Different Shaker family α-subunit genes generate distinct voltage-dependent K+ currents when expressed in heterologous expression systems. Thus it generally is believed that diverse neuronal K+ current phenotypes arise, in part, from differences in Shaker family gene expression among neurons. It is difficult to evaluate the extent to which differential Shaker family gene expression contributes to endogenous K+ current diversity, because the specific Shaker family gene or genes responsible for a given K+ current are still unknown for nearly all adult neurons. In this paper we explore the role of differential Shaker family gene expression in creating transient K+ current (I A) diversity in the 14-neuron pyloric network of the spiny lobster,Panulirus interruptus. We used two-electrode voltage clamp to characterize the somatic I A in each of the six different cell types of the pyloric network. The size, voltage-dependent properties, and kinetic properties of the somaticI A vary significantly among pyloric neurons such that the somatic I A is unique in each pyloric cell type. Comparing these currents with theI As obtained from oocytes injected withPanulirus shaker and shal cRNA (lobsterI shaker and lobsterI shal, respectively) reveals that the pyloric cell I As more closely resemble lobster I shal than lobsterI shaker. Using a novel, quantitative single-cell-reverse transcription-PCR method to count the number of shal transcripts in individual identified pyloric neurons, we found that the size of the somaticI A varies linearly with the number of endogenous shal transcripts. These data suggest that theshal gene contributes substantially to the peak somaticI A in all neurons of the pyloric network.
- transient potassium current
- Shaker family
- potassium channel
- gene regulation
- transcriptional control
- pyloric network
- identified neuron
- noncompetitive PCR
The components of an electrically excitable system, be it a heart or a cortical circuit, possess unique electrophysiological phenotypes that are required for the proper performance of that system. In many instances, differences in the amount and/or properties of the transient K+ current (I A) help to establish these essential cell-specific phenotypes (Connor, 1975; Cassell and McLachlan, 1986;Cassell et al., 1986; Premack et al., 1989; Serrano and Getting, 1989;Hamill et al., 1991; Furakawa et al., 1992; Tierney and Harris-Warrick, 1992; Liu et al., 1993; Banks et al., 1996; Massengill et al., 1997). The functional consequences of I A heterogeneity are evident in the pyloric central pattern generator.
The 14-neuron pyloric network, located in the stomatogastric ganglion of the spiny lobster, Panulirus interruptus, is a model system for neural circuits that generate rhythmic, cyclic movements like locomotion, respiration, and mastication (Selverston and Moulins, 1987; Harris-Warrick et al., 1992; Simmers et al., 1995; Marder and Calabrese, 1996). In these types of systems, muscles must contract in proper succession to perform a motor task correctly. The order and timing of muscle contraction depend on when the different pyloric network neurons fire bursts of action potentials. The burst phase of the various pyloric neurons is partially determined by the amount and specific properties of the I A present in each cell. For example, during an ongoing motor pattern the lateral pyloric (LP) and pyloric constrictor (PY) neurons are simultaneously released from synaptic inhibition and display postinhibitory rebound. The LP rebounds faster and fires first, partly because it has a smallerI A at any given physiological voltage (Hartline, 1979; Graubard and Hartline, 1991; Hartline and Graubard, 1992; Tierney and Harris-Warrick, 1992; Harris-Warrick et al., 1995a,b). Thus, cell-specific differences in the I A strongly influence the order and timing of neuronal firing and muscle contraction.
How is I A heterogeneity established in this system? Constitutive differences in post-translational modifications could generate cell-specific differences in theI A, because theI As in pyloric neurons can be differentially altered by the same neuromodulator. For instance, dopamine shifts the voltages of the somatic I As of half activation in the depolarizing direction in the LP and PY cells (Harris-Warrick et al., 1995a,b) and in the hyperpolarizing direction in the pyloric dilator (PD) cell (Levini et al., 1996; P. Kloppenburg, unpublished data). On the other hand, differential gene expression also might produce I A heterogeneity.
In arthropods, A-channel α-subunits are encoded by twoShaker family genes, shaker and shal(Salkoff et al., 1992; M. Kim et al., 1995, 1996; Tsunoda and Salkoff, 1995a,b; Baro et al., 1996a) (also see Results). A single multimeric A-channel contains either shaker or shal α-subunits, but never a combination of the two (Covarrubias et al., 1991; Li et al., 1992;Sheng et al., 1993; Wang et al., 1993; Deal et al., 1994; Lee et al., 1994; Shen et al., 1995; Xu et al., 1995). In addition to α-subunits, arthropod A-channels may contain β-subunits, γ-subunits, and/or other auxiliary proteins (Zhong and Wu, 1993; Chouinard et al., 1995;Jegla and Salkoff, 1997; Tejedor et al., 1997). For the purposes of this paper, we will define an A-channel by the type ofShaker family α-subunit it possesses. Because all pyloric neurons express both the shaker and shal genes (Baro et al., 1996b) (also see Results), we previously hypothesized that varying mixtures of shaker and shal channels carry the somaticI A in each cell type. Differences in the somaticI A between cell types could be obtained by varying the fraction of shaker versus shal A-channels.
Like most adult systems, the lobster pyloric network is genetically intractable, so it is difficult to judge the extent to which differences in Shaker family gene expression contribute toI A heterogeneity. Voltage-clamp studies presented in this paper indicate that the six different pyloricI As more closely resemble lobsterI shal than lobsterI shaker. To explicate this finding, we developed a quantitative, single-cell-reverse transcription PCR (SC-RT-PCR) method to count the number of shal transcripts in single, identified pyloric neurons. Using this method in conjunction with standard electrophysiological studies, we discovered a strictly linear relationship between shal transcript number and the size of the somatic I A in all pyloric neurons. After considering all of our data, we believe that our earlier hypothesis was incorrect. Large variations in the ratio of somatic shaker to shal channels are not responsible for somatic I Aheterogeneity in the pyloric network.
