Author response:
The following is the authors’ response to the original reviews.
Public Reviews:
Reviewer #1 (Public Review):
Summary:
In this study, Zhang et al., presented an electrophysiology method to identify the layers of macaque visual cortex with high density Neuropixels 1.0 electrode. They found several electrophysiology signal profiles for high-resolution laminar discrimination and described a set of signal metrics for fine cortical layer identification.
Strengths:
There are two major strengths. One is the use of high density electrodes. The Neuropixels 1.0 probe has 20 um spacing electrodes, which can provide high resolution for cortical laminar identification. The second strength is the analysis. They found multiple electrophysiology signal profiles which can be used for laminar discrimination. Using this new method, they could identify the most thin layer in macaque V1. The data support their conclusion.
Weaknesses:
While this electrophysiology strategy is much easier to perform even in awake animals compared to histological staining methods, it provides an indirect estimation of cortical layers. A parallel histological study can provide a direct matching between the electrode signal features and cortical laminar locations. However, there are technical challenges, for example the distortions in both electrode penetration and tissue preparation may prevent a precise matching between electrode locations and cortical layers. In this case, additional micro wires electrodes binding with Neuropixels probe can be used to inject current and mark the locations of different depths in cortical tissue after recording.
While we agree that it would be helpful to adopt a more direct method for linking laminar changes observed with electrophysiology to anatomical layers observed in postmortem histology, we do not believe that the approach suggested by the reviewer would be particularly helpful. The approach suggested involves making lesions, which are known to be quite variable in size, asymmetric in shape, and do not have a predictable geometry relative to the location of the electrode tip. In contrast, our electrophysiology measures have identified clear boundaries which precisely match the known widths and relative positions of all the layers of V1, including layer 4A, which is only 50 microns thick, much smaller than the resolution of lesion methods.
Reviewer #2 (Public Review):
Summary:
This paper documents an attempt to accurately determine the locations and boundaries of the anatomically and functionally defined layers in macaque primary visual cortex using voltage signals recorded from a high-density electrode array that spans the full depth of cortex with contacts at 20 um spacing. First, the authors attempt to use current source density (CSD) analysis to determine layer locations, but they report a striking failure because the results vary greatly from one electrode penetration to the next and because the spatial resolution of the underlying local field potential (LFP) signal is coarse compared to the electrical contact spacing. The authors thus turn to examining higher frequency signals related to action potentials and provide evidence that these signals reflect changes in neuronal size and packing density, response latency and visual selectivity.
Strengths:
There is a lot of nice data to look at in this paper that shows interesting quantities as a function of depth in V1. Bringing all of these together offers the reader a rich data set: CSD, action potential shape, response power and coherence spectrum, and post-stimulus time response traces. Furthermore, data are displayed as a function of eye (dominant or non-dominant) and for achromatic and cone-isolating stimuli.
This paper takes a strong stand in pointing out weaknesses in the ability of CSD analysis to make consistent determinations about cortical layering in V1. Many researchers have found CSD to be problematic, and the observations here may be important to motivate other researchers to carry out rigorous comparisons and publish their results, even if they reflect negatively on the value of CSD analysis.
The paper provides a thoughtful, practical and comprehensive recipe for assigning traditional cortical layers based on easily-computed metrics from electrophysiological recordings in V1, and this is likely to be useful for electrophysiologists who are now more frequently using high-density electrode arrays.
Weaknesses:
Much effort is spent pointing out features that are well known, for example, the latency difference associated with different retinogeniculate pathways, the activity level differences associated with input layers, and the action potential shape differences associated with white vs. gray matter. These have been used for decades as indicators of depth and location of recordings in visual cortex as electrodes were carefully advanced. High density electrodes allow this type of data to now be collected in parallel, but at discrete, regular sampling points. Rather than showing examples of what is already accepted, the emphasis should be placed on developing a rigorous analysis of how variable vs. reproducible are quantitative metrics of these features across penetrations, as a function of distance or functional domain, and from animal to animal. Ultimately, a more quantitative approach to the question of consistency is needed to assess the value of the methods proposed here.
We thank the reviewer for suggesting the addition of quantitative metrics to allow more substantive comparisons between various measures within and between penetrations. We have added quantification and describe this in the context of more specific comments made by this reviewer. We have retained descriptions of metrics that are well established because they provide an important validation of our approaches and laminar assignments.
Another important piece of information for assessing the ability to determine layers from spiking activity is to carry out post-mortem histological processing so that the layer determination made in this paper could be compared to anatomical layering.
