Article Information

• PDF (204 kb)
• Full Text with Large Figures
• Cited by in Scopus (2)

PubMed

• Articles by M. Zaniboni
• Articles by M. Groppi

Related Articles

• Dynamical Mechanisms of Pacemaker Generation in IK1-Downregu...

  Dynamical Mechanisms of Pacemaker Generation in IK1-Downregulated Human Ventricular Myocytes: Insights from Bifurcation Analyses of a Mathematical Model Biophysical Journal, Volume 89, Issue 4, 1 October 2005, Pages 2865-2887 Yasutaka Kurata, Ichiro Hisatome, Hiroyuki Matsuda and Toshishige Shibamoto Abstract Dynamical mechanisms of the biological pacemaker (BP) generation in human ventricular myocytes were investigated by bifurcation analyses of a mathematical model. Equilibrium points (EPs), periodic orbits, stability of EPs, and bifurcation points were determined as functions of bifurcation parameters, such as the maximum conductance of inward-rectifier K current (), for constructing bifurcation diagrams. Stable limit cycles (BP activity) abruptly appeared around an unstable EP via a saddle-node bifurcation when was suppressed by 84.6%. After the bifurcation at which a stable EP disappears, the -reduced system has an unstable EP only, which is essentially important for stable pacemaking. To elucidate how individual sarcolemmal currents contribute to EP instability and BP generation, we further explored the bifurcation structures of the system during changes in L-type Ca channel current (), delayed-rectifier K currents (), or Na/Ca exchanger current (). Our results suggest that 1), is, but or is not, responsible for EP instability as a requisite to stable BP generation; 2), is indispensable for robust pacemaking with large amplitude, high upstroke velocity, and stable frequency; and 3), is the dominant pacemaker current but is not necessarily required for the generation of spontaneous oscillations. Abstract | Full Text | PDF (408 kb)

• Kinetic Properties of the Cardiac L-Type Ca Channel and Its ...

  Kinetic Properties of the Cardiac L-Type Ca Channel and Its Role in Myocyte Electrophysiology: A Theoretical Investigation Biophysical Journal, Volume 92, Issue 5, 1 March 2007, Pages 1522-1543 Gregory M. Faber, Jonathan Silva, Leonid Livshitz and Yoram Rudy Abstract The L-type Ca channel (Ca1.2) plays an important role in action potential (AP) generation, morphology, and duration (APD) and is the primary source of triggering Ca for the initiation of Ca-induced Ca-release in cardiac myocytes. In this article we present: 1), a detailed kinetic model of Ca1.2, which is incorporated into a model of the ventricular mycoyte where it interacts with a kinetic model of the ryanodine receptor in a restricted subcellular space; 2), evaluation of the contribution of voltage-dependent inactivation (VDI) and Ca-dependent inactivation (CDI) to total inactivation of Ca1.2; and 3), description of dynamic Ca1.2 and ryanodine receptor channel-state occupancy during the AP. Results are: 1), the Ca1.2 model reproduces experimental single-channel and macroscopic-current data; 2), the model reproduces rate dependence of APD, [Na], and the Ca-transient (CaT), and restitution of APD and CaT during premature stimuli; 3), CDI of Ca1.2 is sensitive to Ca that enters the subspace through the channel and from SR release. The relative contributions of these Ca sources to total CDI during the AP vary with time after depolarization, switching from early SR dominance to late Ca1.2 dominance. 4), The relative contribution of CDI to total inactivation of Ca1.2 is greater at negative potentials, when VDI is weak; and 5), loss of VDI due to the Ca1.2 mutation G406R (linked to the Timothy syndrome) results in APD prolongation and increased CaT. Abstract | Full Text | PDF (1235 kb)

• A Computational Model of the Human Left-Ventricular Epicardi...

  A Computational Model of the Human Left-Ventricular Epicardial Myocyte Biophysical Journal, Volume 87, Issue 3, 1 September 2004, Pages 1507-1525 Vivek Iyer, Reza Mazhari and Raimond L. Winslow Abstract A computational model of the human left-ventricular epicardial myocyte is presented. Models of each of the major ionic currents present in these cells are formulated and validated using experimental data obtained from studies of recombinant human ion channels and/or whole-cell recording from single myocytes isolated from human left-ventricular subepicardium. Continuous-time Markov chain models for the gating of the fast Na current, transient outward current, rapid component of the delayed rectifier current, and the L-type calcium current are modified to represent human data at physiological temperature. A new model for the gating of the slow component of the delayed rectifier current is formulated and validated against experimental data. Properties of calcium handling and exchanger currents are altered to appropriately represent the dynamics of intracellular ion concentrations. The model is able to both reproduce and predict a wide range of behaviors observed experimentally including action potential morphology, ionic currents, intracellular calcium transients, frequency dependence of action-potential duration, Ca-frequency relations, and extrasystolic restitution/post-extrasystolic potentiation. The model therefore serves as a useful tool for investigating mechanisms of arrhythmia and consequences of drug-channel interactions in the human left-ventricular myocyte. Abstract | Full Text | PDF (561 kb)

• …more

Copyright © 2005 The Biophysical Society. All rights reserved.
Biophysical Journal, Volume 89, Issue 3, 2170-2181, 1 September 2005

doi:10.1529/biophysj.105.062828

Electrophysiology

Effect of Input Resistance Voltage-Dependency on DC Estimate of Membrane Capacitance in Cardiac Myocytes

M. Zaniboni^*, ,  , F. Cacciani^* and M. Groppi^†

^* Dipartimento di Biologia Evolutiva e Funzionale-Sezione Fisiologia, Università degli Studi di Parma, Parma, Italy
^† Dipartimento di Matematica, Università degli Studi di Parma, Parma, Italy

Address reprint requests to M. Zaniboni, Tel.: 05-21-905623.

