INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The gating of cyclic nucleotide-gated (CNG) channels is the final step in signaling cascades in the sensory neurons of the visual and olfactory systems (Yau and Baylor, 1989; Zufall et al., 1994). These CNG channels are also found in testis, kidney, heart, and brain, where they may provide a mechanism for intracellular cGMP and cAMP to directly modulate the electrical state of the cell and levels of intracellular Ca²⁺ (McCoy et al., 1995; Biel et al., 1994; Weyand et al., 1994; Zufall et al., 1997; Rieke and Schwartz, 1994). Like the kinases, distinct CNG channels are differentially activated by cAMP or cGMP. In the rod photoreceptor, cGMP is the physiological ligand (Tanaka et al., 1989; Baylor and Nunn, 1982), and the CNG channels in the rod strongly select for cGMP over cAMP (Ildefonse et al., 1992; Gordon and Zagotta, 1995b). However, in the olfactory receptor neuron, CNG channels are opened equally well by cAMP, produced by odorant-stimulated activation of adenylyl cyclase (Nakamura and Gold, 1987; Anholt, 1993). Insight into the mechanism of ligand specificity of CNG channels has come from the x-ray structures of catabolite gene-activator protein (CAP) (Weber and Steitz, 1987), a dimeric cAMP-regulated transcription factor, and the regulatory subunit of the cAMP-dependent protein kinase (Su et al., 1995). The cyclic nucleotide-binding region (CNBR) of CNG channels exhibits sequence similarity to that of CAP (Kaupp et al., 1989), and this similarity has given us the opportunity to probe the molecular mechanism of ligand specificity. In CAP, the structure of the CNBR consists of a -roll, made from eight  strands, followed by two  helices, the B-helix, and the C-helix. The cAMP molecule is bound inside the -roll, and its purine ring interacts with T127 in the C-helix of the same subunit. Thus, it has been suggested that a region in the putative C-helix in the CNBR of CNG channels plays a pivotal role in determining ligand specificity (Goulding et al., 1994; Varnum et al., 1995). In particular, Varnum et al. (1995) showed that D604 in the rod CNG channel  subunit, at a position equivalent to T127 in CAP, is responsible for the high specificity for cGMP over cAMP in these channels. Mutation of D604 to methionine caused a dramatic decrease in the efficacy of cGMP, and an increase in the efficacy of cAMP, as agonists (Varnum et al., 1995; Sunderman and Zagotta, 1999a).

The cloning of the rod (Kaupp et al., 1989) and olfactory (Dhallan et al., 1990; Ludwig et al., 1990) CNG channel  subunits has demonstrated that these channels are members of the voltage-activated family of channels (Jan and Jan, 1990, 1992). Like voltage-activated channels, CNG channels contain six membrane-spanning segments, a pore-forming P-region, and a tetrameric arrangement of subunits (see Zagotta and Siegelbaum, 1996 for a review). Native CNG channels are heteromultimers of at least two kinds of subunits,  and . When expressed alone, olfactory  (CNG2, CNC3, OCNC1) subunits produce functional homomeric channels, whereas  subunits alone do not produce CNG currents (Chen et al., 1993; Bradley et al., 1994; Liman and Buck, 1994; Korschen et al., 1995; Sautter et al., 1998; Bonigk et al., 1999). Recently, it has been shown that the native olfactory channel contains two types of  subunits. Heteromeric channels containing both types of  subunits behave more like native olfactory channels than do channels containing only one type of  subunit (Sautter et al., 1998; Finn et al., 1998; Bonigk et al., 1999). One type of  subunit (CNG4.3, CNC1b) is an alternatively spliced variant of the rod  subunit and was recently cloned from olfactory epithelium (Sautter et al., 1998; Bonigk et al., 1999). The other type of  subunit (CNG5, CNC4, OCNC2) is expressed at high levels in sensory neurons of the primary olfactory epithelium and vomeronasal organ (Liman and Buck, 1994; Bradley et al., 1994; Berghard et al., 1996) and will be referred to in this study as the olfactory  subunit. This subunit has 52% sequence identity with the olfactory  subunit, and 30% sequence identity with the rod  subunit (Bradley et al., 1994; Liman and Buck, 1994). It has a membrane topology similar to the  subunit but lacks much of the amino-terminal region of the olfactory  subunit. This region of the olfactory  subunit has been shown to be an autoexcitatory/calmodulin binding domain that strongly interacts with the gating machinery of the CNBR (Liu et al., 1994; Varnum and Zagotta, 1997).

