Fig 1.
Human but not mouse fast-spiking basket cells exhibit a robust somatic HCN channel-mediated sag potential.
(A) Somatic sag potentials recorded from human fast-spiking pv interneurons. (A1) Anatomical and immunohistochemical illustrations of human fast-spiking pv-immunopositive basket cells filled with biocytin during whole-cell recording from the soma. (Inset schematics shows whole-cell clamp protocol.) Top: Partial anatomical reconstruction showing the soma and dendrites (red) and axon (black) of cell h16 (visualized in two 60-μm thick merged sections). L2/3: neocortical layer 2–3. Bottom: Confocal images showing pv immunofluorescence in a bioc-filled neuronal axon (cell h16, pv, and bioc merged, costained axon boutons indicated by arrows) and a neuronal soma (cell h19). (A2) Human fast-spiking basket cell membrane potential responses to hyperpolarizing square-pulse current steps delivered from −70 mV. Traces reaching a peak hyperpolarized potential of −90 mV illustrate the hyperpolarization-activated sag potential in three of four sample neurons under control conditions (blue traces). Sample recordings show cell h19 with large sag, cells h35 and h27 with close to median-sized sag, and a cell without sag (cell h16). Black traces show responses of these same cells in the presence of the HCN channel blocker ZD7288 (ZD, 10 μM). Red markings indicate the ZD-sensitive sag potential amplitude (sag) measured during a hyperpolarizing step from −70 mV to −90 mV. (B) Absence of robust somatic sag potential in mouse fast-spiking interneurons. (B1) Left: Partial reconstruction of a mouse fast-spiking pv-expressing cell filled with bioc during whole-cell recording (inset schematic shows whole-cell clamp protocol). Right: Same cell (m18) with fluorescence signal (pvCre-tdTomato) indicating pv expression and visualized with biocytin (+bioc). (B2) Sample membrane potential traces from two mouse fast-spiking pv-expressing basket cells (control conditions, green; in ZD, black) showing a small voltage sag in one (cell m4) but no sag in the other (cell m12). The latter cell is typical of most fast-spiking mouse neurons recorded. (C) The robust somatic sag potential is mediated by HCN channels. Plot summarizes the sag amplitude (ordinate) in 24 fast-spiking human basket cells measured by a membrane potential step to −90 mV from −70 mV under control conditions. Sag potential is blocked by wash-in of ZD (P < 0.001 by Wilcoxon signed-rank test). (D) Somatic HCN sag potential is common and robust in human fast-spiking basket cells compared to corresponding mouse cells. (D1) Cumulative histogram showing HCN sag potential amplitudes in human (blue) and mouse (green) basket cells evoked by a hyperpolarizing somatic step from −70 mV to −90 mV. The HCN sag amplitude is larger in human cells (n = 72) than mouse cells (green, n = 60) (P < 0.001 by Mann–Whitney U test). Sample cells with average of five voltage traces are illustrated in the inset. (D2) Histogram of HCN sag amplitude in human and mouse basket cells evoked by a membrane potential step from −50 mV to −70 mV. The HCN sag amplitude in human basket cells is larger than in mouse basket cells (P < 0.001 by Mann–Whitney U test). Insets show sample voltage traces from a human and a mouse basket cell. (D3) Histogram of HCN sag amplitude measured at the resting membrane potential (Em) following a depolarizing step to −50 mV. The depolarizing step deactivates HCN channels, while repolarization to Em activates these channels, generating a sag potential as illustrated in the inset. The HCN deactivation–activation voltage step cycle applied at Em revealed larger sag potentials in human than mouse basket cells (P < 0.001 by Mann–Whitney U test), although Em values did not differ significantly between species (P = 0.09 by Mann–Whitney U test). (E) Somatic sag potential is a general feature of human fast-spiking basket cells in the neocortex. (E1) The sag potential in human fast-spiking basket cells is observed at all ages from 20 to 82 years, and sag amplitude shows no significant correlation with patient age (R = 0.139, P = 0.245 by Pearson’s correlation analysis). (E2) Robust sag potentials in fast-spiking basket cells from different neocortical areas. Left: Plot showing sag potential amplitudes of cells from temporal (n = 24), frontal (n = 35), and other (n = 13) cortical regions. “Other” cortices include occipital, parietal, ventral, and periventricular areas. There is no difference in sag potential amplitude between areas (P = 0.131 by ANOVA on ranks). Right: Cumulative histograms of these same data (temporal, black; frontal, gray solid; other, gray dotted line). (E3) Prominent sag potentials are also observed in cells from neocortical tissue samples resected due to different primary diagnoses, and amplitudes do not differ among tissue samples resected for tumor, hydrocephalous, or aneurysm treatment (P = 0.345 by ANOVA on ranks). The underlying data supporting Fig 1C–1E can be found in S1 Data. ANOVA, analysis of variance; bioc, biocytin; HCN, hyperpolarization-activated cyclic nucleotide–gated; pv, parvalbumin.
