High spatial density is associated with non-conducting Kv channels from two families

Abstract

Ion channels are well known for their ability to regulate the cell membrane potential. However, many ion channels also have functions that do not involve ion conductance. Kv2 channels are one family of ion channels whose non-conducting functions are central to mammalian cell physiology. Kv2.1 and Kv2.2 channels form stable contact sites between the endoplasmic reticulum and plasma membrane via an interaction with endoplasmic reticulum resident proteins. To perform this structural role, Kv2 channels are expressed at extremely high densities on the plasma membranes of many cell types, including central pyramidal neurons, α-motoneurons, and smooth muscle cells. Research from our lab and others has shown that the majority of these plasma membrane Kv2.1 channels do not conduct potassium in response to depolarization. The mechanism of this channel silencing is unknown but is thought to be dependent on channel density in the membrane. Furthermore, the prevalence of a non-conducting population of Kv2.2 channels has not been directly tested. In this work we make improved measurements of the numbers of conducting and non-conducting Kv2.1 channels expressed in HEK293 cells and expand the investigation of non-conducting channels to three additional Kv α-subunits: Kv2.2, Kv1.4, and Kv1.5. By comparing the numbers of gating and conducting channels in individual HEK293 cells, we found that on average, only 50% of both Kv2.1 and Kv2.2 channels conducted potassium and, as previously suggested, that fraction decreased with increased channel density in the plasma membrane. At the highest spatial densities tested, which are comparable with those found at Kv2 clusters in situ, only 20% of Kv2.1 and Kv2.2 channels conducted potassium. We also show for the first time that Kv1.4 and Kv1.5 exhibit density-dependent silencing, suggesting that this phenomenon has an underlying mechanism that is shared by Kv channels from multiple families.

Significance

Ion flux through potassium channel proteins controls electrical activity in most cells, but these proteins have non-conducting functions as well. At least 50% of Kv2.1 channels do not conduct ions in response to depolarization, but the mechanism and downstream effects of this silencing is unknown. We quantified channel silencing by comparing channel molecule numbers with whole-cell gating currents and conductance and showed that it is not an exclusive property of Kv2.1 channels. We found that Kv channels from the Kv1 family are silenced at high membrane densities, suggesting that the molecular mechanism is more common than previously thought and could exist to ensure that Kv channel overexpression does not electrically silence excitable cells.

Introduction

Kv2.1 is a voltage-gated potassium channel abundantly expressed in many tissue types, including brain, muscle, and pancreas. Ion flux through Kv2.1 channels controls neuronal excitability, especially under conditions of high frequency firing (1, 2, 3). However, Kv2.1 also has a non-conducting function that is intimately related to its plasma membrane (PM) expression and localization. Namely, Kv2.1 channels aggregate into dense clusters on the PM and form endoplasmic reticulum (ER)-PM junctions via direct interaction with ER-resident Vamp-associated proteins (VAPs). This interaction is regulated by the phosphorylation status of a 7-amino-acid motif in the Kv2.1 C-terminus, which imparts dynamic control over the formation of ER-PM junctions (4,5). Although we do not fully understand the molecular events downstream of Kv2-induced ER-PM junction formation, we know that these junctions are important regulatory domains for diverse processes such as endocytosis and exocytosis (6, 7, 8), Ca²⁺ channel localization and function (9, 10, 11, 12), PM lipid composition (4,13), and cell-to-cell communication (14,15).

Our lab, and others, have also shown that the majority of Kv2.1 channels expressed in the PM of rat hippocampal neurons and vascular myocytes do not conduct potassium in response to depolarization (12,16, 17, 18). Early on it was thought that clustered Kv2.1 channels in ER-PM junctions were silent while non-clustered channels carried all Kv2.1 current (19). However, we now know that there are many more non-conducting channels than clustered channels and instead, high spatial densities of Kv2.1 channels in the PM appear to favor the non-conducting state of the channel, whether clustered in ER-PM junctions or not (16,17). To date no stimulus has been identified that alters the fraction of conducting Kv2.1 channels, and therefore the molecular mechanism of silencing remains a mystery. As Kv channel conductance regulates many cellular processes, from neuronal excitability to apoptosis, understanding the elusive silencing mechanism of Kv2.1 channels will improve our understanding of these processes.

Previous investigations of Kv2.1 suggest that non-conducting channels have functional voltage-sensing machinery (12,16,18). However, the evidence to support this claim comes from separate measurements of either Kv2.1 gating charge and conductance under different recording conditions (12,16) or semi-quantitative observations (18). In this work we set out to improve quantitative estimates of conducting and non-conducting Kv2.1 channels by making measurements of gating charge and conductance from individual cells in a single voltage-clamp trace. In parallel, we used single-molecule imaging to estimate the number of fluorescent channel proteins in the PM of the cells we had recorded from. To our knowledge, this is the first attempt to quantify gating charge, ionic conductance, and protein number from individual cells. We used the same approach to expand the investigation of non-conducting channels to additional Kv isoforms.

Our results indicate that on average, 50% of Kv2.1 channels expressed in the PM of HEK293 cells are non-conducting. This percentage was a function of channel density in the membrane, and at the highest expression levels 80% of channels were non-conducting. We found that like Kv2.1, Kv2.2, and Kv1.5 had significant populations of non-conducting channels at physiological densities. Surprisingly, the fraction of non-conducting channels was a function of channel density for all four channel types studied (Kv2.1, Kv2.2, Kv1.5, and Kv1.4), suggesting that a common density-dependent silencing mechanism regulates the conductance of these channels from two major families. Together these results support the idea that diverse Kv channels are sensitive to their density in the PM. This finding has implications for the function of cells that express high densities of channels or that localize channels to dense microdomains.

Materials and methods

DNA constructs

Plasmid DNA containing the coding sequence for GFP-Kv2.1 (rat), GFP-Kv2.2 (rat), GFP-Kv1.4-BAD (rat), or YFP-Kv1.5 (human) was transiently transfected into HEK293 cells as described previously (4,6,20, 21, 22). The N-terminal GFP fusion to Kv1.4 interferes with N-type inactivation, resulting in a slowly inactivating current (23). The Kv1.4 construct differs further from the wild-type channel in that it contains a biotin acceptor domain encoded in the first extracellular loop, which does not alter channel activation or inactivation (6,21).

