Patch Clamp for Peptide Ion Channel Modulators

A peptide that inhibits a voltage-gated channel can act at a site inside the conduction pathway or on a voltage sensor, from the extracellular face or from the cytoplasmic one, and the recording configuration you choose decides which of those possibilities you are even able to test. The four configurations of the giga-seal method differ in exactly the respect that matters here: which side of the bilayer your perfusion solution reaches. Pipette-membrane seals of 10⁹–10¹¹ Ω isolate the patch electrically and physically, and excision yields either an inside-out patch with its cytoplasmic membrane face exposed to the bath or an outside-out patch with its extracellular face exposed ^[1]. Whole-cell recording additionally dialyses the cell interior with the pipette solution, and single-channel resolution is achievable in whole-cell mode when the cell diameter is under about 20 µm ^[1]. Pair the peptide with the wrong configuration and you will record a clean, well-controlled null.

Sidedness is a design decision, not a detail

In cell-attached mode the patch beneath the pipette is sealed off from the bath, so a bath-applied peptide does not contact the channels being recorded; to reach them it has to be in the pipette solution ^[1]. If bath application changes the behaviour of channels in a patch the peptide cannot physically reach, the effect is being carried by something diffusible on the cytoplasmic side rather than by direct contact at the recorded membrane.

For a peptide with an intracellular site, the demonstration has to present it to the cytoplasmic face. The canonical case is the synthetic peptide corresponding to the first 20 residues of the ShB splice variant of Shaker: internal application to non-inactivating mutant channels expressed in Xenopus oocytes restored inactivation in a concentration-dependent manner, and the peptide-induced inactivation was not noticeably voltage-dependent ^[2]. Trypsin-treated peptide and peptides with sequences derived from the first 20 residues of non-inactivating mutants did not restore inactivation ^[2]. Those two negative controls are what separate a specific occluding interaction from a nonspecific one, and both require the same internal-access configuration as the positive result.

The mirror case is a peptide confined to the outer face. Charybdotoxin blocks K⁺ conduction through high-conductance Ca²⁺-activated K⁺ channels by binding to the external side of the channel, with an apparent dissociation constant of approximately 10 nM at physiological ionic strength ^[3]. That measurement was made on single channels from rat skeletal muscle plasma membranes inserted into uncharged planar phospholipid bilayers ^[3], not in a patch; a bilayer gives the same independent control of both faces that an excised patch does, so the design logic transfers even though the preparation does not.

What a voltage protocol has to contain

A single test pulse from a single holding potential cannot distinguish the two ways a peptide can reduce peak current. Hanatoxin inhibits the drk1 voltage-gated K⁺ channel by shifting channel opening to more depolarized voltages, and channels opened by strong depolarization in the presence of the toxin deactivate much faster upon repolarization, indicating that toxin-bound channels can open ^[4]. The conclusion drawn was that hanatoxin inhibits the channel not by physically occluding the ion conduction pore but by modifying channel gating ^[4]. Reaching that conclusion required conductance-voltage relations across a range of depolarizations plus tail-current kinetics on repolarization. Neither is recoverable from a peak-current-versus-concentration series.

The same phenotype appears on sodium channels. ProTx-I and ProTx-II shift the voltage dependence of activation of Na_V1.5 to more positive voltages, and both inhibit multiple Na_V subtypes with IC₅₀ values below 100 nM; ProTx-I additionally inhibits K_V2.1 with 10-fold less potency than on Na_V channels ^[5]. Promiscuity across channels within and across families is a described hallmark of this structural class ^[5]: a shift measured on one subtype constrains nothing about the others.

Assigning a shift to a particular voltage sensor takes mutagenesis, not protocols. ProTx-II and huwentoxin-IV had both been proposed to block hNa_V1.7 activation by trapping the domain II voltage sensor in the resting configuration, but residues in the domain II paddle contribute very unequally between them: the mutation E818C increases the IC₅₀ for block of hNa_V1.7 currents 4-fold for ProTx-II and 400-fold for huwentoxin-IV, while F813G decreases ProTx-II affinity 9-fold and has no effect on huwentoxin-IV affinity ^[6]. ProTx-II, but not huwentoxin-IV, was also found to impede fast inactivation, an effect attributed to trapping the domain IV voltage sensor in the resting configuration ^[6]. The authors present this against what had been proposed as a guiding principle of sodium channel pharmacology, concluding that molecules do not have to exclusively target the domain IV voltage sensor to influence sodium channel inactivation ^[6]. Two peptides producing the same macroscopic phenotype can therefore be engaging different determinants.

Use-dependence, and what it is not

Single-channel charybdotoxin data separate the state-dependence and voltage-dependence that a pulse train would otherwise confound in a macroscopic measurement. The dwell-time distributions of both blocked and unblocked states are single-exponential, the association rate varies linearly with toxin concentration, and the dissociation rate is independent of it, so binding is a bimolecular process ^[3]. The toxin is competent to block both open and closed channels, with association sevenfold faster for the open channel and the same dissociation rate for both conformations; membrane depolarization enhances the dissociation rate e-fold per 28 mV, and if open probability is held constant as voltage varies, the association rate is voltage independent ^[3].