MATERIALS AND METHODS
Pyloric neurons. The protocol used to study pyloric cell I As using two-electrode voltage clamp has been described in detail by Harris-Warrick et al. (1995a,b). Briefly, a stomatogastric ganglion with the appropriate motor nerves and the associated commissural and esophageal ganglia was dissected from the animal (Selverston et al., 1976) and pinned in a dish. The preparation was perfused continually at 16°C with lobster saline containing (in mm): 479 NaCl, 12.8 KCl, 13.7 CaCl2, 3.9 Na2SO4, 10 MgSO4, 2 glucose, and 11.1 Tris, pH 7.35. Pyloric cells were identified electrophysiologically, using standard intracellular and extracellular recording techniques. I As were characterized with a two-electrode voltage clamp. The following drugs were present in the saline to isolate the I A and block synaptic transmission: 0.05 mm picrotoxin, 20 mm TEA, 10−7 m TTX, 5 mm Cs+, and 0.2 mmCd2+. Activation curves were generated by holding each cell at a potential at which the I A largely is inactivated and stepping to depolarized potentials to activate leak-subtracted non-I As. These non-I A records were digitally subtracted from current traces in which the depolarization was preceded by a 200 msec hyperpolarizing prestep to remove resting inactivation ofI A. The resulting subtracted current could be abolished by 4 mm 4-AP and represents pureI A. The inactivation data were generated by varying the amplitude of the prestep while stepping to a fixed, depolarized potential near full activation. In both cases the voltage-dependent peak currents were converted to conductance by usingE Rev = −86 mV (Eisen and Marder, 1982). The average E Rev was determined for each of the six pyloric cell types using tail current measurements of theI A. Tail currents were obtained by a series of hyperpolarizing steps after a 6 msec depolarization to +20 mV (preceded by a hyperpolarizing prepulse) to activate theI A. Non-I As were digitally subtracted, as previously described. We found that the average E Rev did not vary among the six pyloric cell types. Peak conductance was plotted versus the step potential for activation data or the prestep potential for inactivation data. The Boltzmann equation used for fitting was of the form: Equation 1where G max is the maximal conductance,V A is the voltage of half-maximal activation,s is the slope factor, and n = 3 for activation and n = 1 for inactivation. The inactivation kinetics were fit with two exponentials, using the least-squares minimization procedure of pClamp (Axon Instruments, Foster City, CA). The current as a function of time (t) corresponds to the equation: Equation 2where τf and τs represent the time constants of inactivation, and the amplitude of each time constant,I f andI s, represents the relative contribution of each component to the peak. The time constants of activation (τa) were estimated by fitting the entire waveform (as seen in Fig. 2) to Equation 2, using three exponentials, where τf, τs,I f, andI s were fixed to the values obtained previously from the inactivation fits to that waveform, and τa and I awere allowed to vary. All time constants were determined for a depolarizing step to +20 mV (PD, PY, LP, and VD) or +25 mV (AB and IC).
The average cellular input capacitance for each of the six pyloric cell types was determined as previously described by Serrano and Getting (1989).
Xenopus oocytes. Two-electrode voltage clamp was used to study the shaker-evoked I A 2–4 d after injecting an oocyte with shaker RNA [clone K17(I); M. Kim, D. Baro, C. Lanning, M. Doshi, J. Farnham, H. Moskowitz, J. Peck, B. Olivera, and R. Harris-Warrick, unpublished data]. Harvesting, injections, and maintenance of oocytes were as previously described (Baro et al., 1996a). Shaker currents (lobsterI shaker) were elicited by depolarizing steps from a holding potential of −70 mV. Protocols and equations for determining the voltage dependence and inactivation kinetics of lobster I shaker were as described in Baro (1996a), except that a minimum of three exponentials was required to fit the lobsterI shaker inactivation kinetics. A similar characterization of lobsterI shal appeared in Baro et al. (1996a).
Derivation of the correction factor for IAGmax
We have modeled the I A as the sum of a current passing through two A-channels that differ only in their rates of inactivation (Harris-Warrick, 1995a,b; Willms, 1997). The peak conductance, ḡ A, is given by: Equation 3where V is the voltage, E revis the reversal potential, p is a positive integer, f and s are the maximal conductances of the populations of fast and slowly inactivating channels, respectively, m is the activation variable, andh f andh s are the inactivation variables for the fast and slow channels, respectively. Thus, the peak conductance is determined by both the activation and inactivation variables.
Because of inactivation during the rising phase of the current, the peak conductance for an I A is always less than the true maximal conductance (Fig. 1). We will define the true maximal conductance as the conductance obtained when all of the A-channels are open, before any inactivation occurs. An estimate of the true maximal conductance (called the correctedG max) can be obtained by multiplying the measured G max by a correction factor that has been derived by Willms (1997). This correction factor (CF) represents the ratio of the true maximal conductance to the measured peak maximal conductance and is given by: Equation 4where: and are the fractions of the current that inactivate with the fast and slow time constants, and are the ratios of the inactivation time constants to the activation time constant, and the effective time ratio is given by:
When the relative number of A-channels in neurons with markedly different I A inactivation rates is assessed, it is more appropriate to use the correctedG max, rather than the measuredG max, because the correctedG max accounts for differences inI A inactivation kinetics, which the measuredG max does not. Simulated conductance traces based on our kinetic measurements of the PD and VDI As are displayed in Figure 1 along with the time courses for activation and inactivation. The PD peak conductance (Fig. 1 A) is much closer to the trueG max than the VD peak conductance (Fig.1 B), because the VD I A inactivates much more rapidly than the PD I A (Table 1; see Results). When multiplied by the correction factor, the peak conductances of both the PD and VD I As more closely approximate the true maximal conductance (Willms, 1997).
