Paired pulse depression in the somatosensory cortex: Associations between MEG and BOLD fMRI

Abstract

Interpretation of the blood oxygen level dependent (BOLD) response measured using functional magnetic resonance imaging (fMRI) requires an understanding of the underlying neuronal activity. Here we report on a study using both magnetoencephalography (MEG) and BOLD fMRI, to measure the brain's functional response to electrical stimulation of the median nerve in a paired pulse paradigm. Interstimulus Intervals (ISIs) of 0.25, 0.5, 0.75, 1.0, 1.5 and 2.0 s are used to investigate how the MEG detected neural response to a second pulse is affected by that from a preceding pulse and if these MEG modulations are reflected in the BOLD response. We focus on neural oscillatory activity in the β-band (13–30 Hz) and the P35m component of the signal averaged evoked response in the sensorimotor cortex. A spatial separation of β ERD and ERS following each pulse is demonstrated suggesting that these two effects arise from separate neural generators, with ERS exhibiting a closer spatial relationship with the BOLD response. The spatial distribution and extent of BOLD activity were unaffected by ISI, but modulations in peak amplitude and latency were observed. Non-linearities in both induced oscillatory activity ERS and in the signal averaged evoked response are found for ISIs of up to 2 s when the signal averaged evoked response has returned to baseline, with the P35m component displaying paired pulse depression effects. The β-band ERS magnitude was modulated by ISI, however the ERD magnitude was not. These results support the assumption that BOLD non-linearity arises not only from a non-linear vascular response to neural activity but also a non-linear neural response to the stimulus with ISI up to 2 s.

Introduction

A close correspondence has been demonstrated between task specific modulations in the Blood Oxygenation Level Dependent (BOLD) functional Magnetic Resonance Imaging (fMRI) response and neuro-electrical effects detectable by magnetoencephalography (MEG) (Muthukumaraswamy et al., 2009, Stevenson et al., 2011, Winterer et al., 2007). The precise relationship between these two measures, however, remains unclear (Winterer et al., 2007). Despite the disparate nature of the two phenomena, similarities in the spatial distribution and amplitude modulation of haemodynamic and neuro-electrical effects have been attributed to a common source of underlying activity. Synaptically generated Local Field Potentials (LFPs) have been shown to be a good predictor of the BOLD fMRI response (Logothetis et al., 2001), and similarly, the MEG detectable signal is thought to arise from post synaptic current flow in apical dendrites (Hämäläinen et al., 1993). Work by Murakami and Okada (2006) suggests that the MEG signal reflects ensembles of ~ 50,000 synchronously active neurons, with post synaptic current flow in pyramidal cells in layers II/III and V dominating the signal.

Paired pulse stimulus paradigms have been used extensively in invasive electrophysiological studies (Deisz and Prince, 1989, Xu et al., 2009) to systematically manipulate facilitation and inhibition of post synaptic potentials, and consequently may prove informative in understanding what is driving both neuro-electrical and haemodynamic effects.

Intracellular recordings have shown that Inhibitory Post Synaptic Currents (IPSCs) exhibit a strong sensitivity to interstimulus interval (ISI) in paired pulse paradigms, displaying significant GABA mediated depression of the second pulse for ISIs of 100 ms to 1000 ms (Xu et al., 2009). A scalp level correlate of this effect has been reported in MEG (Karhu et al., 1994, Wikström et al., 1996) where increasing the frequency of stimulation of a train of median nerve pulses from 0.2–6.7 Hz resulted in an equivalent reduction in amplitude of the P35m component of the somatosensory signal averaged evoked response in S1, suggesting that the P35m may reflect IPSCs.

In addition to a phase-locked evoked response, which typically lasts ~ 300 ms, electrical stimulation of the median nerve is also known to generate time-locked modulations in μ (8–14 Hz) and β (13–30 Hz) oscillatory power in the sensorimotor cortex that persists for up to ~ 2 s following stimulation (Neuper and Pfurtscheller, 2001, Nikouline et al., 2000, Salmelin and Hari, 1994, Salenius et al., 1997). The characteristic modulation of sensorimotor β oscillatory rhythms in response to finger movement or median nerve stimulation displays two distinct phases: during stimulation a loss in power, termed Event Related Desynchronisation (ERD), is measured and has been associated with active cortical processing (Della Penna et al., 2004); this is followed on stimulus cessation by an increase in β power, referred to here as Event Related Synchronisation (ERS), the quantification of which is thought to provide a measure of inhibition of the motor cortex (Pfurtscheller et al., 2002). The functional role of β oscillations in the sensorimotor cortex is not fully understood, however recent studies (Gaetz et al., 2011, Hall et al., 2011, Jensen et al., 2005) suggest that these oscillations are also governed by GABAergic modulation.

There is increasing evidence to suggest that β ERD and ERS arise from distinct neural generators, although the precise mechanisms giving rise to each of the effects are as yet unclear. Motor studies suggest a spatial distinction between these oscillatory rhythms, with the ERD being localised to the post central gyrus (Della Penna et al., 2004), and the ERS localised to the pre-central gyrus (Jurkiewicz et al., 2006). Functional differences between the modulations of ERD and ERS have also been reported (Hall et al., 2011), movement related ERD was shown to be facilitated by increased GABAergic drive following administration of diazepam, whilst no equivalent effect was seen in the post stimulus ERS. As such, one may expect a differential manifestation of the GABA mediated paired pulse depression effect in the amplitudes of β ERD and ERS respectively.

Ultra-high field (UHF), defined here to be 7 T, affords high SNR and increased BOLD sensitivity (Gati et al., 1997). Furthermore, the BOLD response is weighted to microvasculature with little contribution from the intravascular signal from veins due to the disproportionate shortening of the T₂ of venous blood compared to that of tissue (Yacoub et al., 2001). As such it is expected that at 7 T the BOLD response lies closer to the true site of neural activity. It has been shown (Jueptner, 1995) that both excitatory and inhibitory synaptic processes modulate energy consumption within the brain, and as such both should be detectable with BOLD fMRI. Indeed, Kampe et al. (Kampe et al., 2000) reported an increase in the extent of the BOLD active area in contra-lateral S1 with increasing frequency of median nerve stimulation (5–100 Hz) which they attributed to an increase in underlying inhibitory activity.

In this study we use non-painful median nerve stimulation in a paired pulse paradigm with varying ISI to examine the effects on the phase-locked evoked response (specifically the P35m component), the ongoing oscillatory activity in the β (13–30 Hz) band, and the BOLD fMRI response. We sought to characterise the spatio-temporal signature of the electrodynamic response to a median nerve pulse and whether ISI related power modulations in MEG detected neural responses were detectable in the spatial–temporal distribution of the corresponding BOLD haemodynamic data. We also examine how the oscillatory response to a subsequent pulse was affected by the magnitude and shape of the induced decrease and rebound in β-band power induced by the first pulse, to test whether paired pulse depression effects manifest themselves in the oscillatory activity.

This paper examines both (i) spatial and (ii) temporal aspects of the MEG and UHF BOLD fMRI signals using a paired pulse median nerve paradigm to test the following hypotheses:

• The spatial localisation of ERD and ERS in the β band differs.

• ISI differentially modulates the amplitude of ERD and ERS.

• Non-linearity in β activity and P35m response resulting from variations in ISI will be reflected in both the amplitude and spatial distribution (if ERS and ERD arise from spatially separate neural generators) of the BOLD response.