Cell culture and transfection

HEK293 (ATCC, Manassas, VA; CRL-#1573) cells were maintained in 10 cm cell-culture dishes (CellTreat, Pepperell, MA; #229620) at 37°C under 5% CO₂ in Dulbecco’s modified Eagle’s medium (DMEM) (Corning, Tewksbury, MA; #10-013-CV) supplemented with 10% fetal bovine serum (FBS). Cells were transfected with plasmid DNA (200 ng to 2 μg) via electroporation (BioRad, Hercules, CA; GenePulse Xcell) in a 0.2 cm gap cuvette (Bulldog Biosciences, Portsmouth, NH; #12358-346) and returned to new 10 cm dishes containing DMEM + 10% FBS. After 12–18 h, transfected cells were trypsinized and plated at low density on glass-bottom dishes (MatTek, Ashland, MA; #P35G-1.5-14-C) coated with Matrigel (BD Biosciences, San Jose, CA; #354230) for use in electrophysiological or imaging experiments over the course of a single day.

Electrophysiology

For simultaneous gating and ionic current recordings, transfected HEK293 cells plated at low density on glass-bottom dishes were bathed in High K⁺ external recording solution ( 150 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 10 mM glucose, and 10 mM HEPES; pH 7.37, 340 ± 3 mOsm/L). Patch pipettes were fabricated from thin-walled borosilicate glass with a Sutter P-97 puller. Tip resistances were in the range of 1–3 MΩ when filled with Low K⁺ intracellular recording solution (30 mM KCl, 115 mM NMDG-Cl, 1 mM MgCl₂, 0.5 mM EGTA, 10 mM HEPES, and 2 mM ATP-Mg₂; pH 7.35, 310 ± 3 mOsm/L), resulting in a theoretical potassium equilibrium potential (E_K) of 41.2 mV. Used together, we called this solution set Flipped K⁺, owing to the inverted driving force on potassium. The liquid junction potential between these Flipped K⁺ solutions was measured on each recording day (9.5 ± 1.3 mV, N = 23), and command voltages were corrected accordingly offline.

For noise fluctuation analysis, transfected cells were bathed in either High K⁺ external solution (described above) or physiological external solution containing 140 mM NaCl, 5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 10 mM glucose, and 10 mM HEPES; pH 7.4, 340 ± 3 mOsm/L. The internal solutions varied to facilitate measurements of unitary channel properties in different potassium concentrations. The physiological and symmetrical High K⁺ internal solutions were the same and contained 150 mM KCl, 4 mM NaCl, 1 mM MgCl₂, 0.5 mM EGTA, and 10 mM HEPES; pH 7.4, 310 ± 3 mOsm/L. The internal solution used in the Flipped K⁺ condition was as described above.

Whole-cell voltage-clamp experiments were carried out at room temperature using an Axopatch200b amplifier and a Digidata1550A digitizer using pClamp10.6 software (Molecular Devices, San Jose, CA). Currents were compensated for whole-cell capacitance and series resistance (>85%), digitized at 33 kHz, and low-pass filtered at 5 kHz. Online leak subtraction was applied using the P/-2 (gating assay) or P/-4 (noise fluctuations) method prior to data acquisition. Raw traces were inspected by eye to ensure data quality.

Gating and ionic current recordings

Transfected cells were identified by GFP or YFP fluorescence as described in TIRF microscopy. Gigaohm seals were formed at 0 mV with light suction. After break-in, cells were held at −90 mV until the pipette solution exchanged with cytoplasm. Full solution exchange occurred over the first 2 min after break-in as evidenced by a stabilization of the K⁺ reversal potential near +40 mV in preliminary experiments. Leaky seals were common after break-in with these solutions and cells with more than 1 pA/mV holding current after this time were not used further, ensuring that all cells had seal resistances ≥1 GΩ. Cells were depolarized from −90 mV to E_K for 20–40 ms to activate all available voltage sensors without eliciting significant ionic current. After the step to E_K, four short voltage steps near the reversal potential were applied to evoke small ionic currents, which were plotted offline to generate a truncated membrane current/voltage (I/V) relationship (see Fig. 1 E for graphical protocol). The area under the on-gating current observed at E_K was used to calculate the charge moved during the depolarizing step, and the slope of the I/V relationship around E_K was used to calculate membrane conductance. Membrane current was ohmic in this small voltage window. The truncated I/V relationship was also used to confirm that the reversal potential of each cell was near the theoretical E_K. All cells had an observed E_K within ±5 mV of the theoretical E_K, ensuring voltage control in cells with membrane conductance up to 1400 nS. The same electrophysiological conditions did not elicit detectable currents in untransfected HEK293 cells, which were patched routinely over the course of the study. Although HEK293 cells do express some voltage-gated potassium channels, the magnitude of their conductance was negligible in our recording configuration.

Figure 1.

Recording Kv2.1 gating and ionic currents. (A) Representative whole-cell currents from a HEK293 cell expressing GFP-Kv2.1 bathed in physiological K⁺ gradient (5 mM K⁺_out/150 mM K⁺_in) in response to 250 ms depolarizing steps to the voltages listed on the right. Tail currents were recorded at −40 mV at the end of the current traces. (B) Representative currents from a HEK293 cell expressing GFP-Kv2.1 bathed in Flipped K⁺ conditions (150 mM K⁺_out/30 mM K⁺_in) in response to the same 250 ms steps as in (A). Tail currents were again recorded at −40 mV and are large due to the increased driving force at negative potentials in these ionic conditions. (C) Steady-state current-voltage plots from recordings like those in (A) and (B). Solid symbols are from GFP-Kv2.1 in a physiological K⁺ gradient, and open symbols are from GFP-Kv2.1 in Flipped K⁺. (D) Activation curves generated from tail currents like those recorded in (A) and (B). Tail current magnitudes were normalized to their maximum value. Solid symbols represent the activation of Kv2.1 in physiological K⁺ (midpoint = −9.1 ± 2.2 mV, N = 4) while open symbols represent channel activation in Flipped K⁺ (midpoint = −12.5 ± 0.5 mV, p = 0.26, N = 3). Error bars represent standard error. (E) Top: voltage-clamp protocol used to elicit gating and ionic current in Flipped K⁺. Black trace: response from an untransfected HEK293 cell to the protocol. No currents were observed in these cells. Blue trace: current response to the protocol from a HEK293 cell expressing GFP-Kv2.1 bathed in Flipped K⁺. On-gating current was evident at the onset of the step to +40 mV (dashed box) and square ionic currents were seen in response to the 1 mV steps at the end of the protocol (shaded box).

