| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
MOLECULAR AND GENOMIC |
Departments of
1 Anaesthesiology
3 Pharmacology
2 Center for Neurobiology and Behavior, Columbia University, New York, NY 10032, USA
 |
Abstract |
|---|
 Top Abstract IntroductionMethodsResultsDiscussionReferences |
|---|
(Received 14 May 2007;
accepted after revision 13 June 2007;
first published online 14 June 2007)
Corresponding author G. R. Tibbs: Department of Anesthesiology, Eye Institute Research Annex, EI3-305, 160 Fort Washington Avenue, New York, NY 10032, USA. Email: grt1{at}columbia.edu
|
Introduction |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
IH is encoded by the four member hyperpolarization-activated, cyclic nucleotide-regulated gene family (HCN1–4) with a single channel being composed of a homomeric or heteromeric assembly of four HCN subunits (Robinson & Siegelbaum, 2003). Intriguingly, the actions of both halothane (Sirois et al. 1998; Chen et al. 2005b) and propofol (Cacheaux et al. 2005; Chen et al. 2005a) are dependent on the subunit composition of the channel. Thus, propofol slows and hyperpolarizes activation of HCN1 channels but it has only weak or largely absent effects on HCN2 and HCN4 (Cacheaux et al. 2005; Chen et al. 2005a; Ying et al. 2006 and herein) whereas halothane hyperpolarizes HCN1 but suppresses that maximal current carried by HCN2 channels (Sirois et al. 1998; Chen et al. 2005b).
The voltage dependence of IH activation is regulated by cAMP (Robinson & Siegelbaum, 2003), internal protons (H+i) (Munsch & Pape, 1999) and several signalling lipids (Pian et al. 2006; Zolles et al. 2006; Fogle et al. 2007). Binding of cyclic nucleotides to a C-terminal domain (the CNBD; see online supplemental material Fig. S1) depolarizes gating of HCN1, 2 and 4 channels albeit with greater efficacy in HCN2 and 4 (Robinson & Siegelbaum, 2003). Introduction of a repulsive interaction between the nucleotide's cyclized phosphate and the CNBD (by mutation of a conserved arginine to a glutamate; RE) selectively lowers the affinity for ligand from nanomolar to high millimolar rendering the channels insensitive to endogenous levels of cAMP (Chen et al. 2001; Ulens & Siegelbaum, 2003; Craven & Zagotta, 2005). Similarly, mutation of a conserved histidine at the cytoplasmic end of S4 (Zong et al. 2001) to either an arginine (HR) or glutamine (HQ) has been reported to eliminate sensitivity of HCN2 channels to changes in the [H+]i (Zong et al. 2001). The molecular bases by which lipid messengers alter channel function have not been established. Interestingly, in the case of halothane, HCN isoform selectivity is dependent on the activation status of the cAMP gating ring such that the responses of HCN1 and HCN2 channels are essentially identical when cAMP levels are high or the inhibitory effects of the gating ring are eliminated by deletion (Chen et al. 2005b).
By analysing the behaviour of wild-type and REHR or REHQ HCN1, 2 and 4 channels and that of a truncated HCN1 channels (wherein the divergent amino terminus and the post-S6 carboxy terminus are deleted) within the framework of concerted (Altomare et al. 2001; Wang et al. 2002; Ulens & Siegelbaum, 2003; Craven & Zagotta, 2005; Mannikko et al. 2005; Elinder et al. 2006) versus sequential (Shin et al. 2004; Craven & Zagotta, 2005; Chen et al. 2007) models (wherein opening is voltage dependent or independent, respectively) we show that: (1) HCN1 gating is better described when activation and opening reactions are separated, but this does not involve a measurable occupancy of a deactivated open state at depolarized potentials; (2) propofol interacts preferentially with closed–resting and closed–activated states of the conserved membrane-embedded HCN channel core, (3) weak binding to the activated–open state likely contributes to the response in intact cells under physiological conditions, and (4) unlike halothane, the selective preference of propofol for HCN1 channels is not solely a function of differences in channel opening energetics.
These findings suggest determining the underlying causes for propofol sensitivity and insensitivity in different HCN isoforms will yield insights into the basic operation of HCN channels and suggest ways in which the drug can be modified to enhance or alter its action.
|
Methods |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Murine HCN channels and mutants thereof (made and verified as previously described; Fogle et al. 2007) were subcloned into pGH19 (HCN1 and HCN4) or pGHE (HCN2) expression vectors. One to fifty nanograms of cRNA transcribed from NheI (HCN1 and HCN4) or SphI (HCN2) linearized DNA using T7 RNA polymerase (Message Machine; Ambion, Houston, TX, USA) was injected per oocyte.
