Dendritic inhibition differentially regulates excitability of dentate gyrus parvalbumin-expressing interneurons and granule cells

Fast-spiking parvalbumin-expressing interneurons (PVIs) and granule cells (GCs) of the dentate gyrus receive layer-specific dendritic inhibition. Its impact on PVI and GC excitability is, however, unknown. By applying whole-cell recordings, GABA uncaging and single-cell-modeling, we show that proximal dendritic inhibition in PVIs is less efficient in lowering perforant path-mediated subthreshold depolarization than distal inhibition but both are highly efficient in silencing PVIs. These inhibitory effects can be explained by proximal shunting and distal strong hyperpolarizing inhibition. In contrast, GC proximal but not distal inhibition is the primary regulator of their excitability and recruitment. In GCs inhibition is hyperpolarizing along the entire somato-dendritic axis with similar strength. Thus, dendritic inhibition differentially controls input-output transformations in PVIs and GCs. Dendritic inhibition in PVIs is suited to balance PVI discharges in dependence on global network activity thereby providing strong and tuned perisomatic inhibition that contributes to the sparse representation of information in GC assemblies.

SUPPLEMENTARY FIGURE 3. KCC2 expression in PVI and GC dendrites and effects of KCC2 block on EGABA. a Representative dentate gyrus slice from an adult (P71) rat brain immunolabelled against PV, KCC2 and calbindin (scale bar 10µm). Left image, granule cell layer (gcl); right image, outer molecular layer (oml). Bar graphs summarize the mean normalized KCC2 fluorescence intensity for PV and calbindinexpressing somata and dendrites (two-way ANOVA test with Holm-Sidak pairwise comparison). b Preincubation of slices with the KCC2 blocker VU0240551 (10 µM) induces a significantly more depolarized EGABA in proximal and distal dendrites of GCs (5 GCs) and reduces the EGABA difference between proximal and distal dendrites in PVIs (5 PVIs; compare to Fig. 2). Left, dashed lines represent the corresponding resting membrane potential (RMP; PVIs, -65 mV; GCs -70 mV). Right, individual open circles connected by lines represent individual experiments. Filled circles with lines and bars represent mean ± s.e.m. **, p ≤ 0.01; ***, p ≤ 0.001.

SUPPLEMENTARY FIGURE 4. Divisive and additive interactions of on-and off-path inhibition in
PVIs and GCs. a, b Amplitude of postsynaptic potentials (PSPs) resulting from the interaction between inhibition and excitation were plotted as a function of the EPSP peak amplitude evoked by extracellular stimulation of the medial perforant path (somatodendritic distance ~150 µm). On-and off-path inhibition was evoked by RubiGABA uncaging of 7 randomly chosen spots targeting proximal dendrites on the level of the inner molecular layer (somato-dendritic distance of 25-75 µm) or distal dendrites on the level of the outer molecular layer (somato-dendritic distance > 200 µm) respectively. Data were fit to an extrapolated linear function for individual cells. Lines with shadows represent means ± s.e.m. for on-(bright green) and off-path (dark green) inhibition in PVIs, and on-(gray) as well as off-path (black) inhibition for GCs. Dashed line represent the identity line. Note, EPSPs with peak amplitudes < 9 mV were potentiated but reduced at > 9 mV by on-path inhibition in PVIs, resulting in a homogenization of PSP amplitudes. In contrast, EPSPs were always reduced in amplitude by on-and off-path inhibition in GCs.

SUPPLEMENTARY FIGURE 5. Mild efficiency of off-path inhibition in controlling proximally and distally evoked excitatory signals at GC dendrites. a, b
Extracellular stimulation pipette was placed either in the middle molecular layer (MML excitation; ~150 µm from the soma) or in the outer molecular layer (OML excitation; >200 µm from the soma) to evoke EPSPs in GCs. RubiGABA uncaging spots were placed either close to the soma (on-path, 25-75 µm) or at the distal tips of GC dendrites in the outer molecular layer (off-path, >200 µm). Note, on-path inhibition was always more efficient, independent of the precise EPSP induction site. Circles connected by lines represent individual experiments. Each circle is the mean of 20-50 traces. *, p ≤ 0.05.

