Concurrent Recording of Co-localized Electroencephalography and Local Field Potential in Rodent

Although electroencephalography (EEG) is widely used as a non-invasive technique for recording neural activities of the brain, our understanding of the neurogenesis of EEG is still very limited. Local field potentials (LFPs) recorded via a multi-laminar microelectrode can provide a more detailed account of simultaneous neural activity across different cortical layers in the neocortex, but the technique is invasive. Combining EEG and LFP measurements in a pre-clinical model can greatly enhance understanding of the neural mechanisms involved in the generation of EEG signals, and facilitate the derivation of a more realistic and biologically accurate mathematical model of EEG. A simple procedure for acquiring concurrent and co-localized EEG and multi-laminar LFP signals in the anesthetized rodent is presented here. We also investigated whether EEG signals were significantly affected by a burr hole drilled in the skull for the insertion of a microelectrode. Our results suggest that the burr hole has a negligible impact on EEG recordings.


Introduction
It is generally accepted that LFPs recorded via microelectrodes primarily reflect the weighted sum of synchronized excitatory and inhibitory synaptic activities of local pyramidal neural populations 1,2,3,4 . Our recent research demonstrated that the profile of the LFP signal could be separated into components of excitation and inhibition 5,6 . However, as LFP is normally measured via an invasive procedure, it is not suited for most studies of the human brain.
On the other hand, EEG is a non-invasive technique for measuring the electrical activity of the brain. It is widely used as a diagnostic tool for certain types of neurological diseases such as epilepsy, and as a research tool in human cognitive studies. Despite its popularity, a major limitation of EEG is the inability to interpret its temporal profiles precisely in terms of the underlying neural signals 7,8,9 . Increasingly, mathematical models of EEG are developed to enhance understanding of brain function 10,11,12,13,14,15 . Most of the existing EEG models are developed based on fitting frequency domain characteristics of the model predicted output to the EEG data spectrum during spontaneous activity, and very few EEG models can generate realistic sensory evoked potentials. In this context, concurrent recordings of EEG and LFP will provide important insight and constraints for developing more accurate mathematical models of EEG.
To address this need for concurrent recordings to further explore the neural origin of EEG, we developed a methodology to simultaneously record EEG and multi-laminar LFP signals in the neocortex of the anesthetized rat. The setup is similar to previous concurrent EEG/LFP studies conducted in primates 16,17 . We further investigated the effect of a burr hole drilled into the skull on EEG recordings surrounding the hole, by comparing bilateral EEG recordings (i.e., one hemisphere with a burr hole, the other hemisphere intact) in the absence of sensory stimulation. Our results demonstrate that concurrent EEG/LFP recordings can be conducted simply and effectively, with little EEG signal distortion from the burr hole in the skull. The setup consists of a modified nose cone for ease of whisker pad stimulation under isoflurane anesthesia, two stimulating electrodes inserted into the whisker pad, a spider electrode positioned on the skull above the barrel cortex contra-lateral to the stimulating electrodes, a multi-channel microelectrode inserted into the barrel cortex through the spider electrode, and reference electrodes placed inside an incision at the back of the rat's neck. (B) A view through the microscope of the spider electrode securely positioned onto the skull by EEG paste. The microelectrode is inserted into a burr hole drilled into the skull under the spider electrode. The scalp is held back by surgical thread (suture) tied to the stereotaxic frame. Please click here to view a larger version of this figure.
6. Smear EEG paste onto the reference electrode for the EEG and place it securely inside the incision at the back of the rat's neck. 7. Connect the EEG electrodes to the preamplifier via a passive signal splitter for low impedance signals (Figure 2). Make sure the impedance of the spider electrode is below 5 kΩ. If it is not, check that the EEG paste is in good contact with the skull and the electrode is firmly pressed to the EEG paste. Add more EEG paste if necessary. 8. Mount a micromanipulator arm on the stereotaxic frame. Connect a linear 16-channel microelectrode (100 µm spacing, area of each site 177 µm 2 ) to a 16-channel acute headstage clipped securely onto the micromanipulator arm. 9. Smear EEG paste onto the reference electrodes for the EEG and microelectrode, then securely place them inside the incision ( Figure 1A). 10. Adjust the angle of the micromanipulator arm so that the microelectrode is perpendicular to the cortical surface. This angle is normally between 25-35 °. 11. Lower the microelectrode under a microscope by turning the micromanipulator knobs so that the tip of the microelectrode is aimed at the tiny opening at the bottom of the burr hole until the uppermost electrode just penetrates the cortical surface. Care must be taken to avoid forcing the microelectrode onto the surface of the dura as this would break the electrode. 12. Couple the 16-channel microelectrode to a preamplifier connected to a data acquisition unit via a fiber optic cable (Figure 2). 13. Turn on the preamplifier, the data acquisition unit, and the computer connected to the unit. Turn on the stimulator box. 14. Insert the microelectrode normally to the cortical surface by slowly turning the z-axis knob of the micromanipulator to a depth of 1,500 µm 20 .
15. Micro-adjust the depth by applying a train of stimulus to the whisker pad and observing the 16-channel evoked LFP on a PC monitor using the software of the data acquisition unit installed on the PC. Carefully turn the z-axis knob on the micromanipulator until the highest amplitude of the evoked LFP occurs around channel 7 (as this coincides with layer IV in the cortex). NOTE: Ipsi-lateral EEG electrode setup: For some experiments, a second spider electrode was placed on the ipsi-lateral side of the intact skull above the barrel cortex. This setup allowed bilateral EEG recording during the resting state to investigate the effect of the burr hole on the EEG signal. NOTE: The surgical procedure to set up the EEG electrode is identical to that described above, except that during step 2.6, the temporalis muscle on each side of the head was carefully separated from the skull, sutured back and tied securely to the corresponding side of the stereotaxic frame. NOTE: The concurrent EEG/LFP setup is also identical to that described above, with an additional step that a second spider electrode is loaded with the EEG paste, then pressed firmly to the skull above the ipsi-lateral barrel cortex.    9. Check the impedance of the EEG probe(s) by pressing the middle icon in the '2: EEG' panel. If too high, add more EEG paste to the probe. Press 'OK' to return to the original display. 10. Wait 20 s to avoid recording the initial fluctuation of the EEG recordings. 11. Go back to the PC monitor (after the 20 s wait) and press the 'Return' key on the keyboard. The EEG and the LFP signals will be recorded.