Incorporation of the olfactory  subunit changes the gating properties of the olfactory channels. Channels composed of both  and  subunits (+ channels) are half-activated (K_1/2) by much lower concentrations of cAMP than channels composed of only  subunits ( channels) (Liman and Buck, 1994; Bradley et al., 1994). Since cAMP is the ligand for these channels in olfactory neurons, this raises the possibility that the role of  subunits is precisely to achieve the necessary high apparent affinity of cAMP, making the functional effect of the  subunit of particular physiological significance. In this study we focus on the olfactory  subunit and investigate how this  subunit affects CNG channel gating and pharmacology, and explored the structural determinants of these effects. By comparing the properties of  channels, + channels, and +chimeric / channels, we show that the effects of the  subunit on ligand specificity localize to the CNBR, while the effects on the slope of the dose-response relation, voltage dependence, desensitization, and Mg²⁺ block localize outside the CNBR. Furthermore, we show that a single residue in the C-helix of the CNBR could account for the altered ligand specificity.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

The cDNA for the  subunit (CNG2, CNC3, OCNC1) and the  subunit (CNG5, CNC4, OCNC2) of the rat olfactory CNG channel were kindly provided by the laboratories of R. R. Reed (The Johns Hopkins School of Medicine, Baltimore, MD) and Kai Zinn (California Institute of Technology, Pasadena, CA), respectively. These cDNAs were separately subcloned into a high expression vector, kindly provided by E. R. Liman, that contains the untranslated sequences of the Xenopus -globin gene (Liman et al., 1992). In general, the oocyte expression and electrophysiology were like those previously described (Gordon and Zagotta, 1995a). Briefly, Xenopus oocytes were injected with in vitro transcribed RNA coding for channel subunits, incubated for 3-7 days at 16°C, and then patch-clamped in the inside-out configuration. Intracellular and extracellular solutions contained 130 mM NaCl, 0.2 mM EDTA, and 3 mM HEPES, pH 7.2. For some experiments, niflumic acid (500 µM final concentration) was added to the pipette (extracellular) solution to reduce endogenous Ca²⁺-activated Cl currents. Cyclic nucleotides were added to the internal solution at the concentrations indicated. The cDNAs for the chimeric subunits were generated using a method based on PCR like that previously described (Gordon and Zagotta, 1995a) and were verified by sequence analysis. For the -CNBR chimera, the sequence between C352 and E491 of the  subunit was replaced by the sequence between C460 and S593 of the  subunit. The -C5 chimera had the following mutations in the  subunit: M464L, K467R, L473M, M475E, N476G.

We generated heteromeric channels by co-injecting RNA for the  subunit together with either the wild-type  subunit or a chimeric / subunit. We found that co-injecting RNA for the subunits at a ratio of 4:1 : maximized expression of heteromultimers. The experiments summarized in Fig. 6, showing the results from a range of : RNA injection ratios from 2:1 to 100:1, indicate that an injection ratio of 4:1 produces sufficient  subunits to form almost exclusively heteromeric channels of their preferred subunit composition (Shapiro and Zagotta, 1998). Thus, all the rest of the data from heteromultimers were from RNA injection ratios of ~4:1. Heteromeric channels have a 1:1 stoichiometry (Shapiro and Zagotta, 1998), and so we interpret the optimal 4:1 injection ratio as reflecting a greater translational efficiency of  versus  subunits. Due to the large effect of the  subunit on the apparent affinity of the channel for cAMP, a population of -homomultimers in the coinjection experiments >10% would be easily seen in the dose-response curve. We did not see evidence of -homomultimers with an injection ratio of 4:1.

For patches with homomeric  channels, voltage pulses were applied every 3-5 s. Currents from heteromeric + or +chimeric / channels desensitized. For these channels, cyclic nucleotide-free solution was perfused for a minimum of 20 s before each application of ligand, and voltage pulses to these patches were applied every 1 s. Once ligand was applied, the pulse with the greatest current was used for the measurement. Using this protocol, we estimate that errors from desensitization were seldom greater than 15% for any given measurement.

For the Mg²⁺ experiments (Fig. 7), we added to our usual internal solution various amounts of MgCl₂ to obtain solutions with various free [Mg²⁺], calculated using the MAXC program written by Chris Patton, Stanford University. For solutions containing a free [Mg²⁺] of 120 µM, 811 µM, 2.81 mM, 9.81 mM, and 29.8 mM, the amount of MgCl₂ added was 300 µM, 1 mM, 3 mM, 10 mM, and 30 mM, respectively.

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We studied cloned rat olfactory CNG channels expressed in Xenopus oocytes using the inside-out configuration of the patch-clamp technique. Oocytes injected solely with RNA coding for the olfactory  subunit produced homomeric channels ( channels). Fig. 1 shows currents from  channels in response to various concentrations of cAMP, cGMP, or cIMP. Currents were recorded using successive pulses to 60 and 60 mV from a holding potential of 0 mV (Fig. 1, inset). To elicit CNG current, cyclic nucleotides were applied to the intracellular side of the patch and the leak currents in the absence of cyclic nucleotide were subtracted. For cAMP and cGMP, the properties of these  channels were very similar to those previously described for homomeric channels of the olfactory  subunit (Dhallan et al., 1990).

