Science Update news 2/14/17

Check out this article from the NeuroNexus Science Update, describing exciting research done using NeuroNexus products.
First Author: German Mendoza
Last Author: Hugo Merchant
Other authors: Adrien Peyrache, Jorge Gamez, Luis Prado, Gyorgi Buzsaki

Animal Model:  Monkey daily chronic recording experiment
Focus:  Pre-motor cortex area
Products Used:  Primate probes, penetrating electrode arrays
This paper evaluates a technique to semichronically record the cortical extracellular neural activity in the behaving monkey using NeuroNexus electrodes. Chronic microdrives were used that allow varying the depth of recording locations after implantation surgery. Exracellular unit activity was recorded from diferent depths of the presupplementary motor cortex (pre-SMA) of rhesus monkey that was trained in a tapping task. Cells were classified as putative pyramidal cells or interneurons. This work validated the semichronic technique as a viable option for large-scale parallel recordings of local circuit activity in the cortex of large animals.

Feature: György Buzsáki

Published July 14, 2014

buszaki-profileDrs Antal Berényi and György Buzsáki recently published a paper titled "Large-scale, high-density (up to 512 channels) recording of local circuits in behaving animals."

This paper describes a system that allows high-channel-count recordings from a small volume of neuronal tissue using a lightweight signal multiplexing headstage that permits free behavior of small rodents. The system integrates multishank, high-density recording silicon probes, ultraflexible interconnects, and a miniaturized microdrive. These improvements allowed for simultaneous recordings of local field potentials and unit activity from hundreds of sites without confining free movements of the animal.

NeuroNexus caught up Dr. Buzsáki to discuss recording using probes with very high densities of recording sites, such as the NeuroNexus Buzsaki256 probe.

Q: In the introduction to the paper it is stated that “monitoring a statistically representative fraction of neurons of the investigated circuits in behaving animals is a prerequisite for understanding neuronal computation.” In your opinion, what is the practical lower limit for the number of channels required to reach that “statistically representative fraction”? In a perfect world, is there an upper limit on how many sites you would like to record from?

This is of course a super important question and there is not a good answer to it. If one records from a homogenous population (eg., locus coeruleus), the fraction may be small and perhaps 1-5% is sufficient. On the other hand, in the neocortex, there is an enormous diversity of cell classes, and some types of interneurons are very rare. Thus, if one wants to have a ‘representative’ snapshot of how neurons interact recordings from even the rarest ones are needed. An additional complication for sampling eg., pyramidal neurons of the cortex is that their firing rates show a very strongly skewed (typically log-normal) distribution. While only 10% of the neurons do half of the job, understanding the remaining half requires sampling from very large numbers of neurons. The more the better.

Q: The Buzsaki256 probe has 256 sites within the active recording area that spans 1.6 mm (along the shanks) by 2.1 mm. In the discussion of the paper it is stated that “Denser recording sites without increasing the shank volume is desirable,” and you look on the horizon to probes with “>1000 site counts and 20 um site spacing.” What site density would you consider sufficient for unit recordings and cell clustering? What is the optimal minimum site spacing, beyond which you see no further benefit?

Designing an optimum probe depends on the scientific question. For reliably separating cortical pyramidal neurons our estimate is that 15-20 µm spacing is a good compromise between separability of neurons and probe size. One would like to have more but without increasing probe width and depth. The size of the probe is very critical. Most current NeuroNexus probes are too wide and far from the ideal. For the new 1000 site probes, I specified the 20 µm spacing out of compromise. My favorite probe for exploring cortical circuits would have <60 µm shanks, 20 µm spacing, staggered on the sides, and contain 250-300 sites to cover the entire depth of the neocortex. Multiple shanks with 150µm apart would be desirable, ideally with signal multiplexer integrated in the back end of the probe.

Q: The Buzsaki256 probe is a planar design with 32 sites on each of eight shanks. Do you see advantages to a 3-dimensional probe (with the ability to span both cortical columns and cortical layers) as opposed to the planar approach? Would the 3D approach loosen the density requirements, or would that remain the same?

3-d probes have merits. However, the number one rule in neurophysiology is to record from a healthy brain. Inserting an 8-shank 2-D probe requires quite a bit of surgical experience and despite best intentions, recordings are not always successful. The brain is extremely dense with blood vessels and ‘shooting’ a 3 –D probe into the tissue comes with costs. I have seen histology from many such 3-D arrays from rodents to primates and layer I and often layer 2 are invariably destroyed or compromised. Studying such mutilated cortical tissue is not my intention, although I understand that recording from units even from such compromised tissue can be useful for certain applications.

