Interneurons contribute to the hemodynamic/metabolic response to epileptiform discharges

Abstract

Interpretation of hemodynamic responses in epilepsy is hampered by an incomplete understanding of the underlying neurovascular coupling, especially the contributions of excitation and inhibition. We made simultaneous multimodal recordings of local field potentials (LFPs), firing of individual neurons, blood flow, and oxygen level in the somatosensory cortex of anesthetized rats. Epileptiform discharges induced by bicuculline injections were used to trigger large local events. LFP and blood flow were robustly coupled, as were LFP and tissue oxygen. In a parametric linear model, LFP and the baseline activities of cerebral blood flow and tissue partial oxygen tension contributed significantly to blood flow and oxygen responses. In an analysis of recordings from 402 neurons, blood flow/tissue oxygen correlated with the discharge of putative interneurons but not of principal cells. Our results show that interneuron activity is important in the vascular and metabolic responses during epileptiform discharges.

Keywords: action potentials; animal models; cerebral hemodynamics; electrophysiology; epilepsy; neurovascular coupling; oxygen.

Fig. 1.

A: photograph of the craniotomy and the setup for in vivo simultaneous recordings in a rat model of epileptiform discharges (EDs) induced by bicuculline injection (Bic) in the somatosensory cortex, using a silicon probe (SP) for multiunit activity (MUA), a tungsten electrode for local field potential (LFP), a laser-Doppler probe (LD) for cerebral blood flow (CBF), and an oxygen probe [tissue partial oxygen tension (Po₂)]. B: Nissl-stained slice of the recorded area showing the injection site (blue oval) and the location of 1 shank of the silicon probe (red box).

Fig. 2.

Estimation of the delay of Po₂ probe. EDs were induced by bicuculline in intact hippocampus in vitro. A: an oxygen-related current (blue trace) in hippocampi extracted from 7-day-old mice during EDs induced with 10 μM bicuculline (red trace). B: to evaluate the time delay between each ED and the onset of oxygen consumption, the baseline slope (dashed black line) was subtracted from the oxygen signal. Consumption onset was defined as a moment when the amplitude of the oxygen-related current (after subtraction) reached 5 times that of the baseline noise, measured as root-mean-square (RMS) amplitude (black arrows).

Fig. 3.

Multimodal responses to EDs. A: simultaneous multimodal recording of MUA with a silicon probe, LFP with a field electrode, CBF with LD, and Po₂ with an oxygen probe. Red dotted lines indicate the time 0 of each discharge. a.u., Arbitrary units. B: an ED from the trace in A. Red arrowhead shows the very sharp peak used to set time 0. C: zoom on windows 1 (win1) and 2 (win2), corresponding to rise and decay of the sharp wave, respectively. D: zoom on window 3 (win3), corresponding to the slow wave part of the EDs.

Fig. 4.

Classification of neurons into putative principal cells (PCs) and interneurons (INTs). Left: waveforms for the 4 classes, based on a principal component analysis. Right: corresponding autocorrelograms (ACG).

Fig. 5.

Raster plot of EDs showing the variability of LFP recordings and single-unit firing (silicon probe recordings) across time and neurons in rat 1. A: all EDs measured on the LFP signals recorded with the field electrode. Events are sorted by occurrence time; time 0 corresponds to the sharp deflection on each ED (see Fig. 3). Red arrows mark the time of bicuculline injection. B–G: PCs (B–D) and INTs (E–G); time 0 corresponds to the sharp peak (see Fig. 3). Most raster plots show considerable time-dependent variability in the LFP shape (A) and MUA (B–G) after the injection. Firing patterns varied and did not clearly correspond to the neuron class. Still, there is a close relationship between the LFP and the unit's firing rate.

Fig. 6.

Correlation between parameters of interest for all 6 rats: matrix representing, for each pair of parameters, the number of rats for which the Pearson correlation is significant. Colors indicate the number of significant data sets. On the top of the matrix, hierarchical clustering allows similar variables to be grouped into 1 clusters: 1) LFP in window 1 and window 3 and INT and PC in window 2; 2) LFP in window 2, inter-ED interval (logfreq, in log scale), baseline CBF, and baseline Po₂; 3) INT and PC in window 1; and 4) INT and PC in window 3. This shows that variables are highly correlated, in particular unit activities from the same windows, which justifies the use of stepwise regression.

Fig. 7.

Hemodynamic and metabolic response function (HRF). A: CBF impulse response function obtained by deconvolution in each rat. B: gamma functions HRF1 (red) and HRF2 (blue) fitted on the CBF response (black) from rat 1. C: Po₂ response obtained by deconvolution in all rats. D: gamma functions fitted on the Po₂ response from rat 1. Note the high level of variability across rats, seen both in the amplitude and in the shape of the second part of the response (drift back to baseline in CBF and overshoot for Po₂). The time at which an ED occurred was defined as time 0.

Fig. 8.

CBF and Po₂ signals are well predicted by the general linear model (GLM) (example of rat 1). A and B: prediction of CBF response with all regressors (constant response, parametric regressors, slow fluctuations; A) and after prewhitening (B). Blue, original traces; red, linear model; gray, timing of EDs. Prewhitening allows better fulfillment of the hypothesis of residual white noise and thereby improves the quality of estimation of statistics on regression coefficients. C and D: prediction of Po₂ response with all regressors (C) and after prewhitening (D). On all traces, only 1,000 s is shown.

Fig. 9.

INTs are better predictors of CBF and Po₂ response fluctuations than PCs: t-statistics for regressors of the GLM parametric model of the CBF response (A) and the Po₂ response (B). Only regressors corresponding to the first gamma model are shown (first part of Po₂ and CBF responses). Asterisk indicates parameters that are consistently significant (i.e., in 4 of 6 rats): amplitude of LFP during the fast wave (win1), the baseline inter-ED interval (logfreq), and INT activity. Significant parameters are the same for CBF and Po₂.