The value of membrane capacitance (C_m) is a measure of how much charge is needed to displace cell polarization, how fast cell membrane passively responds to current/voltage changes, and how large is the membrane surface area. C_m and input resistance (R_m) of excitable cells are usually referred to as passive electrical properties and derived referring to an equivalent circuit, parallel combination of the two electrical elements. Microelectrode and patch-clamp techniques which have been developed, under current-clamp (CC) and voltage-clamp (VC) conditions, to solve this parallel RC combination to calculate C_m1,2, can be grouped in two categories: the alternating current (AC) and the direct current (DC) methods. Very accurate estimates of C_m are achieved using AC methods, by means of single or multiple sinusoidal command voltages, used to clamp the cell with two-phase lock-in amplifiers 3,4. Square-wave stimulations are also employed and analyzed in the frequency domain 5, and ramp command voltage protocols have been recently improved 6 and applied to cardiac myocytes 7,8. These techniques have been originally established and largely employed mainly to study changes in membrane area during exocytosis and endocytosis 3,9,10, but they are still too technically demanding for a more general use 6 like the estimate of cell size in cardiac preparations.

The measure of specific C_m (μF/cm²) in cardiac tissue and, since the introduction of enzymatic isolation procedures, of C_m (pF) in single cardiac myocytes, has been pursued from the early days of cardiac electrophysiology 11,12,13, given the pivotal role of this parameter in cardiac excitability and the physiopathological interest in monitoring cell size in the heart. Systematic errors in the measure of C_m will lead, for example, to erroneous evaluation of cardiomyocytes remodeling in studies where this parameter is measured in hypertrophied hearts 7,14,15, failing hearts 16, dilated cardiomyopathy 17, diabetes 18, and myocardial infarction 19. Such errors will also affect studies on intercellular electrotonic interactions 20,21,22 where C_m plays a primary role in determining source-sink properties of interacting cells. Finally, they will affect those studies where C_m is measured to normalize transsarcolemmal ion currents and fluxes through carriers to cell surface and, indirectly, to cell volume 23.

Definitely the more common tool adopted by cardiac electrophysiologists to measure C_m remains the use of time-domain DC methods, where constant current (or voltage) steps are imposed to the cell membrane and the study of the resulting voltage (or current) displacements is performed in terms of mono-exponential functions 1, a procedure which, as recently pointed out 6, has not been modified substantially since its introduction 30 years ago 24. This approach is based on the assumption that C_m is constant for a given cell, and that R_m is fairly constant over membrane potential (V_m) changes 1, at least in the voltage range of application of the mentioned protocols. Since R_m changes have been documented around the resting potential (V_r) (cited below), the second assumption is actually that these changes are negligible with respect to C_m estimate.

The fact that R_m changes during the cardiac action potential follows directly from the Hodgkin-Huxley theory of membrane excitability, has been described by Weidmann in early days 25, and recently measured in guinea pig ventricular myocytes 22. In addition, Weidmann, by means of constant current injections into cardiac fibers, noticed that R_m continuously decreased, starting from threshold V_m toward rest and for increasingly hyperpolarizing current injections. The decrease of R_m as membrane potential hyperpolarizes with respect to threshold potential, is accounted for in cardiac myocytes mainly by inward rectification of I_K126,27 and has already been measured in isolated rat ventricular myocytes 13,28.

In the present work, we show that R_m voltage-changes around V_r are indeed not negligible when standard CC/VC protocols are used to calculate C_m, which therefore results to be voltage-dependent as well. To this end, we measure the R_m(V_m) function around V_r and incorporate it in a mechanistic RC equivalent model, which better resembles membrane passive electrical properties. The numerical solution of such circuit allows accurate estimates of C_m which do not depend on voltage. We suggest that I_K1 rectification, as the main responsible for the input resistance voltage dependency, is also the primary source for the error in C_m estimate through the classical approach. This study, performed in turn, on real isolated rat ventricular myocytes, on five different mathematical models of the cardiac action potential, and on a simple RC mathematical model, also suggests a simplified procedure that allows to derive accurate estimates of C_m with the classical constant CC and VC protocols.

Single cells were obtained by enzymatic dispersion of adult (6 months, 400–500g) male Wistar rat left ventricles. After thoracotomy, the heart was rapidly removed, mounted on a Langendorff apparatus, and perfused at 37°C with the following sequence of solutions: Ca²⁺-free (control, no added calcium) Tyrode solution for 5min to remove the blood, low-Ca²⁺ (0.1mM) solution containing 1mg/ml type 2 collagenase (Worthington, Lakewood, NJ) and 0.1mg/ml type XIV protease (Sigma Aldrich, Milan, Italy) for 20min, and enzyme-free low- Ca²⁺ solution for 5min. The left ventricle was then minced and shaken for 10min in the low-Ca²⁺ solution. Myocytes were stored at room temperature in the control solution with 0.5mM Ca²⁺. All experiments were performed within 2–8h after isolation. All myocytes used in this study had well-defined striations and did not spontaneously contract.