Fig 2.
Immunohistochemical evidence for stronger HCN1 and HCN2 channel localization at somatic cell membrane of human than mouse pv cells.
Confocal microscope study of HCN1 and HCN2 immunofluorescence labeling in somatic region of human and mouse pv cells in L2/3 of neocortex. Panels A-D illustrate similar experiments in human with HCN1 (A1–3) and HCN2 (B1–3), and mouse with HCN1 (C1–3) and HCN2 (D1–3). Immunofluorescence intensity for pv with either HCN1 or HCN2 was measured in a radial pattern of lines diverging from cell soma center and projecting to extracellular space. (A) Analysis of pv and HCN1 double-labeling in human cells. (A1) Confocal immunofluorescence image of a basket cell in human neocortex L2/3 with pv labeled using Alexa 488 as the fluorophore and HCN1 using cy3 as the fluorophore. Immunofluorescence intensity was measured along six 8-μm long radial lines (i to vi). (A2) Immunofluorescence intensity (ordinate) from pv (gray line) and HCN1 (black line) measured along lines i to vi starting from the cell soma center (left) and crossing the soma plasma membrane to the extracellular space (right). The pv fluorescence signal is strong within the intracellular space and disappears along measurement lines crossing the plasma membrane zone into the extracellular spaces. HCN1 immunofluorescence, measured in parallel, shows peak intensity in the plasma membrane zone, while pv immunofluorescence simultaneously vanishes in this zone. Ordinates of intensity traces are normalized and scaled similarly to demonstrate the temporal relationship (fluorescence signal peak value = 1). (A3) Average pv (top) and HCN1 (bottom) immunofluorescence signals along the six measurement lines for the cell shown in A1–2. Lines are aligned to the midpoint of pv signal descent (see Methods) marking the plasma membrane zone (memb., 1 μm-wide region at pv signal descent illustrated by shaded blue background). (B) Analysis of pv and HCN2 double-labeling in human basket cells. (B1) Confocal microscope images of pv within the soma labeled by Alexa 488 and HCN2 labeled by cy3. Immunofluorescence measurement lines are labeled i to vi. (B2) Immunofluorescence intensity of pv (gray) and HCN2 (black) measured along lines i to vi in the cell from B1. Immunoreactivity for HCN2 is strongest along all measurement lines as they cross the extracellular plasma membrane zone, while pv immunofluorescence simultaneously vanishes. (B3) Average immunofluorescence signal intensities of pv (top) and HCN2 (bottom) along the measurement lines shown in B1–2. The plasma membrane zone is indicated by a blue background. (C) Analysis of the pv-associated intracellular fluorescent signal from tdTomato and from HCN1 immunolabeling in a mouse basket cell. (C1) Confocal fluorescence image of a pv-positive (pvCre-tdTom) mouse basket cell immunostained for HCN1 using Alexa 488 as the fluorophore. Radial measurement lines shown as i to vi. (C2) Line analysis of a mouse pv cell expressing tdTomato and also immunostained for HCN1 showing the absence of a clear HCN1 immunofluorescence peak within the plasma membrane zone. (C3) Averaged line analyses of pv-associated fluorescent signals (top) and HCN1 immunofluorescence (bottom) for the cell shown in C1-C2. The plasma membrane zone is indicated by green background. (D) The pv-associated intracellular fluorescence signal in a mouse cell also immunostained for HCN2. (D1) Endogenous intracellular fluorescence driven by the pv promoter (pvCre-tdTomato) and from fluorescence immunostaining for HCN2 (using Alexa 488 as the fluorophore). Measurement lines i to vi are illustrated on image. (D2) Line analyses (i-vi) of tdTomato and HCN2 fluorescence signals. This mouse cell emits an intracellular HCN2 signal without a clear signal peak within the apparent extracellular membrane zone. (D3) Average of line analyses for D1-D2. Membrane zone