and wt/N1201D) channels

It was previously reported that in the native channel, low concentrations of cAMP, which alone induce a very low current, are able to potentiate cGMP-induced currents (Furman and Tanaka, 1989; Ildefonse et al., 1992). The effect is best observed for cGMP concentrations below EC₅₀. This effect was interpreted as an indication that the dissociation constant for cAMP is much lower than the EC₅₀ measured from dose-response curves, which also depends on the capacity of the cAMP-bound channel to open. The question arises whether this phenomenon is also observed with expressed  channels. The fact that the  subunit increases the sensitivity to cAMP suggests that the potentiation by cAMP could be due to binding of cAMP with higher affinity to the  subunit than to the  subunit, which perhaps is due, at least in part, to N1201.

We have studied the potentiation of cGMP-induced currents by a low concentration of cAMP for three different channel compositions: wt, wt/wt, and wt/N201D. Dose-response curves were measured on the same patch in the presence or absence of 100 µM cAMP (which alone produces a very low current; see legend to Table 3). As an increase in apparent affinity for the nucleotide has been reported to spontaneously occur with time (Gordon et al., 1992; Molokanova et al., 1997), the current was measured three times for each cGMP concentration: first without cAMP, then with cAMP, and again without cAMP, to check the reversibility of the change. The curve obtained in the presence of cAMP and the average curve in the absence of cAMP were fitted to the Hill equation; the variations in EC₅₀ (EC₅₀) and n_H (n_H) were calculated for each experiment. The mean values of EC₅₀ and n_H from all experiments are listed in Table 3.

Fig. 4 shows mean data points from all experiments, normalized to the current at saturating cGMP concentration for each dose-response curve; the fits to the Hill equation shown on the graph were calculated with all data points. EC₅₀ and n_H for each channel composition are very similar to those obtained from the mean parameters of individual fits (Table 3).

View larger version (12K): [in this window] [in a new window]   FIGURE 4   cAMP-potentiation of cGMP-induced currents in oocytes injected with mRNAs for wt (a), wt + wt (b), and wt + N1201D (c). , In the absence of cAMP; , in the presence of 100 µM cAMP. Currents induced by each cGMP concentration were measured first in the absence of cAMP, then in the presence of 100 µM cAMP, and again in the absence of cAMP (a: eight experiments; b: 11 experiments; c: nine experiments). The currents induced by cGMP or by cGMP + cAMP were normalized to the value at saturating cGMP concentration. No increase in the current at saturating cGMP concentration was observed in the presence of added cAMP. The curves are the best fits of all of the normalized data to the Hill equation: a (wt): EC50 = 26.3 ± 1 µM, nH = 2 ± 0.14 (without cAMP, ), EC50 = 27.6 ± 1 µM, nH = 1.95 ± 0.1 (with cAMP, - - -); b (wt/wt): EC50 = 35 ± 0.6 µM, nH = 2 ± 0.1 (without cAMP, ), EC50 = 24 ± 0.7 µM, nH = 1.65 ± 0.1 (with cAMP, - - -); c (wt/N1201D): EC50 = 29.9 ± 0.4 µM, nH = 2.15 ± 0.06 (without cAMP, ), EC50 = 26.8 ± 0.4 µM, nH = 1.97 ± 0.06 (with cAMP, - - -). For clarity, only the mean values of data points obtained for each cGMP concentration from all experiments are indicated (± SE). The data in the presence or absence of cAMP for each experiment were also fit individually to the Hill equation (not shown); mean values of the variation in EC50 and in nH obtained for each experiment are listed in Table 3.

Table 3 and Fig. 4 show that potentiation by cAMP is not observed for homomeric wt channels. The presence of cAMP may even slightly inhibit the cGMP-induced current: for each of the eight experiments, the EC₅₀ of the dose-response curve for cGMP in the presence of 100 µM cAMP was slightly increased compared to that in the absence of cAMP. Potentiation by cAMP (reduction of EC₅₀ for cGMP), however, is observed for the two heterooligomeric channels, but the effect is clearly more pronounced for wt than for N1201D.