Isolation solution contained 126mM NaCl, 22mM dextrose, 5.0mM MgCl₂, 4.4mM KCl, 20mM taurine, 5mM creatine, 5mM Na pyruvate, 1mM NaH₂PO4, and 24mM HEPES (pH=7.4 adjusted with NaOH). The solution was gassed with 100% O₂. Control solution for cell bathing during experiments contained 126mM NaCl, 11mM dextrose, 5.4mM KCl, 1.0mM MgCl₂, 1.08mM CaCl₂, and 24mM HEPES (pH-adjusted to 7.4 with NaOH). The pipette filling solution contained 113mM KCl, 10mM NaCl, 5.5mM dextrose, 5mM K₂ATP, 0.5mM MgCl₂, and 10mM HEPES (pH-adjusted to 7.1 with KOH). A drop of cells was placed in the experimental chamber (∼2.5ml) and superfused by gravity at a flow rate of ∼2ml/min. The temperature of the solutions in the cell bath was 37°C.

Suction pipettes were made from borosilicate capillary tubing (Harvard Apparatus, Edenbridge, United Kingdom) and had a resistance, when filled, of 2–4MΩ. Transmembrane potential and current (i) were recorded by means of an Axoclamp 2B amplifier (Axon Instruments, Union City, CA) adopting the whole-cell configuration of the patch-clamp technique.

Here we summarize time-domain DC methods for the estimate of C_m which have been adopted and criticized in this study. For the sake of clarity, we will consider the resting membrane potential V_r set to zero.

Cells were held in whole-cell configuration (amplifier in Bridge mode) and injected at a frequency of 2Hz, with 200-ms constant current pulses (i_p) usually starting from −0.260nA and incrementing with 5-pA steps up to 0.235nA. Subthreshold voltage deflections were sampled at 5kHz, digitized (Digidata 1200 Series Interface, Axon Instruments, Union City, CA), and best fitted with mono-exponential functions (Kaleidagraph, Synergy Software, Reading, PA) as solutions

(1)

(2)

relative to the linear equivalent parallel RC circuit. It is straightforward to derive the plateau value of V_m(t) as V_∞=i_p×R_m. The description does not include any additional series pipette resistance R_s, which was electrically compensated with bridge balance. Passive electrical properties R_m and C_m were derived as best-fitting parameters of experimental voltage traces with Eq. (1).

To minimize filtering effects due to pipette capacitance, pipettes tips were coated with a hydrophobic compound (Sylgard, Dow Corning, Midland, MI) and the cell bath was maintained at a low level (1mm). Protocol features were analogous to current clamp; with the amplifier in discontinuous single-electrode voltage-clamp mode, cells were initially clamped at their V_r and then stepped with 40-ms constant voltage-clamps (V_c) usually starting from V_r−20mV and incrementing with 1-mV steps up to V_r+15mV. Current traces were best fitted with mono-exponentials as linear functions

(3)

(4)

(5)

relative to the linear equivalent parallel RC circuit in series with an additional access resistance R_s. R_m, C_m, and R_s were derived as best fitting parameters. Membrane potential V_∞, corresponding to the plateau value of the current (i_∞=V_c / (R_m+R_s)), was derived as V_c−i_∞×R_s, being the additional R_s uncompensated during the protocol.

In the following discussion, we will use the notation ΔV_m=V_∞−V_r=V_∞ for the steady-state value reached by V_m during CC and VC protocols.

An alternative VC method to measure C_m, to which we refer in this work, is to measure the charge under the current transient and divide it by the ΔV=V_c−V_r. This is often approximated as

(6)

(7)

The electrical properties of single resting cardiomyocytes were simulated, in turn, using membrane equations from the mathematical models listed in Table 1, or solving Eq. (2) and Eq. (5) for a simple RC circuit including a series resistance R_s, which was set to zero in CC simulations. Equations were numerically solved using an adaptive Euler method for the LR91 model, and a fifth-order Runge-Kutta method for PD01 model and for the RC circuit. Simulations on the LR94, PB01, and RM00 were performed on the Cell Electrophysiology Simulation Environment (CESE Version 1.3.4, available on http://cese.sourceforge.net). All simulations were performed on a Pentium IV processor and codes for LR91, PD01, and for the RC circuit implemented in MatLab language (The MathWorks, Natick, MA).

Results are presented as mean±SE. Statistical analysis was performed using the unpaired Student’s t-test with SPSS software (SPSS, Chicago, IL). Statistical significance was set at P<0.05.

The membrane input resistance (R_m) of single rat ventricular myocytes was measured within ∼±10mV around resting potential (V_r) through CC and VC protocols (see Materials and Methods). Figure 1A shows voltage deflections from a CC protocol and the function R_m(ΔV_m) well fitted by a parabola with c₀=58.35, c₁=3.64, and c₂=0.14. The same analysis performed on 12 myocytes gave an average parabolic fitting of c₀=55.99±3.67, c₁=2.37±0.25, and c₂=0.06±0.01. Current traces from a VC protocol are reported in Figure 1B with the function R_m(ΔV_m), which is similar to the one measured in CC, and well fitted by a parabola with c₀=54.22, c₁=4.51, and c₂=0.12. The same analysis performed on 12 myocytes gave an average fitting of c₀=58.76±6.95, c₁=5.22±0.59, and c₂=0.15±0.02. When the two protocols were numerically simulated on a LR91 model, they returned a qualitatively analogous function R_m(ΔV_m), well fitted by a parabola with c₀=21.71, c₁=1.10, and c₂=0.027 . Similar R_m(ΔV_m) behavior was found in all five mathematical models employed (see below).