shown with green background. (E) Histograms summarizing HCN1 and HCN2 immunofluorescence intensity levels at the soma plasma membrane for 390 pv cells (analyzed like the sample cells in A-D). Bin size is 0.25. (E1) Top: HCN1 immunofluorescence intensity at the plasma membrane zone versus the extracellular space for human pv-positive basket cells (blue bars) and mouse pv-expressing basket cells (green). Both human and mouse cells show higher fluorescence intensity at the membrane zone than the extracellular space, but human basket cells also show a significantly higher ratio of membrane-bound signal compared to mouse basket cells (P < 0.001 between species). Bottom: HCN1 immunofluorescence intensity at the soma plasma membrane compared to the intracellular space. Human cells show higher HCN1 immunofluorescence at the plasma membrane than in the intracellular space (ratio > 1), whereas mouse cells exhibit stronger HCN1 immunofluorescence within the intracellular space (cytoplasm) than at the membrane (ratio < 1) (P < 0.001 for the ratio between species by Mann–Whitney U test). Ratio of 1 at abscissa is indicated in red and with vertical dotted line. The n values indicate the numbers of analyzed cells, patients, or mice. (E2) Top: The HCN2 immunofluorescence intensity at the cell membrane zone versus extracellular space for human pv-positive basket cells (blue bars) and mouse pv-expressing basket cells (green). Both human and mouse pv cells show stronger HCN2 signals at the cell membrane than in the extracellular space, but human cells show a higher ratio of membrane-bound signal (P < 0.001 ratio between species). Bottom: HCN2 immunofluorescence intensity at the cell membrane versus intracellular space. Human cells show stronger signals at the cell membrane than inside the cell. Mouse cells show stronger HCN2 immunofluorescence in the intracellular space than at the cell membrane (ratio < 1) (P < 0.001 ratio between species by Mann–Whitney U test). Ratio of 1 at abscissa is indicated in red and with vertical dotted line. (E3) Comparison of HCN1 and HCN2 immunofluorescence localization results for pv cells between species. Top: Cumulative histogram summarizing the membrane versus extracellular site measurements in human (blue line) and mouse (green line) for HCN1 (solid line) and HCN2 (dotted line) (P < 0.001 between species). Bottom: HCN1 and HCN2 immunofluorescence intensity measured at the cell membrane versus intracellular space (P < 0.001 between species). (ANOVA on ranks with post hoc Dunn’s pairwise test). The data shown in Fig 2E1–2E3 histograms are available in S1 Data. (F-G) Verification of the plasma membrane zone in human (F) and mouse (G) images by immunofluorescence labeling for Kv3.1 potassium channels, which are characteristically enriched at the somatic membrane. (F1) Top: Confocal immunofluorescence image of pv immunoreactivity in a human L2/3 cell (cy3 as the fluorophore) costained for Kv3.1 (Alexa 488 as the fluorophore). Immunofluorescence intensity was measured along radial lines i-vi. Bottom: Line analysis of lines i-vi for pv and Kv3.1 immunofluorescence. Kv3.1 signal shows peak intensity at the plasma membrane zone, whereas pv immunofluorescence disappears within this zone. (Intensity traces normalized to the fluorescence signal peak value.) (F2) Average of line analyses from F1. (G1) Top: Confocal fluorescence image of a mouse basket cell showing fluorescence signals from pv (pvCre-tdTomato) and Kv3.1 (Alexa Fluor 488). Also shown are radially patterned intensity measurement lines i-vi. Bottom: Line analyses (i-vi) and (G2) average of aligned fluorescence intensity traces showing the Kv3.1 signal peak at the plasma membrane zone.
Fig 3.
Visualization of somatic HCN1, HCN2, and Kv3.1 channels in human and mouse pv-immunopositive interneurons by dSTORM super-resolution immunofluorescence microscopy.