Pre-process the evoked LFP and EEG signals on a trial-by-trial basis using the following steps.
1. Shift back the neural data in time by 20 samples (equivalent to 0.82 ms). This is the delay produced by the circuit used to collect neural data in TDT itself. By shifting the data, the time zero point is aligned to the onset of the stimulus. 2. Remove the stimulus artifact by replacing the neural data from 0 to 1 ms with a straight line connecting the data point at 0 ms with the data point at 1 ms. 3. Zero-mean each trial by subtracting the mean value of the neural signal 200 ms prior to stimulus onset. 4. Low-pass filter the data below 800 Hz using a 4 th order Butterworth IIR type filter in both directions to avoid introducing any temporal shift in the data. 5. Align the multi-laminar data across animals. For each animal's LFP data, apply the inverse Current Source Density (spline iCSD, source radius R = 0.5 mm) analysis 21 with a Gaussian filter (λ = 50 µm) to locate the layer IV sink 1 , which is given by the largest negative peak occurring at a cortical depth below the pial surface within the first 15 ms of stimulus onset. The CSD, and the corresponding LFP, data are then aligned according to their sink locations across animals. The common sink is located in layer IV, 6 00 µm below the pial surface. 6. After alignment, use channels 2, 7, and 12 of the realigned LFP as representatives of neural responses of the supragranular, granular, and infragranular layers, respectively in the barrel cortex.
2. Calculate the mean evoked LFP and EEG by averaging the pre-processed data over 100 trials. 3. To investigate the effect of the burr hole on the EEG, down-sample the EEG signals to 1,000 Hz, and compute the power spectral density (PSD) for the contra-lateral (with a hole in the skull) and ipsi-lateral (intact skull) spider electrode recordings over a 250 s period of resting state. PSD is computed from 0.1-100 Hz in Matlab using the function 'pmtm' which is based on the multitaper method where P c and P i are the average PSD of the contra-and ipsi-lateral EEG, respectively in the frequency band of interest. 6. Within each frequency band, perform a one-sample t-test to test the hypothesis that there is no significant difference (at 0.05 significance level) between the PSD of the EEG signal recorded from the two hemispheres.