Animal procedures
Xenopus oocytes were harvested from frogs anesthetized by immersion in ice-cold 0.05% Tricane buffered to pH 7 with sodium phosphate. After suturing of the incision, frogs were allowed to recover slowly by being supported at the surface of an ice/water bath while it thawed or were euthanized by transfer from the anaesthetic bath to one containing Tricane 0.65%. All procedures were according to a Columbia University IACUC approved protocol (PI no. 366G, CU no. 2928).
Electrophysiology
Unless otherwise indicated recordings in two-electrode voltage clamp (TEVC) and inside-out patch clamp (IOPC) were made as previously described (Fogle et al. 2007). For TEVC, oocytes were bathed in a recording solution of (mM) 107 NaCl, 5 KCl, 2 MgCl2, 10 Hepes-free acid pH 7.4 (NaOH) or, if indicated, 87 NaCl, 25 KCl, 2 BaCl2, 10 Hepes-free acid pH 7.4 (NaOH). For IOPC, the extracellular solution was (mM) 107 KCl, 5 NaCl, 1 MgCl2, 1 CaCl2, 10 Hepes-free acid pH 7.4 (KOH) and the intracellular solution was (mM) 107 KCl, 5 NaCl, 1 MgCl2, 1 EGTA-free acid, 10 Hepes-free acid pH 7.4 (KOH). In IOPC, data acquisition was not begun until at least 6 min after patch excision into the Mg2+-containing solution to allow gating to largely equilibrate to the cell free level. Signals were normally filtered at 1 and 2.5 kHz and sampled at 2 and 5 kHz (TEVC and IOPC, respectively) except for non-stationary fluctuation analysis (NSFA) when frequencies of 10 and 50 kHz were used.
Paradigms and analysis
Steady-state gating parameters were determined by fitting tail current activation curves with the Boltzmann equation:
|
| (1) |
In TEVC studies, after determination of an initial activation curve, cells were transferred to a 20 ml glass scintillation vial containing L-15 and allowed to recover for 10 min. Following drug, vehicle or control incubations (20 min in recording solution on a 3-D rotator from Laboratory Line, Melrose Park, IL, USA), cells were then transferred back to the recording chamber and the post-treatment activation curve determined in the continuous presence of the appropriate drug, vehicle or control solution. No cell was exposed to more than one condition.
To follow changes in channel activation with better temporal resolution than could be achieved from serial determinations of complete activation curves, we determined 
V1/2,APPARENT from records collected with a two-step protocol (e.g. Fig. 5A) using:
|
| (2) |
V1/2,APP.
|
For NSFA, currents recorded at VINT and VMAX (see paradigm in Fig. 4A) were monitored to ensure stability of gating while IMAX was analysed according to the methods of Heinemann and colleagues (Heinemann & Conti, 1992; Steffan & Heinemann, 1997). Briefly, the variance (
2; obtained from 0.5 times the mean of the squared difference between sequential pairs of sweeps, 113–590 per patch) was plotted against the mean current and fitted by:
|
| (3) |
2b the background variance when channels were not open. PMAX was determined with:
|
| (4) |
|
|
| (5) |
Data analysis was performed in PulseFit (HEKA Elektronik) or using custom analysis routines written in IgorPro v5.0.4.8 (Wavemetrics Corporation, Lake Oswego, OR, USA).
Gating models
Discrimination between potential gating models was initially carried out by inspection. Subsequently, we analysed the equilibrium and kinetic behaviour of HCN channels in two linear (II-2 and II-3) and two cyclic (III-2 and III-4) models (Fig. 8) using global analysis in IgorPro and Fminsearchbnd in MatLab v7.1.0.246(R14) (Mathworks, Natick, MA, USA).