Display large version of this figure

Figure 1

Measure of R_m voltage-dependency. Left column: (A) voltage deflections from a CC protocol on a rat myocyte. Only one in every four traces is shown for clarity. (B) Current traces from a VC protocol on a different myocyte. Only one in every second trace is shown. (C) Analogous CC and VC protocols simulated on the LR91 model (R_s=5MΩ in VC). Right column: R_m, calculated from plateau values of each voltage and current trace, is reported (solid circles) as a function of ΔV_m, fitted by parabolic functions (dotted lines).

A constant R_m around resting potential V_r (which we will refer to as the Ohmic assumption in the following discussion) is required in classical DC measurements of C_m. If, as shown above, R_m changes up to 30–50MΩ within ±10mV around V_r, the solution of a non-Ohmic circuit that includes the R_m(ΔV_m) function (which we will call the non-Ohmic assumption) should be more appropriate.

In Fig. 2 we show examples in which C_m was calculated (see Materials and Methods) by mono-exponential fittings of the voltage/current traces reported in Fig. 1. The measured C_m varies with ΔV_m, decreasing with membrane polarization. The same analysis performed on seven myocytes led to C_m changes, as measured from a linear fitting for −10mV<ΔV_m<0mV, of 2.95±0.74pF/mV in CC and 1.35±0.54pF/mV in VC. Analogous results are shown, for example, on the more general LR91 model. Analyzing a non-Ohmic RC as if it was Ohmic clearly leads to an error in C_m estimate, which increases with the displacement of membrane potential from V_r, doing more so in CC than in VC. The CC protocol was applied also on the other mathematical models listed in Table 1, and results reported in Fig. 3. In the lower panel of the same figure, the relationship between the relative error in C_m estimate and R_m rectification is also reported for the five mathematical models and for the experimental data. To better isolate this effect, we solved two simple mathematical models of a resting myocyte in whole-cell configuration, which included only R_s, C_m, and R_m (Fig. 4). R_m was kept constant in a first case (Figure 4A) and voltage-dependent in a second case (Figure 4B). Figure 4A shows the symmetrical behavior (with respect to the resting potential/current) of voltage and current traces obtained when the Ohmic circuit (Figure 4A) was solved, in turn, in CC and VC conditions. When the constant R_m of the circuit was replaced (as in Figure 4B) with an experimental R_m(ΔV_m) function of the type shown in Fig. 1, the voltage and current deflections lost their symmetry and appeared more like those measured in real cells. Traces were all analyzed in terms of mono-exponential fittings (see Materials and Methods), assuming each time, for R_m, the constant value measured at plateau, and C_m values derived and reported as solid circles in Figure 4C. The Ohmic assumption, in the case of a physiologically non-Ohmic RC circuit, leads to an error in the estimate of C_m which qualitatively reproduces what we found in real myocytes and in the mathematical model cells (Fig. 2 and Fig. 3), with C_m varying with a sigmoidal-law around the resting potential.

Display large version of this figure

Figure 2

Estimated C_m as a function of V_m displacement from V_r. (Top) Each solid circle represents a DC estimate of C_m from experimental traces of Figure 1AB. (Bottom) Same for the traces from the LR91 model in Figure 1C, where C_m was set to 153.44pF (horizontal lines) on the basis of the geometrical considerations of Luo and Rudy 43. Dotted lines in each panel represent linear fittings of the experimental data, 1.6, 1.1, 3.1, and 1.7pF/mV, respectively.

Display large version of this figure

Figure 3

Voltage-dependency of R_m and of the estimated C_m in five mathematical models of cardiomyocytes. (Top) R_m(ΔV_m) as obtained through the CC protocol applied to the five mathematical models listed in Table 1. (Middle) C_m(ΔV_m) obtained with the same protocol. Results were normalized to the extrapolated resting values of R_m () and C_m (), which were respectively set to 1 for comparison. (Bottom) The above traces were linearly fitted within the −10/0mV ΔV_m range, to plot the slope of C_m(ΔV_m) versus the slope of R_m(ΔV_m) for the five models. The cross symbol represents the corresponding averaged estimate made on seven current-clamped rat ventricular myocytes (1.89±0.39mV⁻¹ vs. 3.26±0.25mV⁻¹), also cited in the Results.

Display large version of this figure

Figure 4

DC estimate of C_m in Ohmic and non-Ohmic RC circuits. (A) Current-clamped (from −0.25 to 0.25nA, step 0.025nA) voltage displacements and voltage-clamped (from −20 to +20mV, step 2mV) current traces, mathematically simulated on the circuit reported on the left, with R_m=56.5MΩ and C_m=250pF. R_s was set to 0MΩ in CC and 5MΩ in VC. (B) Identical protocols simulated on the same circuit were . (C) C_m versus ΔV_m, as calculated from mono-exponential fittings of the traces in B (solid circles).