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSIONS REFERENCES

Comparison between the functional properties of homomeric wt channels, heteromeric channels from coexpressed wt and wt subunits, and native channels

While this work was in progress, Shammat and Gordon (1999) published a study of the coexpression of the wild-type bovine rod  subunit and human rod  subunit, in which they compare the I_max(cAMP)/I_max(cGMP) ratio for the two channel types. Our results with coexpressed wild-type bovine  and  subunits are totally consistent with their results: the value of the I_max(cAMP)/I_max(cGMP) ratio (15%, Fig. 1, compared to 13% in Shammat and Gordon, 1999) is significantly higher than for homomeric  channels and comparable to that previously obtained from native channels (Tanaka et al., 1989; Gavazzo et al., 1996; Picco et al., 1996). Although Scott and Tanaka (1998) report that coexpression of the  subunit with the  subunit does not restore the native I_max(cAMP)/I_max(cGMP) ratio, it should be noted that, as shown in Fig. 1, the effect may be unnoticed if currents are measured at negative voltage, because of the very low amplitude of the currents.

Similar EC₅₀ values for cGMP for homomeric  and heteromeric / channels were previously reported by Chen et al. (1993) (60-80 µM at +60 mV, human rod), Körshen et al. (1995) (40 µM at +80 mV, bovine rod), Scott and Tanaka (1998) (80 µM), and Shammat and Gordon (1999) (77 ± 31 µM and 63 ± 30 µM at +100 mV), although there is some variation in the value itself. Altenhofen et al. (1991) report a value for expressed bovine  subunits (32 ± 13 µM at +80 mV) that is closer to our value.

The value of the open probability obtained for wt channels (P_0max(cGMP) = 0.97 from Ni²⁺ potentiation, or 0.96 from single-channel analysis; Table 2) is consistent with previous reports: P_0max = 0.9 from noise analysis (Goulding et al., 1994); 0.78 from single-channel measurements (Bucossi et al., 1997); 0.96 ± 0.01 from the ratio of I_max(cGMP) in the presence or absence of Ni²⁺; or 0.95 from single-channel recordings (Sunderman and Zagotta, 1999). No report concerning the open probability of expressed heteromeric / channels is as yet available. Torre et al. (1997) published a single-channel study of coexpressed bovine  and  subunits, where they describe three channel types with different properties, but they do not give any estimate of P_0max. Our estimates of P_0max(cGMP) for wt/wt channels from single-channel analysis (0.85 ± 0.03, +80 mV) and Ni²⁺ potentiation (0.88 ± 0.01, +80 mV) are higher than previous measurements on native rods; from a single-channel kinetic analysis Taylor and Baylor (1995) obtained values of P_0max = 0.56 (+50 mV) and 0.30 (50 mV), consistent with the report of Matthews and Watanabe (1988) (0.30 ± 0.05 at 71 mV). These works, however, were both performed on amphibian rods, with Na⁺ as the permeating cation. Our results suggest that Ni²⁺ potentiation is a reliable method for estimating the open probability of heteromeric / channels, as previously demonstrated for homomeric  channels by Sunderman and Zagotta (1999).

Because  subunits alone can form functional channels and  subunits alone cannot, a mixed population of homomeric  and mixed heteromeric / channels (₂₂ and ₃) could be expected when  and  subunits are coexpressed (as well as for native channels). (With a : mRNA ratio = 2 (3), assuming that the two messengers are translated with the same efficiency, that the stabilities of the two proteins are equivalent, and that channels with three or four  subunits are not functional, the probabilities of forming homomeric , heteromeric ₃, and heteromeric ₂₂ would be, respectively, 3 (1.5)%, 24 (18)%, and 73 (80)%.) In this case, the values measured from macroscopic currents for I_max(cAMP)/I_max(cGMP) and P_0max(cGMP) would be intermediate between those for / channels and those for  channels. The results of Shammat and Gordon (1999), however, suggest that when  and  mRNAs are coinjected, even at a 1:1 ratio, only heteromeric ₂₂ channels are formed. Our single-channel records of patches from oocytes coinjected with  and  mRNA in the presence of L-cis-diltiazem suggest that, for an : mRNA ratio of 3, only a few homomeric  channels may coexist with heteromeric channels.