Read that as a protocol warning. A stimulus train that holds the membrane depolarized for a larger fraction of each cycle drives more unbinding for a blocker with voltage-dependent dissociation, so the apparent IC₅₀ can move with the duty cycle rather than with anything about the peptide. Current decline over a pulse train is a composite of state-dependent association, voltage-dependent dissociation, and whatever accumulation of inactivation the protocol itself produces. Attributing it to state-dependent binding alone requires that those terms have been separated, which is what varying pulse duration, interpulse interval and holding potential independently is for. The charybdotoxin rate constants above came from bilayers where open probability was under direct experimental control ^[3]; whole-cell recording offers no such handle, so the separation has to be built out of protocol design instead.

Series resistance and the shift you did not cause

For a gating modifier the readout is a voltage, which makes series resistance error an error in the reported quantity rather than a second-order concern. Series resistance can be determined in the amplifier's current-clamp mode with an active bridge circuit adjusted by a phase-sensitive method: short sine-wave currents at 3–7 kHz are injected, complete capacitance neutralization is indicated by disappearance of the phase lag between current and voltage, and correct bridge balance by a minimized voltage response. The value obtained then supports compensation in voltage-clamp mode, with an accuracy confirmed to 2–3% on single-compartment model cells and compensation of up to 80–90% achievable even in cells with low input resistance ^[7]. That study also showed that significant series resistances directly affect the determination of membrane conductances as well as the kinetic properties of small-amplitude spontaneous synaptic currents ^[7].

The arithmetic correction applied afterwards deserves more suspicion than the compensation does. Dual patch-clamp recordings from brainstem motoneurons of adult frogs, one electrode voltage-clamping K⁺ currents and the other reading membrane potential directly, found voltage errors averaging under 5 mV for currents of about 7–13 nA and under 10 mV for currents of 25–30 nA; in most cases Ohm's-law-based correction overpredicted the measured errors by roughly 2.5-fold, and using it to correct led to erroneous current-voltage relationships, with the greatest distortion for inactivating currents ^[8]. Recordings with low electrode series resistance moderately compensated by the amplifier circuitry appeared to have smaller voltage errors than those with larger series resistance and high compensation, despite the same effective value ^[8]. That is one preparation and one current type, and it is not a licence to stop compensating; it is a reason to prefer a low-resistance electrode over aggressive post hoc arithmetic when the number you intend to defend is a half-activation voltage.

The concentration you actually applied

Peptide channel modulators are often worked at concentrations where the total mass in the fluid path is small enough that surface losses are material. ProTx-II inhibits Na_V1.7 with an IC₅₀ of 0.3 nM, against IC₅₀ values of 30 to 150 nM for other heterologously expressed Na_V1 subtypes, and that subtype selectivity was abolished by a point mutation in DIIS3 ^[9]. A selectivity ratio spanning that range is only as trustworthy as the delivered concentrations at both ends of it.

Adsorptive loss has been measured for peptides in the relevant range: recovery of three different peptide solutions at 1.5–12 nM was enhanced on plasma-treated ultra-low-binding microplates relative to standard polypropylene microplates ^[10]. Within automated patch clamp itself the labware effect has been characterised mostly with small molecules. Up to 27 compounds were tested across three instruments; on one of them, the IC₅₀ values of 21 compounds correlated well with published values regardless of voltage protocol or labware when the compounds were used immediately after preparation, but the potency of astemizole decreased when it was prepared in Teflon or polypropylene compound plates two to three hours before the experiment; the recommendation drawn was caution with Teflon, polystyrene or polypropylene vessels if experiments are not performed immediately after preparation ^[11]. Those are small molecules on hERG and the magnitudes do not transfer to peptides. What transfers is that the interval between preparing a solution and delivering it is an experimental variable, not a logistics detail.

There is also a sink that is not labware. ProTx-II membrane binding affinity correlates directly with its potency as an hNa_V1.7 channel inhibitor, in a model where a hydrophobic patch on the peptide surface anchors it at the cell surface in a position that optimises interaction with the binding site on the voltage sensor domain ^[12]. Peptide partitioned into lipid is peptide that is not in free aqueous solution. A nominal bath concentration is best treated as an upper bound on the free concentration at the membrane unless something independent of the electrophysiology has verified it.

Slow wash-off and apparent irreversibility

Failure to recover on wash is weak evidence for irreversible binding. Because the interaction of ProTx-II with lipid membranes is reported as a prerequisite for its channel inhibition ^[12], a peptide of this class sits partly in the bilayer, and exchanging the bath does not by itself remove it. Incomplete recovery is therefore compatible with ordinary reversible binding supplied from a membrane-associated pool, and does not on its own establish a covalent or otherwise irreversible step.