(A-C) dSTORM super-resolution images showing high-resolution immunofluorescence labeling of HCN1, HCN2, and Kv3.1 channels in human pv-immunopositive cells. Pv immunoreactivity is shown in paired confocal images. (A) dSTORM image of HCN1 immunolabeling (A1) using Alexa 647 as the fluorophore paired with a confocal image of pv immunoreactivity (A2) using Cy3 as the fluorophore. HCN1 immunoreactivity is concentrated at apparent membrane zone around the cell soma and proximal dendrites. (B) dSTORM image of HCN2 immunolabeling (B1) using Alexa 647 as the fluorophore paired with a confocal image of pv immunoreactivity (B2) using Cy3 as the fluorophore. Strongest HCN2 immunofluorescence is localized at the apparent cell soma membrane zone. (C) dSTORM image of Kv3.1 potassium channel immunostaining using CF568 as the fluorophore (C1) paired with a confocal image of pv immunoreactivity (C2) using Alexa 488 as the fluorophore. Kv3.1 immunoreactivity is concentrated at the apparent membrane zone. (D-E) Mouse pv cell showing weak or absent HCN1 or HCN2 immunofluorescence at the cell soma membrane. dSTORM images of HCN1 (D1) and HCN2 (E1) immunoreactivity using Alexa 647 as the fluorophore are paired with confocal images of pv-associated fluorescence (tdTomato) (D2 and E2, respectively). (F) dSTORM image of Kv3.1 immunolabeling (F1) using Alexa 647 as the fluorophore paired with a confocal image of pv-associated intracellular fluorescence signal (F2). Kv3.1 immunofluorescence is localized around the cell soma.
Fig 4.
Evidence for stronger HCN channel activity at the somatic compartment of human fast-spiking interneurons compared to mouse fast-spiking interneurons from outside-out patch recordings.
(A) Outside-out patch recordings obtained from human fast-spiking cell somata. (A1) First, the sag potential was measured in somatic whole-cell clamp mode (blue traces) by applying hyperpolarizing current steps from −60 mV. Sag potential amplitude during a voltage step to −90 mV is indicated by the red marking. Next, outside-out nucleated patches were established by pulling out the pipette. Gray traces show membrane potential changes from current steps applied to the same cell but in the somatic outside-out patch configuration. Schematic insets on left show the experimental design. (A2) Input resistance (ordinate) of the same cell measured by membrane potential steps to hyperpolarized potentials (abscissa) from −60 mV in the somatic whole-cell clamp mode (blue symbols) and outside-out patch configuration (gray symbols). (A3) Input resistance in 13 fast-spiking human basket cells measured by hyperpolarizing step to −90 mV from −60 mV in whole-cell clamp mode (blue symbols) and the outside-out patch configuration (gray symbols) (P < 0.001 by Wilcoxon signed-rank test). (B) Sag amplitude is largely preserved in human fast-spiking cell somatic outside-out recordings. (B1) Sag amplitude (ordinate) in one fast-spiking cell measured during membrane potential steps to hyperpolarizing potentials (ordinate) from −60 mV in somatic whole-cell recording mode (blue symbols) and the outside-out patch configuration (gray symbols). Left: Plot shows incremental increases in sag amplitude at more hyperpolarized membrane potentials in both whole-cell mode (blue) and the outside-out patch configuration (gray). Right: Traces show hyperpolarizing membrane potential steps from −60 mV in the two modes. Sag amplitude during steps to −90 mV is indicated by red marking. (B2) Sag amplitude measured during hyperpolarizing steps to −90 mV from 13 fast-spiking human cells in whole-cell clamp mode (blue) and the outside-out patch configuration (gray). Sag amplitude is only moderately reduced in outside-out patches (P < 0.001 by Wilcoxon signed-rank test). (B3) Sag amplitude change (ordinate) does not correlate with the relative increase in soma input resistance (Rin, abscissa) due to the change from somatic whole-cell clamp (whc) to outside-out patch (o-o) (R = −0.178, P = 0.616, Spearman’s rank order correlation). (C) Outside-out patch recordings from mouse fast-spiking pv cell somata. (C1) Small somatic sag potential measured in whole-cell clamp (green traces) in response to hyperpolarizing