|
|
| (6) |
|
| (7) |
|
| (8) |
|
| (9) |
|
| (10) |
To explore the kinetic behaviour of these models, a Q-matrix was entered and the time course of the state vector solved according to the matrix equation X(t) = eQ(t)X(0). For Model II-2, X(t) is the 5 x 1 time dependent state vector (i.e. CR1, CA2, O3, CR*4, CA*5) at time t, Q(t) is the 5 x 5 Q-matrix, and X(0) is the 5 x 1 initial state vector equal to [1 0 0 0 0]T. The larger models were encoded equivalently using appropriate expansions of the vectors and matrices. The durations of activation and deactivation phases were the same as those used to collect experimental records. Within each phase, all rates within the Q-matrix were held constant. For a given parameter set, the initial state-distribution at the holding potential (–40 mV) was taken as the sum of the row entry of the spectral matrix corresponding to the zero eigenvalue (Colquhoun & Hawkes, 1995) with the channels equilibrated with the appropriate propofol concentration. Activation potentials were set equal to the mean of the observed step voltage minus V1/2 voltage bins plus the mean V1/2 of the observed data analysed for activation kinetics. For deactivation, channels were opened at –142.9 mV and –151 mV (absence and presence of propofol, respectively) then stepped to potentials equal to the mean of the observed deactivation step voltage minus V1/2 voltage bins plus the mean V1/2 of the observed deactivation data.
Rate constants within the Q-matrix were adjusted by nonlinear function minimization until the summed squared error between parameters describing the simulated currents (extracted by analysis as described for observed data) and the observed parameters was minimized. To ensure that all elements in the data contributed equivalently to the solution we performed two levels of error normalization: (1) within each data set the errors from small and large amplitude elements were weighted (for time constants and lags this was achieved by multiplying each contribution to the global error by 
MIN/i where MIN is the smallest value in the observed data set and i is the ith entry and for Boltzmann parameters we multiplied the slope error by the observed V1/2/slope ratio); and (2) we multiplied the error for each model parameter by its own empirically adjusted ‘strength factor’ until the error from each parameter set contributed approximately equivalently to the global error measure. A parameter set was excluded from contributing to the fit by setting its strength to zero.
Dose–response data were fitted using the Hill equation:
|
| (11) |
Statistical analysis
SigmaStat V3.1 (Systat Software, Point Richmond, CA, USA) was used to perform Student's t tests (differences between populations), Student's paired t tests (differences before and after treatment) and one-way ANOVA with post hoc Holm–Sidak analysis (multiple comparisons to a single control population). Data are presented as means ± S.E.M.
Reagents
Propofol (TCI America, Portland, OR, USA) solutions were prepared fresh each day by dilution of a 500 mM stock solution (DMSO) that was stored at –20°C. Other electrophysiology reagents were of the highest purity from Sigma.
|
Results |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
The principal effects of propofol on wild-type (wt) HCN1 are recapitulated in Fig. 1A–D. Inspection of the TEVC current families (Fig. 1A, left) and the corresponding tail currents (Fig. 1A, right) obtained in the absence (Fig. 1A, top) and presence (Fig. 1A, bottom) of 20 µM propofol shows that the drug slows opening (see also Fig. 1B), accelerates closing and shifts gating to more hyperpolarized potentials (see also Fig. 1C). A fit of the Hill equation to the results from a number of such experiments (Fig. 1D) shows that, under these conditions, both the apparent affinity (EC50
13 µM) and maximal efficacy (V1/2
–45 mV) of the propofol-mediated hyperpolarization of gating are similar to those reported previously (EC50
6.7 µM and V1/2
–35 mV; Cacheaux et al. 2005). However, our results diverge from the published findings in one interesting respect: we find concentrations of propofol as high as 100 µM do not significantly reduce the maximal current (Fig. 1H). Variation in the efficacy of an antagonist on the current amplitude is a property of the model that we find best describes the action of the anaesthetic on HCN channels (see Fig. 10 and discussion thereof) and we suspect that differences in the recording conditions used in the various studies account for this discrepancy (see Discussion).
|
|
To ascertain if alterations in coupling of the CNBD and S4 proton sensor to channel gating are necessary or sufficient for selective inhibition of HCN1, we generated mutant channels wherein we concatenated the arginine to glutamate mutation in the CNBD (RE) with the histidine to glutamine (HQ) or arginine (HR) in S4. The resulting channels, HCN1-REHQ, HCN1-REHR, HCN2-REHR and HCN4-REHR activate and deactivate with time courses similar to their cognate wild-type counterparts (supplemental material Figs S2 and S4) but in each background, the mutant channels require more negative voltages to open (supplemental material Table S1) as a result of the designed loss of the positive influence of basal cAMP (Chen et al. 2001) and introduction of an inhibitory pseudo-protonation in the HR background (Zong et al. 2001). Importantly, we find that while introduction of these mutations weakens the effect of propofol on the V1/2 of HCN1 and leads to the emergence of an anaesthetic mediated inhibition of the maximal current (supplemental material Fig. S4) they do not eliminate the sensitivity of HCN1 nor do they introduce a marked increase in propofol responsiveness in HCN2 and 4 (supplemental material Figs S2 and S3). These findings are consistent with the cAMP insensitivity of the propofol inhibition of IH in thalamic relay cells and of HCN2 heterologously expressed in HEK293 cells (Ying et al. 2006).