SUPPLEMENTARY FIGURE 6. Different distributions of GABAAR-mediated conductances and EGABA
at PVI and GC apical dendrites. a, b Somatic whole-cell voltage-clamp recordings were performed from PVIs (green) and GCs (black) during RubiGABA uncaging at different positions along a single apical dendrite starting at 50 µm distance from the soma. The holding potential was systematically changed from -80 to -40 mV and the slope conductance (GGABA) and the reversal potential of evoked GABAAR-mediated signals (EGABA) were calculated from the current-voltage relationship. The obtained GGABA (a) and EGABA (b) are plotted for PVIs (green) and GCs (black) as a function of somatic distance (6 PVIs and 5 GCs). Note, that during conditions of whole-cell recordings over similar periods of time (30.0 ± 1.13 and 28.0 ± 3.0 min for on-and off-path inhibition, respectively) using the same intracellular chloride-containing pipette solution, a gradient in EGABA was maintained in PVIs. *, p ≤ 0.05; **, p ≤ 0.01.

SUPPLEMENTARY FIGURE 7. Attenuation of distally evoked GABAergic signals in PVIs depends
on the dendritic Rm gradient. a PVI single cell models were equipped with exponentially increasing membrane resistance (Rm) from 10 to 100 KΩ cm 2 . This gradient was modified by systematically varying the constant  from 1 to 10 (color code; 3 PVI models). Attenuation is depicted as the ratio of somatically measured GGABA for signals induced at 250 and 50 µm somato-dendritic distance (14 nS at 50 µm). Note, that only under conditions of  > 5 and with distal GGABA of > 40 nS we reproduced GGABA ratios observed in vitro (green dashed line). b Same as (a) but with =5 and Rm at the dendritic tips was systematically increased to the color-coded values. Only under conditions of high distal GGABA of > 40 nS and distal Rm > 100 KΩ, the measured GGABA ratio reached values observed during in vitro experiments. Green dashed line with shaded area represent the experimentally defined mean ± s.e.m. GGABA ratio. Squares with lines represent mean ± s.e.m.

SUPPLEMENTARY FIGURE 8. The relation between inhibitory efficiency and the amplitude of excitatory signals is shaped by EGABA. a
In PVI models, an excitatory conductance (Gexc) was added at a somato-dendritic distance of 150 µm and step-wise increased to evoke EPSPs of varying amplitude. GABAergic inputs were activated either on-(~50 µm) or off-path (> 200 µm) relative to the EPSP induction site. EGABA was changed from a linearly decreasing somato-dendritic gradient constrained by our in vitro data (gray traces) to uniform values of -75, -65 and -55 mV (blue, green and red, respectively; 3 PVI models). Black line depicts in vitro data (Fig. 3b, d). Note, the model with a linearly decreasing EGABA could qualitatively reproduce our in vitro data for on-and off-path inhibition (black vs gray). b Same as in (a) for 3 GC models. Introducing our experimentally defined EGABA gradient (Fig. 2e) or a constant EGABA value of -80 mV qualitatively reproduced our in vitro data for on-and off-path inhibition (gray and blue vs black line). Squares with lines represent mean ± s.e.m. 9. The distribution of Rm along the somato-dendritic axis has a small influence on inhibitory efficiency in PVIs and GCs. Inhibitory effect (see Fig. 6e, legend for details) is plotted against GGABA and EGABA for on-(left, 50 µm from soma) and off-path inhibition (right, 250 µm from soma) in PVIs. The exponentially increasing Rm from the soma to distal dendrites used as basic condition in our single cell simulations (green surfaces; Rm = 10 -100 KΩ cm 2 ,  = 5, 3 model PVIs) was changed to a model with constant Rm value (gray surfaces; Rm = 10 KΩ cm 2 ). Inhibitory effects were measured in relation to EPSPs evoked at somato-dendritic distances of 150 µm (excitatory conductance 5 ns; EPSP amplitude of 5.1 mV, measured at the soma). Note, Rm had a negligible influence on inhibitory effects.

SUPPLEMENTARY FIGURE 10. Proximal inhibition is more efficient than distal inhibition in controlling action potential generation evoked by distal inputs in GCs and attenuation of inhibitory signals in GCs and PVIs. a
Morphology of a representative GC used for single cell simulations. Five excitatory inputs where located at the tips of apical dendrites of GC models (red filled circles). GGABA was applied at seven distal distributed either close to the soma (gray filled circles) or at the distal regions of dendrites (black filled circles). b Action potentials were generated by systematically rising the magnitude of excitatory conductances (Gexc), which induced a continuous increase in action potential probability (red squares). When GABAergic inputs close to the soma were activated, action potential generation was completely abolished (gray squares). In contrast, when distal GGABA were introduced near Gexc locations (5 µm distance) only a mild effect on the generation of action potentials was observed (black squares). c Attenuation of distally evoked IPSPs was evaluated in GCs and PVIs (3 and 5 model cell, respectively) using fast unitary decay time constants of the GABAA receptor-mediated conductance (GC: 4 ms, PVI: 2 ms 43 ). Squares with lines represents mean ± s.e.m. Lines in (b) represent sigmoid fits to the mean action potential probability.