We are not aware of any approach that would provide such information at sufficient resolution. For example, it is well known that electrolytic lesions often do not match to the locations expected from electrophysiological changes observed with single electrodes. As noted above, our observation that the laminar changes in electrophysiology precisely match the known widths and relative positions of all the layers of V1, including layer 4A, provides confidence in our laminar assignments.
On line 162, the text states that there is a clear lack of consistency across penetrations, but why should there be consistency: how far apart in the cortex were the penetrations? How long were the electrodes allowed to settle before recording, how much damage was done to tissue during insertion? Do you have data taken over time - how consistent is the pattern across several hours, and how long was the time between the collection of the penetrations shown here?
Answers to most of these questions can be found within the manuscript text. We have added text describing distance between electrode penetrations (at least 1mm, typically far more) and added a figure which shows a map of the penetration locations. The Methods section describes electrode penetration methods to minimize damage and settling times of penetrations. Data are provided regarding changes in recordings over time (see Methods, Drift Correction). The stimuli used to generate the data described are presented within a total of 30 minutes or less, minimizing any changes that might occur due to electrode drift. There is a minimum of 3 hours between different penetrations from the same animal.
The impact of the paper is lessened because it emphasizes consistency but not in a consistent manner. Some demonstrations of consistency are shown for CSDs, but not quantified. Figure 4A is used to make a point about consistency in cell density, but across animals, whereas the previous text was pointing out inconsistency across penetrations. What if you took a 40 or 60 um column of tissue and computed cell density, then you would be comparing consistency across potentially similar scales. Overall, it is not clear how all of these different metrics compare quantitatively to each other in terms of consistency.
As noted above, we have now added quantitative comparisons of consistency between different metrics. It is unclear why the reviewer felt that we use Figure 4A to describe consistency. That figure was a photograph from a previous publication simply showing the known differences in neuron density that are used to define layers in anatomical studies. This was intended to introduce the reader to known laminar differences. At any rate, we have been unable to contact the previous publishers of that work to obtain permission to use the figure. So we have removed that figure as it is unnecessary to illustrate the known differences in cell density that are used to define layers. We have kept the citation so that interested readers can refer to the publication.
In many places, the text makes assertions that A is a consistent indicator of B, but then there appear to be clear counterexamples in the data shown in the figures. There is some sense that the reasoning is relying too much on examples, and not enough on statistical quantities.
Without reference to specific examples we are not able to address this point.
Overall
Overall, this paper makes a solid argument in favor of using action potentials and stimulus driven responses, instead of CSD measurements, to assign cortical layers to electrode contacts in V1. It is nice to look at the data in this paper and to read the authors' highly educated interpretation and speculation about how useful such measurements will be in general to make layer assignments. It is easy to agree with much of what they say, and to hope that in the future there will be reliable, quantitative methods to make meaningful segmentations of neurons in terms of their differentiated roles in cortical computation. How much this will end up corresponding to the canonical layer numbering that has been used for many decades now remains unclear.
Reviewer #3 (Public Review):
Summary:
Zhang et al. explored strategies for aligning electrophysiological recordings from high-density laminar electrode arrays (Neuropixels) with the pattern of lamination across cortical depth in macaque primary visual cortex (V1), with the goal of improving the spatial resolution of layer identification based on electrophysiological signals alone. The authors compare the current commonly used standard in the field - current source density (CSD) analysis - with a new set of measures largely derived from action potential (AP) frequency band signals. Individual AP band measures provide distinct cues about different landmarks or potential laminar boundaries, and together they are used to subdivide the spatial extent of array recordings into discrete layers, including the very thin layer 4A, a level of resolution unavailable when relying on CSD analysis alone for laminar identification. The authors compare the widths of the resulting subdivisions with previously reported anatomical measurements as evidence that layers have been accurately identified. This is a bit circular, given that they also use these anatomical measurements as guidelines limiting the boundary assignments; however, the strategy is overall sensible and the electrophysiological signatures used to identify layers are generally convincing. Furthermore, by varying the pattern of visual stimulation to target chromatically sensitive inputs known to be partially segregated by layer in V1, they show localized response patterns that lend confidence to their identification of particular sublayers.
The authors compellingly demonstrate the insufficiency of CSD analysis for precisely identifying fine laminar structure, and in some cases its limited accuracy at identifying coarse structure. CSD analysis produced inconsistent results across array penetrations and across visual stimulus conditions and was not improved in spatial resolution by sampling at high density with Neuropixels probes. Instead, in order to generate a typical, informative pattern of current sources and sinks across layers, the LFP signals from the Neuropixels arrays required spatial smoothing or subsampling to approximately match the coarser (50-100 µm) spacing of other laminar arrays. Even with smoothing, the resulting CSDs in some cases predicted laminar boundaries that were inconsistent with boundaries estimated using other measures and/or unlikely given the typical sizes of individual layers in macaque V1. This point alone provides an important insight for others seeking to link their own laminar array recordings to cortical layers.