Our results agree with those of Scott and Tanaka (1998) concerning the restoration of potentiation of cGMP-induced currents by low concentrations of cAMP by coexpression of the  subunit; the potentiation by cAMP of cGMP-induced currents observed with coexpressed  and  subunits is similar to that previously described for native channels (Furman and Tanaka, 1989; Ildefonse et al., 1992). Because evolution of the channel characteristics has been reported to occur spontaneously with time (Gordon et al., 1992; Molokanova et al., 1997), seemingly because of dephosphorylation of the channel, we have been very careful to check that the effect of cAMP on the cGMP-induced currents is reversible. Moreover, it should be noted that in our potentiation experiments, preliminary control measurements (leak current, current induced by 100 µM cAMP, current at saturation of cGMP and cAMP) are performed before measurements for dose-response curves and take several minutes (usually more than 5 min). The decrease in EC₅₀ for  channels expressed in oocytes described by Molokanova et al. (1997) is a fast process, which becomes negligible ~5 min after patch excision.

Role of the charge of residue 604 in the  subunit and 1201 in the  subunit in the sensitivity to cAMP, cAMP potentiation of cGMP-induced currents, and gating efficacy of cGMP

We have studied the role of the charge of residue 604/1201 in three aspects of channel function: sensitivity to cAMP (EC₅₀ and I_max(cAMP)/I_max(cGMP) ratio), gating efficacy of cGMP (P_0max(cGMP)), and cAMP potentiation of cGMP-induced currents.

The results show that, as previously reported (Varnum et al., 1995), replacing D604 in the  subunit by the uncharged residue N, which is present at the corresponding place in the  subunit, increases the I_max(cAMP)/I_max(cGMP) ratio. We also show that coexpressing the wt subunit with the wt subunit produces qualitatively similar (but smaller) effects, as is expected if the effect is due to the presence of an uncharged residue in position 604/1201, because the heteromeric channel has both D and N in position 604/1201, instead of 4N in D604N. However, coexpressing the wt subunit with D604N further increases the sensitivity to cAMP (I_max(cAMP)/I_max(cGMP) ratio and EC₅₀), and replacing N1201 by D in the  subunit does not restore the characteristics of the  homooligomer: wt/N1201D, which has 4 D in positions 604/1201, is intermediate between wild-type wt/wt (in which both D and N are present) and  channels (which also have 4 D in position 604).

Similarly, although N1201D is less efficient than wt, significant potentiation by cAMP of cGMP-induced currents is observed when the mutated  subunit is coexpressed with the  subunit.

Another effect of the D604N mutation is to reduce the channels' open probability for cGMP compared to wt channels (P_0max(cGMP) = 0.35 from Ni²⁺ potentiation or 0.25 from single-channel analysis; Table 2). These values are much higher than those reported by Sunderman and Zagotta (1999) with the same methods; as noted in the Results, the discrepancy could be due to the combined effect of the nature of the permeating cation, and the slow response of D604N channels to cGMP, the low values (0.07-0.08) measured by Sunderman and Zagotta (1999) being closer to the values that we usually obtain during the first minute in the presence of cGMP (Fig. 3 c). The P_0max(cGMP) value measured for D604N channels by Varnum et al. (1995) also seems higher than that reported by Sunderman and Zagotta (1999) (see below). The reduction of the gating efficacy of cGMP is also observed, although less marked, with the coexpression of the wt and wt subunits (P_0max(cGMP) = 0.88 or 0.85, according to the method used).