View larger version (36K): [in this window] [in a new window]   FIGURE 1   Currents in homomeric  channels, heteromeric + channels, and heteromeric +-CNBR channels activated by a range of concentrations of cAMP, cGMP, and cIMP. The pulse protocol used is shown at the bottom. Selected concentrations near half-maximal activation are labeled. For  channels, the cAMP concentrations were (µM) 10, 20, 50, 100, 200, 1000; the cGMP concentrations were (µM) 0.5, 1, 2, 5, 10, 20, 50, 100; and the cIMP concentrations were 50 µM, 100 µM, 200 µM, 500 µM, 1 mM, 2 mM, 10 mM. For + channels, the cAMP concentrations were (µM) 1, 2, 5, 10, 20, 50, 100, 200, 1000; the cGMP concentrations were (µM) 0.5, 1, 2, 5, 10, 20, 50, 100; and the cIMP concentrations were (µM) 10, 20, 50, 100, 200, 1000. For +-CNBR channels, the cAMP concentrations were (µM) 5, 10, 20, 50, 100, 200, 1000; the cGMP concentrations were (µM) 0.2, 0.5, 1, 2, 5, 10, 20, 50, 100; and the cIMP concentrations were 10 µM, 20 µM, 50 µM, 100 µM, 200 µM, 1 mM, 10 mM. For  channels, currents are from three different patches. For + channels, the currents in cGMP and cIMP are from the same patch, and the currents in cAMP are from a different patch. For +-CNBR channels, currents in cAMP and cGMP are from the same patch, and the currents in cIMP are from a different patch.

Oocytes injected only with RNA for the  subunit (CNG5, CNC4, OCNC2) did not express functional CNG channels. However, oocytes co-injected with RNA coding for both the  subunit and the  subunit expressed heteromeric channels formed from both the  and  subunits (+ channels) with gating and pharmacological properties different from  channels (Liman and Buck, 1994; Bradley et al., 1994). Compared to  channels, + channels were activated by much lower concentrations of cAMP and cIMP, but similar concentrations of cGMP. In addition, the currents exhibited a slow relaxation to steady state after a voltage step. This may reflect greater voltage dependence or slower gating of the heteromeric channels, such that more channels are open at depolarized versus hyperpolarized potentials. These currents also exhibited greater rectification at saturating concentrations of ligand than the currents from  channels. Finally, + channels, but not  channels, expressed in oocytes desensitized after the application of ligand over a period of several seconds (Fig. 2). Thus, when recording from heteromultimers, every application of ligand was preceded by >20 s in cyclic nucleotide-free control solution, which was sufficient time for full recovery (data not shown). Collectively, these identifying characteristics clearly distinguish between channels composed of  subunits and those composed of both  and  subunits, and represent the signature effects of the  subunit.

View larger version (17K): [in this window] [in a new window]   FIGURE 2   Desensitization. Currents from patches containing homomeric  channels (A), heteromeric + channels (B), or heteromeric +-CNBR channels (C), activated by 1 mM cAMP. Patches were held at 0 mV, and voltage pulses were applied to 60 mV every 1 s. In each case, the first pulse shown was given within 1 s of application of ligand. In A-C, the first five, seven, and seven pulses are shown, respectively.

To localize the region of the  subunit responsible for the altered properties, we first constructed a chimeric  subunit with the  subunit cyclic nucleotide-binding region (-CNBR). Like the wild-type  subunit, the chimeric subunit did not yield functional CNG channels when expressed alone, but when co-expressed with the  subunit yielded heteromeric channels (+-CNBR channels) with some properties like  channels and some properties like + channels. Like + channels, +-CNBR channels also exhibited a slow relaxation to steady state after a voltage step and rectification at saturating concentrations of cyclic nucleotides (Fig. 1). In addition, they also desensitized like + channels (Fig. 2). However, +-CNBR channels are activated by concentrations of cAMP, cGMP, and cIMP similar to  channels. We also generated the inverse chimera, consisting of the  subunit with the CNBR of the  subunit, but that chimera did not produce functional channels, either alone or as a heteromultimer with the  subunit.

Dose-response curves for activation of channels by cAMP, cGMP, and cIMP from patches expressing  channels, + channels, or +-CNBR channels are shown in Fig. 3. There were two robust effects of co-expression of the wild-type  subunit on the dose-response relation: a marked ligand-specific increase in the apparent affinity of the channel for cAMP and cIMP, and a ligand-nonspecific decrease in the slope of the dose-response curve. We quantified CNG channel gating by fitting dose-response data with the Hill equation. For cAMP (Fig. 3 A), the concentration that gave half-maximal current (K_1/2) at 60 mV in patches with  channels was 83 ± 3 µM (mean ± SEM, n = 26), but about eightfold less, 10.1 ± 0.7 µM (n = 22) for patches with + channels. For cGMP, however (Fig. 3 B), K_1/2 was similar for the two channel types: 2.8 ± 0.1 µM (n = 10) for  channels and 4.8 ± 0.8 µM (n = 6) for + channels. These results for activation of + channels by cAMP and cGMP are very similar to those reported (Liman and Buck, 1994; Bradley et al., 1994). As for cAMP, the presence of the  subunit had a large effect on the apparent affinity of the channel for cIMP (Fig. 3 C). For  channels, K_1/2 was 350 ± 34 µM (n = 10), but for + channels, K_1/2 was 86 ± 21 µM (n = 3). For all three ligands, the slope of the dose-response curve was considerably less for + channels (Hill coefficients = 1.3-1.5) than for  channels (Hill coefficients = 2.1-2.6).