To ask whether the selective action of propofol on HCN1 was encoded within the poorly conserved distal amino and carboxy termini or by elements within the C-terminal cAMP-gating ring, we analysed the response HCN1-NvC, a channel wherein both the variable amino terminal region (Nv) and all of the cytoplasmic C-terminus including the CNBD and C-linker (C) were removed (see supplemental material Fig. S1). Inspection of the TEVC currents (Fig. 1E left) and the corresponding tail currents (Fig. 1E right) in the absence (Fig. 1E top) and presence (Fig. 1E bottom) of 20 µM propofol shows that HCN1-NvC retains the essential behaviour of wtHCN1. Thus, opening of both wtHCN1 (Figs 1A and B) and HCN1-NvC (Fig. 1E and F) are relatively fast and are well described by a delayed single exponential function (see Methods) and anaesthetic slows opening (Figs 1E and F), accelerates closing (Fig. 1E) and hyperpolarizes gating (Fig. 1E–G) of the truncated construct.
To what extent does HCN1-NvC recapitulate the quantitative responses of wtHCN1 in the absence and presence of propofol? Inspection of Fig. 2 reveals that propofol slows opening of both channels by increasing the durations of both the time constant and lag in a manner that is largely (wtHCN1) or completely (HCN1-NvC) accounted for by the shift in the midpoint in activation gating. A fit of the Hill equation to a concentration response plot (Fig. 1D) reveals that propofol's actions on gating of HCN1-NvC are similar to the effect of the drug on wtHCN1 although shifted to somewhat higher affinity and lower efficacy (EC50
3.3 µM and V1/2
–32 mV versus
13 µM and
–45 mV for HCN1-NvC and wtHCN1, respectively). In addition to these modest quantitative differences, propofol concentrations that maximally hyperpolarize HCN1-NvC suppress the maximal current (Fig. 1H), an effect not seen in wtHCN1. These differences are similar to those seen upon introduction of the REHR and REHQ mutations. Moreover, as observed for the point mutations, opening of the activation gate of HCN1-NvC is harder (as revealed by the inherently slower kinetics and hyperpolarized V1/2
– compare Fig. 1A and B with Fig. 1E and F and see also Fig. 2 and supplemental material Table S1). The finding that distinct changes which render channels harder to open are correlated with a weakening of the effect of propofol on the V1/2 and emergence of a sensitivity of the maximal current to anaesthetic suggests that both these changes in efficacy are mediated through the energetics of the gating reactions. This is explored further in the discussion of Fig. 10.
|
NvC as this channel, unlike wtHCN1, does not cluster in membranes (authors' unpublished observations – see also Gravante et al. 2004).
Propofol inhibition of HCN1-NvC is membrane delimited
Representative IOPC recordings (Fig. 3A) and current–voltage, current inhibition and normalized steady-state activation curves obtained therefrom (Fig. 3B–D, respectively) reveal intracellular application of propofol modifies HCN1-NvC channel function in a manner that is qualitatively identical to the effect of the drug applied externally in TEVC. Thus, gating is slowed and shifted to more hyperpolarized potentials while the maximal current is suppressed. Interestingly, propofol suppression of the current is itself voltage sensitive (Fig. 3C) such that the current is robustly suppressed at relatively depolarized potentials but this inhibition is reduced at voltages where opening was saturated (see arrows in Fig. 3B and D). Importantly, the relief of current suppression is well represented by a scaled version of the equilibrium activation curve suggesting the voltage dependence of this arises from drug modification of gating and not unblock (which would not be expected to saturate at potentials negative to channel activation). Furthermore, the finding that propofol hyperpolarizes gating of HCN channels irrespective of the face of the membrane to which the drug is applied indicates that its action is not mediated via alteration of the surface charge (Veintemilla et al. 1992).
|
Saturation of the voltage-dependent relief of current suppression does not, by itself, establish that the action of propofol is solely determined by modification of gating. Thus, the data are not inconsistent with propofol also reducing the single channel current via a voltage-independent mechanism. This distinction has important consequences with regard to our ability to discriminate between gating models. As analysis of single HCN channels is rendered problematic by their small conductance (1 pS; DiFrancesco, 1986; DiFrancesco & Mangoni, 1994; Dekker & Yellen, 2006 see also below), here we explored the effect of propofol on the maximal open probability (PMAX) and single channel conductance by means of non-stationary fluctuation analysis (NSFA).