Pyloric neurons were identified electrophysiologically, the glial caps were removed, and single neurons were isolated physically and used in shal RT-PCRs, as previously described (Baro et al., 1996b), with the following modifications. The α-tubulin primers were excluded and an RNA standard was added to the RT master mix (see below). 32P end-labeled primers (Baro et al., 1996b) were added to the PCR master mix (105 cpm/90 μl of mix) and the [MgCl2] was 1.5 mm; the PCR cycle was 1× at 95°C for 5 min; 25× at 94°C for 1 min, → 68°C for 1 min, → 72°C for 30 sec; and 5–10× at 94°C for 1.5 min, → 68°C for 1 min, → 72°C for 30 sec + 10 sec extension/cycle. The completed SC-RT-PCRs were electrophoresed on a 10% polyacrylamide gel. The gel was dried, and the PCR products were imaged with a PhosphorImager (Molecular Dynamics, Sunnyvale, CA) and stored on a Dell Dimension XPS 450V computer. The digitized 32P signals were quantitated with ImageQuant software (version 3.3, Molecular Dynamics). The bands usually were positioned in the center of boxes (but see Results) for which the dimensions did not vary, and the relative amount of 32P within each box was calculated automatically using a volume integration procedure.
The RNA standard was made by deleting a 45 bp segment (nucleotides 1282–1326) from the shal cDNA clone K/S10 (Baro et al., 1996a), using a modified, nested deletion method (Henikoff, 1987) in which the deletion extended bidirectionally from a BspEI restriction enzyme site. The deleted shal clone (Δshal) was linearized with HindIII in a standard restriction digest (Sambrook et al., 1989). The linearized Δshal clone then served as a template in a transcription reaction using T3 RNA polymerase and a Ribomax kit (Promega, Madison, WI). The transcripts were DNased (Life Technologies, Gaithersburg, MD), a small amount of 32P-dCTP was added, and free nucleotides were removed with a Nuctrap column (Stratagene, La Jolla, CA). Fractions containing no radioactivity were phenol/CHCl3extracted immediately, ethanol precipitated, and resuspended in dH2O. The concentration of the RNA standard was determined with a spectrophotometer. The concentration of the RNA standard was ∼109-fold greater than the final concentration in a SC-RT-PCR. Cloned DNA and RNase contamination were detected by using small aliquots of the concentrated RNA standard as the template in a PCR or in an overnight incubation in 1× superscript buffer at 37°C, followed by denaturing gel electrophoresis. An RNA standard was used only if both DNA and RNase were absent and the RNA appeared as a discrete band of the appropriate size. The DNA- and RNase-free concentrated RNA standard was stored at −70°C in 5 μl aliquots in siliconized tubes for up to 1 year. One aliquot was used per experiment and then discarded. At the time of the experiment an aliquot of the RNA standard was diluted with dH2O, using siliconized tubes to prevent the RNA from sticking. Carrier RNA (MS2, Boehringer Mannheim, Indianapolis, IN) also was added during the dilution series (final MS2: RNA standard = 106, w/w). The diluted RNA standard was heated to 95°C for 5 min and quick-frozen on dry ice. The RNA standard was thawed, spun, and added to the RT master mix (which was stored immediately on ice) right before aliquotting the mix into the tubes containing the cells. Three different preparations of the Δshal RNA standard were used in the quantitative SC-RT-PCR experiments described in this paper. All three preparations gave the same results.
IA is unique in each pyloric cell type
The 14 neurons of the pyloric network fall into six identified cell types (Fig. 2). Each cell type possesses a unique, unambiguous electrophysiological phenotype (Hartline and Graubard, 1992). To determine the extent ofI A heterogeneity in this network, we characterized the I A in each cell type with two-electrode voltage clamp from the cell soma. Using this method,Hartline et al. (1993) demonstrated that the maximal amplitude, activation threshold, voltage dependence, and inactivation kinetics of the I A were the same in an intact pyloric neuron as in a ligated soma. Thus, the I As we measure from these intact neurons primarily reflect channels in the soma and initial length of the monopolar neurite, with little contribution from the current in unclamped distal neurites. We will refer to this current as the somatic I A.
Figure 2 demonstrates that the somatic I A in each cell type is unique under the same recording conditions. The upper panels show the somatic I As obtained by depolarizing pyloric cells to nearly the same membrane potential (+20 or +25 mV). These traces demonstrate that at a given membrane potential both the size and the inactivation kinetics ofI A vary significantly between cell types. The peak amplitudes at these voltages vary by up to sevenfold. TheI A inactivation was fit by the sum of two exponentials. The I As in the VD and AB cells inactivate much more rapidly relative to the other four cell types (Fig. 2, Table 1). This is attributable to two factors: (1) the time constants of inactivation (τfast and τslow) are up to 10-fold faster in these cells, and (2) a greater fraction of the channels inactivates with the fast, relative to the slow, time constant (Table1). The lower panels in Figure 2 display the voltage dependence of theI As. The activation and inactivation curves are shifted in different cell types, with the V 1/2s for activation and inactivation varying by up to 14 mV (Table 1). Consequently, the steady-state “window” I Ais active over a different voltage range in different cells (Fig. 2). Finally, the maximal conductance (G max), obtained from Boltzmann fits to the peak conductance/voltage relation, varies between cell types by a factor of eight (Table 1). All of these data indicate that the properties of the somaticI A are distinct in each cell type under the same recording conditions. Because synaptic input is blocked by Cd2+ and picrotoxin and neuromodulators are not present in the bath, intrinsic differences in the baseline currents must be responsible for the observed I Aheterogeneity.