View larger version (18K): [in this window] [in a new window]   FIGURE 3   Dose-response data. Data are the amplitude of CNG currents at 60 mV elicited by a range of cyclic nucleotide concentrations, normalized to the maximum current, for  channels (squares), + channels (circles), or +-CNBR channels (triangles) for cAMP (top), cGMP (middle), and cIMP (bottom). Superimposed on the data are fits to Hill equations of the form I = Imax([cNMP]n/K1/2n + [cNMP]n), where [cNMP] is the concentration of ligand, K1/2 is the concentration that produces half-maximal current, and n is the Hill coefficient. For activation of  channels by cAMP, cGMP, and cIMP, K1/2 = 85 µM, 2.8 µM, 363 µM and n = 2.2, 2.7, 2.4, respectively. For activation of + channels by cAMP, cGMP, and cIMP, K1/2 = 8.8 µM, 4.3 µM, 61 µM and n = 1.4, 1.3, 1.7, respectively. For activation of +-CNBR channels by cAMP, cGMP, and cIMP, K1/2 = 143 µM, 6.3 µM, 579 µM and n = 1.2, 1.2, 1.2, respectively.

Dose-response relations for +-CNBR channels indicated that their specificity for the three ligands reverted to be like  channels (Fig. 3). Data for activation of channels by cAMP, cGMP, and cIMP were fit by Hill equations with K_1/2 values of 119 ± 9 µM (n = 9), 7.0 ± 0.5 µM (n = 8), and 598 ± 152 µM (n = 3), respectively. However, Hill coefficients for activation of +-CNBR channels by the three ligands (1.1-1.4) were similar to those for activation of + channels. Thus, substitution of the  subunit CNBR into the  subunit nearly restores the K_1/2 values to be like  homomultimers; however, the shallow slopes of the dose-response curves remain. We conclude that the ligand-specific shift of the apparent affinities caused by the  subunit localizes to the CNBR, but the ligand-nonspecific shallowing of the dose-response relation, slow gating, rectification, and desensitization localize to a different part of the protein.

Structural basis for ligand specificity in the CNBR

To further localize the structural basis for the alterations in ligand specificity produced by the -CNBR chimera, we made more restricted chimeras within this domain. We focused on amino acid differences that were predicted from the CAP structure to be within 5-10 Å from the purine ring, the portion of the cyclic nucleotide that differs between cAMP, cGMP, and cIMP. A cluster of such residues was found in the putative C-helix of the CNBR, and we replaced five residues in the C-helix of the  subunit (M464L, K467R, L473M, M475E, N476G) with the corresponding residues of the  subunit (-C5). Replacement of these five residues dramatically increased the apparent affinity for cGMP and decreased the apparent affinity for cAMP of +-C5 channels compared to + channels (Fig. 4, A and B). For activation of + channels by cGMP, K_1/2 was 4.8 ± 0.8 µM (n = 6), but for +-C5 channels, K_1/2 decreased by >10-fold to 0.34 ± 0.03 µM (n = 5). The K_1/2 for activation of +-C5 channels by cAMP, however, was actually increased by almost twofold from 10.1 ± 0.7 µM (n = 22) for + channels to 18.5 ± 1.9 µM (n = 7) for +-C5 channels.

View larger version (31K): [in this window] [in a new window]   FIGURE 4   Effect of chimeras in the C-helix on cAMP selectivity. Data are the amplitude of CNG currents at 60 mV elicited by a range of concentrations of cAMP and cGMP for + channels (A), +C5 channels (B), or +-M475E channels (C). Superimposed on the data are fits to Hill equations, as described in the legend to Fig. 3. For activation of + channels by cAMP and cGMP, K1/2 = 6.38 µM and 3.85 µM and n = 1.7 and 1.7, respectively. For activation of +C5 channels by cAMP and cGMP, K1/2 = 20 µM and 0.34 µM and n = 1.5 and 1.5, respectively. For activation of +-M475E channels by cAMP and cGMP, K1/2 = 48 µM and 0.69 µM and n = 1.7 and 1.7, respectively. Plotted in D are box plots for the ratio of K1/2 values for cAMP to cGMP for these three channels. The vertical line in the middle of each box marks the median of the data. The box shows the middle half of the data, between the 25th and 75th percentiles. The "whiskers" show the data between the 5th and 95th percentiles. The circles are the extreme points in the data.