Mitigation of electrophysiological error

Non-ideal voltage-clamp conditions can lead to inaccurate measurements of membrane conductance, and it is of paramount importance that these potential errors are taken into consideration when attempting to draw quantitative conclusions. We made several accommodations to ensure the best possible clamping conditions in this work. First, we minimized the tip resistance of our electrodes to between 1 and 3 MΩ and we only recorded from cells with less than 7 MΩ series resistance after break-in. In addition, we used at least 85% series resistance compensation for every recording so that our effective series resistance was at most 1 MΩ. Because we measured ionic currents near the reversal potential, their magnitudes were relatively small (<3 nA), thereby reducing the potential for voltage error in our experiments. Nonetheless, we monitored the reversal potential of each cell to ensure we maintained voltage control, and cells with observed E_K more than 5 mV away from the theoretical E_K were discarded. We applied this cutoff even though we believe the major source of error in E_K was due to slight deviations in the liquid junction potential from recording to recording, as opposed to voltage error. We believe this for two reasons. First, measured liquid junction potential often differed by ±3 mV on different recording days, which was similar to the magnitude of measured differences in reversal potential. Second, there was no relationship between reversal potential and membrane conductance, as would be expected if voltage error were the major determinant of the measured value of reversal potential.

Some cells had uncompensated capacitive spikes remaining despite the use of capacitance compensation (see Fig. 1 E, insets). These signals were much faster than the signals of interest from channel gating, which were only observed in transfected cells. We found that uncompensated membrane capacitive spikes were over within a fraction of a millisecond. We ignored currents between the onset of the depolarizing step and 0.24 ms in order to minimize the inclusion of capacitive transients while maximizing the detection of gating charge. However, we also analyzed gating currents starting from 0.36 ms and found only a 2.7% reduction in the overall detected gating charge. This suggested that the original analysis not only avoided capacitive artifacts but also captured the maximum gating charge.

Non-stationary noise fluctuation analysis

We performed non-stationary noise fluctuation analysis to estimate the single-channel conductance and open probability of each Kv channel under the Flipped K⁺ conditions. For Kv2.1 and Kv2.2 we repeatedly depolarized cells to +40 mV, then repolarized to −55 mV and analyzed the noise fluctuations in the deactivating tail currents. We performed similar experiments with Kv1.4 and Kv1.5 to determine their open probabilities at +40 mV. Estimates of Kv1.4 and Kv1.5 single-channel conductance were made by measuring noise fluctuations at −20 mV or +20 mV, respectively. See supporting material for more details on estimating channel unitary conductance. For all channels under all conditions, we collected between 50 and 300 consecutive sweeps from each cell. Data were inspected by eye for quality, and traces with sources of noise not due to channel gating were discarded (e.g., sudden changes in leak, oscillations, or seal deterioration). The remaining traces showed variable amounts of channel rundown and therefore were analyzed following the well-accepted subtraction method (24, 25, 26) as described below.

The average current, $\overline{I}\left(t\right)$ during either the activation or deactivation phase of each experiment was calculated from some number (M) of sweeps:

$$\overline{I}\left(t\right)=\frac{1}{M}\sum _{m=1}^M{I}_m\left(t\right).$$ (1)

We obtained the difference records, $y_M\left(t\right)$, by subtracting subsequent sweeps from one another:

$$y_m\left(t\right)=\frac{y_m-y_{m-1}}{2}.$$ (2)

Non-stationary noise was calculated as the variance in the difference records according to

$${\sigma }_y^2=\frac{2}{M-1}\sum _{m=1}^M{(y_m\left(t\right)-\overline{Y_m}\left(t\right))}^2.$$ (3)

The variance in the difference records was plotted as a function of the average current during the recording and fitted with a second-order polynomial with offset, k:

$${\sigma }_y^2\left(I\right)=i\times \overline{I}-\frac{1}{N}\times \overline{I^2}+k.$$ (4)

The single-channel current ($i$) was obtained directly from this fit and converted to single-channel conductance ($g$) using

$$i=g\ast \left(V_m-E_{\text{K}}\right),$$ (5)

where E_K is the theoretical reversal potential in each solution pair (Physiological = −87 mV, Symmetrical High K⁺ = 0 mV, Flipped K⁺ = 41.2 mV). In experiments using the Flipped K⁺ solutions, the single-channel conductance obtained from noise analysis was adjusted by a rectification index (see Figs. 4 A and S3) to estimate the single-channel conductance near +40 mV.

Figure 4.

Kv2.1 unitary conductance measured from noise fluctuations in Flipped K⁺. (A) Open channel I/V curve from GFP-Kv2.1 expressed in HEK293 cells bathed in Flipped K⁺ revealed inward rectification due to the strong inward gradient on potassium. The plotted current is the normalized instantaneous current at the given voltage after activating channels at +50 mV for 100 ms. Values were normalized to the current at +80 mV to allow for averaging (N = 6). The slope of the I/V at +40 mV (orange dashed line) was 39% that of the slope at −55 mV (blue dashed line). (B) Example tail current traces used for noise fluctuation analysis in Flipped K⁺. Channels were opened at +40 mV (not shown) and tails were recorded at −55 mV. (C) Current variance versus average current plot for the data shown in (B). The data were fit with Eq. 4. Single-channel current is given by the coefficient of the first term of the polynomial and was −1.6 ± 0.01 pA for this cell. (D) Absolute values of Kv2.1 single-channel conductance obtained from parabolas like that in (C) by converting single-channel current to conductance. At −55 mV the average single-channel conductance was 17.7 ± 0.90 pS (blue). After correcting for rectification, the average single-channel conductance at +40 mV was 7.44 ± 0.36 pS (orange). N = 9. Kv2.1 open probability measured from noise fluctuations was 0.70 ± 0.03.

TIRF microscopy

Total internal reflection fluorescence (TIRF) images were acquired on a Nikon Eclipse Ti fluorescence microscope equipped with Nikon Perfect Focus and 100 mW diode lasers at 405, 488, 561, and 640 nm. Light was collected using a 100× PlanApo 1.49 NA TIRF objective in conjunction with appropriate band-pass filters housed in a Sutter Lambda 10-3 filter wheel. Images were acquired with an Andor iXon DU-897 EMCCD camera (512 × 512). Cells expressing Kv α-subunits were identified by GFP or YFP fluorescence.

Fluorescent channel counting

HEK293 cells were transfected with either 200 ng or 2 μg of plasmid DNA encoding one of the four Kv α-subunits. Cells expressing the two amounts of DNA were mixed after transfection, plated on glass-bottom dishes, and bathed in external solution. First, each dish was searched for cells expressing low enough levels of transfected protein to resolve single channels. These cells were imaged at 20 Hz for 10 s with 15% 488 laser power for offline analysis of single-channel fluorescence. Offline, single channels were identified by their mobility (6,27) and selected by hand with 3 × 3 pixel regions of interest (ROIs). The sum fluorescence of each ROI was measured to build a distribution of single-channel fluorescence values. The measurements were made from the first frame of the time series to reduce effects of photobleaching. The average unitary fluorescence (F₀) in each dish was calculated from the distribution of single-channel fluorescence values from 3–10 cells per dish. F₀ was used to estimate the number of fluorescent channels from brighter cells in the same dish (those yielding gating currents) as outlined below.

