Parametric variation of gamma frequency and power with luminance contrast: A comparative study of human MEG and monkey LFP and spike responses

Highlights

• Parametric effects of contrast on gamma oscillation frequency in human MEG

• Comparable effects of contrast on gamma oscillation frequency in human MEG, monkey LFP and single unit spike trains

• Comparable effects of contrast on a novel metric of gamma oscillations; spectral asymmetry, in the three data sources

• Different effects of contrast on gamma power in MEG data compared to LFP and single unit spike trains

Abstract

Gamma oscillations contribute significantly to the manner in which neural activity is bound into functional assemblies. The mechanisms that underlie the human gamma response, however, are poorly understood. Previous computational models of gamma rely heavily on the results of invasive recordings in animals, and it is difficult to assess whether these models hold in humans. Computational models of gamma predict specific changes in gamma spectral response with increased excitatory drive. Hence, differences and commonalities between spikes, LFPs and MEG in the spectral responses to changes in excitatory drive can lead to a refinement of existing gamma models. We compared gamma spectral responses to varying contrasts in a monkey dataset acquired previously (Roberts et al., 2013) with spectral responses to similar contrast variations in a new human MEG dataset. We found parametric frequency shifts with increasing contrast in human MEG at the single-subject and the single-trial level, analogous to those observed in the monkey. Additionally, we observed parametric modulations of spectral asymmetry, consistent across spikes, LFP and MEG. However, while gamma power scaled linearly with contrast in MEG, it saturated at high contrasts in both the LFP and spiking data. Thus, while gamma frequency changes to varying contrasts were comparable across spikes, LFP and MEG, gamma power changes were not. This indicates that gamma frequency may be a more stable parameter across scales of measurements and species than gamma power. The comparative approach undertaken here represents a fruitful path towards a better understanding of gamma oscillations in the human brain.

Introduction

In monkey visual cortex, gamma oscillations have been studied for many years, and are thought to mediate a number of computational functions (Buzsáki and Wang, 2012, Engel et al., 1999, Fries, 2009, Fries et al., 2007, Tiesinga and Sejnowski, 2010). Gamma oscillations are thought to be related to local inhibition. Because of the inhibition, excitatory neurons show low excitability in general (Buzsáki, 2006). However, the inhibition is thought to vary over time leading to a rhythmic modulation of neuronal excitability, thereby defining windows of opportunity for neurons to fire (Buzsáki and Draguhn, 2004, Csicsvari et al., 1999, Lakatos et al., 2005). The length of the information–integration phases, the spacing between them, and the duration of the firing phases, is governed by the frequency of the oscillation. Hence, communication among different neural populations is enabled by an appropriate and consistent relationship in time among these windows of opportunity (Fries, 2005), which requires a sufficient match in the oscillatory frequencies between communicating populations (Roberts et al., 2013). More generally, the theory of weakly coupled oscillators predicts that similar frequencies are a prerequisite for synchronization of nearby neuronal populations (Lowet et al., 2015). Thus, oscillation frequency is a network parameter of utmost functional relevance and theoretical interest (Barardi et al., 2014, Cohen, 2014, Colgin et al., 2009). This is supported by correlations between gamma peak frequency and relevant parameters including perceptual performance (Edden et al., 2009), age (Gaetz et al., 2012), and cortical structure (Schwarzkopf et al., 2012). Power is the other highly relevant feature of spectral responses. Studies in monkeys (Jia et al., 2013, Ray and Maunsell, 2010, Roberts et al., 2013) have shown that power tends to be maximized around a mid-level of contrast or excitatory drive, and a reduction of power at much higher or lower levels of drive is observed suggesting an optimization of resonance at those contrasts. At the local network level, such phenomena require a specific architecture of the network, and hence the study of power modulations as a function of excitatory drive provides another important probe into the underlying gamma generating mechanisms. Further, in spatially-aggregate signals power provides an indirect measure of the strength of the underlying synchronization (Hadjipapas et al., 2009, Musall et al., 2014).

