EEG-based visual deviance detection in freely behaving mice

The mouse is widely used as an experimental model to study visual processing. To probe how the visual system detects changes in the environment, functional paradigms in freely behaving mice are strongly needed. We developed and validated the first EEG-based method to investigate visual deviance detection in freely behaving mice. Mice with EEG implants were exposed to a visual deviant detection paradigm that involved changes in light intensity as standard and deviant stimuli. By subtracting the standard from the deviant evoked waveform, deviant detection was evident as bi-phasic negativity (starting around 70 ms) in the difference waveform. Additionally, deviance-associated evoked (beta/gamma) and induced (gamma) oscillatory responses were found. We showed that the results were stimulus-independent by applying a “flip-flop” design and the results showed good repeatability in an independent measurement. Together, we put forward a validated, easy-to-use paradigm to measure visual deviance processing in freely behaving mice.

Male C57BL/6J mice (n=13) were used to implement and validate the newly developed 88 vMMN paradigm. Animals were single-housed in individually ventilated cages for at least 89 one week prior to surgeries and maintained on a 12:12 light-dark cycle with ad libitum 90 access to food and water. All experiments were approved by the Animal Experiment  Europe, Almere, the Netherlands) were used to attach electrodes to the skull. Post-107 operative pain relief was achieved by a subcutaneous injection of Carprofen (5 mg/kg). 108 EEG recordings started after a 14-day recovery period.

Analysis
No animals had to be excluded on the basis of low signal quality as judged from the 158 baseline assessment of stimulus responses. For two animals, positive-negative inverted 159 signals were evident on one of the visual cortex electrodes (once right V1 and once left 160 V1); these electrodes were excluded from analysis. Next, recordings were manually 161 checked to exclude recording periods with artefacts, as well as periods of sleep, as 162 deviance detection is known to be attenuated or even absent in non-REM sleep 163 (Sculthorpe et al., 2009 waves of the right and left V1 electrode (using cluster-based permutation analysis) did 188 not reveal any time windows of significant differences (data not shown). In subsequent 189 analyses VEPs from the right and left electrode were averaged. Averaging over trials, 190 electrodes and recordings was performed for the data from individual animals before 191 performing any statistical analysis for the data-sets across animals.

192
To test whether vMMN was significantly different from zero cluster-based 193 permutation analysis was used as previously described (Maris and Oostenveld, 2007). In   However, in these cases permutations were performed by randomly exchanging the data 215 between the two conditions in the comparison. As small numbers of clusters were found 216 for the comparison of the first and second recording, for these data all clusters that were 217 found in the bootstrap were pulled together; for the deviants, no bootstrap was used but 218 all 63 deviants were simply compared between the first and second recording.  Hz linear steps. The number of cycles increased from 2 to 10 with increasing frequency. 234 Next, power was converted to a log10 scale and an absolute baseline correction was 235 performed using a window from 200 until 100 ms before stimulus onset as the baseline.       contained additional early latency components between 20 and 60 ms that were not 375 evident in the VEP in response to an intensity decrease (Fig. 3A).         Oscillatory activity can be divided into evoked power, which is the direct 518 frequency representation of the VEP waveform response, and induced power, which is 519 the oscillatory activity that is non-phase-locked to the stimulus and thus not found in the 520 25 VEP waveforms (Jones, 2016). To asses which oscillatory clusters in our analysis 521 represented induced power, the time-frequency analysis was also ran after subtracting 522 the average VEP waveform from every single trial per condition (Park et al., 2018). 523 Clusters 3 and 4 in the TFR and cluster A in the difference plot were not present when 524 using this analysis (Fig. 6B)    thus interpretations related to these specific frequency bands should be drawn carefully.

738
In conclusion, we developed the first, robust and repeatable vMMN paradigm

981
response of the light decrease, concern a response to an increase in light intensity. The On-response of the 982 light decrease, as well as the Off-response of the light increase, concern a response to a light intensity 983 decrease. Presented data show the responses to a light increase and decrease, averaged over all mice for 984 right and left V1 responses and 2 recordings on separate days.