They next offer a set of measures based on analysis of AP band signals. These measures include analyses of the density, average signal spread, and spike waveforms of single- and multi-units identified through spike sorting, as well as analyses of AP band power spectra and local coherence profiles across recording depth. The power spectrum measures in particular yield compact peaks at particular depths, albeit with some variation across penetrations, whereas the waveform measures most convincingly identified the layer 6-white matter transition. In general, some of the new measures yield inconsistent patterns across penetrations, and some of the authors' explanations of these analyses draw intriguing but rather speculative connections to properties of anatomy and/or responsivity. However, taken as a group, the set of AP band analyses appear sufficient to determine the layer 6-white matter transition with precision and to delineate intermediate transition points likely to correspond to actual layer boundaries.
Strengths:
The authors convincingly demonstrate the potential to resolve putative laminar boundaries using only electrophysiological recordings from Neuropixels arrays. This is particularly useful given that histological information is often unavailable for chronic recordings. They make a clear case that CSD analysis is insufficient to resolve the lamination pattern with the desired precision and offer a thoughtful set of alternative analyses, along with an order in which to consider multiple cues in order to facilitate others' adoption of the strategy. The widths of the resulting layers bear a sensible resemblance to the expected widths identified by prior anatomical measurements, and at least in some cases there are satisfying signatures of chromatic visual sensitivity and latency differences across layers that are predicted by the known connectivity of the corresponding layers. Thus, the proposed analytical toolkit appears to work well for macaque V1 and has strong potential to generalize to use in other cortical regions, though area-targeted selection of stimuli may be required.
Weaknesses:
The waveform measures, and in particular the unit density distribution, are likely to be sensitive to the criteria used for spike sorting, which differ widely among experimenters/groups, and this may limit the usefulness of this particular measure for others in the community. The analysis of detected unit density yields fluctuations across cortical depth which the authors attribute to variations in neural density across layers; however, these patterns seemed particularly variable across penetrations and did not consistently yield peaks at depths that should have high neuronal density, such as layer 2. Therefore, this measure has limited interpretability.
While we agree that our electrophysiological measure of unit density does not strictly reflect anatomical neuronal density, we would like to remind the reader that we use this measure only to roughly estimate the correspondence between changes in density and likely layer assignments. We rely on other measures (e.g. AP power, AP power changes in response to visual stimuli) that have sharp borders and more clear transitions to assign laminar boundaries. Further, as noted in the reviewer’s list of strengths, the laminar assignments made with these measures are cross validated by differences in response latencies and sensitivity to different types of stimuli that are observed at different electrode depths.
More generally, although the sizes of identified layers comport with typical sizes identified anatomically, a more powerful confirmation would be a direct per-penetration comparison with histologically identified boundaries. Ultimately, the absence of this type of independent confirmation limits the strength of their claim that veridical laminar boundaries can be identified from electrophysiological signals alone.
As we have noted in response to similar comments from other reviewers, we are not aware of a method that would make this possible with sufficient resolution.
Recommendations for the authors:
Reviewing Editor (Recommendations For The Authors):
The reviewers have indicated that their assessment would potentially be stronger if their advice for quantitative, statistically validated comparisons was followed, for example, to demonstrate variability or consistency of certain measures that are currently only asserted. Also, if available, some histological confirmation would be beneficial. It was requested that the use and modification of the layering from Balaram & Kaas is addressed, as well as dealing with inconsistencies in the scale bars on those figures. There are two figure permission issues that need to be resolved prior to publication: Balaram & Kaas 2014 in Fig 1A, Kelly & Hawken 2017 in Fig. 4A.
Please see detailed responses to reviewer comments below. We have added new supplemental figures to quantitatively compare variability among metrics. As noted above, the suggested addition of data linking the electrophysiology directly to anatomical observations of laminar borders from the same electrode penetration is not feasible. The figure reused in Figure 1A is from open-access (CC BY) publication (Balaram & Kaas 2014). After reexamining the figure in the original study, we found that the inferred scale bar would give an obviously inaccurate result. So, we decided to remove the scale bar in Figure 1A. We haven’t received any reply from Springer Nature for Figure 4A permission, so we decided to remove the reused figure from our article (Kelly & Hawken 2017).
Reviewer #1 (Recommendations For The Authors):
Figure 4A has a different scale to Figure 4B-4F. It is better to add dashed lines to indicate the relationship between the cortical layers or overall range from Figure 4A to the corresponding layers in 4B to 4F.
The reused figure in Figure 4A is removed due to permission issue. See also comments above.