Data for NSFA were acquired with the paradigm shown in Fig. 4A. Monitoring the current amplitudes at both VINT and VMAX (e.g. Fig. 4B) allowed us to establish that gating was stable and VMAX was adequate to maximally activate the channels throughout each recording. The variance (obtained from the mean squared difference between sequential sweeps acquired at VMAX
– e.g. Fig. 4C) was plotted versus the mean current and fit with eqn (3) (e.g. Fig. 4D). From five such patches, we determined PMAX to be 0.86 ± 0.02 (consistent with the observation that the variability is greatest during the rising phase of the current in Fig. 4C). A plot of mean variance versus mean current for HCN1-NvC in the presence of 10 µM propofol does not reach the apex of the parabola suggesting that the PMAX is less than 0.5 in the presence of anaesthetic. Consistent with this, a fit of eqn (3) suggests that PMAX is 0.43.
Fits of a straight-line function to the initial phase of distributions such as those shown in Fig. 4D and E (Fig. 4F and G) yielded estimates of the single channel conductance of 0.96 ± 0.05 pS (n
= 7) and 1.15 ± 0.05 pS (n
= 5) in the absence and presence of propofol, respectively. The finding that propofol does not reduce but, rather, appears to modestly increase the observed single channel conductance (P
0.026) indicates that propofol does not act as a fast ‘flickery’ blocker of the HCN1 pore. However, from these data we cannot exclude the possibility that at least part of the decrease in PMAX may involve a slow pore block. Thus, given that current suppression occurs with a relatively low affinity (13 µM in IOPC, see below), if we assume that propofol can access the pore via the aqueous phase at a diffusional rate (108
M–1 s–1) we can calculate that the mean block time (0.8 ms) would be resolvable within the bandwidth of our analysis.
The propofol mediated hyperpolarization of opening occurs with a higher apparent affinity than the drug mediated suppression of the maximal current
How fast does propofol inhibition of gating and maximal current develop and reverse and with what concentration dependence? Slow gating of HCN channels precludes use of serial activation curves to follow development of inhibition with adequate temporal resolution and raises the possibility that propofol mediated changes in channel function may be convolved with drug-independent changes in gating. To circumvent these problems we utilized the V1/2,APP paradigm.
Figure 5A shows a typical protocol (top) and the voltage paradigm (bottom) used in these recordings. Figure 5B and C shows superimposed current records obtained before, during and after exposure of a patch containing HCN1-NvC channels to 20 µM propofol at the times indicated in Fig. 5D. At both times, application of propofol suppressed the current elicited at both VINT and VMAX (a voltage that will saturate channel opening in the presence or absence of propofol) but the inhibition was more pronounced at VINT.
Inspection of the determined values of V1/2,APP and IMAX/IMAX,INITIAL as a function of time reveals that current suppression develops more slowly than the hyperpolarization of gating during propofol wash-on but the maximal current amplitude recovers more quickly than V1/2,APP when propofol is washed off (Fig. 5D). These findings suggest that the affinities of the two processes for propofol are different with the current suppression being less sensitive to anaesthetic. Serial application of propofol supports this hypothesis by showing that anaesthetic concentrations at or below 1 µM hyperpolarized gating but had no obvious effect on the maximal current. Construction of full dose–response relationships from a number of such recordings (Fig. 5E) shows that the negative shift in the V1/2 occurred with an EC50 of 3.7 µM and a relatively shallow response (Hill coefficient of 1.3) while the decrease in IMAX required higher concentrations of drug (EC50
= 12.9 µM) and developed with a steeper dependence with respect to propofol concentration (Hill coefficient = 1.7).
The data presented in Fig. 5B and C offer a further indication of apparent complexity in the effect of propofol on the V1/2 and IMAX. Thus, inspection of the current records acquired early in the recording (Fig. 5B) compared to those acquired later (Fig. 5C) shows that 20 µM propofol elicits a more pronounced suppression of IMAX and weaker suppression of IINT at the later time. This change is clearly seen in the plots of V1/2,APP and IMAX/IMAX,INITIAL (Fig. 5D) where it is reported as a weakening of the action of the drug on the V1/2 and a strengthening of its action on maximally activated channels.
