Completed on 21 Nov 2017 by Timo Van Kerkoerle . Sourced from https://www.biorxiv.org/content/early/2017/11/18/202564.
Login to endorse this review.
Author response is in blue.
Congratulations on the great paper! Very impressive work.
There are some inaccuracies which I thought would be good to point out. I can imagine that some people thought this represents the first 'solid evidence' about the brain mechanisms of alpha because it is stated in the manuscript that previous papers use a global reference? This is not the case for our 2014 PNAS paper. For the laminar recordings we referenced to the metal shaft of the probe, which is right next to the contact sites, giving something like a bipolar LFP. We compared this with a silver/silver chloride wire in the recording chamber (as also noted in the paper), which didn't change the results. Furthermore, the crucial finding that implicated the involvement of layer 5 in the alpha rhythm (figure 4 of the PNAS paper) are based on CSD and MUA signals, for which the reference is not relevant. The V1-V4 data (not V2 as mentioned in the discussion) used bipolar LFP signals as well.
Also, it might be good to note that only in early visual areas alpha has been shown to be strongest in the deep layers. In particular, Bollimunta, Chen, Schroeder and Ding J. Neuroscience 2008 showed that alpha was only strongest in the deep layers for V1 and V4, but in the superficial layers in IT.
Hi Prof. Kerkoerle,
Thank you for your comments! Your PNAS paper was often drawn on while writing ours; thank you for catching our confusion of V4 with V2.
However, I don’t think that referencing to the shank is equivalent to a bipolar reference – in fact, it seems most similar to an average reference, as the shank will be sensitive to dipoles throughout the cortical depth. Therefore, contacts referenced to the shank will reflect a mix of both local and global signals. Furthermore, your observation that shank and monopolar (wire) referencing yielded equivalent results speaks against shank referencing being the same as a bipolar, as recent work has shown that bipolar and monopolar references yield very different alpha profiles in V1 (see Haegens 2016). Lastly, the ‘bipolar’ reference used in the V1-V4 portion of your paper is not a standard bipolar referencing scheme (i.e. you do not reference adjacent contact pairs to each other); instead, you reference all of the contacts to a single (unspecified) channel on the array. This makes the spatial properties of the data difficult to interpret, as you’re then integrating currents between each channel and your reference, rather than measuring local activity between pairs of adjacent contacts.
While it’s true that some of your results used CSD, this was only after aligning the data to infragranular alpha peaks derived from the non-local LFP. I found it unclear whether or not the CSD-MUA coherencies were derived from monopolar-aligned data, but even if they aren’t, the difference in the coherence of superficial and deep alpha CSD with MUA appears relatively small (Fig. S9) and isn’t (that I could find) reported to be statistically different. I’m also concerned about the effect of averaging MUA across several channels prior to finding the CSD-MUA coherence, rather than reporting the CSD-MUA coherencies for each channel pair (see our Fig. 3). Lastly, I think it’s more difficult to draw conclusions about alpha generation from CSD-MUA coherencies rather than the most straightforward metric, alpha power versus depth.
While Bollimunta’s results are definitely interesting, I’m not sure how to directly compare them to our findings for several results. First, they aligned their data to the monopolar LFP prior to further analysis, thus biasing their CSD profiles and CSD-MUA coherence to sources which are maximum in that (volume-conduction prone) configuration. While their second major analysis (Granger Causality between bipolar LFPs) was interesting, I think it’s trickier to interpret than just CSD alpha power across the cortical depth (which is not reported), especially when causality between only one contact pair in each laminar compartment is measured. I’d also note that Haegens 2016 (using bipolar and CSD derivations) found that V1 alpha is strongest in superficial layers, consistent with our data indicating that alpha is also supragranular in associational cortices.
Thank you for the feedback!