Reviewer #2 (Recommendations For The Authors):
General comments
This paper demonstrates that voltage signals in frequency bands higher than those used for LFP/CSD analysis can be used from high-density electrical contact recording to generate a map of cortical layering in macaque V1 at a higher spatial resolution than previously attained.
My main concern is that much of this paper seems to show that properties of voltage signals recorded by electrodes change with depth in V1. This of course is well known and has been mapped by many who have advanced a single electrode micron-by-micron through the cortex, listening and recording as they go. Figure 4 shows that spike shapes can give a clear indication of GM to WM borders, and this is certainly true and well known. Figures 5 and 6 show that activity level on electrodes can indicate layers related to LGN input, and this is known. Figure 7 shows that latencies vary with layer, and this is certainly true as we know. A main point seems to be that CSD is highly inconsistent. This is important to know if CSD is simply never going to be a good measure for layering in V1, but it would require quantification and statistics to make a fair comparison.
We are glad to see that the reviewer understands that changes in electrical signals across layers are well known and are expected to have particular traits that change across layers. We do not claim that have discovered anything that is unexpected or unknown. Instead, we introduce quantitative measures that are sensitive to these known differences (historically, often just heard with an audio monitor e.g. “LGN axon hash”). While the primary aim of this paper is not to show that Neuropixels probes can record some voltage signal properties that cannot be recorded with a single electrode before, we would like to point out that multi-electrode arrays have a very different sampling bias and also allow comparisons of simultaneous recordings across contacts with known fixed distances between them. For example our measure of “unit spread” could not be estimated with a single electrode.
We’ve added Figure S3 to show quantitative comparison of variation between CSD and AP metrics. These figures add support to our prior, more anecdotal descriptions showing that CSDs are inconsistent and lack the resolution needed to identify thin layers.
Some things are not explained very clearly. Like achromatic regions, and eye dominance - these are not quantified, and we don't know if they are mutually consistent - are achromatic/chromatic the same when tested through separate eyes? How consistent are these basic definitions? How definitive are they?
The quantitative definitions of achromatic region/COFD and eye dominance column can be found in our previous paper (Li et al., 2022) cited in this article. The main theme of this study is to develop a strategy for accurately identifying layers, the more detailed functional analysis will be described in future publications.
Specific comments
The abstract refers to CSD analysis and CSD signals. Can you be more precise - do you aim to say that LFP signals in certain frequency bands are already known to lack spatial localization, or are you claiming to be showing that LFP signals lack spatial resolution? A major point of the results appears to be lack of consistency of CSD, but I do not see that in the Abstract. The first sentence in the abstract appears to be questionable based on the results shown here for V1.
We have updated the Abstract to minimize confusion and misunderstanding.
Scale bar on Fig 1A implies that layers 2-5 are nearly 3 mm thick. Can you explain this thickness? Other figures here suggest layers 1-6 is less than 2 mm thick. Note, in a paper by the same authors (Balaram et al) the scale bar (100 um, Figure 4) on similar macaque tissue suggests that the cortex is much thinner than this. Perhaps neither is correct, but you should attempt to determine an approximately accurate scale. The text defines granular as Layer 4, but the scale bar in A implies layer 4 is 1 mm thick, but this does not match the ~0.5 mm thickness consistent with Figure 1E, F. The text states that L4A is less then 100 um thick, but the markings and scale bar in Figure 1A suggests that it could be more than 100 um thick.
We thank the reviewer for pointing out that there are clearly errors in the scale bars used in these previously published figures from another group. In the original figure 1(Balaram & Kaas 2014), histological slices were all scaled to one of the samples (Chimpanzee) without scale bar. After reexamining the scale bar we derived based on figure 2 of the original study, we found the same problem. Since relative widths of layers are more important than absolute widths in our study, we decided to remove the scale bar that we had derived and added to the Figure 1A.
Line 157. Fix "The most commonly visual stimulus"
Text has been changed
Line 161. Fix "through dominate eye"
Text has been changed
Line 166. Please specify if the methods established and validated below are histological, or tell something about their nature here.
The Abstract and Introduction already described the nature of our methods
Line 184. Text is mixing 'dominant' and 'dominate', the former is better.
Text has been changed accordingly
Line 188. Can you clarify "beyond the time before a new stimulus transition". Are you generally referring to the fact that neuronal responses outlast the time between changes in the stimulus?
That is correct. We are referring to the fact that neuronal responses outlast the time between changes in the stimulus. We have edited the text for clarity.
Line 196. Fix "dominate eye" in two places.
Text has been changed
Line 196. The text seems to imply it is striking to find different response patterns for the two eyes, but given the OD columns, why should this be surprising?