current steps from −60 mV. Sag potential amplitude during voltage step to −90 mV is indicated by red marking. Gray traces show membrane potential changes from current steps applied to the same cell after establishing the somatic outside-out patch configuration. Schematic shows the experimental design. (C2) Input resistance (Rinput) of 8 fast-spiking mouse cells measured by hyperpolarizing step to −90 mV from −60 mV in whole-cell clamp mode (green symbols) and the outside-out patch configuration (gray symbols) (P = 0.008 by Wilcoxon signed-rank test). (C3) Sag amplitude (measured during steps to −90 mV) of 8 fast-spiking mouse pv-expressing fast-spiking cells in whole-cell clamp mode (green) and outside-out patch configuration (gray). Sag amplitude is preserved although reduced in outside-out patches (P < 0.0016 by Wilcoxon signed-rank test). (D) On average, the sag amplitude was similarly preserved in outside-out patches compared to the preceding whole-cell recording in both mouse (green symbols) and human cells (blue symbols) (P = 0.971 by Mann–Whitney U test). Box plots show median and upper and lower quartiles. (E-F) Voltage-clamp recording at the cell soma showing a greater reduction of leak conductance (Gleak) in human than mouse fast-spiking basket cells during application of the HCN channel blocker ZD7288 (ZD, 10 μM). Cells were voltage-clamped at the resting membrane potential of baseline control condition throughout experiment (average of −59.0 mV for human and −70.6 mV for mouse cells). (E1) Sample traces (averages of five) showing somatic whole-cell voltage-clamp current evoked by −10 mV steps (10 ms in duration) from Em in a human fast-spiking neuron. Blue trace is the response under control conditions and black trace is that in the presence of ZD7288 (10 μM). Voltage-clamping of a fast-spiking basket cell at −65 mV. Trace holding-current levels in the figure are aligned for comparison of the step amplitude change induced by ZD. Zoomed image illustrates the current response to a −10-mV voltage-clamp step under control conditions and in the presence of ZD. The current amplitude (shown as Ictrl) is smaller during the same voltage step in the presence of ZD (shown as IZD) as indicated by brackets. Ictrl or IZD amplitude is proportional to somatic leak conductance. Inset shows a schematic of the clamp protocol in human cell. (E2) Plot showing the current amplitudes evoked by −10 mV voltage steps before and during ZD wash-in (indicated with gray bar). (E3) Somatic input resistance, measured as 1/Gleak, in five human fast-spiking cells at baseline and following ZD7288 wash-in. In four of five cells (red lines), the change is significant (P < 0.05 by Wilcoxon signed-rank test). One cell without significant change is indicated by black line. Values are averages from 1-min time windows before ZD application (baseline, bl) and after 5 min in the presence of ZD (ZD). (F) Mouse fast-spiking cells regularly fail to show a change in leak conductance at the cell soma under HCN channel blockade. (F1) Sample traces (average of five) showing whole-cell voltage-clamp current responses to −10 mV voltage steps (10 ms) from Em in a mouse fast-spiking pv neuron. Green trace is under baseline control conditions, and black trace is in the presence of ZD7288 (10 μM) at −69 mV. Zoomed image shows the current responses to a −10-mV step at baseline and in the presence of ZD. There is a negligible change in current amplitude following ZD wash-in. The schematic inset indicates whole-cell recording from mouse cell. (F2) Somatic current amplitudes in response to −10 mV voltage steps before and during ZD wash-in (indicated by gray bar). (F3) Somatic input resistance (1/Gleak) in five mouse fast-spiking pv cells at baseline (bl) and following ZD7288 wash-in (ZD) (averages from 1-min time windows before ZD application and after 5 min in the presence of ZD). Input resistance is significantly changed in only one cell (red line) of five (P < 0.05 by Wilcoxon signed-rank test). The data shown in Fig 4E and 4F can be found in S1 Data.
Fig 5.
HCN channel blockade lengthens the somatic time constant and somatic spike generation delay in human fast-spiking interneurons.