IA density varies significantly among pyloric neurons
Cell-specific phenotypes can be brought about by changing the biophysical properties and/or the total amount of theI A in a given cell type. Table 1 demonstrates that pyloric neurons differentially regulate the properties of the somatic I A. Next, we set out to determine whether the somatic I A density also varies among cell types or whether the different current amplitudes seen in Figure 2merely reflect the different sizes of pyloric neurons. To obtain the somatic I A densities, we needed a measure of the size of both the soma and the maximal somatic I Afor each cell type. We estimated the average soma surface area for each cell type, using input capacitance as a gauge (Table2). The average input capacitance for each cell type indicates that the sizes of pyloric somata vary considerably. If somatic I A density is constant, then the maximum size of the somatic I A should be positively correlated with soma size. Conversely, if the six pyloric cell types differentially regulate somatic I Adensity, then the maximum size of the somatic I Ashould vary in a manner that is independent of soma size. TheG max, calculated from peak current measurements (Table 1), is used often as a measure of the size of theI A in a cell. If we normalize theG max for soma size (averageG max/average capacitance), somaticI A density varies by a factor of 6.9 (Table 2).
The G max values in Table 1 were derived from peak current measurements and thus underestimate the true maximum size of the I A in a cell, because not all of the channels are open during the peak current because of channel inactivation during the rising phase of the current (Fig. 1, Materials and Methods). This is not a problem when neurons are compared with similar rates of I A inactivation; however, if the I A in one cell inactivates much more rapidly than the others, as is the case with VD, the underestimate is disproportionately greater for that cell (Fig. 1). To compare more accurately the maximum size of the I A among cell types, we multiplied the measured G max by a correction factor (Willms, 1997) that represents the ratio of the maximal conductance before any inactivation occurs to the measured conductance at the peak current (see Materials and Methods). The resulting value, which we will term the correctedG max, is shown in Table 2. The effect of the correction factor can be seen in Figure3. In most cases the average correctedG max is not significantly different from the average measured G max. However, the average corrected G max for the rapidly inactivating VD cell is more than twice the average measuredG max.
Using the average corrected G max as the measure of the maximum size of the somatic I A in each cell type and normalizing for cell size (average correctedG max/average capacitance), we find that the corrected somatic I A density varies between cell types by up to a factor of 3.4 (Table 2). Therefore, with either the corrected or uncorrected G max, the size of the I A does not simply increase or decrease with pyloric cell size. This finding is consistent with the idea that unique electrophysiological phenotypes are established by varying both the properties and the density of A-channels in a cell.
Comparison of the pyloric cell IAs to lobster Ishal and lobsterIshaker
Neurons could alter the properties and the amount ofI A by differentially regulating A-channel gene expression. Like their Drosophila homologs, thePanulirus shaker and shal genes both encode α-subunits for rapidly inactivating A-type channels, although with somewhat different properties than for the Drosophilachannels (Fig. 2, Table 1; M. Kim et al., 1995, 1996; Baro et al., 1996a). We compared the I As obtained from overexpressing shaker and shal cRNA inXenopus oocytes (lobsterI shaker and lobsterI shal) with the six pyloricI As (Fig. 2, Table 1). We discovered that the variations in pyloric I As were not consistent with the idea that distinct pyloric I As result from different mixtures of shaker and shal A-channels. Instead, we found that the pyloric cell I As qualitatively resemble lobster I shal more than lobster I shaker; however, no pyloricI A was identical in all parameters to lobsterI shal.
The voltage dependence of the six pyloric I As was quite variable but generally resembled lobsterI shal more than lobsterI shaker. The voltages of half activation (V 1/2act) for pyloric cellI As range from −33 to −45 mV. The lobsterI shal V 1/2actis approximately in the middle of this range (−40 mV), whereas the lobster I shaker V 1/2act lies below the lower limit of this range (−46 mV). The slopes of the activation curves are similar for allI As except the LP. The pyloricI A voltages of half inactivation (V 1/2 inact) range from −71 to −57 mV. The lobster I shal V 1/2inact (−71 mV) is identical to the VDI A and marks the lower bound of the range. In contrast, the lobster I shaker V 1/2inact (−44 mV) is significantly more depolarized, and the slope of the inactivation curve is significantly steeper than any of the six pyloric I As. The pyloric I A voltages of half activation and inactivation are not identical to either lobsterI shal or lobsterI shaker, nor do they vary in a manner that would suggest the pyloric I A is a mixture of lobster I shaker and lobster I shal. For example, theV 1/2act of the VD I Acurrent is more similar to lobsterI shaker, whereas itsV 1/2inact is identical to lobsterI shal.
The inactivation kinetics for all six pyloricI As are also more similar to lobsterI shal than lobsterI shaker or a mixture of the two channel types. First, lobster I shalwas fit with a double exponential relation, like all six pyloricI As, whereas lobsterI shaker could be fit only with a third-order equation. Second, lobsterI shaker contains a large noninactivating component that is not present in the six pyloricI As or lobsterI shal (Fig. 2, Table 1). The fast time constants of inactivation (τfast) for the PD, PY, LP, and IC I As are very similar to each other and to lobster I shal, but they are significantly slower than lobsterI shaker. The slow time constants (τslow) of these pyloric neurons are approximately two times faster than lobster I shal, but 5–17 times faster than lobsterI shaker (Table 1). The time constants of inactivation for the AB and VD I As are significantly different from both lobsterI shal and lobsterI shaker (Fig. 2, Table 1).
A comparison of the eight different I As shown in Figure 2 and Table 1 is not sufficient to ascertain which A-channels carry the pyloric I As. However, the overall similarity of the neuronal I As to lobsterI shal suggested that shalmay be an important contributor to the pyloric cellI As. Therefore, we developed a method to quantitate shal gene expression in single identified neurons, using noncompetitive RT-PCR (Ferre, 1992; Foley et al., 1993;Gause and Adamovicz, 1994; Sucher and Deitcher, 1995).
Quantitating shal gene expression in single identified neurons
In our method, RNA from a single cell is reverse-transcribed and amplified along with 103 Δshal RNA standard molecules in an RT-PCR containing 32P-labeledPanulirus shal-specific primers. The Δshal RNA standard is identical to the endogenous shal transcript, except that it lacks the distal-most portions of the 5′ and 3′ untranslated regions and it contains a very small deletion in the region between the two PCR primers. This minor deletion allows the separation of the cellular shal and the standard Δshal RT-PCR products on the basis of size. The number of cellular transcripts is determined by normalizing the cellularshal RT-PCR product against the standard ΔshalRT-PCR product.