In the same dish, brighter cells were identified that were likely to contain enough channels to resolve gating currents from electrophysiological noise. Static images of these cells were acquired with 2% 488 laser power under the same TIRF illumination conditions used to acquire movies of single channels. A reduced laser power was needed to image these cells to prevent oversaturation of camera pixels. Offline, the fluorescence from the entire cell footprint was background-subtracted and multiplied by a conversion factor (7.8 ± 0.5) to account for the decreased laser power used to image bright cells. This conversion factor was determined from a separate experiment in which TIRF images of GFP-expressing cells were acquired at several laser powers. The response from our system was linear, and cells imaged at 15% laser were 7.8 ± 0.5 times brighter than those imaged at 2% laser on average. After this conversion, the corrected basal fluorescence (F_B) was divided by the single-channel fluorescence (F₀) to obtain the number of channels in the basal TIRF image. This value was divided by the area of the basal cell surface (A_B, in μm²) to estimate the basal surface fluorescent channel density (d_B):

$$d_{\text{B}}=\frac{\frac{F_{\text{B}}}{F_0}}{A_{\text{B}}}.$$ (6)

Finally, d_B was multiplied by the total cell area in square microns (A_tot), which we determined from the cell's capacitance (1.11 μF/cm² (28)), to estimate the total number of fluorescent channels in each patched cell (#_Fluor):

$${\#}_{\text{Fluor}}=d_{\text{B}}\times A_{\text{tot}}.$$ (7)

This calculation assumed that all of the footprint fluorescence came from channels localized in the PM. Although Kv2.1 traffics efficiently from the ER to the PM (4), the other three channels studied in this work have variable levels of protein remaining in internal membranes. We estimated the magnitude of this type of artifact by expressing an ER-localized protein, CB5-GFP, in HEK293 cells and imaging the basal surface of the cells using the same TIRF illumination settings as for the bright cells we had patched. We measured the total basal fluorescence from the CB5-GFP and divided it by the average unitary fluorescence of a single Kv2.1 channel molecule and found that an ER full of GFP contributed, on average, 5.2 channels (ch)/μm² to the TIRF image. Because this density was negligible compared with the fluorescent channel densities in our patched cells, we did not correct for it in our calculations.

Analysis and statistics

Error bars represent standard error. In some cases, error bars are smaller than symbols. Fitting was performed in Origin2018b using least-squares methods. Statistical significance was defined as p < 0.05 in t-tests comparing sample means.

Results

Resolving gating and ionic currents from a single-channel population

Our goal was to design a voltage-clamp protocol that would report the maximal gating charge movement and membrane conductance from Kv channels expressed in HEK293 cells. Measuring these two parameters from the same population of channels is non-trivial because gating currents are at least an order of magnitude smaller than ionic currents. Furthermore, because they are so small, gating current magnitudes are most easily measured from extremely large populations of channels. At the same time, this requirement for high expression compromises the ability to accurately measure the large ionic currents from that population due to inherent limitations of a single-electrode voltage clamp. Compounded with these limitations, we wanted to avoid the use of pharmacological agents that could induce artifacts associated with non-uniform channel block (29). Below, we describe an approach to these challenges that builds on earlier work in our lab where we used optical and electrophysiological methods to count Kv2.1 channels (17), and in other labs where Kv channel gating currents and ionic currents were used to estimate channel properties (12,18,29,30).

Like most Kv family members, the threshold of Kv2.1 channel activation is more depolarized than the normal E_K. In physiological solutions, whole-cell Kv2.1 currents activated at a command potential of −20 mV and were outward (Fig. 1 A). Indeed, under these recording conditions, Kv2.1 only conducted in the outward direction due to the depolarized potential relative to E_K (Fig. 1, A and C, solid symbols). By altering the potassium concentrations of our recording solutions, we were able to shift E_K to a voltage where Kv2.1 channel activation was maximal. In a solution set we called “Flipped K⁺” (150 mM K⁺_out/30 mM K⁺_in), whole-cell Kv2.1 currents reversed at a command potential of about +40 mV (Fig. 1, B and C, open symbols). Despite the dramatic change in the polarity of Kv2.1 ionic currents recorded in Flipped K⁺, the voltage dependence of Kv2.1 activation was not different from that measured in the physiological K⁺ gradient (Fig. 1 D, p = 0.26). Furthermore, it is noteworthy that in both conditions the activation curves were almost completely saturated at +40 mV (Fig. 1 D), suggesting that at this voltage all available Kv2.1 channels were maximally activated. Thus, the Flipped K⁺ conditions represent an ideal system to record gating and ionic currents from Kv2 channels pseudo-simultaneously, as described below.

We designed a single voltage-clamp protocol to capture both gating and ionic currents from individual HEK293 cells (Fig. 1 E and materials and methods). There was no evidence of gating current or ionic conductance when this protocol was applied to untransfected cells (Fig. 1 E, black trace). Uncompensated capacitive spikes were observed in some untransfected cells at the onset of depolarization, and care was taken to avoid analyzing the range of times where these signals were observed (see materials and methods). In contrast, when the same voltage-clamp protocol was applied to HEK293 cells expressing GFP-Kv2.1, the step to E_K elicited an outward current that resembled on-gating current (Fig. 1 E, blue dashed inset) while the four square pulses generated ionic currents that reversed at E_K (blue shaded area). Three pieces of evidence suggest that the transient current at the start of the +40 mV step is indeed gating current. First, these putative gating currents were not observed in recordings from untransfected HEK293 cells. Second, the magnitude of the charge moved in the putative gating current increased linearly with GFP-Kv2.1 expression (see Fig. 3 D). Third, the time constant of the decay of these currents was consistent with previous recordings of Kv2.1 gating currents (1.6 ± 0.5 ms, N = 7 (31,32)).

Figure 3.

Using unitary fluorescence of GFP-KV2.1 to identify its unitary gating charge. (A) Still image from a movie of single, diffusing GFP-Kv2.1 channels expressed in a HEK293 cell. Scale bar, 10 μm. Panels on the right show a 3 s time series of the boxed region and illustrate diffusive behavior of a single channel in this cell. Scale bars, 1 μm. (B) Histogram of fluorescence intensities measured from putative single GFP-Kv2.1 channels as identified in (A). The average fluorescence from the mobile single Kv2.1 channels in this dish was 2817 ± 701 AU. (C) Plot of fluorescence versus time showing two GFP bleach events taken from a time-lapse movie as in (A). The inset shows a histogram of all GFP-Kv2.1 bleach events analyzed. The average fluorescence of the smallest bin was 656 ± 28 AU and corresponds to the fluorescence of a single GFP molecule. (D) The relationship between whole-cell gating charge and fluorescent channel number from a subset of cells where both values were quantified. The relationship was linear with a slope equivalent to the unitary charge of the channel. Slope = 6.66 ± 0.79 e₀. N = 14. The shaded region around the fitted line represents the 95% confidence interval of each fit.