(A) The HCN channel blocker ZD7288 slows the time course of membrane potential changes in human fast-spiking interneuron soma. (A1) Somatic membrane apparent time constant (τ) measured from equal-amplitude hyperpolarizing membrane potential steps from −70 mV to −90 mV in whole-cell clamp mode (illustrated in inset schematic) under control conditions (blue, average of five traces from one cell) and in the presence of ZD7288 (black). Istep shows time course of the hyperpolarizing square-pulse current step. (A2) Plot summarizing τ as measured using standard hyperpolarizing membrane potential steps for human and mouse fast-spiking cells (A1) at baseline control (bl) and following ZD7288 (ZD) infusion. τ is significantly lengthened by ZD in human (blue, P < 0.001, n = 24 cells) but not mouse (green) fast-spiking cells (n = 12, P = 0.470 by Wilcoxon signed-rank test). (A3) The effect of ZD7288 on somatic membrane τ is strongest in cells showing the largest HCN sag potential amplitude at baseline. Plot shows the baseline-normalized membrane τ value in the presence of ZD (ordinate) against the sag amplitude (abscissa) measured before ZD wash-in. The two parameters show a strong correlation (n = 24, R = 0.73, P < 0.001 by Spearman’s rank order correlation analysis). Gray line shows regression. (B) ZD prolongs the somatic action potential generation delay in human fast-spiking cells. (B1) Excitatory postsynaptic currents (EPSCs) resembling natural EPSCs recorded from the somata of human fast-spiking interneurons were generated at the soma by dynamic clamp to elicit EPSPs reaching firing threshold. Traces illustrate EPSPs with spikes (blue) elicited by EPSCs (gray, dynamic clamp) at −70 mV. Dotted vertical lines and the bracket show EPSC-to-spike lag (from EPSC onset to spike peak). Schematic inset illustrates the experimental setup for dynamic clamp. (B2) Dynamic clamp parameters were set to generate incrementally greater EPSC amplitudes to evoke EPSPs with spike probabilities <1. Traces show EPSPs starting from subthreshold and reaching beyond the firing threshold by applying dynamic clamp input in increments of 0.1 nS from 5.8 nS to 6.1 nS (dynamic clamp input strengths indicated below EPSPs) at an interval of 4 s. Traces illustrate three consecutive cycles for one cell at baseline with EPSP–spike probability <1 in response to the second weakest dynamic clamp input (5.9 nS). Resting membrane potential in the experiment was kept at −70 mV. The cycle of four incremental EPSPs was repeated 30 times. (B3) Recording from the same cell following 5 min ZD wash-in (resting membrane potential kept at −70 mV). Cell input resistance increased, so EPSC strengths were readjusted to elicit an EPSP–spike probability <1. Traces show three consecutive cycles of incrementally increasing EPSP amplitude with EPSP–spike probability <1 in response to 5.5 nS input to dynamic clamp. (B4) EPSP–spike probability plotted against the spike delay measured during 30 cycles in one fast-spiking human basket cell. Plot shows spikes generated by three EPSP strengths with spike probability <1 at baseline (blue dots) and following 5 min ZD wash-in (black symbols). P values show the difference between spike time delay values evoked at a similar spike probability under baseline conditions and in the presence of ZD. (B5) Superimposed EPSP–spike traces illustrated in the same cell at baseline (blue) and in the presence of ZD (black). EPSP–spike probabilities shown in the inset. (B6) Histograms summarizing EPSC-to-spike delay (from EPSC onset to spike peak) values for the spikes in B5. (C1) Plot showing averaged EPSC-to-spike responses in six human cells (blue) and four mouse cells (green) at baseline and following ZD wash-in (black). Identifying experimental codes are shown on the ordinate (see S1 Table). Red arrows indicate direction of a significant change (P < 0.05 by Wilcoxon signed-rank test). (C2) The EPSP–spike probabilities for the EPSC-to-spike delay data in C1 at baseline (blue, green) and in the presence of ZD (black). (C3) The effect of ZD on spike lag is largest in cells showing the greatest increase in membrane time constant under ZD treatment (R = 0.685 P = 0.025 by Spearman’s rank order correlation, n = 10 cells). Blue, human cells; green, mouse cells. Gray line indicates the regression. The underlying data supporting Fig 5B4`5B6 and 5C1–2, and data illustrated in Fig 5A–5C can be found in S1 Data.
Fig 6.
A computational model demonstrates the effects of perisomatic HCN conductance modulation on human basket cell excitability and input–output function.