The results of a typical experiment are shown in Figure4. Neurons were identified electrophysiologically. Glial caps were removed because theshal gene is expressed in glial cells (Baro et al., 1996b), and individual neurons were physically isolated and used in RT-PCRs containing 103 Δshal RNA standard molecules. The RT-PCR products were size-separated, using polyacrylamide gel electrophoresis, and phosphorimaged. The upper band in each lane represents the product of the endogenous shaltranscripts present in a single cell. The lower band represents the product of the 1000 Δshal RNA standard molecules. The number of shal transcripts in each cell was calculated from: Equation 5where X is the relative amplification efficiency per cycle of a Δshal to a shal DNA template, andn is the number of cycles in the PCR.
Shorter DNA molecules often are amplified more efficiently than longer molecules in a PCR. To determine whether the 262 bp ΔshalPCR product was amplified more efficiently than the 307 bpshal PCR product, we added equal numbers of shaland Δshal DNA templates to the same PCR (Fig.5). The PCR products were electrophoresed and phosphorimaged, and the digitized 32P signals were quantitated as described in Materials and Methods. The amplification efficiency per cycle of a Δshal relative to ashal DNA template was determined from the following equation: X = (cpm Δshal/cpmshal)1/n, where Xand n are described above. We found that ΔshalDNA molecules are amplified on average 1.029 ± 0.002 (n = 103) times more efficiently than an equivalent number of shal DNA molecules per PCR cycle. So, for a 30-cycle PCR, X n = (1.029)30 = 2.4.
To ensure that our measurements of the relative amplification efficiency were accurate, we used several different DNA template preparations, and we varied the number of starting molecules and PCR cycles within the linear range of amplification (see below); otherwise, the conditions of the PCR were identical to the quantitative SC-RT-PCR. In those experiments with a large number of template molecules and PCR cycles, the shal and Δshal products tended to bleed together along the edges of the lane (Fig. 5 C). Because the bands “smile” (Figs. 4, 5, 6), a smeared/streaked signal along the edge of the lane belongs to the band just below the smear/streak. Thus, in the few cases in which bleeding occurred, the phosphorimager measuring boxes (see Materials and Methods) were positioned so that the smear/streak between the bands went with the lower band. The average amplification efficiency of Δshalrelative to shal did not vary significantly with the number of starting molecules or PCR cycles.
RNase is the bane of the quantitative SC-RT-PCR experiments. If an RNase is introduced when the cellular transcripts and the RNA standard are both present, they should be degraded equally, and the ratio of the signals will not change, just their intensity. However, if an RNase acts preferentially on either the endogenous transcript or the standard, there will be errors in our measurement. To detect and control for trace RNase contamination, we carried at least two blanks per experiment (RT-PCRs containing 1000 Δshal RNA standard molecules but no cell; Fig. 4). We used the data from an experiment only if the counts per minute in the standard bands of the blanks varied by less than a factor of 2. We used the data from an individual cell within an experiment only if the counts per minute in the standard band of that SC-RT-PCR were within or above the range of the blanks. For example, in Figure 4 the starred PD cell failed this criterion, so the data from this cell were not used.
Demonstrating that input is proportional to output in our SC-RT-PCRs
For our SC-RT-PCR method to be quantitative, we have to demonstrate that input is proportional to output. In a typical PCR the product increases exponentially with cycle number until eventually a plateau is reached. The PCR product is proportional to the number of starting molecules only if the PCR remains within the exponential phase (for review, see Ferre, 1992; Foley et al., 1993; Gause and Adamovicz, 1994). Several factors determine when the plateau is reached, including the number of starting molecules: everything else being equal, the larger the number of starting molecules, the sooner the PCR enters the plateau phase. For a given cycle number the linear range of amplification is defined as the range of starting template molecules over which the PCR remains within the exponential phase (for review, see Ferre, 1992). We determined the linear range of amplification for a 35 cycle RT-PCR under our quantitative SC-RT-PCR conditions (Fig.6). RT-PCRs containing 50–50,000 Δshal RNA molecules were performed for 35 cycles, and the amount of 32P incorporated into the ΔshalRT-PCR product was quantitated (Fig. 6 A). Figure6 B shows the relationship between the number of starting molecules and the amount of product. Each data point represents the average of nine different RT-PCR experiments. As Figure6 B demonstrates, the log of the product increases linearly as a function of the log of the starting template over the range from 50 to at least 10,000 Δshal RNA molecules. The data point at 50,000 molecules is slightly below the line. This suggests that an RT-PCR containing 50,000 Δshal starting molecules enters the beginning stages of the plateau phase by 35 cycles and input may no longer be proportional to output. However, when the RT-PCR contains fewer starting molecules, and in particular <104, input is still proportional to output after 35 cycles. Thus, the linear range of amplification for a 35 cycle RT-PCR under the present SC-RT-PCR conditions includes at least 50–10,000 Δshal RNA template molecules. Preliminary experiments indicated that the number of endogenous shaltranscripts in a pyloric neuron never exceeded 4000. Because we add 1000 Δshal RNA standard molecules to a SC-RT-PCR, each reaction has between 1000 and 5000 starting molecules, which is well within the linear range of amplification for a 35 cycle SC-RT-PCR (Fig.6 B). In some experiments we reduced the SC-RT-PCR cycle number to 30, and this did not change our results. This is what we would predict, because the upper limit of the linear range of amplification increases with decreasing cycle number. We should point out that the level of nonspecific RNA does not change significantly when a cell is added to the RT-PCR, because we include 20 ng of carrier RNA in each RT-PCR and a neuron most likely contributes <100 pg of nonspecific RNA to a reaction. Thus, adding a cell to the RT-PCR will not affect the linear range of amplification [see Gause and Adamovicz (1994) for a discussion of this point].