In Fig. 5 two additional examples of C_m derived through mono-exponential fittings in two real ventricular myocytes are shown. In addition, the R_m(ΔV_m) parabolic functions, derived from both cells, were, in turn, included in a mathematical non-Ohmic model equivalent to the one in Figure 4B. The value R_s was also derived from the VC protocols, and was included in the model. Each voltage and current trace was then fitted with numerical solutions of the non-Ohmic form of Eq. (2) and Eq. (5) including the experimentally derived R_m(ΔV_m), to obtain new estimates of C_m (see Appendix for details on the algorithm), also reported in the figures. Whereas C_m, measured in the Ohmic assumption, is voltage-dependent and varies ∼2.1pF/mV in CC and 1.9pF/mV in VC, the values derived upon considering the non-Ohmic assumption were fairly constant within the voltage range under study, being 128.4±0.6pF in CC and 150.2±1.1pF in VC. Same result (voltage-independent estimates of C_m) was obtained when all five mathematical model cells were studied within the non-Ohmic assumption (not shown). For an additional control, we numerically simulated the CC and VC real experiment on two non-Ohmic model circuits like the one in Figure 4B, including, respectively, the non-Ohmically calculated two mean values of the C_m, the R_m(ΔV_m) functions and R_s. When we fitted the numerically integrated voltage and current traces with mono-exponentials, we found a continuous C_m(ΔV_m) function that well fits the C_m values Ohmically derived from experimental data. The same type of analysis was performed with analogous results on seven current-clamped and seven voltage-clamped ventricular myocytes where C_m values from different cells were normalized for comparison to the constant value found each time with non-Ohmic assumption, which was set to 1. Average error in normalized C_m estimate was 0.0189mV⁻¹ in CC and 0.0097mV⁻¹ in VC. Analogous results were obtained when C_m was measured in the LR91 model (Fig. 6). In summary, when, instead of using mono-exponential fittings, we solved a mechanistic model including the R_m(ΔV_m) derived from the simulated traces, we were able to measure, at least in a 10–15-mV range below V_r, a value of C_m that tightly resembles the actual one.

Display large version of this figure

Figure 5

Comparison of C_m estimate with Ohmic and non-Ohmic assumption in real ventricular myocytes. C_m, first calculated through mono-exponential fittings of CC-elicited voltage traces (top), and VC-elicited current traces (bottom), plotted as a function of ΔV_m (solid circles). C_m is also reported as derived through the algorithm which assumes non-Ohmicity of the circuit (open circles). Solid line represents the average value of C_m in the non-Ohmic assumption. This value, together with calculated R_s and R_m(ΔV_m), is incorporated in an equivalent circuit whose solutions, in CC and VC simulations, are fitted with mono-exponentials to derive C_m for each ΔV_m (dotted lines).

Display large version of this figure

Figure 6

Comparison of C_m estimate with Ohmic and non-Ohmic assumption in the LR91 model cell. C_m, as calculated adopting the Ohmic assumption (solid circles) and the non-Ohmic assumption (open circles) through a CC and a VC protocol on the LR91 model. Horizontal lines represent the actual C_m of the model. A series resistance of 10MΩ was added in VC simulations.

Another way, used in early works, to study the time course of voltage deflections in constant step-CC experiments, was to plot the logarithm of the voltage displacement as a function of time 13,1, which should result in a straight line in the case of mono-exponential rise. In Figure 7A we show, in a real cell, the time course of a voltage displacement from a CC protocol, together with the time course of its logarithm, respectively fitted with a mono-exponential and a straight line (C_m derived as 152pF). Although other traces from the same experiment revealed a 10-MΩ difference between R_m measured at V_∞=−8mV and V_∞=−4mV (Fig. 7, arrows), which in turn resulted in a 20-pF difference in the C_m estimate at the two potentials, both fittings showed a very high (R=0.999) correlation with experimental curves. A good mono-exponential fitting correlation or a quasi-linear relation between ln(ΔV_m) and time do not guarantee, along with the associated experimental noise, the Ohmicity of the tested RC and thus the independency of R_m from V_m. When the same voltage deflection was analyzed with the least-squares algorithm (see Appendix ) based on the non-Ohmic circuit (Figure 7B), the best-fitting solution corresponded to a C_m=193pF, which changed very little (mean±SE=0.7) in all the analyzed traces from V_∞=−1 to V_∞=−10mV.

Display large version of this figure

Figure 7

Ohmic and non-Ohmic analysis of a current-clamped voltage deflection. (A) Mono-exponential fitting (dotted line) (τ=5.7ms) of a current-clamped (−0.24nA) voltage displacement (solid line) from a ventricular myocyte. (Inset) Linear fitting (dotted line) of the logarithm of ΔV_m (solid circles; only one in every four experimental samples is plotted). (B) Dotted lines are solutions of Eq. (2) with , experimentally derived from the same cell, and for C_m values from 100pF to 280pF (only 1 in every 20 solutions is reported); solid line is the experimental voltage deflection.

The assumption of a constant R_m for the cell membrane leads to errors in the estimate of C_m as shown above. To investigate whether this error, described in Fig. 4 as a sigmoidal C_m(ΔV_m) function centered at the resting potential, depends on the actual C_m value of the cell, we performed simulations on a simple RC model including an experimentally derived parabolic R_m(ΔV_m), and, in turn, three different values for C_m. Simulated traces were fitted with mono-exponentials in the Ohmic assumption and measured C_m(ΔV_m) functions were reported as solid lines in Figure 8AB. A typical non-Ohmic R_m(ΔV_m) function, like the one used here, led to a −11.8% (CC) and −2.4% (VC) errors in the estimate of C_m at ΔV_m=−10mV in a 265pF cell. When cell capacitance was increased by 50%, this error decreased to −10.5% in CC and did not change in VC, whereas a 50% decrease of C_m led to an increase of error up to −12.7% in CC and again no changes in VC. We then repeated the test for a series of C_m from 5 to 1000pF. Interestingly, we found that, as the cell capacitance increased, the error in C_m estimate monotonically decreased from an initial value of 13.3% in CC, whereas it remained fairly constant at 2.8% in VC (with R_s=5MΩ). When R_s was increased to 20MΩ, the error increased in VC up to 5.6% for a C_m=5pF, and monotonically decreased to 4.2% for a C_m=1000pF. To summarize, the percentage error in C_m estimate decreases with the increasing of C_m, being less important in voltage than in current clamp. The error in VC-estimate of C_m depends on the uncompensated series resistance R_s.