Onset carries the same ambiguity in reverse. In the same multi-site automated patch clamp benchmarking study, slow-acting small molecules on hERG required extended incubation of at least 6 min to reach steady state and therefore stable IC₅₀ values, and a minimum incubation time of 5 min was recommended ^[11]. Reading potency before steady state understates it; declaring irreversibility before the wash has cleared the compartment overstates it. Both are addressed by measuring the time course rather than assuming it.

Where dissociation is voltage-dependent there is a lever. Depolarization enhances the charybdotoxin dissociation rate e-fold per 28 mV ^[3], so for a blocker with that property a wash protocol that holds the membrane depolarized can accelerate recovery. It will do nothing for a peptide whose unbinding is voltage-independent, and that difference is itself informative.

Reading the signature: gating modifier versus pore blocker

The observations worth collecting, roughly in the order they discriminate:

• Conductance-voltage relation. A shift of activation towards more depolarized potentials points towards gating modification: hanatoxin shifts channel opening to more depolarized voltages ^[4], and ProTx-I and ProTx-II shift the voltage dependence of Na_V1.5 activation to more positive voltages ^[5].
• Tail currents. Channels opened by strong depolarization in the presence of hanatoxin deactivate much faster upon repolarization, showing that toxin-bound channels can open ^[4]. A peptide occluding the conduction pathway cannot produce that observation.
• Single-channel kinetics. Charybdotoxin produces discrete blocked and unblocked dwell times, each single-exponentially distributed, with association rate linear in concentration and dissociation rate independent of it ^[3].
• Competition. The pore blocker agitoxin-2 and the gating modifier hanatoxin can bind simultaneously to individual K⁺ channels ^[13]. Non-competition between a candidate peptide and a pore blocker is therefore not, on its own, evidence that the peptide lacks a site.
• Ionic strength. Increasing external solution ionic strength from 20 to 300 mM reduced the charybdotoxin association rate by two orders of magnitude with little effect on the dissociation rate ^[3]. For a basic peptide, buffer composition is a potency variable and has to be held constant across a comparison.

What a shift does not do is locate a binding site. Alanine scanning of the region extending from the C-terminal end of S3 through S4 identified residues that likely contribute to the hanatoxin sites, which were placed near the outer edges of S3 and S4, eccentrically located at least 15 Å from the central pore axis ^[13]. Separately, the concentration dependence of equilibrium occupancy together with the kinetics of onset and recovery indicated that multiple hanatoxin molecules can bind simultaneously to a single channel, consistent with a model of four equivalent sites ^[4]. Stoichiometry above one also means a Hill fit to a macroscopic inhibition curve is not reporting a simple one-to-one affinity.

Automated planar patch versus manual

High-throughput instruments with gigaohm seal data quality are relatively recent, and experiments screening a large number of compounds are described as ideally suited to them, particularly when the amount of compound available is low ^[14] — a condition that often applies to peptide work. Thirty-six arthropod venoms were tested against Na_V1.9 on a single 384-well plate, and the same platform class supports voltage- and current-clamp recording from primary and stem-cell-derived cells and offers intracellular solution exchange, which allowed effects of intracellular Ca²⁺ on TRPC5 and cAMP on HCN2 to be investigated without excised patches ^[14]. That last capability partly answers the sidedness problem on a format that has no excised-patch configuration.

The cost sits in what the platform does to obtain the seal. Fluoride has been used in the internal recording solution for decades because it improves seal resistance and promotes longer-lasting recordings, and because a planar aperture cannot be repositioned the way a pipette can, automated platforms use internal fluoride to raise the proportion of gigaohm seals; internal fluoride can, however, affect the voltage dependence of activation and inactivation as well as internal second messenger systems ^[15]. For a gating modifier, whose whole readout is the position of those voltage dependences, that is not a nuisance parameter. A fluoride-free approach on a 384-well system reached gigaohm seal success rates above 40%, demonstrated with hERG expressed in HEK cells and with Na_V1.5, Na_V1.7 and K_Ca3.1 expressed in CHO cells ^[15].

The recombinant background is part of the measurement too. In patch-clamp recordings, ProTx-II slowed the rate of Na_V1.7 activation while OD1 reduced the rate of fast inactivation and accelerated recovery from inactivation; the β3 subunit partially abrogated these effects, and the hyperpolarising shift in the half-activation voltage of steady-state activation that OD1 induced was not observed in the presence of β3 ^[16]. A potency or a shift measured in a line expressing the pore-forming subunit alone is a statement about that construct, not about the channel in any other assembly.

Keeping the claim inside the measurement

A whole-cell IC₅₀ obtained at one holding potential licenses the statement that the peptide reduced current under those conditions, and nothing about mechanism. A depolarizing shift of the conductance-voltage relation with faster tail deactivation licenses the statement that the data are consistent with gating modification, and still says nothing about where the peptide binds. Sidedness follows from the configuration used, not from the potency measured. Each of those statements also carries an implicit qualifier — at the free concentration actually present at the membrane — which is the term you are least likely to have measured directly.