No evidence for entrainment: endogenous gamma oscillations and responses to rhythmic visual stimulation coexist in visual cortex

Motivated by the plethora of studies associating gamma oscillations (∼30-100 Hz) with various neuronal processes, including inter-regional communication and neuroprotection, we asked if endogenous gamma oscillations in the human brain can be entrained by rhythmic photic stimulation. The photic drive produced a robust Magnetoencephalography (MEG) response in visual cortex up to frequencies of about 80 Hz. Strong, endogenous gamma oscillations were induced using moving grating stimuli as repeatedly shown in previous research. When superimposing the flicker and the gratings, there was no evidence for phase or frequency entrainment of the endogenous gamma oscillations by the photic drive. Rather – as supported by source modelling – our results show that the flicker response and the endogenous gamma oscillations coexist and are generated by different neuronal populations in visual cortex. Our findings challenge the notion that neuronal entrainment by visual stimulation generalises to cortical gamma oscillations.


Introduction 23
Neuronal cell assemblies have long been known to synchronise their discharges with millisecond preci-24 sion (Buzsáki et al., 1992;Traub et al., 1996;Singer, 1999;Varela et al., 2001). This synchronisation has 25 been linked to oscillatory activity in the gamma-frequency band (∼30-100 Hz) in various brain regions 26 and species, e.g in rodents and primates (e.g. Eckhorn et al., 1988;Gray & Singer, 1989; Engel et al.,   Photic drive induces responses up to ∼80 Hz 103 We next set out to quantify the rhythmic response to the flicker as a function of frequency in the flicker 104 condition, in which stimulation was applied to an invisible patch. Figure 2 A Figure 1. C Grand-average of the responses to the photic drive for each flicker frequency. On average, the magnitude of the flicker response decreases with increasing frequency, and is identifiable for stimulation below 80 Hz. Figure 3A displays the power spectra in the flicker condition, estimated from the TFRs as explained 115 above, averaged over all participants, as a function of stimulation frequency. These are equivalent to 2C.

116
Diagonal values indicate the magnitude of the oscillatory responses (relative to baseline) at the stimulation 117 frequencies, reaching values of up to 300% and decreasing monotonically with frequency. This confirms 118 an upper limit for the stimulation of around 80 Hz. Off-diagonal values indicate oscillatory activity at 119 frequencies different from the stimulation frequency. Figure 3B shows the same spectra after aligning to the 120 individual IGFs, prior to averaging. Figure 3C and D display the spectra in the flicker&gratings condition 121 7 (averaged in the 2.25 -3.75s interval), during which the photic drive was applied to the moving grating 122 stimulus (see Figure 10B). The induced gamma band activity can be observed as the horizontal yellow band 123 at ∼60 Hz. When aligning the spectra to the IGF ( Figure 3D), we observe a decrease in the flicker response 124 but no evidence for an amplification at or close to the IGF. The flicker condition after the spectra were aligned to the IGF. C The flicker&gratings condition. All spectra demonstrate both the flicker response and induced gamma oscillation (observed as the yellow/orange horizontal band). Again, the amplitude of the rhythmic stimulation response appears to decrease with increasing frequency in both conditions. D The spectra for the flicker&gratings condition now aligned to the IGF. There is no indication that the rhythmic flicker captures the endogenous gamma oscillations.
Magnitude of flicker response decreases as a function of frequency 126 The averaged TFRs of power in Figure 3 Figure 4A,B depicts the phase-locking value (PLV) between the photodi-132 ode and the MEG signal at the SOI (planar gradiometers, not combined). This measure reveals a systematic 133 decrease in phase-locking with increasing flicker frequency for both the flicker (orange) and flicker&gratings 134 (blue) condition (A). The observed relationship is preserved when aligning the frequencies to the IGF (B, 135 also see Table 1). Note the absence of increased phase-locking at the IGF. The magnitude of the flicker  Table 1 and dotted lines in Figure 4). We then identified 146 the individual peak frequencies, eliciting the strongest response to the flicker in the flicker&gratings condi-147 tion 4E, and related those to the IGF, as seen in Figure 4F. Importantly, the frequency inducing the strongest 148 response to the rhythmic drive was below the IGF in the majority of participants, whereby the frequencies 149 turned out to be uncorrelated (r(21)=-0.15, p =0.5, flicker&gratings condition).  The phase-locking values between the photo-diode and the MEG signals as a function of frequency after the spectra were aligned to IGF. Again, the phase-locking decreases with increasing frequency (see Table 1 for a statistical quantification of the simple linear regression models). C Relative power change with respect to baseline as a function of frequency. Generally, the power decreased with frequency, however, in the flicker&gratings there is an apparent peak at ∼56 Hz; yet, the shaded errors (SE) indicate considerable variance between participants. D The relative power spectra as a function of frequency after the individual spectra were aligned in frequency according to the IGF, demonstrating that responses to a photic drive at the IGF are not amplified. E Relative power change as a function of frequency for each individual subject (N = 22), indicates that the peak at ∼56 Hz in C is driven by comparably high power in that frequency range in just a few individuals. F Flicker frequency inducing highest power values versus IGF, demonstrating no systematic relationship (r(21) = −0.15, p = .5). Instead, the frequencies inducing maximum power change were below the IGF in the majority of participants. 10 Gamma oscillations and flicker response coexist 151 We initially hypothesised that entrainment of the gamma oscillations in the flicker&gratings condition would

156
The IGF (58 Hz for this subject) and the respective stimulation frequencies are indicated by dashed lines.

157
The endogenous gamma oscillations, induced by the moving grating stimulus, are observed as the sustained 158 power increase from 0 -6 s whereas the flicker response is demonstrated by a power increase at 2 -4 s.

159
The plots reveal that gamma oscillations persist at the IGF and coexist with the response to the photic drive, frequency. These phase slips will be observed as steps between the phase plateaus. We implemented the 208 phase plateau analysis to complement the PLV analysis in Figure 4, which quantifies the average synchrony 209 between photodiode and neuromagnetic signal over trials, but is not able to identify intermittent plateaus.  Figure 4A and B.   Figure 9: Source estimates using the LCMV beamformer approach mapped on a standardised MNI brain. A Source estimation of the visually induced gamma oscillations, with the peak of the source identified at MNI coordinates [0mm -98mm -7mm]. B Source estimation of the flicker response, with the average peak source at [3mm -96mm 2mm]. C Coordinates of the peak sources for all participants (small scatters) and grandaverage (large scatters) for the flicker&gratings and flicker condition (blue and orange, respectively), indicating that the gamma oscillations peak in brain areas inferior to the flicker response. D Difference between the z-coordinates (inferiorsuperior axis) of the peaks of the sources in both conditions, demonstrating an average difference of 8.5mm. A dependent sample t-test confirms this distance to be significant, t(21) = −5.12, p = 2.29e − 5***, r = .55,95% CI = [−Inf − 5.67].

269
In this MEG study, we explored resonance properties and entrainment of the human visual system to a rapid   in response to the stimulation, i.e. the moving grating and/or the photic drive, was calculated as: with P stim being the power during stimulation and P base being the power in the baseline interval. The where θ(t, n) = φ m (t, n) − φ p (t, n) is the phase difference between the MEG (m) and the photodiode (p) 503 signal at time bin t in trial n (see Lachaux et al., 1999, p.195 and Figure 4 and 8). whom there was no anatomical image available, was aligned to a standardised template brain.