Display large version of this figure

Figure 8

Effect of C_m and R_m(ΔV_m)-slope on the accuracy of C_m estimate in a simple RC circuit. (A and B) CC and VC estimate of C_m through Ohmic assumption on an RC model including an and an R_s=5MΩ (in VC), for three different values of C_m (horizontal dotted lines). (C) Error in C_m estimate (percent underestimate from the actual C_m), as measured at ΔV_m=−10mV, for different values of C_m, in CC (solid line) and VC (dotted lines) simulated tests, and for two values of R_s. (D) Three R_m(ΔV_m) parabolic functions differing for the slope. (E and F) CC and VC estimates of C_m by Ohmic assumption on an RC model including, in turn, the three functions. The double-arrow symbols represent the typical experimental noise level in C_m estimate.

We performed additional simulations to investigate how the slope of the R_m(ΔV_m) function affects the error in C_m estimate. We chose three different R_m(ΔV_m) parabolic functions (Figure 8D) differing in the first-order coefficient (∼±50% from a central value of 1.5) and all ranging within measured experimental values. We then run CC and VC simulations for an RC circuit including, in turn, the three R_m(ΔV_m) functions and having a C_m=265pF (Figure 8EF). With the increasing of the slope of the R_m(ΔV_m) function, the error in the estimate of C_m, as measured at ΔV_m=−10mV, increased both in CC and VC from values of 4.4% and 0.7%, respectively (lower slope function) up to 22.4% and 7.5% (higher slope function).

From the simulations performed on the non-Ohmic RC model in Fig. 4, it appears that the sigmoidal function of the estimated C_m(ΔV_m), derived assuming a constant R_m, has the property that it crosses the horizontal line corresponding to the actual value of C_m in the resting membrane potential (ΔV_m=0). Moreover the slope of this function is fairly constant in the V_m range of interest (−10mV<ΔV_m<0mV), where it can be approximated with a straight line. It follows that the proper value of cell capacitance could be, in principle, extrapolated with only two current injections (or VC pulses) experiments, using mono-exponential fittings, and ignoring therefore the voltage-dependency of R_m. Histograms in Fig. 9 show, for both CC and VC, average estimates of C_m, taken ignoring R_m voltage-dependency and adopting mono-exponential fittings (first and second columns in each panel), deriving R_m voltage-dependency and using the least-squares fitting procedure (third column), or assuming the function C_m(ΔV_m) to be linear and extrapolating the zero-potential value from two measures, taken approximately at ΔV_m=−10mV and ΔV_m=−5mV, for each cell (fourth column). The average value obtained with the extrapolation procedure does not significantly differ in CC and VC from the one measured with the non-Ohmic assumption.

Display large version of this figure

Figure 9

Histograms comparing CC and VC estimate of C_m with Ohmic and non-Ohmic assumption in real ventricular myocytes. (First and second bars) Estimate made through the Ohmic assumption at ΔV_m=−10mV and ΔV_m=−5mV, respectively. (Third bar) Estimate made through the non-Ohmic assumption (). (Fourth bar) Estimate made through extrapolation from pairs of Ohmic-assumption measurements. All bars are normalized to third bar, which is set to 1 (the asterisk symbol indicates significant differences).

Among the several ways developed to measure C_m, we focus the present study on the time-domain analysis of square-wave stimulations (as defined in Ref. 2 and the limitations we found in applying this technique to measure C_m in single rat ventricular myocytes. The use, adopted within this approach, of the mono-exponential solutions 1 and 3 of Eq. (2) and Eq. (5) in a subthreshold V_m range, coincides with the assumption that R_m is constant in this range or, at least, its changes do not appreciably affect C_m calculation. We show here that R_m voltage-dependency around V_r makes C_m estimates, through the classical exponential fittings, also voltage-dependent, and that such dependency can be removed by using a mechanistic RC model that includes R_m changes. As an alternative, a better estimate of C_m can be extrapolated from at least two Ohmic measurements on the same cell. The finding of a voltage-dependency also when C_m was measured in a very general model of the ventricular action potential like LR91, as well as in other more complex mathematical models, rules out the possibility that such effect could be, even only partially, due to experimental artifacts like time- or voltage-changes in R_s or changes in experimental parameters other than those (R_m, C_m, R_s) included in the simple representation of Figure 4B.

It should be noted that a frequently adopted variant of constant step VC protocol to measure C_m is the one (see Materials and Methods) based on the calculation of the charge underlying the capacitive current transient. An earlier version of this approach simply calculated C_m as in Eq. (6). Such calculation is not accurate and has been more recently replaced by Eq. (7), where the mono-exponentially derived time constant τ appears explicitly, and therefore makes this protocol suffer as well from the voltage-dependency of R_m. Indeed, errors in C_m calculations using Eq. (7) have been derived for the voltage-clamp simulation reported in Figure 4B without finding any difference from those derived with the classical protocol (data omitted for the sake of clarity). Therefore we did not discuss in a separate section errors in C_m estimate derived with the numerical integration of the current transient during voltage steps.