One of the mutations made in the -C5 chimera was M475E. This residue is at a position equivalent to D604 in the rod  subunit that has previously been shown, for rod  channels, to have a dramatic effect on cyclic-nucleotide specificity (Varnum et al., 1995). To test whether a difference at this residue alone might be able to account for the effects of the  subunit on cyclic nucleotide specificity, we constructed the point mutation M475E in the  subunit. Co-expression of this -M475E subunit with olfactory  subunits also yielded channels with a greatly increased apparent affinity for cGMP, and decreased apparent affinity for cAMP (Fig. 4 C). For activation of these +-M475E channels by cGMP and cAMP, K_1/2 was 0.55 ± 0.05 µM (n = 4) and 37 µM ± 4 µM (n = 4), respectively. Thus, the -M475E mutation produces large, ligand-specific alterations in the apparent affinities of the olfactory heteromeric channels, similar to those seen in rod  channels (Varnum et al., 1995; Varnum and Zagotta, 1996). In addition, comparison of +-CNBR channels with +-M475E channels suggests that the CNBR of the  subunit contains amino acid differences at positions other than 475 that produce an increase in the apparent affinity for each ligand.

We calculated the ratio of K_1/2 values for cAMP and cGMP (K_1/2,cAMP/K_1/2,cGMP) as a measure of the cAMP-to-cGMP selectivity (Fig. 4 D). For olfactory -homomultimers, cGMP is a more potent agonist than cAMP; K_1/2,cAMP/K_1/2,cGMP for these channels is nearly 30. The effect of the  subunit in the channel is to greatly increase the apparent affinity of cAMP relative to cGMP, decreasing K_1/2,cAMP/K_1/2,cGMP to ~2. Replacing the CNBR of the  subunit with that of the  subunit almost completely reversed this effect, so that K_1/2,cAMP/K_1/2,cGMP for +-CNBR channels was more similar to that for  channels. Furthermore, replacing just the C-helix (-C5) or only a single residue within the C-helix (-M475E) was sufficient to produce high cGMP-to-cAMP specificity. We conclude that the residue at position 475 in the olfactory  subunit plays a significant role in ligand discrimination, and that, like in rod  subunits, an acidic amino acid at this position induces high cGMP-to-cAMP selectivity. A methionine residue at this position in the olfactory  subunit allows for activation of the heteromultimeric olfactory channels by cAMP, its physiological ligand.

The residue at position 475 in the  subunit also plays a key role in determining the cIMP-to-cGMP specificity in olfactory heteromeric channels. Both +-C5 channels and +-M475E channels exhibited an apparent affinity for cIMP only slightly higher than the apparent affinity of cIMP for + channels (Fig. 5, A-C). For +-C5 channels, the K_1/2 for activation by cIMP was 37 ± 9 µM (n = 3), and for +-M475E channels, the K_1/2 was 53 ± 7 µM (n = 4). However, the K_1/2,cIMP/K_1/2,cGMP ratios for these channels were much more similar to those of the  channels than to those of + channels (Fig. 5 D). This result reflects the fact that the -C5 chimera and -M475E point mutation are having a much larger effect on activation by cGMP than on activation by cIMP. Since cGMP and cIMP differ at only the 2 position of the purine ring, this result suggests that, like in the  subunit, the amino acid at 475 is able to interact with this region of the cyclic nucleotide molecule.

View larger version (31K): [in this window] [in a new window]   FIGURE 5   Effect of chimeras in the C-helix on cIMP selectivity. Data are the amplitude of CNG currents at 60 mV elicited by a range of concentrations of cGMP and cIMP, for + channels (A), +C5 channels (B), or +-M475E channels (C). Superimposed on the data are fits to Hill equations, as described in the legend to Fig. 3. For activation of + channels by cGMP and cIMP, K1/2 = 4.3 µM and 61 µM and n = 1.5 and 1.5, respectively. For activation of +C5 channels by cGMP and cIMP, K1/2 = 0.41 µM and 55 µM and n = 1.4 and 1.4, respectively. For activation of +-M475E channels by cGMP and cIMP, K1/2 = 0.46 µM and 55 µM and n = 1.5 and 1.5, respectively. Plotted in D are box plots for the ratio of K1/2 values for cIMP to cGMP for these three channels. The vertical line in the middle of each box marks the median of the data. The box shows the middle half of the data, between the 25th and 75th percentiles. The "whiskers" show the data between the 5th and 95th percentiles.

For  and + channels, all three cyclic nucleotides activated the same maximal current. For  channels, the ratio of the current activated by saturating concentrations of cAMP to that activated by saturating concentrations of cGMP (I_{cAMP, sat}/I_{cGMP, sat}) was near one (Table 1). The ability of these ligands to activate the same maximal current in  channels arises from the energetically favorable opening transition of this channel (Gordon and Zagotta, 1995b). For + channels, I_{cAMP, sat}/I_{cGMP, sat} was also near one. That these ligands can activate the same maximal current for + channels as well suggests, but does not prove, that opening is energetically favorable for + channels also (Table 1). In theory, the increased time-dependence and rectification of + channels versus  channels could reflect a less favorable opening transition in + channels. However, currents from three heteromeric channels with different cyclic nucleotide selectivities (+ channels, +-M475E channels, and +-C5 channels) all displayed time-dependent rectification and I_{cAMP, sat}/I_{cGMP, sat} values near one at both 60 mV and +60 mV (data not shown). This suggests either the unlikely possibility that the opening transition is unfavorable and cyclic nucleotide-independent in these heteromeric channels, or more likely, that the opening transition is favorable and the time-dependent rectification comes from additional closed states apart from those leading to opening. Voltage-dependent occupancy of these additional closed states could produce a greater voltage dependence in maximal open probability, such that fully liganded heteromeric channels spend more time open at 60 mV than at 60 mV. Comparison of currents obtained by stepping patches held at 0 mV directly to 60 mV or 60 mV suggests that both this mechanism and open-channel rectification may contribute to increased steady-state rectification. Using this protocol, the instantaneous current at the start of the voltage step, which should reflect open-channel properties, was greater at 60 mV than at 60 mV (data not shown).