Q: Your lab has utilized the Buzsaki256 probe design in a published study entitled “Spatially Distributed Local Fields in the Hippocampus Encode Rat Position” that helped illustrate the need for large spatial coverage. Ultimately, what types of questions do high-channel, high-site-density probes allow you to answer that couldn’t otherwise be addressed?

Silicon probes have the great advantage for studying the mesoscopic local field potential signals. The 256-site probe was designed to accomplish this, allowing to record simultaneously the LFP signals from multiple regions, simultaneously with units (even if the unit yield with such probes is less powerful than from higher density version). Imaging methods have only one type of signal. Electrically one can record both neuronal output with spikes and their inputs with high resolution LFP, as well as looking at the relationship between individual components and their aggregate responses.

Feature: Chou P. Hung

Published June 1, 2014

ChouDrs Chou P. Hung, Chia-pei Lin, and Yueh-peng Chen recently published a paper titled "Tuning and spontaneous spike time synchrony share a common structure in macaque inferior temporal cortex."

NeuroNexus caught up Dr. Chou to discuss his work, techniques, and potential applications for humans.


Q: Your study was performed on primates. Do you think the findings are directly applicable for humans? If not, what are the differences that you might expect?

The short answer is, "yes." Previously, it was thought that the more differently cells behave, the more information they carry. Our study says that instead, it is the cells that behave more similarly, the ‘choristers,’ that carry the useful information. The other cells, the ‘soloists,’ may fine-tune the patterns. The dense spacing of these arrays, about 32 channels per cortical column, was critical to measuring this correlated activity. For ethical reasons, we can't make the same recordings in humans, but our preliminary data indicate that this activity in primates is linked to coarser signals in human functional imaging and to human psychophysics. Understanding these issues is important to unraveling how we learn and what goes wrong in mental diseases, where studies have found altered correlated activity in functional magnetic resonance imaging (fMRI) but the link to single neurons has not been made.

Q: Your study found that “tuning and spike synchrony were linked by a common spatial structure that is highly efficient for Object representation.” What is the next step in this research?

In autism and other brain diseases, brain imaging signals have different patterns, but it has been difficult to link these coarser patterns to mechanisms in single cells. It would be helpful if we had a better understanding of how different signal types are linked in the same animal. Also, how the brain recognizes visual information is an extremely challenging problem. It is much harder than reasoning. Understanding how the brain processes information is key to understanding how the brain creates intelligence. To do this, we would like to better understand what are the roles of the choristers versus the soloists. If our hypothesis is correct, the choristers and soloists should have different roles in learning and behavior. We would like to find the underlying principles and to apply them, at a suitable level of abstraction, in computational models. These are among the goals outlined in the Brain Initiative, and the dense spatiotemporal sampling enabled by these arrays is critical to meeting these goals.

Q: You were recording with 64-channel probes that had a planar design. In your perfect world, how many sites would you like to be able to record from simultaneously to get the optimal contrast/area coverage for similar visual studies?

Lin 2014 Tuning and spontaneous spike time synchrony share a common structure in macaque inferior temporal cortex


Having 64 channels in two cortical columns is already very good, because it lets us hear correlated single-unit activity that would be missed at coarser resolutions. Theoretically, the effects of correlated activity are amplified in densely connected, homogeneous populations. But, because neurons are very heterogeneous, even within a cortical column, dense sampling is necessary to hear the correlated activity. Our preliminary data indicate that at least ~8 channels per cortical column are needed to measure an effect of noise correlation on object coding, and the effect size increases at higher density. This threshold might vary across cortical areas - in our studies, about half the V1 neuronal pairs in an array were correlated, versus only about 6% in inferior temporal cortex, during spontaneous activity. This may have to do with the increasing complexity of the representation along the ventral visual pathway. In a perfect world, it might be good to increase the range of depth, because 1.4 mm isn't quite enough to sample all cortical layers simultaneously. But, there is a tradeoff in the signal quality as you increase the number of contacts.

Q: Would a 3-dimensional probe (with the ability to span both cortical columns and cortical layers) benefit your research? If so, in what way?

Definitely, a 3D array would help. Having a 3D array would give us a clearer picture of phenomena such as surround inhibition, and it would aid in efforts to link spiking activity to functional imaging in the same animal. Going to 3D may also improve chronic stability, so that we can track changes in the activity patterns during learning. Also, 3D would help with the issue of having sufficient spatial density, to 'hear' the neurons that are strongly interconnected and that 'care' about the animal's behavioral task.