Discussion
We have described an experimental procedure for concurrent recording of co-localized EEG and LFP signals of an isoflurane anesthetized rat in response to whisker pad stimulation. A microelectrode was inserted into the neocortex through an opening in the EEG spider electrode which was aligned with a burr hole drilled into the skull. The electrode was secured to the skull by a conductive and adhesive EEG paste 23 . The nose cone used for the administration of isoflurane was modified so that stimulating electrodes could be inserted into the whisker pad with ease.
The EEG paste was effective at mounting the spider electrode securely to the skull, while providing excellent electrical conductivity throughout the experimental day without the need for additional application of paste. It replaced the undesirable use of glue to fix the periphery of the spider electrode to the skull, as glue is non-conductive and can increase the impedance of the electrode if it runs between the skull and the electrode. EEG paste has a number of advantages over EEG gel, which is difficult to shape around the burr hole and can dry out during experiment, resulting in poor EEG signals.
Several critical steps in the protocol need special attention. The first is the insertion of the microelectrode through the burr hole. As the dura is otherwise intact, the precision of the insertion is crucial. A slight resistance at the tip of the electrode usually means the electrode is not positioned correctly. It must be raised, position adjusted, and re-inserted. The second is the position of the nose cone on the rat. It must not be too loose, as isoflurane will escape from the cone. It also must not be too tight, as this can obstruct the nostrils of the rat and cause difficulty breathing. Special attention is also required to ensure that the amplitude of the EEG recording is much smaller (usually 5 to 10 times smaller) than the LFP top channel recording. If they are similar, it is an indication that the EEG probe has come into direct or indirect contact with the microelectrode. An indirect contact is usually through the cerebral spinal fluid (CSF) that sometimes fills the hole drilled in the skull. The conductivity of CSF is typically 100 times that of the skull 24,25 . Thus, if the level of CSF inside the burr hole is sufficiently high, it can make contact with the spider electrode. To avoid this, the hole should be frequently cleaned with super absorbent cotton sponges such as the absorption spears.
The effect of a burr hole (diameter <2 mm) in the skull on the EEG recording surrounding the hole was studied by placing another spider electrode on the intact skull atop the ipsi-lateral barrel cortex so that bilateral EEG recordings could be compared. The results shown in Figure  9 and Figure 10, suggest the effect to be insignificant at the 0.05 level of significance. Other factors affecting the amplitude of the EEG include how well the EEG paste was in contact with the skull, how firm the electrode was pressed to the paste, and the spatial extent of the EEG paste on the skull.
It is also worthwhile to note that the protocol described here recorded skull EEG, which is different from scalp EEG used in human EEG studies. The scalp acts like a resistor or a low-pass filter, which will reduce the signal-to-noise ratio of the EEG recording further.
Finally, comparison of the temporal dynamics of the ERP and that of the evoked LFP across cortical layers suggest that somatosensory evoked potential reflects better the LFP in the supragranular layer of the cortex than that in the granular and infragranular layers. This is in agreement with our earlier work 6 , demonstrating that the initial segment (P1) of the ERP is related to the return current arising from the inflow of the excitatory synaptic current occurring in the granular layer, while the subsequent decrease (N1) in ERP may be related to the delayed arrival of thalamic afferent to cortical layers II/III and/or feedforward signals from deeper cortical layers. In conclusion, concurrent recordings of EEG/LFP can enhance understanding of the neural genesis of EEG, and facilitate the mathematical modeling of EEG in terms of neural signals across cortical layers.

Disclosures
The authors have nothing to disclose.