For +-CNBR channels, saturating concentrations of cAMP produced only ~75% of the current produced by saturating concentrations of cGMP. This smaller current produced by saturating cAMP almost certainly reflects a decreased ability of cAMP to induce opening once bound, suggesting that the opening of +-CNBR heteromultimers is less energetically favorable than the opening of either  channels or the other heteromeric channels. It is likely that the CNBR of the  subunit contains amino acid differences at positions other than 475 that affect the free energy of the opening transition.

Properties not localized to the CNBR

While the ligand-specific shift of the apparent affinities caused by the  subunit localized to the CNBR, other effects of the  subunit did not. Shallowing of the dose-response relation and desensitization were seen with all of the chimeric / subunits and with all three cyclic nucleotides. In addition, the gating of wild-type + channels was more voltage dependent than that of  channels. This is apparent both in the larger relaxation to steady state after a voltage step, and in the larger voltage dependence to the apparent affinities for cyclic nucleotide. Values of K_1/2 were ~60% greater at 60 mV than at 60 mV for + channels, but <10% greater for  channels (Table 1). This increased voltage dependence was seen with all of the chimeric / subunits and with all three cyclic nucleotides. Thus, the ligand-nonspecific effects of the  subunits, including shallowing of the dose-response relation, increased voltage dependence, rectification, and desensitization, seem to localize outside of the CNBR.

The data from + channels presented so far were obtained by co-injecting RNA in the oocytes at a ratio of 4:1 :. We wished to verify that the 4:1 co-injection ratio produces sufficient expressed  subunits to form a uniform population of heteromeric channels of their preferred stoichiometry and arrangement (Shapiro and Zagotta, 1998). Therefore, we systematically varied the : RNA injection ratio and examined the properties of the channels formed. Fig. 6 summarizes data from currents produced by injection of RNA for  and  subunits at ratios ranging from 2:1 to 100:1 (:). We focused on three parameters that distinguish  channels from + channels: the K_1/2 for cAMP, the Hill coefficient (n), and rectification. We found that the K_1/2 for cAMP and the Hill coefficient parameters were nearly the same for channels produced by injection ratios varying from 2:1 to 20:1 (Fig. 6, A and B). K_1/2 data for channels from a ratio of 100:1 were intermediate between those at a lower ratio and those of  channels. The rectification parameter was nearly the same at 2:1 or 4:1, somewhat higher at 20:1, and still higher at 100:1. In each case the effects of the  subunit appear to be saturating at an injection ratio of 4:1. Similar results were found for :-CNBR co-injections (data not shown). These experiments suggest that an injection ratio of 4:1 is more than sufficient to produce ample  subunits, and that the channels formed are not the result of a limiting supply of the  subunit. The saturation of the reduction in the Hill coefficient parameter at injection ratios up to 100:1 suggests that the reduced slope is an intrinsic feature of + channels, and not the result of a mixture of different channel populations, although we cannot completely exclude this possibility. The intermediate behavior of channels from the 100:1 ratio could be due to a mixed population of  channels and normal + channels, or the result of a stoichiometry or arrangement of subunits, distinct from that preferred, caused by the scarcity of  subunits. The difference in the ratio where each of the three parameters saturated probably reflects different sensitivities of the parameters to a mixed population of channels. We predict that functional heteromeric channels have a 1:1 : stoichiometry (Shapiro and Zagotta, 1998), and interpret these injection-ratio data as resulting from a greater efficiency by the oocyte in expressing  subunits, relative to  subunits.

View larger version (12K): [in this window] [in a new window]   FIGURE 6   Varying the ratio of injected subunits. Data from heteromultimers from co-injections of RNA for  and wild-type  subunits at different ratios, and from -homomultimers. Shown are box plots for the K1/2 values (A), Hill coefficients (B), and steady-state rectification of the current (C), for activation of channels by cAMP. The values were taken from fits of dose-response data to a Hill equation, as in Fig. 3, from each patch. Dose-response data at 60 mV and 60 mV were separately fit to Hill equations. The Imax values are the steady-state currents at saturating cAMP. The horizontal line in the middle of each box marks the median of the data. The box shows the middle half of the data, between the 25th and 75th percentiles. The "whiskers" extending from some of the boxes show the data between the 5th and 95th percentiles. The circles show the extreme points in the data.