Quantification of macroscopic gating charge and ionic conductance

To quantify macroscopic gating charge, we obtained the area under the on-gating current over the first 3 ms after depolarization (Fig. 2 A, dashed box). From the same trace, the average current during each of the four 1 mV steps was plotted against its respective command voltage to yield the I/V relationship of the membrane (Fig. 2 A, shaded box). Note that the I/V relationship was linear in this voltage range with a slope equal to the membrane chord conductance, which was calculated according to $I_{\text{Mem}}=G_{\text{Mem}}(V_{\text{M}}-E_{\text{K}})$. A cross-plot of membrane conductance and gating charge from cells expressing a wide range of GFP-Kv2.1 levels revealed a positive relationship between the two parameters (Fig. 2 B). However, at the highest expression levels the relationship between membrane conductance and gating charge appeared to plateau. The data were fit well by a single exponential decay function, supporting the visual inspection of the data. This result suggests that membrane conductance per gating charge decreased as the expression of Kv2.1 in the PM increased.

Figure 2.

Quantification of macroscopic gating charge and ionic conductance. (A) Representative current response to the voltage-clamp protocol from a HEK293 cell expressing GFP-Kv2.1 bathed in Flipped K⁺. On-gating current was evident at the onset of the step to +40 mV (dashed box), and square ionic currents were seen in response to the 1 mV steps at the end of the protocol (shaded box). Whole-cell gating charge was measured as the mathematical area under the curve of the on-gating current (dashed inset). In this cell, 663 nC was moved at the onset of depolarization. Membrane conductance was calculated as the slope of the current response from the four voltage steps at the end of the protocol (shaded inset). In this cell, membrane conductance was 1280 nS. The observed reversal potential was 42.1 mV. (B) Cross-plot of Kv2.1 membrane conductance and whole-cell gating charge acquired from 25 cells as described in (A). At the highest expression levels the relationship begins to plateau, as evidenced by the relatively good fit to an exponential decay function (solid curve). (C) TIRF image of a HEK293 cell expressing GFP-Kv2.1 that was patched to produce data as shown in (A).

To determine the numbers of conducting and non-conducting channels represented by the data in Fig. 2 B, the macroscopic charge and conductance measurements needed to be converted to channel numbers. To this end, we performed two independent experiments to estimate the Kv2.1 unitary gating charge and unitary conductance in Flipped K⁺. First, to estimate Kv2.1 unitary charge, we made an independent measure of the number of channels expressed in the patched cells like that in Fig. 2 C using TIRF microscopy and our expertise in single Kv channel imaging (6,21,33). Second, we performed noise fluctuation analysis on Kv2.1 in Flipped K⁺ to estimate single-channel conductance and open probability. The results of these experiments are described below.

Measuring Kv2.1 unitary charge by fluorescent channel counting

In a subset of the patched cells, a TIRF image of the basal cell surface was acquired during patching (Fig. 2 C) to enable fluorescent channel counting offline as described in materials and methods and summarized in Fig. S1. In brief, we estimated the unitary fluorescence of GFP-tagged channels and used that value to estimate the number of fluorescent channels in the bright cells we had patched. More specifically, the unitary fluorescence of GFP-tagged Kv2.1 was determined from time-lapse TIRF movies of channels expressed at low density in the PM (Fig. 3 A and Video S1). Some channels were mobile and some were immobile over the course of time-lapse imaging. The mobile channels exhibited characteristic PM diffusion behavior, which has been studied extensively in our lab (6,27). The average unitary fluorescence of the mobile channels was calculated for each dish of patched cells and was 2817 ± 701 arbitrary units (AU) for the data shown in Fig. 3 B. The variability in the measurements was likely due to uneven adhesion of the basal cell membrane to the glass or the less-than-perfect folding efficiency of GFP proteins (17). To further convince ourselves that this value represented the fluorescence of a single Kv channel, we measured individual GFP bleach steps from the immobile channels in the TIRF movies (Fig. 3 C). The smallest bleach steps were 656 ± 28 AU on average (Fig. 3 C, inset), indicating that a tetrameric channel would have a fluorescence intensity of approximately 2600 AU. This value was similar to the fluorescence associated with the mobile channels identified by diffusion (Fig. 3 B), supporting the idea that the mobile spots were indeed individual GFP-tagged Kv2.1 channels.

Video S1. TIRF time-lapse of the basal surface of a HEK293 cell expressing a low density of GFP-Kv2.1

Mobile channel molecules can be seen moving laterally in the plane of the membrane. Immobile channels are visible as well. Data were collected at 20 Hz and played at 20 Hz.

Using the unitary fluorescence determined above and the footprint fluorescence from patched cells, the number of fluorescent channels in patched cells like that in Fig. 2 C could be estimated. We now had an independent estimate of total fluorescent channels from a subset of cells whose macroscopic gating charge was known. A cross-plot of the macroscopic gating charge number of fluorescent channels in individual cells is shown in Fig. 3 D. The relationship was linear across the entire expression range, suggesting that all fluorescent Kv2.1 channels had functional voltage-sensing machinery. The relationship had a slope of 6.66 ± 0.79 elementary charges per fluorescent channel, which is the unitary charge per Kv2.1 channel that moved within the first 3 ms after depolarization. This result is consistent with a report that about half of all Kv2.1 gating charge moves in the first 10 ms after depolarization (34).

Measuring Kv2.1 unitary conductance

Non-stationary noise fluctuation analysis is one method to determine the unitary properties of channels underlying macroscopic current recordings. Ideally, we would measure non-stationary current noise fluctuations at the same voltage where we measured macroscopic membrane conductance. However, the small magnitude of single-channel currents near E_K precluded this approach. Instead, we chose to record tail current noise fluctuations at −55 mV, a potential where unitary currents would be large due to the strong inward potassium gradient (Fig. 4 A). Non-stationary fluctuations of GFP-tagged Kv2.1 channels were analyzed from repeated tail current recordings obtained at −55 mV (Fig. 4 B). The variance across repeated recordings was calculated from difference records (see materials and methods) and plotted as a function of average current (Fig. 4 C). The resulting relationship was fit well with a parabola whose coefficients corresponded to single-channel current and number of channels. The single-channel current was converted to conductance according to Eq. 5, and on average the Kv2.1 unitary conductance at −55 mV was 17.72 ± 0.90 pS (Fig. 4 D, blue). These values were scaled by a rectification index to estimate the unitary conductance at +40 mV, where macroscopic membrane conductance had been measured. The index was simply the ratio of the slope of the open channel I/V at +40 mV over the slope at −55 mV, which was 0.39 as measured from Fig. 4 A. After this adjustment, the single-channel conductance of Kv2.1 at +40 mV was found to be 7.44 ± 0.38 pS (Fig. 4 D, orange).