shal transcript number varies significantly among cell types
We performed a number of SC-RT-PCR experiments to determine the average number of shal transcripts in each pyloric cell type. The mean number of shal transcripts is plotted for each cell type in Figure 7. There are several points to be made. First, all pyloric cells expressshal. Second, the number of shal transcripts within a given cell type was consistent between individuals. Third, we observed significant differences in the average number ofshal transcripts among cell types, with shaltranscript levels varying by a factor of 2.8. Fourth, there is no positive correlation between the average number of shaltranscripts and the average input capacitance for a given cell type, as seen by our calculations of the shal transcript density, which varied from cell type to cell type (Table 2). Thus, pyloric cells differentially regulate shal gene expression at the level of the transcript. Pyloric neurons may differentially modulate transcript levels by varying rates of transcription, transcript processing, and/or transcript turnover. The fact that transcript levels are regulated does not exclude additional translational and post-translational regulation of shal gene expression in pyloric neurons as well.
The maximum size of the somatic IA varies as a linear function of shal gene expression
If shal underlies a major fraction of the somaticI A in pyloric neurons, then it might be possible to correlate the maximum size of the somatic I Awith the number of shal transcripts in a given pyloric cell type. Plotting the mean number of shal transcripts in each cell type versus the average measured G maxreveals a remarkably strong positive correlation (Fig.8 A). A linear regression fit to these data has an R 2value of 0.95, demonstrating that the maximum size of the somaticI A in each cell type varies as a linear function of shal transcript levels (p < 0.001). The VD data point is significantly below the line in Figure8 A. We suggest this is attributable to an underestimate of the VD G max calculated from peak current measurements because of the more rapid inactivation of the VD I A relative to other pyloric neurons (see Fig. 1). As described above, we can compensate for this underestimate by plotting the corrected G max versusshal transcript number for each cell type (Fig.8 B). In this case the VD more closely approximates the line so that R 2 becomes 0.98 andp < 0.0002.
The simplest interpretation of our data is that shal is an α-subunit for the majority of somatic A-channels in all 14 neurons of the pyloric network. Research on flies and mammals has shown that K+ channel α-subunits from the shaker and shal subfamilies cannot coassemble to form a heteromeric channel, and one never finds an A-channel composed of shaker and shal α-subunits (Covarrubias et al., 1991; Li et al., 1992; Sheng et al., 1993; Wang et al., 1993; Deal et al., 1994; Lee et al., 1994; Shen et al., 1995; Xu et al., 1995) (but see Shahidullah et al., 1996). Subfamily-specific assembly is mediated via the NAB domain in the N-terminal regions of K+ channel subunits (Xu et al., 1995). NAB domains are conserved in a subfamily-specific manner. The amino acid identity among different NAB regions within a subfamily is generally >70%, but between subfamilies NAB identity drops to ∼30% (Xu et al., 1995). Because the NAB domains of Panulirus shaker and shal are 94 and 97% identical to their Drosophila homologs, respectively, we believe that Panulirus shaker and shal α-subunits do not form heterotetramers. Thus, if we consider onlyShaker family α-subunit genes for the moment, three possibilities exist: (1) the somatic I As are carried by shaker channels alone; (2) the somaticI As are carried by two different populations of A-channels, one containing shal α-subunits and the other containing shaker α-subunits; (3) the somatic I As are carried by shal channels alone. Because the size of the somaticI A varies as a linear function ofshal transcript number with p < 0.001, and pyloric somatic I As qualitatively resemble lobster I shal but not lobsterI shaker, we can rule out the first possibility. With regard to the second possibility, the extremely highR 2 value for theshal–I A correlation (Fig. 8) suggests that any significant contribution to the somaticI A from the shaker gene must either (1) remain fairly constant among cell types or (2) vary among cell types in a manner that is essentially identical to shal. If, on the one hand, the shaker gene produced a significant, constant number of somatic A-channels in every cell type, then there should be a sizable I A even when shaltranscripts are absent. In other words, when x is zero in Figure 8, the y-intercept should be positive. Because the extrapolated y-intercept in Figure 8 is negative, we can discard this possibility. If, on the other hand, the ratio of somatic shaker to shal channels is constant among the six different cell types, then shaker and shal gene expression must be completely coregulated in these six different cell types. However, strict coregulation of shaker and shal A-channel gene expression has not been described in previous studies in other systems (Roberds and Tamkun, 1991; Kues and Wunder, 1992; Lesage et al., 1992; Sheng et al., 1992; Tsaur et al., 1992; Dixon and McKinnon, 1994, 1996; Maletic-Savatic et al., 1995; Brahmajothi et al., 1996;Serôdio et al., 1996) (for review, see Chandy and Gutman, 1995). Because we have not yet quantified shaker expression in pyloric neurons, we cannot reject the possibility of coregulation categorically. Nevertheless, because the pyloric somaticI As resemble lobsterI shal more than lobsterI shaker, we suggest that the third possibility is the simplest and most likely: in pyloric neurons theshal gene encodes most or all of the Shakerfamily α-subunits for somatic A-channels. This point eventually could be confirmed by demonstrating a causal relationship betweenshal and I A, usingshaker and shal knock-out techniques that use expression of antisense oligonucleotides (Chung et al., 1995) or dominant-negative mutations (Ribera, 1996).
The previous argument involved Shaker family α-subunits only. This argument did not consider the formation of heterotetramers between Shaker family α-subunits and other proteins.Drosophila mutant analysis indicates that Shakerfamily proteins might form heterotetramers with non-Shakerfamily K+ channel proteins such as EAG (Warmke et al., 1991; Zhong et al., 1991, 1993; Warmke and Ganetzky, 1994), and shaker and EAG have been shown to form heterotetramers in an oocyte expression system (Chen et al., 1996). Similarly, heterotetramers can form between shal α and γ subunits (Jegla and Salkoff, 1997). Our data do not rule out the possibility that some fraction, or even all, of the somatic A-channels are heterotetramers between α-subunits and EAG, γ-subunits, or other as yet unidentified subunits.