Since we didn’t find systematic comparison for CSD depth profiles of dominant/non-dominant eyes, or black/white in the past studies, we just describe what we saw in our data. The rational for testing each eye is that it is known that LGN projections from two eyes remain separated in direct input layer of V1, so comparing CSDs from two eyes could potentially help identifying input layers, such as L4C. Here we provide evidence showing that CSD profiles from two eyes deviate from naive expectations. For example, CSDs from black stimulus show less variation between two eyes, whereas CSDs from white stimulus could range from similar profile to drastically different ones across eyes.
Line 198. Text like, "The most consistent..." is stating overall conclusions drawn by the authors before pointing the reader specifically to the evidence or the quantification that supports the statement.
We’ve adjusted the text pointing to Figure S2, where depth profiles of all penetrations are visualized, and a newly added Figure S3, where the coefficients of variation for several metric profiles were shown.
Line 200. "white stimulus is more variable" - the text does not tell us where/how this is supported with quantitative analysis/statistics.
We’ve adjusted the text pointing to Figure S2, S3
The metric in 4B is not explained, the text mentions the plot but the reader is unable to make any judgement without knowledge of the method, nor any estimate of error bars.
The figure is first mentioned in section: Unit Density, and text in this section already described the definition of neuron density and unit density. We’ve also modified the text pointing to the method section for details.
Line 236. The text states the peak corresponds to L4C, but does not explain how the layer lines were determined.
As described early in the CSD section, all layer boundaries are determined following the guide which layouts the strategy for how to draw borders by combining all metrics.
At Line 296 the spike metrics section ends without providing a clear quantification of how useful the metrics will be. It is clear that the GM to WM boundary can be identified, but that can be found with single electrodes as well, as neurophysiologists get to see/hear the change in waveform as the electrode is advanced in even finer spatial increments than the 20 um spacing of the contacts here.
The aim of this study is to develop an approach for accurately delineating layers simultaneously. The metrics we explored are considered estimation of well-known properties, so they can provide support for the correctness we hope to achieve. Here we first demonstrate the usefulness and later show the average across penetrations (Figure 9C-F). We are less concerned in quantification of how different factors affect precision and consistency of these metrics or how useful a single metric is, but rather, as described in the guide section, whether we can delineate all layers given all metrics.
Line 302-306. Why this statement is made here is unclear, it interrupts the flow for a reason that perhaps will be explained later.
This statement notes the insensitivity of this measure to temporal differences, introducing the value of incorporating a measure of how AP powers changes over time in the next section of the manuscript.
Line 311. What is the reason to speculate about no canceling because of temporal overlap? Are you assuming a very sparse multi unit firing rate such that collisions do not happen?
Here we describe a simple theoretical model in which spike waveforms only add without cancelling, then the power would be proportional to the number of spikes. In reality, spike waveform sometimes cancels causing the theoretical relationship to deteriorate to some degree.
Lines 327-346. There is a considerable amount of speculation and arguing based on particular examples and there is a lack of quantification. Neuron density is mentioned, but not firing rate. would responses from fewer neurons with higher firing rate not be similar to more neurons with lower firing rates?
According to the theoretical model we described, power is proportional to numbers of spikes which then depend on both neuron density and firing rate. So fewer neurons with higher firing rate would generate similar power to more neurons with lower firing rate. We’ve expanded the explanation of the model and added Figure S4 about the depth profile of firing rate. Text has also been adjusted pointing to the Figure S2, S3 about quantitively comparisons of variability.
Line 348 states there is a precise link between properties and cortical layers, but the manuscript has not, up to this point, shown how that link was determined or quantified it.
Through our generative model of power and the similarity between depth profile of firing rate and depth profile of neuron density (Figure S4), depth profile of power can be used to approximate depth profile of neuron density which is known to be closely correlated to cortical layering.
Line 350. What is meant by "stochastic variability"?
The text essentially says distances from electrode contact to nearby cell bodies were random, so closer cells have higher spike amplitudes and in turn result in higher power on a channel.
The figures showing the two metrics, Pf and Cf, should be shown for the same data sets. The markings indicate that Fig 5 and Fig 6 show results from non-overlapping data sets. This does not build confidence about the results in the paper.
Here we use typical profiles to demonstrate the characteristics of the power spectrum/coherence spectrum because of the variation across penetrations. We show later, in the guide section, all metrics for one penetration (another two cases in supplemental figures) and how to combine all metrics to derive layer delineations.
Line 375 the statement is somewhat vague, "there are nevertheless sometimes cases where they can resolve uncertainties," can you please provide some quantitative support?
We provided 3 examples in Figure 6, and more examples are shown in Figure 8, Figure S5, S6.