As controls for the approach, noise fluctuations were also analyzed from GFP-Kv2.1 currents recorded in two additional electrophysiological solution sets (Fig. S2). Noise fluctuation analysis of Kv2.1 currents recorded in physiological K⁺ revealed a single-channel conductance that matched well with previously published values of this parameter (16,29). Additionally, Kv2.1 currents recorded in high-symmetrical K⁺ had larger single-channel conductance than those bathed in the physiological K⁺ concentration, in accordance with Kv2.1's sensitivity to potassium concentration (29,35). Furthermore, noise fluctuations measured from activating step currents and tail currents yielded similar values of single-channel conductance in cells bathed in symmetrical K⁺. Table S1 contains a summary of Kv2 channel unitary conductance measured in this work and the historical literature (36, 37, 38, 39, 40, 41). Together, these results indicate that the analysis of noise fluctuations from tail currents in Fig. 4 yielded reasonable values of Kv2.1 single-channel conductance in Flipped K⁺.

The number and density dependence of non-conducting Kv2.1 channels in HEK cells

The unitary gating charge and conductance values measured above were used to estimate the numbers of channels contributing to the macroscopic gating charge and membrane conductance values obtained from voltage-clamp traces recorded in Flipped K⁺. Fig. 5 A relates the number of conducting and gating channels from each cell. As expected, most cells had fewer conducting Kv2.1 channels than gating channels and, on average, only 0.49 ± 0.05 of the gating channels were also conducting (Fig. 5 B). In a previous investigation of the Kv2.1 non-conducting state, the fraction of non-conducting channels in HEK293 cells was not a constant. Instead, increased channel density in the membrane was associated with higher fractions of non-conducting channels (17), leading to the hypothesis that high membrane density favored a non-conducting state of the channel. In accordance with this suggestion, the fraction of conducting channels in this work also decreased as a function of channel density in the membrane (Fig. 5 C).

Figure 5.

The fraction of non-conducting Kv2.1 channels is density dependent in HEK293 cells. (A) Comparison of the number of conducting and gating channels in individual HEK293 cells expressing GFP-Kv2.1. These values were obtained from the macroscopic gating charge, membrane conductance, unitary gating charge, unitary conductance, and open probability as described in the text. The number of conducting channels was consistently lower than the number of gating channels in individual cells. Note the logarithmic scale. (B) The same data in (A) shown as the fraction of gating channels that were conducting. On average, about 49% of Kv2.1 channels were conducting. Asterisk indicates that the sample mean was significantly different from a mean of 1 (one-sample t-test, p < 0.005). (C) The fraction of conducting Kv2.1 channels plotted as a function of the overall density of gating channels. Data were smoothed using a moving average with window = 3.

Kv2.1 single-channel conductance and open probability are not density dependent

The presence of non-conducting channels is evidenced by a mismatch in the number of channels that sense voltage and the number that conduct. The number of channels that conduct is a function of single-channel conductance and open probability. Therefore, a reduction in single-channel conductance, open probability, or both at high expression levels could underlie the non-conducting fraction of channels. We explored our current noise fluctuation data to look for evidence of expression dependence in these two Kv2.1 parameters. Fig. 6 shows single-channel conductance and open probability detected in current noise fluctuation experiments as a function of the maximum membrane conductance, which is a proxy for the expression level of conducting channels. Neither single-channel conductance nor open probability was a function of channel expression, as evidenced by the low R² values of the fits to the data (Fig 6). This suggests that the reduction in conducting fraction at high densities seen in Fig. 5 C is not due to a graded decrease in either single-channel conductance or open probability of conducting channels. Instead, the reduction in conducting fraction seen at high expression levels likely represents a change in the partitioning of channels into two distinct populations, conducting and non-conducting.

Figure 6.

Unitary properties of Kv2.1 are not density dependent. (A) Kv2.1 single-channel conductance from noise fluctuation analysis in cells expressing a range of Kv2.1 densities. Each data point represents one cell. A line fit to the data had an R² value of −0.05, indicating that there was no relationship between the two parameters. (B) Kv2.1 open probability from the same cells in (A) as a function of the peak conductance. There was no relationship between these parameters across the expression range tested.

Macroscopic gating charge and ionic conductance from Kv2.2, Kv1.4, and Kv1.5

The method described above to quantify conducting fraction does not use any pharmacological agents; therefore, it can be applied to any voltage-gated potassium channel. Kv2.2 shares many properties of Kv2.1, including efficient PM expression and localization to dense clusters (4,5). However, it is unknown whether Kv2.2 also has a significant non-conducting fraction at high densities. The Kv1 family of channels is more distantly related to Kv2 channels and not thought to exhibit a non-conducting fraction when expressed in HEK293 cells (17). We set out to compare the behavior of these Kv channels with Kv2.1 in the Flipped K⁺ gating assay to determine whether density-dependent channel silencing is a common phenomenon among Kv channels.

Kv2.2, Kv1.4, and Kv1.5 expressed in HEK293 cells were all amenable to the gating assay described above. TIRF imaging revealed that Kv2.2 had a punctate localization in the HEK293 cell membrane (Fig. 7 A), in accordance with its ability to bind to ER VAPs and form ER-PM junctions. In contrast, Kv1.4 and Kv1.5 channels were uniformly distributed across the HEK293 cell PM (Fig. 7 A). Using the same recording conditions described in Fig. 1, small, transient currents resembling on-gating currents from all three channel types were observed when the membrane potential was stepped to E_K in Flipped K⁺ (Fig. 7 B, dashed boxes). Ionic currents that reversed near E_K were also observed in response to the four variable command voltages at the end of the protocol (Fig. 7 B, shaded boxes). The macroscopic gating charge and membrane conductance from individual cells are shown on cross-plots for each channel isoform (Fig. 7 C). The relationships between conductance and gating charge were positive, as expected, since membrane conductance is due to the expressed channel subunits.

Figure 7.

Macroscopic gating charge and ionic conductance from three Kv channels. (A) TIRF images of HEK293 cells expressing GFP-Kv2.2 (top), GFP-Kv1.4 (middle), or YFP-Kv1.5 (bottom). Note the difference in Kv2.2 localization compared with the other two channels. Scale bars, 10 μm. (B) Example recordings of gating (dashed boxes) and ionic currents (shaded boxes) from Kv2.2 (top), Kv1.4 (middle), or Kv1.5 (bottom) in Flipped K⁺. (C) Relationship between raw membrane conductance and gating charge in three Kv channels. The macroscopic conductance is shown as a function of macroscopic gating charge from HEK293 cells expressing Kv2.2 (top), Kv1.4 (middle), or Kv1.5 (bottom).