Internal Mg²⁺ block

We wondered whether replacing the CNBR of the  subunit with that of the  subunit would have any effect on channel pharmacology. We focused on the voltage-dependent pore blocker Mg²⁺ (Colamartino et al., 1991; Root and MacKinnon, 1993; Zufall and Firestein, 1993; Kleene, 1993; Dryer and Henderson, 1993; Zimmerman and Baylor, 1992; Karpen et al., 1993), and characterized the differences in internal Mg²⁺ block between  channels and + channels. Superimposed in Fig. 7 A are current-voltage curves for  channels activated by a saturating concentration of cGMP in Mg²⁺-free solution or in the presence of various concentrations of free Mg²⁺. Mg²⁺ block is generally greater with increasing depolarization, as expected for a positively charged blocker acting from the inside. At lower concentrations, the current-voltage relation is biphasic, with best block at potentials near 0 mV, but relief of block at more positive potentials, as though Mg²⁺ were weakly permeant at positive potentials. Fig. 7 B shows similar current-voltage curves for a patch with + channels. Block here is also more pronounced at more positive potentials, but the voltage dependence of the block seems much weaker. Current-voltage curves for Mg²⁺ block of +-CNBR channels were similar to those for + channels (Fig. 7 C), showing again voltage-dependent block weaker than that for  channels. Quantifying the voltage dependence of Mg²⁺ block of homo and heteromultimers confirmed that they were dramatically different. Fig. 7 D plots the block of the three types of channels as a function of voltage at a fixed concentration of 2.81 mM Mg²⁺. The voltage dependence of block of  channels was very high, with a z value of 2.71e ± 0.15 (n = 4). In contrast, the voltage dependence of block of + channels was fairly modest, with a z value of 0.85e ± 0.14 (n = 7). The voltage dependence of block of +-CNBR channels was similarly modest, with a z value of 1.1e ± 0.1 (n = 3). The dose-response data for Mg²⁺ block of  channels, + channels, and +-CNBR channels at 60 mV were fit to Hill equations (Fig. 7 E), indicating very similar affinities at 60 mV for wild-type homo and heteromultimers, and a somewhat lower affinity for +-CNBR channels (Table 2). For all three channels, the slope of the Hill equation was near one, suggesting that one Mg²⁺ ion in the pore is sufficient to block the channel. Thus, the block of both types of heteromeric channels was very similar but the voltage dependencies of block of homo and heteromeric channels are strikingly different.

View larger version (28K): [in this window] [in a new window]   FIGURE 7   Block by internal Mg2+. Current-voltage curves for a patch with  channels (A), + channels (B), or +-CNBR channels (C) in the absence or presence of a range of internal Mg2+ concentrations. For all three panels, the Mg2+ concentrations are: (circles) 0, (squares) 120 µM, (triangles) 811 µM, (inverted triangles) 2.81 mM, (bows) 9.81 mM, (inverted bows) 29.8 mM. In (A), the data are the steady-state currents from a family of potentials from 80 mV to 80 mV at each Mg2+ concentration. Because of desensitization, the data in (B) and (C) were obtained by applying pseudo-ramps at each concentration, at 1/s. Each pseudo-ramp was a series of 30-ms steps progressing from 60 mV to 60 mV in 20-mV steps, and each measurement was made at the end of each step. Thus, any errors due to desensitization should be comparable to those using our usual pulse protocol (Fig. 1, inset). (D) Normalized current-voltage relations for  channels, + channels, and +-CNBR channels at a fixed Mg2+ concentration of 2.81 mM. Superimposed on the data are fitted Boltzmann relations of the form I = (Imax  Is)/(1 + exp[z(V  V1/2)/kT]) + Is, where V1/2 is the voltage at which the current is half-blocked, z is the valence of the blocker (2 for Mg2+),  is the fraction of the transmembrane electrical field sensed by the blocker, Is is the current remaining at very positive potentials, and Imax, R, and T have their usual meaning. For  channels, z = 2.8e and V1/2 = 56 mV; for + channels, z = 1.1e and V1/2 = 3 mV; for +-CNBR channels, z = 0.90e and V1/2 = 17 mV. (E) Normalized dose-response relations for Mg2+ block of the three types of channels at a fixed potential of 60 mV. Superimposed on the data are fitted Hill equation curves of the form I = ImaxK1/2n/(K1/2n + [Mg2+]n), where I is the steady-state current, K1/2 is the concentration of Mg2+ that produces half-block, and n is the Hill coefficient. We constrained Imax to be unity. For  channels, K1/2 = 425 µM and n = 1.06. For + channels, K1/2 = 469 µM and n = 0.88. For +-CNBR channels, K1/2 = 927 µM and n = 0.91.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