The trivial interpretation of the above observations is that separate sites with different affinities for propofol mediate the two events. However, this simplistic conclusion does not necessarily follow as the observed separation of apparent affinities and reciprocal changes in efficacy can arise from the behaviour of even simple drug models as a function of modest changes in the energetics of gating (see red lines in Fig. 5E and see Fig. 10 and discussion thereof).
V1/2 coupled and uncoupled effects of propofol on HCN gating kinetics
Figure 6 shows that activation (Fig. 6A and B) and deactivation (Fig. 6C and D) of HCN1-NvC in both the absence (grey) and presence (black) of 10 µM propofol are well fitted by a delayed single exponential function. It should be noted that the duration of steps we employed to open the channels (3 s to explore the kinetics of activation at different voltages and 1.5 s at either –145 or –155 mV to investigate the kinetics of deactivation) are such that any slow open state transitions should have fully developed (Mannikko et al. 2005; Elinder et al. 2006). In accord with this, we did not observe an unfitted slow component of channel opening at any voltage where active current was discernible from baseline (note the stability of the residuals plotted in green in Fig. 6A). Similarly, when deactivation was measured at a fixed potential of –40 mV following activation at different potentials, the lag and time constant of current decay were independent of the voltage at which the channels were opened (data not shown). Furthermore, we found that the relationship between the deactivation test potential (between –135 and –95 mV) and the amplitudes of the lag and time constant of channel closing in the absence of anaesthetic was not dependent on whether the channels were opened at –145 or –155 mV (data not shown; analysis of the deactivation kinetics in the presence of propofol was always studied following activation at –155 mV). It is not possible to distinguish whether the appearance of a weak dependence of the deactivation lag on the deactivation test voltage (the red dots migrate closer to zero time at the most negative potentials) arises as a consequence of poor separation of the exponential and lag terms at these voltages or as a result of true complexity in deactivation following opening under the indicated conditions.
|
NvC recorded under TEVC. Thus, in this representative recording, where 10 µM propofol hyperpolarized gating by 18 mV, the opening trajectory at any voltage in the absence of propofol is qualitatively similar to the opening observed at a 20 mV more hyperpolarized potential in the presence of anaesthetic. This can best be seen by inspection of the rate of rise of the exponential component and the shortening of the pre-exponential lag (the red dots progress towards earlier times with increasingly hyperpolarized voltages) in the plot of the early waveform (Fig. 6B). Similarly, inspection of the deactivation records (Fig. 6C and D) suggests that the exponential component of channel closing is also well coupled with the midpoint of equilibrium gating. Thus, in this recording (where 10 µM propofol hyperpolarized the V1/2 by 14 mV; data not shown), channel closing in the absence of anaesthetic is most similar to the deactivation trajectory obtained at a 10 mV more hyperpolarized potential when propofol was present.
As suggested from the qualitative inspection of the opening and closing records we find that channel opening at hyperpolarized potentials is slowed in the presence of propofol (Fig. 7A left) while the anaesthetic accelerates deactivation across the entire voltage range we explored (Fig. 7A right). Similarly, Fig. 7C shows that propofol lengthened the lag that precedes opening at each tested voltage but had no clear impact on the essentially voltage-independent lag that precedes channel closing. However, it is also apparent from these plots that the time constants of the slowest relaxations in the presence of propofol are faster than those recorded in the absence of anaesthetic (Fig. 7A) while there is little difference in the maximum amplitude of the lags (Fig. 7C). These findings suggest that there are two effects of propofol on gating kinetics as measured in IOPC. In an attempt to deconvolve these components, we binned the time constants and lags with respect to the applied voltage relative to the V1/2 (indicated by 0 mV in Fig. 7B and D). It is apparent from these plots that, at voltages far from the V1/2, the time constants describing opening and closing are essentially independent of anaesthetic (Fig. 7B). It is also clear that, when corrected for the shift in equilibrium gating, there is no effect of anaesthetic on the lag that precedes opening while the lag that precedes closing is independent of both voltage and the presence or absence of propofol (Fig. 7D). Thus, the effect of propofol on gating kinetics determined in IOPC is the sum of two components – a left shift in the
versus voltage and activation lag versus voltage curves (changes which are strictly coupled with the hyperpolarization of equilibrium gating) and an acceleration of basal gating that is independent of changes in the midpoint of activation.
|
A number of models have been used to describe aspects of voltage-dependent activation of HCN channels. These models can be divided into two general classes shown as Scheme I and Schemes II and III in Fig. 8A.