Counting non-conducting Kv2.2, Kv1.4, and Kv1.5

TIRF microscopy was used to measure the unitary fluorescence of the GFP-Kv2.2, GFP-Kv1.4, and YFP-Kv1.5 channels expressed in HEK293 cells (Fig. S3, A–C). Similar to Kv2.1, the relationship between macroscopic gating charge and unitary fluorescence was linear for the three channel types, and the slope of each line represented an estimate of the unitary charge moved in 3 ms after depolarization (Fig. S3 D). The single-channel conductance and open probability of Kv2.2, Kv1.4, and Kv1.5 were measured by noise fluctuation analysis from currents recorded in Flipped K⁺ (Fig. S4 and text in the supporting material). These unitary values were used to estimate the numbers of gating and conducting Kv2.2, Kv1.4, and Kv1.5 channels in individual HEK293 cells (Fig. 8 A). These paired measurements showed that in some cells the estimates of conducting and gating channel numbers were quite different. Indeed, the fraction of the gating channels that conducted was variable for all three α-subunits (Fig. 8 B). On average, Kv2.2-expressing cells had a conducting fraction of 0.44 ± 0.02, Kv1.4-expressing cells had an average conducting fraction of 0.96 ± 0.05, and Kv1.5 had an average conducting fraction of 0.26 ± 0.01. One-sample t-tests revealed that the average conducting fraction of Kv1.4 was not significantly different from a test mean of 1, supporting the previously published finding that Kv1.4 does not have a significant population of non-conducting channels. However, the conducting fractions of Kv2.2 and Kv1.5 were significantly different from 1, indicating that these two channels had significant non-conducting populations. To our knowledge these are the first results to suggest that Kv2.2 and Kv1.5 exhibit non-conducting states when expressed in HEK293 cells. As observed for Kv2.1, the variability in conducting fraction was also related to channel density in the membrane for these three channels. Indeed, the conducting fraction tended to decrease as a function of the density of gating channels in the membrane for Kv2.2, Kv1.4, and Kv1.5 (Fig. 8 C). This phenomenon was apparent even for Kv1.4, which on average did not have a non-conducting state but at the highest membrane densities developed one.

Figure 8.

Density dependence in three Kv channels. (A) Summary of the number of conducting versus gating channels for Kv2.2 (left), Kv1.4 (middle), and Kv1.5 (right). (B) The same data in (A) shown as the fraction of conducting channels in each cell. On average, 0.44 ± 0.02% of Kv2.2 channels, 0.96 ± 0.05% of Kv1.4 channels, and 0.26 ± 0.01% of Kv1.5 channels were conducting. The fraction of conducting Kv2.2 and Kv1.5 channels were significantly different from 1 (one-sample t-test, p < 0.05 and p < 0.0005, respectively). (C) Conducting fraction from (B) plotted as a function of the density of gating channels in the membrane. The data shown represent a moving average of the data from all cells (window = 3). Note that the fraction of conducting Kv1.4 channels does not go below 0.6.

Discussion

Summary

We investigated the prevalence of silent Kv channels from multiple families expressed in HEK293 cells using a simple voltage-clamp protocol. We measured both the macroscopic charge associated with channel activation as well as the membrane conductance associated with pore opening from the same population of channels in the same voltage-clamp trace. We used empirical values of unitary charge, conductance, and open probability to estimate the numbers of channels underlying the macroscopic charge and conductance measurements and compared those values to reveal non-conducting channels. The results showed that on average, only half of Kv2.1 and Kv2.2 channels function canonically. The other half are non-conducting, despite having functional voltage-sensing domains. This is the first study to count gating and conducting Kv2.1 channels from the same cell and the first to ask whether Kv2.2 also has a non-conducting population of channels. In addition, we observed a non-conducting population of channels from a member of the Kv1 family. Indeed, on average only 25% of Kv1.5 channels expressed in HEK293 cells conducted potassium after voltage sensor activation. In contrast, on average, 95% of Kv1.4 channels expressed in the cell membrane conducted potassium after voltage sensor activation. Surprisingly, for all four α-subunits studied here, the fraction of conducting channels was sensitive to the density of those channels in the membrane. These findings suggest that density-dependent silencing is a common mechanism of Kv conductance regulation that occurs at the PM.

Important assumptions

Several assumptions were made to arrive at the estimates of non-conducting channel numbers quoted in this work. First, we assumed that the density of fluorescent channels on the basal surface of each cell was representative of the channel density across the entire cell membrane. This is a reasonable assumption for Kv1.4 and Kv1.5, whose expression patterns are generally uniform throughout the entire cell membrane. We have previously shown that Kv2.1 clusters cover an equal portion of the basal and apical surfaces of HEK293 cells (17), and we have not noted any deviations from this pattern for Kv2.2. Therefore, we believe our estimation of fluorescent Kv2 channels is reasonable.

We also assumed in this work that channel gating occurred stochastically upon depolarization delivered in the whole-cell patch-clamp configuration. In other words, we assumed that every channel was as likely to sense depolarization of the membrane as any other channel, and that the area under the on-gating current was a sum of all channels' gating behavior. Space-clamp artifacts could undermine this assumption. The relatively small size of the cells used in this study coupled with our attempts to maintain voltage control (see materials and methods) minimized the risk of these artifacts. However, since we only integrated 3 ms of the on-gating current, we cannot rule out the possibility that for Kv2.1, Kv2.2, and Kv1.5 some of the channels in the membrane did not gate at all. More gating charge was not detected when we integrated 10 ms of Kv2.1's on-gating current, suggesting that most of the charge did move in the first 3 ms (not shown). In addition, two previous studies of non-conducting Kv2.1 channels further support the idea that all channels in the membrane sense voltage and contribute to on-gating charge measurements (12,16). However, the simultaneous detection of channel conformational changes and macroscopic gating charge could confirm this for all the channels studied in this work.

Furthermore, we assumed that conducting and non-conducting channels have similar unitary gating charge magnitudes. An inability to gate fully could explain why some Kv channels are non-conducting, and a corresponding decrease in the gating charge magnitude would be expected if this were the case, especially at high channel densities. While the data in Fig. 3 D suggest that this is not the case, we cannot rule out the possibility that subtle differences in gating charge magnitude or kinetics between the two populations could have been masked in our analysis.