Line 379. I believe the change you want to describe here is a change associated with a transition in the visual stimulus. It would be good to clarify this in the first several sentences here. Baseline can mean different things. I got the impression that your stimuli flip between states at a rate fast enough that signals do not really have time to return to a baseline.
We rephrased the sentence to describe the metric more precisely. A pair of uniform colors flipping in 1.5 second intervals is usually long enough for spiking activities to decay to a saturated level.
This section (379 - 398) continues a qualitative show-and-tell feel. There appears to be a lot of variability across the examples in Figure 7. How could you try to quantify this variability versus the variability in LFP? And, in this section overall, the text and figure legend don't really describe what the baseline is.
Text adjustments are made to briefly describe the baseline window and point to the Method section where definitions are described in detail. We’ve added Figure S3 together with Figure S2 to address the variability across penetrations, stimuli, and metrics.
Line 405 - 415. The discussion here does not consider that layers may not have well defined boundaries, the text gives the impression that there is some ultimate ground truth to which the metrics are being compared, but that may not be accurate.
Except for a few layers/sublayers, such as L2, L3A, L3B, most layer boundaries of neocortex are well defined (Figure 1A) and histological staining of neurons/density and correlated changes in chemical content show very sharp transitions. The best of these staining methods is cytochrome oxidase, which shows sharp borders at the top and bottom of layer 4A, top and bottom of layer 4C, and the layer 5/6 border. There is also a sharp transition in neuronal cell body size and density at the top and bottom of layer 4Cb. The definition and delineation of all possible layers are constantly being refined, especially by accumulated knowledge of genetic markers of different cell types and connection patterns. In our study, we develop metrics to estimate well known anatomical and functional properties of different layers. We have also discussed layer boundaries that have been ambiguous to date and explained the reason and criteria to resolve them.
Line 423. The text references Figure 1A in stating that relative thickness and position is crucial, but FIgure 1A does not provide that information and does not explain how it might be determined, or how much of a consensus there is. Also, the text does not consider that the electrode may go through the cortex at oblique angles, and not the same angle in each layer, and the relative thickness may not be a dependable reference.
There are numerous studies that describe criteria to delineate cortical layers, the referenced article (Balaram & Kaas 2014) is used here as an example. We are not aware of any publication that has systematically compared the relative thickness of layers across the V1 surface of a given animal or across animals. Nevertheless, it is clear from the literature that there is considerable similarity across animals. Accordingly, we cannot know what the source of variability in overall cortical thickness in our samples is, but we do see considerable consistency in the relative thickness of the layers we infer from our measures. We illustrate the differences that we see across penetrations and consider likely causes, such as the extent to which the coverslip pressing down on the cortex might differentially compress the cortex at different locations within the chamber.
The angle deviation of probe from surface will not change the relative thickness of layers, and the rigid linear probe is unlikely to bend in the cortex.
Line 433. The term "Coherence" is used, clarify is this is you Cf from Figure 6. The text states, "marked decrease at the bottom of layer 6". Please clarify this, I do not see that in Figure 6.
Text has been adjusted.
In Figure 6, the locations of the lines between L1 and 2 do not seem to be consistent with respect to the subtle changes in light blue shading, across all three examples, yet the text on line 436 states that there is a clear transition.
We feel that the language used accurately reflects what is shown in the figure. While the transition is not sharp, it is clear that there is a transition. This transition is not used to define this laminar border. We have edited the text to clarify that the L1/2 border is better defined based on the change in AP power which shows a sharp transition (Figure 7).
The text states that the boundary is also "always clear" from metrics... and sites Figure 5, but I do not see that this boundary is clear for all three examples in Figure 5.
Text has been adjusted.
Line 438. The text states that "it is not unusual for unit density to fall to zero below the L1/2 border (Figure 8E)", but surprisingly, the line in Figure 8 E does not even cover the indicated boundary between L1 and L2.
At this point, the number of statements in the text that do not clearly and precisely correlate to the data in the figures is worrisome, and I think you could lose the confidence of readers at this point.
We do not see any inconstancy between what is stated in our text and what is noted by the reviewer. The termination of the blue line corresponds to the location where no units are detected. This is the location where “unit density falls to zero”. In this example, no units resolved through spike sorting until ~100mm beneath the L1/L2 boundary, which is exactly zero unity density (Figure 8E). That there are electrical signals in this region is clear from the AP power change (Figure 8C) which also shows the location of the L1/L2 border.
Line 448. Text states that the 6A/B border is defined by a sharp boundary in AP power, but Figure 8A "AP power spectrum" does not show a sharp change at the A/B line. There is a peak in this metric in the middle to upper middle of 6A, but nothing so sharp to define a boundary between distinct layers, at least for penetration A2.