I A diversity in the 14-neuron pyloric network of the spiny lobster, Panulirus interruptus, is established by varying the density and/or the properties of A-channels between cells. Arthropod A-channels are multimeric proteins containingShaker family α-subunits encoded by either theshaker or shal gene in addition to other subunits (Salkoff et al., 1992; M. Kim et al., 1995, 1996; Tsunoda and Salkoff, 1995a,b; Baro et al., 1996a) (M. Kim, D. Baro, C. Lanning, M. Doshi, J. Farnham, H. Moskowitz, J. Peck, B. Olivera, and R. Harris-Warrick, unpublished data). Our voltage-clamp studies demonstrated that the pyloric I As are more similar to lobsterI shal than lobsterI shaker. Quantitative SC-RT-PCR experiments show that there is an exceptionally strong linear correlation between the magnitude of the somaticI A and the number of shal transcripts in pyloric neurons. The two most likely interpretations of these data are either that shal channels underlie the major component of the somatic I A or that an unchanging ratio of shaker to shal channels underlies the somatic I A in all pyloric neurons.
Somatic IA heterogeneity in the pyloric network
Our results suggest that although lobster shal α-subunits or a constant ratio of shaker to shal α-subunits underlies the somaticI A in pyloric neurons, theI As vary dramatically in their biophysical properties, depending on which cell is expressing the gene or genes. For example, the LP cell has more shal transcripts than the PY cell and more somatic A-channels (as determined from the average corrected or uncorrected G max; Fig. 3). However, the somatic I A is smaller in the LP cell at all submaximal activating voltages. This is because the LP activation curve is shifted in the depolarizing direction relative to the PY so that at any physiological membrane potential fewer A-channels will be open in the LP cell (Fig. 2, Table 1).
Qualitative and quantitative differences in Shaker family α-subunit gene expression do not seem to underlie the biophysical differences in the LP and PY I As. Stable cell-specific variation in the post-translational modifications of A-channels could be partially responsible for I Adiversity in the pyloric network (Levitan, 1994; Holmes et al., 1996;Jonas and Kaczmarek, 1996; Villarroel and Schwarz, 1996; Harris-Warrick et al., 1997; Keros and McBain, 1997). Because a single lobster shal α-subunit has at least 31 putative protein kinase sites (Baro et al., 1996a) and the same neuromodulator can alter pyloric somaticI As differentially (Harris-Warrick et al., 1995a,b; Levini et al., 1996; P. Kloppenburg, unpublished data), it is possible that cell-specific differences in constitutive cycles of α-subunit phosphorylation–dephosphorylation reproducibly alter pyloric cell I As.
Other mechanisms also could contribute to pyloricI A diversity. For instance, it has been shown recently that the Shaker gene family encodes both classical α-subunits and regulatory γ-subunits. Classical α-subunits like shaker (Kv1), shab (Kv2), shaw (Kv3), shal (Kv4), and Kv5 are membrane-spanning subunits that can form functional homotetramers and contain a subset of 44 amino acids that are totally conserved across subfamilies and species (Chandy and Gutman, 1995). Although the definition of γ-subunits is still evolving, it seems that γ-subunits like Kv6, Kv7, Kv8, and shalγ1 are regulatory membrane-spanning subunits that cannot form functional homotetramers and contain alterations in a few of the 44 amino acids that are totally conserved in all classical α-subunits (Drewe et al., 1992; Hugnot, 1996; Jegla and Salkoff, 1997). γ-Subunits can form heteromultimers with classical α-subunits and change the properties of the K+ channel (Hugnot et al., 1996; Jegla and Salkoff, 1997). Jegla and Salkoff (1997) have shown recently that shal α and γ subunits interact in a constant stoichiometry. Thus, the number of α-subunit transcripts might help to determine the number of A-channels in a pyloric cell, whereas different γ-subunits in different cell types could alter the biophysical properties of those A-channels. Cell-specific alternate splicing of the α-subunit transcripts (Mottes and Iverson, 1995; Rogero and Tejedor, 1995) and/or β-subunit or other protein interactions with the α-subunits could alter I As, although to date no shal β-subunits have been reported (Rudy et al., 1988; Chabala et al., 1993; Rettig et al., 1994; Serôdio et al., 1994; Chouinard et al., 1995; England et al., 1995; E. Kim et al., 1995; Majumder et al., 1995; McCormack et al., 1995; Morales et al., 1995; Rhodes et al., 1995, 1996; Cohen et al., 1996; Nakahira et al., 1996; Sewing et al., 1996; Shi et al., 1996; Yu et al., 1996; Tejedor et al., 1997). Finally, the same α-subunits could produce different I As if the composition of the cellular membranes were distinct in different cells (Coronado et al., 1984; Barrantes, 1993; Bhushan and McNamee, 1993;Chang et al., 1995).
The VD IA
The VD I A is significantly different from the other pyloric neurons: it is much smaller, more rapidly inactivating, and both the activation and inactivation curves are more hyperpolarized (Fig. 2, Tables 1, 2). With uncorrectedG max values, the VD data point does not lie on or near the line relating I A G max to shal transcript number, as do the data points from all the other pyloric neurons (Fig.8 A). There are at least three possible interpretations of these data, and they are not mutually exclusive. The first interpretation is that shaker contributes more substantially to the VD I A than to other pyloricI As. The second interpretation is that most VD somatic A-channels contain shal α-subunits, but a greater fraction of VD shal channels are “silenced” under control conditions relative to other pyloric cells. The third interpretation is that the VD falls off the line in Figure 8 A because, as shown in Figure1, the more rapid inactivation of the VD I Aproduces a significant underestimate of the true VDG max relative to the other pyloric cell types. To compensate for this error, we have used the correctedG max as a measure of I Amagnitude. As seen in Figure 8 B, plotting the corrected G max versus transcript number causes the VD to lie much closer to the line, suggesting that shaldoes underlie I A in the VD cell as well. Unfortunately, because of the more rapid inactivation, the correction procedure will be the most inaccurate for the VD cell (Willms, 1997). Thus this issue is not resolved and we cannot state conclusively which, if any, of these three possibilities is correct.