Advances in MEG beamformer source reconstruction have enabled the study of gamma oscillations localized to human visual cortex (Adjamian et al., 2004, Hall et al., 2005, Hoogenboom et al., 2006). Using these methods, several MEG studies reported human gamma oscillations to exhibit marked stimulus specificity (Adjamian et al., 2004, Duncan et al., 2010, Hadjipapas et al., 2007, Hall et al., 2005, Koelewijn et al., 2010, Muthukumaraswamy and Singh, 2008, Perry et al., 2013, Swettenham et al., 2009), paralleling neurophysiological findings in the macaque (Berens et al., 2008, Friedman-Hill et al., 2000, Frien et al., 2000, Gieselmann and Thiele, 2008, Gray and Singer, 1989, Henrie and Shapley, 2005). Although these stimulus-dependent effects in monkeys and humans were recorded in experiments that are difficult to compare, they suggest that gamma may be generated by neural mechanisms, which if not in common, are at least partially shared across the two species and across recording modalities. Yet, it is still not clear whether characteristics of gamma-band activity observed in invasive macaque recordings can be linked in a straightforward manner to characteristics observed from human non-invasive recordings. More specifically, can observed modulations in gamma frequency and power in a human MEG experiment be thought of as equivalent to frequency and power modulations in monkey electrophysiological recordings? This is a non-trivial question of crucial importance for non-invasive EEG/MEG studies on human brain functions that base their hypothesis on prior monkey electrophysiology studies.

Here we investigated whether the functional behavior of gamma oscillation frequency and power in the human is consistent with the behavior observed in the macaque and further whether it is a function of the scales of measurement (spikes, LFP, MEG). To this end, it was necessary to identify a stimulus parameter with robust and well-understood effects on gamma frequency and manipulate this parameter in both human and non-human primates. In computational studies, gamma is often modeled in terms of an interaction between excitation (E) and inhibition (I). In Pyramidal Interneuron Gamma (PING) networks (Tiesinga and Sejnowski, 2009), spikes from E-neurons quickly lead to spiking in I-neurons, which in turn briefly shut down the network. The recurrence of this process leads to oscillations, which are in the gamma range due to the time constants of the neurons involved (Börgers and Kopell, 2005). Increasing E-drive leads to a faster recovery from inhibition, resulting in a higher oscillation frequency (Brunel and Wang, 2003, Buia and Tiesinga, 2006, Tiesinga and Sejnowski, 2009, Traub et al., 1996). The gamma frequency increase predicted to follow from increased E-drive has been confirmed in neurophysiological studies in the macaque, in which gamma oscillation frequency as measured in the LFP recorded from early visual cortex was shown to increase strongly with stimulus contrast (Jia et al., 2013, Ray and Maunsell, 2010, Roberts et al., 2013). Recently contrast-induced frequency shifts have also been shown in human MEG (Perry et al., 2014). Moreover, PING models typically yield an optimal frequency at which gamma power is maximized, with lesser power for frequencies below or above the optimal frequency (Lowet et al., 2015, Roberts et al., 2013). Stimulus contrast can be thought of as an experimentally accessible proxy for E-drive to primary visual cortex (Contreras and Palmer, 2003, Sclar et al., 1990). Hence, parametric variation of contrast is a paradigm that is theoretically well understood such that it may permit the identification of biophysical mechanisms.

We applied this rationale to a human MEG experiment, which was designed for comparison with a previous study in macaque V1 conducted in our group (Roberts et al., 2013). Designing the MEG experiment for comparative purposes was not trivial: on the one hand, one needs to match the monkey and human experiments as closely as possible; at the same time due to the low gamma SNR in the MEG for a number of experimental conditions, some parameters in the MEG experiment need to be adapted such that sufficient SNR could be obtained. Despite the resulting limitations in achieving a perfect comparison, the study benefited from the use of a visual parameter in both studies of which the effects on gamma are well-understood, and from identical spectral analyses of human and macaque data. In addition, in the macaque data we also analyzed the gamma response in single-unit spike trains as a function of visual contrast. We found that increases in visual contrast led to similar shifts in gamma frequency in monkey single unit spiking, in monkey LFP, and in human MEG. Further, we found spectral asymmetries, which were consistently modulated by contrast in all recording modalities (single unit spikes, LFP, MEG). Alongside the similarities between MEG, LFP and spikes, we also observed a major difference: whereas there was a linear increase in the gamma oscillation power with stimulus contrast in the MEG data consistent with previous observations (Hall et al., 2005, Perry et al., 2014), in the LFP data, saturation and even reduction of power were observed at high contrasts.

Overall, the pronounced similarity of these findings across measurement scales and across human and macaque suggests major similarities in the generative model of gamma-band activity in macaque monkeys and humans. However, the pronounced difference in contrast-dependent power modulation indicates that MEG-measured gamma power cannot be easily equated to LFP gamma power. These results as well as theoretical considerations indicate that oscillation frequency might be a more robust parameter for comparing findings across species and spatial scales of measurement. At the same time it will be necessary to explore whether power differences are related to differences in gamma generating mechanisms between macaque monkeys and humans, or to potential differences in signals picked up during invasive neurophysiological recordings and MEG measurements.