It can therefore be concluded that the charge of residue 604/1201 plays an important role in the sensitivity to cAMP, the gating efficacy for cGMP, and cAMP potentiation of cGMP-induced currents; however, other part(s) of the proteins participate in these effects. In addition, only a moderate (if significant) increase in the EC₅₀ for cGMP is observed when wt and wt are coexpressed, whereas a 15-fold increase is observed with D604N channels, also suggesting a role of other parts of the protein in the sensitivity to cGMP. It can be proposed that other residues in the binding site or in another domain (for example, in the C-linker; Zong et al., 1998; Paoletti et al., 1999) are determinants of the action of the nucleotide.

Interpretation of the data in terms of affinity and gating efficacy of the ligand

The open probability at the saturating concentration of nucleotide, i.e., when all of the binding sites are occupied (P_0max), is an indication of the gating efficacy of the nucleotide. Independently of the model, as long as unliganded (or partially liganded) channel openings are negligible compared to fully liganded channel openings, the constant for the gating transition,
can be calculated from the experimental estimates of P_0max:

Using the estimates of P_0max from Table 2 for the three channel types (wt, D604N, and wt/wt), we calculate that the free energies of gating (G_op = RT ln[K_op]) upon coexpression of the  subunit with the  subunit are intermediate between those for wt channels and those for D604N channels. The decrease in the free energy of gating by cAMP for D604N channels compared to wt (Table 5) is larger than (but less than twice) that for wt/wt channels, and the increase in the free energy of gating by cGMP for D604N channels compared to the wt channel is more than twice that for the heteromeric channel compared to the wt channel. If the effect on gating was only due to the charge of residue 604/1201, and if the  and  subunits assemble as an oligomer, as proposed by Shammat and Gordon (1999), a factor of 2 would be expected between G_op for D604N and wt/wt channels compared to wt. Our results may indicate that improved gating by cAMP upon coassembly of the  subunit is not due solely to residue N1201 (consistent with the conclusions drawn from the I_max(cAMP)/I_max(cGMP) ratios measured for heteromeric wt/N1201D and D604N/wt channels), and that reduced gating by cGMP of  D604N involves residues other than N604. Our results with D604N channels are consistent with those of Varnum et al. (1995), who calculated that the free energy of gating for cAMP was reduced by ~1.3 kcal/mol by mutations of D604 to neutral residues, whereas that for cGMP was increased by 2 kcal/mol for D604N. This indicates that the P_0max(cGMP) measured by these authors for D604N channels in this work is closer to our estimate (0.30 ± 0.05) than to the estimate given by Sunderman and Zagotta (1999) (0.07-0.08), which would produce a much larger G.

Using a simplified linear scheme with two binding sites, Varnum et al. (1995) find that the free energy of initial binding is not substantially altered by mutations at position 604. They conclude that interaction of the purine ring of the nucleotide with D604 in the C helix is important for the conformational change leading to channel opening rather than for initial binding. However, the value of the binding constant depends on the model chosen to calculate it. Although the linear model has proved useful in interpreting the effects of mutations in the  subunit (Gordon and Zagotta, 1995; Varnum et al., 1995), a fundamental aspect of CNG channel function is the existence of spontaneous openings in the absence of ligand (Picones and Korenbrot, 1995; Goulding et al., 1994; Tibbs et al., 1997; Ruiz and Karpen, 1997), which is not compatible with simple linear models of activation, in which channels can only open after binding of the ligand. Several allosteric models, including the Monod-Wyman-Changeux (MWC) concerted model (Monod et al., 1965; used by Goulding et al., 1994; Varnum and Zagotta, 1996; Tibbs et al., 1997; Paoletti et al., 1999), a coupled-dimer model in which two independent dimers undergo a concerted allosteric transition (Liu et al., 1998), and a complete scheme including intermediate states (Ruiz and Karpen, 1999) have been recently shown to be better adapted for the description of CNG channel function. In the case of heteromeric channels, the models should in fact be modified to include different characteristics for the two types of subunits, but this would increase the number of parameters and increase the uncertainty in the fitting operation. We have therefore used the simple MWC model to fit the results, as an approximation that should be closer to the real mechanism of activation than the linear model used by Varnum et al. (1995), to obtain the binding constants for the nucleotide. Fitting the data with the coupled dimer model (Liu et al., 1998) gives qualitatively similar results (not shown).