Scheme I represents concerted allosteric models wherein the opening transition is voltage dependent as defined by the equilibrium constant KV. In this simple rendition, the channels exist in only two states, a closed conformation wherein the voltage sensors, the pore and any drug binding sites are in ‘closed’ channel conformations and an open conformation. In such a model, opening of the pore occurs in concert with transition of the voltage sensors and drug binding sites to their ‘open channel’ conformations. Although extensions of this formalism have been widely, and successfully, used to describe aspects of HCN gating (Altomare et al. 2001; Wang et al. 2002; Ulens & Siegelbaum, 2003; Mannikko et al. 2005; Elinder et al. 2006) no model of this type can account for a gating-mediated suppression of the channel current because if the voltage is made sufficiently permissive, the channels will open. In light of our evidence that current suppression arises from a propofol modification of gating, we reject class I models as appropriate descriptions of basal gating of HCN channels and the basis of HCN1 sensitivity to propofol (but see Discussion for further consideration).
Scheme II represents sequential models wherein activation of the voltage sensors precedes a voltage-independent opening reaction defined by the equilibrium LA. In this class of models, changes in the architecture of the protein related to interaction with allosteric modifiers (such as propofol) can be modelled as occurring to any or all of the states. This class of models has recently been used to describe a discrete action of cAMP modulation, notably the ability of cAMP to overcome a latent inactivation (not shown in Scheme II for simplicity) that can account for both the enhancement of channel current by nucleotide and an inverse Cole–Moore effect (Shin et al. 2004) and as a basis for understanding the action of the volatile anaesthetic, halothane (Chen et al. 2005b).
Within Scheme II, we can envision three separate families of models: (1) single-state binding models – those where propofol binds to only one of the three states (CR, CA or OA); (2) two-state binding models – those where propofol can bind to two states (CA and CR, CA and OA and CR and OA); and (3) three-state binding models – where propofol can bind to CR, CA and OA. Two- and three-state binding models can be further subdivided depending on whether the propofol bound states communicate or not.
All single-state binding models can be immediately discarded. Thus, binding to CR alone will hyperpolarize voltage dependence without changing the maximal current, binding to OA alone will facilitate voltage-dependent opening and enhance the maximal current, while binding to CA will suppress the maximal current but will facilitate voltage-dependent opening. This analysis further shows that the only two-state binding model that can describe the action of the anaesthetic is that where propofol binds to CR and CA. Thus, by inspection and comparison to our data, we conclude that of class II models only II-2 and II-3 (where drug also binds to the open state OA) can account for the response of HCN1 to propofol.
Scheme III represents a recently proposed variant of the class II models discussed above. This model is attractive as it retains the voltage-independent isomerization between closed and open states but it restores the symmetry associated with allosteric models by allowing channels to open irrespective of the position of the voltage sensors and the voltage sensors to move irrespective of the status of the activation gate. The motivation for development of this model was the finding that class II models could not simultaneously account for the speed of HCN2 activation and deactivation (Chen et al. 2007). Inspection of class III models reveals that, as observed with class II models, only III-2 and variants where propofol can additionally bind weakly to the open state can account for the response of propofol on HCN1 channels.
Determination of the strength of the activated and deactivated opening isomerization reactions
The critical step in parameter estimation in class II and class III models lies in the estimation of the voltage-independent opening reaction(s) LA and LR. Thus, if channel opening were highly unfavourable (LA is small) propofol could only hyperpolarize activation by binding more tightly to CR than to CA. However, if, as we observe, LA for HCN1-NvC in IOPC is relatively large (6.6 ± 0.8, based on NSFA and eqn (9)), opening will, under drug free conditions, couple to, and draw forward, the voltage-dependent step (CR to CA) such that preferential stabilization of either CR or CA will hyperpolarize gating.