Text has been adjusted.
In Figure 8, the layer labels are not clear, whereas they are reasonably clear in the other figures.
This is a technical problem regarding vector graphics that were not properly converted in PDF generation. We will upload each high-quality vector graphics when we finalize the version of record.
The text emphasizes differences in L4B and L4C with respect to average power and coherence, but the transition seems a bit gradual from layer 3B to 4C in some examples in Figure 6. And in Figure 5, A3, there doesn't appear to be any particular transition along the line between 4B and 4C.
In this guide section, we pointed out early that some metrics are good for some boundaries and variation exists between penetrations. We’ve expanded text emphasizing the importance of timing differences in DP/P for differentiating sublayers in L4. Lastly, in case of several unresolvable boundaries given all the metrics, the prior knowledge of relative thickness should be used.
Line 466 provides prescriptions in absolute linear distances, but this is unwise given that cortex may be crossed at oblique angles by electrodes, particularly for parts of V1 that are not on the surface of the brain. Other parts of the text have emphasized relative measurements.
Text has been changed using relative measurements.
Line 507. The text says 9C and 4A are a good match, but the match does not look that good (4A has substantial dips at 0.5 and 0.75, and substantial peaks), and there is no quantification of fit. The error bars on 9C do not help show the variability across penetrations, they appear to be SEM, which shows that error bars get smaller as you average more data. It would seem more important to understand what is the variance in the density from one penetration to the next compared to the variance in density across layers.
We have replaced “good match” with “roughly corresponds”. We note that we do not use unit density as a metric for identification of laminar borders and instead show that the expected locations of layers with higher neuronal density correspond to the locations where there are similar changes in unit density. It should be noted that Figure 9C is an average across many penetrations so should not be expected to show transitions that are as sharp in individual penetrations. Because of the figure permission issue, we have removed Figure 4A, and changed the text accordingly.
Figure 9C-F show a lot of variability in the individual curves (dim gray lines) compared to the overall average. Does this show that these metrics are not reliable indicators at the level of single penetration, but show some trends across larger averages?
In the beginning of the guide, we emphasized that all metrics should be combined for individual penetration, because some metrics are only reliable for delineating certain layer boundaries and the quality of data for the various measures varies between penetrations. The penetration average serves the same purpose explained in the previous question as an indicator that our layer delineation was not far off.
The discussion mentions improvements in layer identification made here. Did this work check the assignments for these penetration against assignments made based on some form of ground truth? Previous methods would advance electrodes steadily, and make lesions, and carry out histology. Is there any way to tell how this method would compare to that?
Even electrolytic lesions do not necessarily reveal ground truth and can be quite misleading. And their resolution is limited by lesion size. Lesions are typically variable in size, asymmetric and have variable shape and position relative to the location of the electrode tip, likely affected by the quality and location of electrical grounding and variations in current flow due to locations of blood vessels. A review of the published literature with electrode lesions shows that electrophysiological transitions are likely a far more accurate indicator of recording locations than post-mortem histology from electrolytic lesions. It is extremely rare for the locations of lesions to be precisely aligned to expected laminar transitions. See for example Chatterjee et al (Nature 2004). Also see several manuscripts from the Shapley lab. The lone rare exception of which we are aware is Blasdel and Fitzpatrick1984 in which consistently small and round lesions were produced and even these would be too large (~100 microns) to accurately identify layers if it were not for the fact that the electrode penetrations were very long and tangential to the cortical layers.
Reviewer #3 (Recommendations For The Authors):
- The authors say (lines 360-362) that "Assuming spikes of a neuron spread to at least two adjacent recording channels, then the coherence between the two channels would be directly proportional to number of spikes, independent of spike amplitude." Has this been demonstrated? Very large amplitude spikes should show up on more channels than small amplitude spikes. Do waveform amplitudes and unit densities from the spike waveform analyses show consistent relationships to the power and/or coherence distributions over depth across penetrations?
This part of the manuscript is providing a theoretical rational for what might be expected to affect the measures that we have derived. That is why we begin by stating that we are making an assumption. The answers to the reviewer’s questions are not known and have not been demonstrated. By beginning with this theoretical preface, we can point to cases where the data match these expectations as well as other cases where the data differ from the theoretical expectations.
Coherence, by definition, is a normalized metric that is insensitive to amplitude. Spike amplitude mainly depends on how close the signal source is to electrode, and spike spread mainly depends on cell body size and shape given the same distance to electrode. Therefore, a very large spike amplitude could stem from a very close small cell to electrode, but would result in a small spike spread, especially axonal spikes (Figure 4B, red spike). Spike amplitudes on average are higher in L4C which matches the expectation that higher cell density would result, on average, closer cell body to electrode (Figure S4A). Nonetheless, the high-density small cell bodies in L4C result in a small spike spread (Figure 9D).