In the MWC and derived models, the protein can exist in two states (T, corresponding to the closed state, and R, corresponding to the open state of the channel); the T state has a lower affinity than the R state for the ligand (c = K_R/K_T 1, where K_R and K_T are the dissociation constants for the open and closed states). The transition between T and R is disfavored in the absence of ligand (L = [T]/[R] 1) and is increasingly favored upon ligand binding because of the higher affinity for the R state. The MWC model for a tetramer is represented by the following scheme (see also Materials and Methods):

The function represents the proportion of channels in the R (open) state, and its variation as a function of ligand concentration simulates the variation of the open probability. For clarity, we will use below the parameter c for cGMP (c = K_RG/K_TG, where K_RG and K_TG are the dissociation constants of the open and closed states for cGMP) and the parameter d for cAMP (d = K_RA/K_TA, where K_RA and K_TA are the dissociation constants of the open and closed states for cAMP).

Comparison of data obtained with wt and data obtained with wt/wt or with D604N, which show a lower P_0max for cGMP and a higher I_max(cAMP)/I_max(cGMP) for the wt/wt heterooligomer and for D604N than for wt, suggests that the difference is not due to a modification of L, which should produce similar modifications for the two ligands (see Materials and Methods), but rather to modifications of c and d, although it cannot be excluded that L is also modified, but that the modification due to L is masked by larger modifications due to c and d. This was also proposed by Varnum and Zagotta (1996) for mutations of D604 in the  subunit from interpretation of their data with the MWC model. We have therefore searched for values of K_RG and c (cGMP), K_RA and d (cAMP) that are consistent with the data for a fixed value of L (L = 7999, calculated from the value for spontaneous open probability P_sp = 1.25 ×10⁴ reported by Tibbs et al. (1997)). The experimental data that were used in Fig. 2 for three channel types (wt, wt/wt, and D604N), corrected for the P_0max values indicated in Table 2, were fitted to the function of the MWC model. As in Fig. 2, fits were performed using all of the data points from all experiments. Fig. 5 only shows the fits corresponding to the P_0max values obtained from single-channel analysis for the three channel types. Parameters that give the best fit of the data for both estimates of the P_0max(cGMP) are indicated in Table 4.

View larger version (14K): [in this window] [in a new window]   FIGURE 5   Fits of dose-response curves according to the MWC model (Monod et al., 1965) for wt (a), wt/wt (b), and D604N (c) channels. , cGMP; , cAMP. Normalized data points from all experiments used for the dose-response curves shown in Fig. 2 were fitted to the function of the MWC model (see Materials and Methods), after the currents were corrected at saturating nucleotide concentrations for the P0max values from Table 2. The data in the figure were corrected for the P0max(cGMP) estimated from single-channel analysis. For clarity, only the mean values of all data points obtained for each ligand concentration are shown in the figure (± SE). A fixed value of L, calculated from the value of spontaneous channel openings Psp = 1.25 ×104 measured by Tibbs et al. (1997), was used: L = 1/Psp  1 = 7999. The free parameters were KRG and c (cGMP), KRA and d (cAMP). Fits for wt channels (from a) are indicated in b and c (- - -). Parameters that give the best fit are given in Table 4. Note that the slopes of the different dose-response curves (reduced nH for cAMP compared to cGMP for wt-containing channels, and reduced nH for cGMP for D604N compared to wt channels, Table 1) are well simulated by the MWC model, in which an increase in the parameter c (which reflects the gating efficacy of the ligand) results in a reduction of the Hill number (Rubin and Changeux, 1966).

Dose-response curves measured upon coexpression of wt with wt are fitted with reduced K_RA and K_TA, compared to the values that fit dose-response curves of wt, the effect being more marked for K_RA, which corresponds to a decrease in d (increasing the gating efficacy of cAMP); coexpression of