Lastly, we assumed that increasing extracellular potassium and lowering intracellular potassium did not perturb the mechanism underlying the non-conducting state of the channels. The similar estimate of the Kv2.1 non-conducting fraction in this work compared with our previous work suggests this was not a major issue. Indeed, our previous studies used physiological potassium gradients and found similar fractions of non-conducting Kv2.1 channels (50–98%) (16,17). The current work did reveal density-dependent silencing in Kv1.4, which deviates from our earlier finding that the Kv1.4 conducting fraction was uniform across expression levels (17). However, we attribute this difference to the wider range of expression levels tested in this work rather than the different potassium concentrations used. Here we measured conductance from cells expressing up to 300 Kv1.4 channels per square micron, while the previous study only explored conducting fraction up to 120 ch/μm². Indeed, the density dependence of the Kv1.4 conducting fraction seen in this work was only apparent at densities higher than about 150 ch/μm².

The Kv non-conducting state exists at physiologically relevant channel densities

Kv2 channels are found at relatively high densities in many cell types in vivo, especially in clusters where they form ER-PM junctions. Fox et al. measured densities of Kv2.1 in rat hippocampal neurons using quantitative immunolabeling and found that the density of Kv2.1 channels in clusters was about 62 ch/μm², while channels outside of clusters had an average density of about 10 ch/μm² (17). O'Dwyer and colleagues counted the number of gating Kv2 channels in mesenteric arterial smooth muscle cells and, although they did not distinguish between clustered and non-clustered channels, they found overall densities of about 36 ch/μm² for males and 85 ch/μm² for females (12). As shown in Figs. 5 C and 8 C, our data from HEK293 cells indicate that 40–50% of Kv2.1 and Kv2.2 channels are silent at channel densities between 50 and 100 ch/μm², consistent with the detection of non-conducting Kv2 channels in situ (12,17).

Whether the Kv1.5 non-conducting state exists at physiological expression levels is somewhat difficult to predict. As seen in Fig. 8 C, 70% of Kv1.5 channels were non-conducting at a density of ∼100 ch/μm². However, measurements of Kv1.5 membrane densities in situ are lacking. Noteworthy is that Kv1.5 is known to be enriched at the intercalated disk of cardiac ventricular myocytes (42,43), and this enrichment may result in channel silencing. In support of this theory, the ultra-rapid potassium current (IK_UR) carried by Kv1.5 channels in the heart is undetectable in ventricular myocytes where these concentrations of Kv1.5 have been observed (42). Conversely, IK_UR is readily detected in atrial myocytes, which express Kv1.5 channels at similar levels (44). An investigation of non-conducting Kv channels could be undertaken in cardiac tissue using a combination of the methods described in this and our past work (16). In contrast to Kv1.5, Kv1.4 largely lacked a non-conducting population even at densities up to 150 ch/μm². Therefore, the Kv1.4 non-conducting state is only likely to be physiologically relevant at extremely high densities, which to our knowledge have not been reported for Kv1.4 in any endogenous system.

The role of density-dependent silencing

The silencing of Kv channels at high densities is probably not a coincidence. Instead, the mechanism may be protective against electrical paralysis of the cell membrane. For example, cultured rat hippocampal neurons and arterial myocytes express between 100,000 and 200,000 Kv2 channels endogenously (12,17). If each of these channels conducted potassium, massive potassium efflux would occur in response to moderate depolarizing stimuli. In this paradigm, it would be difficult for cells to generate the large swings in membrane potential that underlie so much of excitable cell signaling, effectively silencing electrical communication. The data in this work suggest that silencing of Kv channel conductance is due to the channels' sensitivity to their spatial density in the PM. For Kv2.1 and Kv2.2, this property allows the channel protein to be expressed at extremely high levels and utilized for additional cellular functions, such as ER-PM junction formation, without electrically silencing the cell. The non-conducting functions of the other Kv channels studied here will be a key area of future exploration.

Potential mechanisms

Our data suggest that high membrane density favors an inactive state of Kv channels. What might the mechanism be that determines entry into this state? Membrane lipids are thought to be integral to ion channel behavior. Indeed, crystallization of a mammalian Kv channel required the presence of phospholipids, and the X-ray structure of Kv1.2 revealed multiple contacts between membrane lipids and channel transmembrane domains (45). Kv7.1 channels are the most well-known lipid-regulated Kv channels to date, whose voltage sensor-pore coupling requires phosphatidylinositol 4,5,-bisphosphate (PIP2), an anionic membrane lipid (46). There are mixed reports on the effects of PIP2 on other Kv channels' function, including Kv2.1 (47, 48, 49). However, even if PIP2 does not control the gating of the channels studied in this work, other membrane lipids could be involved. For example, sphingomyelin supports Kv channel gating through direct interactions with voltage sensors (50,51). Enzymatic removal of the phosphocholine headgroup of sphingomyelin dramatically reduces Kv2.1 currents in Xenopus oocytes (50). The density dependence of channel conductance observed in this work may arise from a limiting amount of sphingomyelin available in the PM. The variable sensitivities of the channels studied here to this mechanism are reasonable in light of the non-uniform effects of sphingomyelin lipases on Kv2.1, Shaker, Kir1.1, and KcsA (50). Sphingomyelin and its metabolites have been shown to regulate the function of diverse potassium channels, including HERG (52), Kv7.1 (53), and Kv1.3 (54), making this a plausible mechanism for the results shown here. It would be interesting to map the abundance of sphingomyelin across tissues and its subcellular localization in certain compartments in situ, such as the intercalated disk in the case of Kv1.5.

Conclusion

We counted non-conducting channels from two families of Kv channels expressed in HEK293 cells by comparing the numbers of channels contributing to whole-cell gating charge and whole-cell conductance. In all four channels studied here, all PM channels were sensitive to voltage, i.e., generated gating currents, but with increasing density fewer of these gating channels conducted K⁺, a phenomenon that had only been observed in Kv2.1 in the past (17). The four α-subunits we investigated had different sensitivities to membrane density. While Kv2.1, Kv2.2, and Kv1.5 had significant numbers of non-conducting channels at physiological densities, Kv1.4 only developed a large non-conducting population at the highest, perhaps non-physiological, PM densities. These differences may arise from each channel's specific interaction with membrane lipids, although further studies are needed to test this hypothesis and identify the Kv subunit moieties involved. We propose that the non-conducting states exist to prevent electrical silencing of excitable cells in cases where ion channel proteins have structural roles, e.g., formation of ER/PM junctions, and as such are expressed at high cell-surface densities.

Author contributions

E.E.M. and M.M.T. designed the research. E.E.M. carried out all experiments and analyzed the data. E.E.M. and M.M.T. wrote the article.

Editor: Brad Rothberg.