We find that incorporation of an olfactory  subunit (CNG5, CNC4, OCNC2) in a heteromeric channel with the olfactory  subunit produces a large ligand-specific shift in the apparent affinities of cAMP, cGMP, and cIMP for the channel. For all three ligands, the slopes of the dose-response curves were also more shallow. Our results with cAMP and cGMP are similar to those previously reported (Bradley et al., 1994; Liman and Buck, 1994). We also show that the presence of the  subunit shifts the apparent affinity for cIMP much like for cAMP. This was somewhat surprising, given that cIMP differs from cGMP only in lacking the amino group on the 2-position of the purine ring. We localized the altered ligand discrimination to a single amino acid, M475, in the putative C-helix of the CNBR. Replacement of M475 with glutamic acid (E) was fully sufficient to transform the ligand specificity of +-M475E heteromultimers to be like -homomultimers. This single point mutation increased the apparent affinity for cGMP (and to a lesser extent cIMP), and decreased the apparent affinity for cAMP. The residue at position 475, then, could account for the ability of the  subunit to promote activation of olfactory channels by cAMP, their physiological ligand.

Varnum et al. (1995) have shown that an important residue for ligand discrimination in rod  channels is the acidic residue D604 in the CNBR, which is the analogous residue to M475 in the olfactory  subunit. They showed that replacement of an aspartic acid with a methionine (D604M) decreased the efficacy of cGMP (and to a lesser extent cIMP) and increased the efficacy of cAMP for rod channels (Varnum et al., 1995; Sunderman and Zagotta, 1999a). Thus, alterations in this amino acid in the C-helix produce nearly identical effects in the rod  subunit and the olfactory  subunit. Based on their results, Varnum et al. (1995) proposed a molecular mechanism to explain the cGMP specificity of the rod channels.

A similar mechanism can explain how the  subunit confers greater cAMP efficacy to the olfactory channel. Fig. 8 shows a cartoon depicting + channels bound by either cGMP or cAMP. For simplicity, we show only one  subunit and one  subunit; the heteromeric channels are thought to have two of each type of subunit (Shapiro and Zagotta, 1998). Illustrated are two kinds of interactions between a ligand and the C-helix of the CNBR: a strong energetically favorable interaction (shown by a yellow star), and a weak or repulsive interaction. In the  subunit, the glutamic acid at position 593 (analogous to D604 in the rod  subunit) can form hydrogen bonds with the N1 and N2 hydrogens of cGMP, making the interaction strong and very energetically favorable. A methionine at position 475 in the  subunit, however, will produce a weak interaction with cGMP. In the case of cAMP, E593 in the  subunit will be electrostatically repelled by the unshared pair of electrons at N1 of cAMP. Neutralization of this acidic residue, as in the  subunit, increases the affinity for cAMP by eliminating electrostatic repulsion. Thus, when a + channel is bound by cGMP, the  subunits contribute a strong interaction and the  subunits a weak one, and overall the apparent affinity for cGMP is reduced. When the channel is bound by cAMP, the  subunits contribute a strong interaction and the  subunits a weak one, and overall the apparent affinity for cAMP is increased. The net result is to make a channel with rather similar affinities for cGMP and cAMP.

View larger version (45K): [in this window] [in a new window]   FIGURE 8   Schematic diagram showing how the  subunit alters ligand specificity in a heteromeric + channel. Only one  subunit and one  subunit per channel are shown for simplicity. The CNBR of the  subunits is shown in blue and of the  subunit in green. A cGMP (top) or cAMP (bottom) molecule, shown as its chemical structure, is depicted bound to the CNBR of each subunit. The purine ring of the cGMP or the cAMP is shown interacting with the C-helices (shown as cylinders). A yellow star connotes a strong interaction (see text). The red region in the amino termini of the  subunits is the autoexcitatory region of olfactory CNG channels thought to bind Ca2+/calmodulin (Liu et al., 1994) and to interact with the CNBR to facilitate activation (Varnum and Zagotta, 1997). The  subunit lacks this autoexcitatory domain.

All of our chimeric heteromultimers retained a number of the signature properties of wild-type + channels, indicating that these effects of the  subunit localize to a part of the channel outside of the CNBR (region depicted in black in Fig. 8). These properties include 1) a more shallow slope to the dose-response relations, 2) slow relaxation of the current to steady state after a voltage step, 3) greater rectification at saturating cyclic nucleotide concentrations, 4) greater voltage dependence to the apparent affinity for cyclic nucleotides, 5) desensitization in the presence of maintained agonist, and 6) less voltage dependence to internal Mg²⁺ block. Unlike the effects on the apparent affinity for cyclic nucleotides, these effects were ligand independent. Thus, these signature properties of this olfactory  subunit tell us that the chimeric subunits were expressed, and their persistence in chimeric heteromultimers indicates that they do not localize to the CNBR of the  subunit.

Although the rod  subunit and the olfactory  subunit studied here have a low sequence homology (30% identity), their effects on channel properties seem similar in numerous respects. Both increase the effectiveness of cAMP as an agonist (Fodor et al., 1998; Gordon et al., 1996), both make single-channel currents more "flickery," both increase the voltage dependence of gating and rectification, both weaken divalent block, and neither forms CNG channels by itself (Chen et al., 1993; Liman and Buck, 1994; Korschen et al., 1995). Compared to  channels, expres