A three-compartmental model neuron with standard Hodgkin–Huxley-type spike-generating Na+ and K+ currents, slowly activating M-type K+ current (M), inward-rectifying K+ current (Kir), HCN current, and a physiological GABAergic autaptic conductance at the somatic compartment [35] reproduces the physiological behavior and EPSP–spike coupling of a human fast-spiking pv interneuron. (A) Membrane potential features recorded in human fast-spiking cell h35. (A1) Membrane potential responses to 500-ms current steps starting at −100 pA (blue trace) and reaching +20 pA (red traces) applied at Em (−57 mV). The bottom voltage traces are in the presence of the HCN channel blocker ZD7288 at Em = −62 mV. ZD7288 abolishes sag amplitude (sag) and rebound firing (blue dots show timing of rebound spikes for one membrane potential step). (Inset schematic indicates whole-cell recording from biological cell). (A2) Plot showing the peak negative voltage (Vm) against injected hyperpolarizing current-step amplitude at baseline (open symbols) and following ZD7288 wash-in (solid symbols). (A3) Voltage sag amplitude (Vsag) generated by incrementally increasing hyperpolarizing current steps (amplitude shown on abscissa) at baseline (open symbols). Black symbols illustrate blockade of sag by ZD7288. (B) Membrane potential features of the h35 cell-based computational model. (B1) Top: The computational model closely replicates the electrophysiological response features of fast-spiking interneuron h35. Bottom: Removing HCN channels abolishes sag amplitude (sag) and rebound firing (indicated on top in one trace by blue dots) and reproduces the negative shift in Em. (Inset schematic indicates simulation with model cell.) (B2) Plot of peak negative voltage (Vm) versus injected current from the resting membrane potential for the computational model based on cell h35 with physiological gHCN activity (open symbols) at soma and in 200 μm-long primary dendrite, and when gHCN = 0 nS (black symbols). (B3) Sag amplitude (Vsag) during hyperpolarizing current steps with (open) and without gHCN (black symbols) at cell soma and in primary dendrite (see Methods and S2 Table for model cell details). (C) Somatic EPSP-evoked action potential time lag for the model with different HCN conductances (gHCN). (C1) EPSPs and EPSP-evoked action potentials in the model cell primary dendrite (200 μm long) and soma when gHCN = 0. Top: Superimposed traces show EPSPs elicited in the model cell soma by EPSCs of incrementally increasing conductance (gEPSC) from 1 to 6 nS. Red: An EPSP–spike waveform is evoked at 6 nS. Resting membrane potential is set at −62 mV. Middle: Plot showing EPSP-evoked spike delay measured from EPSC onset (EPSC-to-spike delay) as a function of somatic EPSC conductance (gEPSC) strength (abscissa) from 0.5 nS to 20 nS. Red dot indicates the EPSP–spike response evoked by a 6-nS gEPSC. Inset schematic indicates the generation of somatic EPSCs in the total somatodendritic compartment when gHCN = 0. Bottom: Curves show steady-state activation profile of the M-current (brown) and Kir-current (blue) conductances included in the models producing the voltage responses above. (C2) EPSP and EPSP–spike waveforms at physiological gHCN in the primary dendrite (200 μm) and at 50% of the somatic gHCN estimated for cell h35. Schematic inset shows full HCN at the dendrite and weak gHCN around the soma. Top: Traces show EPSP and EPSP–spike responses to EPSCs evoked by conductances from 1 nS to 6 nS at Em = −59 mV. Note action potentials are evoked at 5 nS and 6 nS (red). Middle: Spike delay changes with incremental increases in gEPSC. Red dot = 6 nS. Bottom: Activation profiles of the simulation showing M- and Kir- conductances and gHCN separately for dendritic (gray, d-gHCN) and somatic (black, s-gHCN) compartments. (C3) EPSPs and EPSP–spikes at full physiological gHCN estimated in the primary dendrite and soma (indicated in schematic inset with full HCN at dendrite and around soma). Top: EPSPs and EPSP–spike responses elicited at Em = −58 mV. Action potentials are evoked by the three input strengths (red = 6 nS). Middle: Spike delay changes with incremental increases in gEPSC. Red dot = 6 nS. Bottom: M- and Kir- conductances and full gHCN in both dendritic (gray) and somatic (black) compartment. (C4) EPSPs and EPSP–spike responses at full dendritic gHCN and 150% of h35 cell somatic gHCN (shown in schematic inset with strengthened perisomatic staining). Top: EPSPs and spikes elicited at Em = −57 mV. Action potentials are generated by the three input strengths (red = 6 nS). Middle: Spike delays at different gEPSC values (Red dot = 6 nS). Bottom: M-, Kir-, and HCN conductances with full dendritic gHCN (gray) and pronounced somatic gHCN (black) strength. The underlying data for Fig 6A–6C are available in S1 Data.