R_m voltage-dependency for subthreshold potentials is known in cardiac tissue at a cellular level 13,27, and has a part in explaining subthreshold behavior of extracellular membrane polarization 29. We limited the majority of our experimental work and analysis on R_m changes in the ∼10mV hyperpolarized V_m-range of well polarized (V_r=−73.45±0.69mV) rat left-ventricular cells (n=24). With this, we deliberately wanted to avoid eliciting the time-dependent processes that develop in cardiac membrane for depolarizing potentials (e.g., activation of inward sodium current, inward L-type calcium current, calcium-independent transient outward potassium current, steady-state outward potassium current; Ref. 30 and for more hyperpolarized potentials (e.g., I_K1 time-dependent block and I_f activation; Ref. 31. In fact, voltage traces always reached a plateau within CC steps, as well as current traces never showed time-dependent components other than the capacitive peak during VC steps on tested ventricular myocytes.

The 40–50% decrease of R_m when membrane potential was current- or voltage-clamped from V_r to (see Fig. 1) is reflected into the progressive compression/broadening of the elicited voltage/current traces. In the case of the mathematical LR91 model, the 40% decrease of R_m for ΔV_m=−10mV, corresponds to a 72% increment of g_K1 whereas other ionic conductances included in the model do not change (carried simulations not shown). Indeed, this is also the case for the other more complex mathematical models, where, in this subthreshold region, the I-V relationship with the highest degree of nonlinearity is that of I_K1 which, moreover, overwhelms all the others in absolute intensity. In fact, the simple analysis of published steady-state I_K1-V curves recorded in rat and other mammals ventricular myocytes (e.g., Refs. 32,33,34,35 shows that I_K1 starts rectifying at potentials that are at least 10- or 15-mV hyperpolarized, with respect to the typical V_r for this cell type, which has to be reflected in some R_m voltage-dependency. From our simulation work, we can rule out the direct contribution to R_m rectification of the electrogenic sodium-potassium pump, whose I-V relationship is linear in the subthreshold region of interest of this study (see, for example, I_NaK equations in PD01, LR94, RM00, and Ref. 36. For the same reason, we can exclude calcium, sodium, and potassium background currents which, although flowing in this voltage range, are usually formulated as linear leakage currents therefore not possibly contributing directly to any R_m rectification (see, for example, related equations in PD01 and LR94). It is tempting to hypothesize a significant role of the sodium-calcium exchanger current, whose IV relationship slightly deviates from linearity in the same region 30,37, but this would need further mathematical and experimental work to be confirmed and quantified. Furthermore, it is possible that some other rectifying mechanisms would be present that are still not included in the models and, therefore, in our knowledge of membrane electrical properties.

A major point of the present work is to show that the measure of C_m within the Ohmic assumption is voltage-dependent in a non-negligible manner. Indeed such dependency was always present in tested ventricular myocytes (Figure 2 and Figure 5), being less dramatic in VC than in CC, and only slightly depending on the actual value of C_m (see experiment in Fig. 8). In this regard, it is interesting to note the 26-pF difference, reported by Tseng and co-authors in canine ventricular myocytes, between average C_m estimates when measurements were performed in CC (voltage displacements up to −10mV) and VC (−5 or −10mV steps) by microelectrode impalement 38. According to our findings, both values underestimate the actual C_m, being the CC error at 54pF worst than the VC error at 80pF.

A voltage-dependency in the estimate of C_m can actually become hard to detect and therefore negligible, for example, in hyperkalemic conditions where I_K1-V curve shifts horizontally toward depolarized potentials and vertically toward more positive currents, becoming therefore more linear around its reversal potential (see [K⁺]_o-dependency of I_K1 equation in Ref. 33, and in Ref. 30. The same [K⁺]_o-dependency can be viewed in the total steady-state I-V relationship (e.g., Ref. 38. When [K⁺]_o increases, the slope of R_m(ΔV_m) function becomes smaller and errors in the estimate of C_m will be negligible compared with experimental noise. If we consider, for example, the experiment in Figure 8D, a cardiac myocyte with an R_m(ΔV_m) function a, characterized by a very weak voltage-dependency, will bring about, when analyzed in VC through the Ohmic assumption (Figure 8F), an underestimate of C_m, as measured at ΔV_m=−10mV, <1%, which will be masked within the typical noise level (∼5–6%) of our experiments. On the other hand, an increase in series resistance, especially in cells with higher R_m(ΔV_m) slope, can produce appreciable differences also in VC, whereas in CC, differences are always measurable practically in all physiological conditions. This is noteworthy, particularly if we consider that, although the great majority of recent cardiac cellular electrophysiological studies adopts VC constant pulses to measure C_m (e.g., Refs. 39,40), many still use the CC approach 21,22, which will then lead to much larger errors.

C_m measurements within homogeneous cell populations suffer frequently for a consistent dispersion (SD often up to 50% of the average) which could be attenuated with the non-Ohmic assumption, allowing better resolution of C_m changes following different pathological conditions or pharmacological treatments. Moreover, errors in C_m evaluation can complicate calculations on intracellular ion dynamics. For example a 10% underestimate of C_m (pF) leads to an 11% overestimate of current density (pA/pF), a 4% underestimate of cell volume in rat ventricular cardiomyocyte 23, and therefore an analogous overestimate of intracellular ion concentration changes. Cell volume and intracellular ion concentration changes will be further underestimated (up to 10%) in species like rabbit, where volume/surface relation is steeper 23.