To investigate the potential contribution of channel opening via a deactivated opening isomerization, we used propofol inhibition to estimate the upper bound of the equilibrium constant, LR. Inspection of sweeps from a representative patch (Fig. 9A and B) and group data from four such high expression recordings (Fig. 9C) reveal that 200 µM propofol is sufficient to essentially abolish the current carried by HCN1-NvC channels activated at –155 mV (IHCN,–155mV was decreased to 0.4% of the pre-drug current amplitude) but this is not accompanied by any evidence of a propofol sensitive component of the leak current at either the holding potential of –60 mV or immediately after stepping to –155 mV. Substituting the measured values of LA and KVC,0, the calculated occupancy of OA and estimated upper bound value of OR occupancy at the holding potential into eqn (10), we determine that LR must be below 0.0015–0.00065. Accordingly, in subsequent modelling we fixed LR to LA
x 10–4.
|
NvC channels and their modification by propofolTo estimate the rate constants of gating and drug binding steps we fitted models II-2 and III-2 to the observed data in a two-step process. First, we fitted the drug-independent aspects of each of model (front faces of the schemes in Fig. 8B and D and 8C and E) to the activation and deactivation time constants and the Boltzmann parameters determined in the absence of propofol. These fits were constrained by setting k32 equal to k23/LA and k71 equal to k17/LR. Where included, k23, k17, k12_0mV, k21_0mV, k73_0mV and their associated gating charges (k12_Z, k21_Z, and k73_Z) were free variables while k37 was controlled by microscopic reversibility.
We next determined the rate constants of the drug binding reactions (paths leading to the back plane of Fig. 8B–E) and the transitions between drug bound states (back plane of the schemes in Fig. 8B–E) by simultaneously optimizing fits of the models to PMAX and V1/2 concentration–response curves (by use of eqns (6), (7) and (8) as appropriate) as well as to activation and deactivation kinetics and Boltzmann parameters determined in the presence of 10 µM propofol. For fitting, we converted the IMAX data shown in Fig. 5E to PMAX based on our measure of PMAX of HCN1-NvC in the absence of anaesthetic. Thus, we assume that the observed IMAX is fully described by a decrease in the open probability that arises from drug modification of gating. In these fits the drug binding and unbinding rates and k45_0mV and its associated gating charge (k45_Z) were free variables but the drug independent rates were constrained to the optimized values determined from the modelling undertaken in the absence of propofol. k54 was constrained by microscopic reversibility.
Qualitatively, the simulated activation and deactivation currents obtained from the optimized fits showed good correspondence the observed data (e.g. supplemental material Fig. S5). To better compare the fits of models II-2 and III-2 we superimposed the predicted time constants and lags on the observed kinetic data (Fig. 7) and the predicted concentration response for both V1/2 and IMAX on the observed concentration response data (Fig. 5). The optimized values of the fit of model II-2 are shown in Table 1. It is apparent from these fits that the simple three-state model II-2 can largely account for the behaviour of the channels in the absence and presence of drug with two exceptions. First, it fails to accommodate the observed lags in that it predicts neither the amplitudes of the lags nor the discontinuity that follows from the voltage dependence of the activation lag and voltage independence of the deactivation lag. Second, the concentration–response relations are somewhat shallow. Somewhat surprisingly, permitting the deactivated channels to open (model III-2) did not offer any obvious qualitative enhancement in the fits. Inclusion of the observed lags in the function minimization for either model did not overcome these limitations (data not shown).
|
|
Discussion |
|---|
|
TopAbstractIntroductionMethodsResultsDiscussionReferences |
|---|
Adequacy of the sequential class II models
To what extent do deviations between observed and model derived kinetic parameters reflect limitations of the experimentally determined data as opposed to limitations in the models?
As our rejection of Scheme I models is only valid if current suppression is a function of gating and not the result of some gating-independent process such as channel block, it is relevant to consider the robustness of the data that support this interpretation. Here we have shown that: (1) current suppression is voltage dependent across the range where channels activate but is not fully relieved at more hyperpolarized potentials (Fig. 3C); (2) the magnitude of current suppression changes reciprocally with respect to the ability of propofol to hyperpolarize gating during prolonged IOPC recording (Fig. 5B–D); (3) introduction of mutations that are far from the pore (REHR and REHQ) alter the relative strengths of the propofol-mediated changes in the V1/2 and IMAX (Fig. 1 and supplemental material Fig. S4); (4) the kinetic description of propofol's action determined when the drug concentration was stationary (Table 1) is consistent with the differential rate of change in the V1/2 and IMAX when the anaesthetic concentration is changing (Fig. 5D); and (5) NSFA demonstrates propofol does not induce a flickery-block of the pore (Fig. 4). Although we canno