- I suggest clarifying what is defined as the baseline window for the ΔP/P measure - is it the entire 10-150 ms response window used for the power spectrum analysis?
Text adjustments are made in the Methods where the time windows are defined at the beginning of the CSD section. Only temporal change metrics (ΔCSD and ΔP/P) use the baseline window ([-40, 10]ms). The other two spectrum metrics (Power and Coherence) use the response window ([10, 150]ms).
- Firing rate differs by cell type and, on average, differs by layer in V1. Many layer 2/3 neurons, for example, have low maximum firing rates when driven with optimized achromatic grating stimuli. To the extent that the generative models explaining the sources of power and coherence signals rely on the assumption that firing rates are matched across cortical depth, these models may be inaccurate. This assumption is declared only subtly, and late in the paper, but it is relevant to earlier claims.
Text adjustments are made to explicitly describe the possibility that uneven depth profile of firing rate could counteract the depth profile of neuron density, resulting distorted or even a flat depth profile of power/coherence that deviates far from the depth profile of neuron density. In a newly added Figure S4, we first show the average firing rate profile during a set of stimuli (uniform color, static/drifting, achromatic/chromatic gratings), then specifically the PSTHs of the same stimuli shown in this study. It can be seen that layers receiving direct LGN inputs tend to fire at a higher rate (L4C, L6A). Firing rates in the PSTHs either roughly match across layers or are much higher in the densely packed layers. Therefore, the depth profile of firing rate contributes to rather than counteracting that of neuron density, enhancing the utility of the power/coherence profile for identification of correct layer boundaries.
- Given the acute preparation used for recordings, I wonder whether tissue is available for histological evaluation. Although the layers identified are generally appropriate in relative size, it would be particularly compelling if the authors could demonstrate that the fraction of the cortical thickness occupied by each layer corresponded to the proportion occupied by that layer along the probe trajectory in histological sections. This would lend strength to the claim that these analyses can be used to identify layers in the absence of histology. Furthermore, variations in apparent cortical thickness could arise from different degrees of deviation from surface normal approach angles, which might be apparent by evaluation of histological material. I would add that variation in thickness on the scale shown in Fig. S4 is more likely to have an explanation of this kind.
To serve other purposes unrelated to this study (identification of CO blobs), we cut the postmortem tissue in horizontal slices, so the histological comparison suggested cannot be made. The cortical thickness measured in this study had been affected not only by the angle deviation from the surface normal but also the swelling and compression of cortex. Nevertheless, evaluating the absolute thickness of cortex is not the main purpose of this study.
Text and figure suggestions:
- Fig 1A has been modified from Balaram & Kaas (2014) to revert to the Brodmann nomenclature scheme they argue against using in that paper; I wonder if they would object to this modification without explanation. Related, in the main text the authors initially refer to layers using Brodmann's labels with a secondary scheme (Hassler's) in parentheses and later drop the parenthetical labels; these conventions are not described or explained. Readers less familiar with the multiple nomenclature schemes for monkey V1 layers might be confused by the multiple labels without context, and could benefit from a brief description of the convention the authors have adopted.
Throughout our article, we only used Brodmann’s naming convention because it has historically been adopted for old world monkey which we use in our study, whereas Hassler’s naming convention is more commonly used for new world monkey. Different naming conventions do not change our result, and it is out of scope for our study to discuss which nomenclature is more appropriate.
- References to "dominate eye" throughout the text and figure legends should be replaced with "dominant eye."
It has been changed throughout the article.
- It is a bit odd to duplicate the same example in Fig. 2C and 2E. Perhaps a unique example would be a better use of the space.
Here we first demonstrate the filtering effect, then compare profiles across different penetrations. The same example bridges the transition allowing side-by-side comparison.
- The legend for Fig. 3 might be clearer if it simply listed the stimulus transitions for each column left to right, i.e. "black to white (non-dominant eye), white to black (non-dominant eye), black to white (dominant eye), ..."
We feel that the icons are helpful. Here we want to show the stimulus colors directly to readers.
- The misalignment between Fig. 4A vs. 4B-F, combined with the very small font size of the layer labels in Fig. 4B-F, make the visual comparison difficult. In Figs. 7 and 8, layer labels (and most labels in general) are much too small and/or low resolution to read easily. Overall, I would recommend increasing font size of labels in figures throughout the paper.
The reused figure in Figure 4A is removed due to permission issue. Font sizes are adjusted.
- Line 591 "using of high-density probes" should be "using high-density probes"
Text has been changed accordingly