The correlation, found in the four ventricular models and the one atrial model analyzed, between the slope of the relative error in C_m estimate and the slope of R_m(ΔV_m) (Fig. 3, bottom panel), furthermore emphasizes the generality of the principle: the input resistance of resting cardiomyocytes rectifies in the subthreshold voltage range under study, and such rectification brings about a proportional error in C_m estimate. This proportionality is fully satisfied by our experimental data, as shown by their position on the correlation line of Fig. 3 (bottom panel). Interesting to note is the failure of the PD01 model to exactly reproduce rat data in this correlation plot, most likely due to the particular I_K1 equation chosen in the original work 30, where, on the other hand, the authors themselves recognize the nonuniform properties of I_K1 across the ventricle as one of the potential limitations of their mathematical description (see also Ref. 41.

We call the model described in Figure 4B “mechanistic” because it is based only on properties of an experimental observable (R_m), independently on its biological determinants (ion channels, pumps, exchangers, etc.). This simple model shows that the property of R_m to vary parabolically with V_m is enough, by itself, to qualitatively explain voltage and current asymmetry in response to symmetric current-/voltage-clamp protocols. It also accounts for the voltage-dependency of C_m estimate in vivo and in the mathematical model cells. Finally it shows that, even though mono-exponential functions fit quite well the solutions of Eq. (2) and Eq. (5) for the non-Ohmic circuit, nevertheless they are solutions of a wrong model of the cell electrical properties and they lead therefore to wrong C_m estimates. If, instead, numerical solutions of Eq. (2) and Eq. (5) including the R_m(ΔV_m) relation are used, the C_m measure becomes independent from V_m and gives estimates that are only a little dispersed around a mean value (Fig. 5), as can also be appreciated in the cell model simulations (e.g., Fig. 6). A further validation of the consistency of the least-square algorithm (see Appendix ) to measure C_m comes from the experiment in Fig. 5. A mechanistic mathematical model of a real rat ventricular myocyte, built with experimentally derived R_m(ΔV_m), R_s, and a constant C_m calculated with the least-square protocol on the same cell, was challenged with CC and VC constant steps and analyzed in terms of mono-exponentials. The fact that the resulting C_m values fit well those derived with the Ohmic assumption on the real cell, demonstrates that the voltage-dependency of R_m is enough to explain the error in C_m estimate, and no other effects like leaky seals, additional uncompensated stray capacitances, series resistances, or different equivalent circuits 11, in principle, need to be considered.

It could be argued that the least-square algorithm presented in this study as a tool to calculate C_m in a voltage-independent manner is too complicated and demanding, especially for what concerns off-line numerical processing of the data, making it therefore not suitable for the general purpose of monitoring passive electrical properties, which is often ancillary in most cardiac cellular electrophysiological studies. To answer this concern, we propose a procedure where an algebraic property of the C_m calculation with the Ohmic assumption is used to extrapolate the real value of C_m without need of any further experimental or mathematical work, as summarized in Fig. 9. Some authors use protocols where symmetrical (depolarizing and hyperpolarizing) VC steps are imposed to the membrane and C_m averaged from the analysis of the two opposite current transients (e.g., Ref. 15. It should be noted that this procedure is quite different from the one we are suggesting here and can also lead to mis-estimate of C_m, because C_m error in the Ohmic assumption is not symmetrical with respect to V_r (see curve c in Figure 8F).

The typical first-order linear differential equation generating mono-exponential functions as solutions is of the type

(8)

In the case of Eq. (5) of VC, this can be conveniently rewritten as

(9)

This explains qualitatively why, assuming the same voltage-dependency of R_m, VC-derived current traces deviate less from mono-exponentials than CC-derived voltage displacements. In other words, provided R_s is kept much lower than R_m, VC gives best estimates of C_m, as we demonstrated in this study with both experimental and numerical approaches.

A limitation of this work is that, although the very general mechanism of subthreshold R_m rectification is presented with its implications on C_m estimate, the relative role of the ionic mechanisms underlying this property have not been unambiguously defined in vivo nor in silico. Although, as in previous studies 26,27, a primary contribution of I_K1 is strongly suggested, additional experimental and numerical work will help to better define the possible role of sodium-calcium exchanger or other ionic mechanisms in R_m rectification. Also, we do not report here on any intervention made to modify the different ionic mechanisms of R_m voltage-dependency. Work in this direction will be required, for example, to investigate whether there are conditions where the membrane of resting cardiomyocytes behaves Ohmically, helping to better explain its subthreshold electrical properties in physiological and pathological states.

This study emphasizes that attention should be paid to the subthreshold voltage-dependency of R_m if C_m is to be measured with classical DC protocols. Large errors always arise from the CC approach, but VC can also be substantially affected. Accurate measures can be achieved by taking R_m(ΔV_m) into account in the analysis of CC and VC experimental results or by ignoring this voltage-dependency, and extrapolating results obtained from at least two different ΔV_m values to ΔV_m=0.

It is straightforward to hypothesize that the voltage-dependency of R_m would also play a role in other techniques to measure C_m, especially those based on complex impedance analysis, widely adopted in the literature and based on the AC study of the same Eq. (2) and Eq. (5).