If shal underlies the somaticIA, where are the shakerA-channels?
We have shown with nonquantitative methods that shakeris expressed in all pyloric neurons and that shaker produces A-channels (Fig. 2; Baro et al., 1996b), yet in our more favored interpretation of the data, shal accounts for the majority of somatic A-channel α-subunits. One possible explanation for this apparent discrepancy is that most shaker A-channels may be localized to axons and nonclamped regions of the neuropil. In Drosophila, shaker channels are more highly concentrated in the axons and neuropil of adult brains and are not the major component of somaticI As in most neurons (Solc et al., 1987; Baker and Salkoff, 1990; Schwarz et al., 1990; Tsunoda and Salkoff, 1995a,b). Similarly, in the mammalian brain, shal channels (Kv4.2) are concentrated in the somatodendritic compartment, whereas shaker A-channels (Kv1.4) are localized to axons and terminals (Sheng et al., 1992; Maletic-Savatic et al., 1995; Veh et al., 1995).
Electrophysiological analyses of Drosophila mutants indicate that shal encodes the somatic I A in most embryonic neurons (Tsunoda and Salkoff, 1995a,b). Moreover, hybrid arrest studies on rat brain mRNA expressed in Xenopusoocytes have shown that shal (Kv4.2), but notshaker (Kv1.4), mRNA underlies the somatic transient K+ current (I SA) recorded from rat thalamic and cerebellar neurons (Serôdio et al., 1994). Finally, Dixon and McKinnon (1996) suggest that members of the shal (Kv4) subfamily are likely to underlie the low-threshold somatic I A in sympathetic neurons, whereas shaker A-channels (Kv1.4) do not make a significant contribution to this current.
Note that our proposal does not imply that shal channels are localized only to the soma and therefore are excluded from the distal regions of the neuropil. Shal channels may be distributed over the entire surface of a pyloric neuron, in which case they certainly can affect neuronal activity and firing patterns. We have shown previously that dopamine modulation of the measured somatic I A can explain quantitatively the alterations in the postinhibitory rebound characteristics of the LP and PY neurons (Harris-Warrick et al., 1995a,b). Thus, either the somatic I Acontributes to neuronal firing patterns, and/or the A-channels present in the soma are also likely to be found in the neuropil. Interestingly, recent work from Drosophila suggests that shal channels are present in the membrane of Type III synaptic boutons (Martı́nez-Padrón and Ferrús, 1997).
shal transcripts and the somaticIA
It was somewhat surprising to find such a simple relationship between shal transcript number and the I A G max. Any number of phenomena could have masked this relationship. If there were large cell-specific differences in the translational regulation of shal such that some cells produced 103 functional proteins per transcript whereas others produced one, then the size of I Awould not correlate with the number of shal transcripts. Similarly, if two cells had the same number of shaltranscripts and proteins, but the first cell concentrated all of its shal channels in the soma whereas the second localized them to the unclamped regions of the neuropil, we would detect anI A in the first cell, noI A in the second cell, and no positive correlation between transcript number and I Amagnitude. By the same token, if the average conductance of somatic A-channels varied significantly among pyloric neurons, then theR 2 values in Figure 8 would be reduced dramatically even if shal encoded the α-subunits of the somatic I A. The fact that we can demonstrate an unequivocal linear relationship between shal transcript number and the size of the somatic I A suggests that translational regulation, shal channel localization (the fraction of shal channels in voltage-clamped vs nonvoltage-clamped regions), and average somatic A-channel conductance do not vary substantially among the 14 neurons of the pyloric network under control conditions.
The fact that the x-intercept in Figure 8 is not zero suggests that there is a sizable pool of unused shal RNA in these cells. Some portions of the transcripts we measure almost certainly are processed incompletely or incorrectly (Kramer, 1996) or partially degraded (Jacobson and Peltz, 1996) and therefore would yield no functional protein. In addition, shal gene expression could be regulated post-transcriptionally so that some of the transcripts are being stored rather than actively translated (Curtis et al., 1995; Decker and Parker, 1995; Hentze, 1995; Jansen et al., 1995;Wymore et al., 1996).
The most parsimonious explanation of the linear relationship between shal gene expression and I A G max is that shal encodes theShaker family α-subunits for somatic A-channels in most or all of the pyloric neurons. If shaker plays a significant role, its expression must be coregulated exactly with shal. In either case, it does not appear that pyloricI A heterogeneity is attributable to varying ratios of shaker to shal channels in different cell types. Stable differences in Shaker family α-subunit gene expression correlate with the variations in somatic A-channel density among cell types; however, differences in Shaker family α-subunit gene expression do not underlie the differences in the biophysical properties of the six pyloric I As. Other phenomena, including post-translational modifications and auxiliary subunits, must be responsible for the variety of differentI As seen in pyloric neurons.
This work was supported by the Human Frontier Science program; National Institutes of Health Grants NS25915, NS35631, and NS17323; Office of Naval Research Grant N00014-95-1-0292 to R.H.-W.; and a Hughes undergraduate fellowship to H.E.R. We thank our anonymous reviewers, Ole Kiehn, Scott Hooper, Thomas Podleski, Jack Peck, Jenifer Levini, Amir Ayali, Ronald Hoy, Bruce Johnson, and David Deitcher for helpful comments on this manuscript.
Correspondence should be addressed to Dr. Deborah J. Baro, Section for Neurobiology and Behavior, Cornell University, Seeley G. Mudd Hall, Ithaca, NY 14853.